Human Developmental Neurotoxicology

Human Developmental Neurotoxicology

edited by

David C. Bellinger Children’s Hospital Boston Harvard Medical School Harvard School of Public Health Boston, Massachusetts, U.S.A.

New York London

Taylor & Francis is an imprint of the Taylor & Francis Group, an informa business

Published in 2006 by Taylor & Francis Group 270 Madison Avenue New York, NY 10016 © 2006 by Taylor & Francis Group, LLC No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number-10: 0-8247-2988-9 (Hardcover) International Standard Book Number-13: 978-0-8247-2988-2 (Hardcover) Library of Congress Card Number 2005046676 This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. No part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC) 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data Human developmental neurotoxicity / edited by David C. Bellinger. p. ; cm. Includes bibliographical references and index. ISBN-13: 978-0-8247-2988-2 (alk. paper) ISBN-10: 0-8247-2988-9 (alk. paper) 1. Neurotoxicology. 2. Nervous system--Diseases. 3. Fetus--Diseases--Complications. 4. Infants (Newborn)--Effect of drugs on. 5. Neurotoxic agents. I. Bellinger, David. [DNLM: 1. Neurotoxicity Syndromes--etiology. 2. Child Development--drug effects. 3. Embryonic and Fetal Development--drug effects. WL 140 C63992 2006] RG627.C65 2006 618.3'2071--dc22

Taylor & Francis Group is the Academic Division of Informa plc.

2005046676

Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com

Preface

Know anything? I don’t even suspect anything! —Yogi Berra, when, following a particularly dismal performance on a test, a teacher asked, “Yogi, don’t you know anything?” As we know, There are known knowns. There are things we know we know. We also know There are known unknowns. That is to say We know there are some things We do not know. But there are also unknown unknowns, The ones we don’t know We don’t know. —Donald H. Rumsfeld, U.S. Secretary of Defense, news briefing, Feb. 12, 2002

Those who investigate the impacts of chemical exposures on children’s neurodevelopment are not quite in the same dire epistemological straits as Yogi Berra. We at least suspect quite a lot of things. Rumsfeld’s musings, although resembling a Zen koan in their obtuseness, are, nevertheless, more apt. Some questions have answers that most (although rarely all) observers would endorse. For example, if asked, “Is children’s neurodevelopment adversely affected at levels of lead exposure that are not high enough to cause an overt encephalopathy?” most investigators would probably answer affirmatively. In Rumsfeld’s system for classifying knowledge, this would be a “known known.” An increase in the specificity with which a question is phrased will, however, often result in responses that are more variable, both in content and the respondent’s level of certainty, suggesting that the issue in question would be better classified as a “known unknown.” “What is the functional form of the dose-response relationship?”, “To what extent do the expressions of nervous system toxicity depend on the age at which exposure occurs?” and “What is the natural history of lead-associated neurodevelopmental deficits under different degrees of postnatal environmental enrichment?” are examples of such questions. And then there are the “unknown unknowns.” These bear on issues that we won’t even think to ask questions about until we learn enough to appreciate the existence of a mystery that, heretofore, lay unrecognized. These are not issues solely of academic interest. In the past two decades, children’s abilities to process information, to reason, to learn, and to achieve a positive psychosocial adjustment have emerged as critical endpoints in risk assessments of chemical exposures. It is recognized that children with even subtle impairments of these skills, who cannot function near the peak of their potential, will not fare as well as those who do in a technological marketplace that places high value on analytic and communication skills and the ability to adapt quickly and effectively to shifting demands and opportunities. It is iii

iv

Preface

no longer adequate to make public health decisions based on risk assessments that use as critical endpoints frank neurological disease or mental retardation (to say nothing of endpoints such as the LD50–the dose at which 50% of animals in the “exposed” group die). Exposure standards will be insufficiently protective if based on using approaches that do not acknowledge the potential importance, both to the individual and to society, of “sub-clinical” effects, that is adverse impacts that are not severe enough to meet diagnostic criteria for a disease or that do not correspond to a pattern that is sufficiently common to have been canonized as a “diagnosis.” The increasing recognition of “subtle” effects of neurotoxicant exposures and their acceptance as worthy of concern has resulted in a steady evolution over the past 30 years in the study methodologies applied in human developmental neurotoxicology studies. Case series describing the severe neurological deficits of children who developed Congenital Minamata Disease after prenatal exposure to methyl mercury or who presented with a fulminant lead encephalopathy were sufficient to convince us of the serious neurotoxicities caused by high dose exposure to these metals. Studies that were more analytic, using a case-control design, became necessary when the goal was to determine whether the prevalence of clinically-defined childhood morbidities, for example diagnoses such as learning disabilities or attention deficit hyperactivity disorder, differed between children who were considered “exposed” or “unexposed” to some chemical. The effect measures applied in case-control studies were typically the odds ratio and relative risk, expressing the extent to which the morbidity was more likely to be present among children in the “exposed’ than the “unexposed” group. When concern began to shift to possible subclinical impairments, it was necessary to mount cohort studies, in which participants were selected on the basis of exposure status rather than outcome status, and outcomes were often represented dimensionally rather than as diagnoses that were noted as being simply present or absent. Prospective studies were recognized as preferable to cross-sectional studies because they permitted an investigator to establish the temporal precedence of exposure vis a vis outcome, to characterize the natural history of exposure-outcome associations, and to identify age-dependent variations in susceptibility. In contrast to the odds ratio or relative risk statistics calculated in case-control studies, the effect measures calculated in cohort studies were more often the rate of change in the dimensional outcome per unit increase in an exposure index, providing new options for modeling dose-response and dose-effect relationships and developing points of departure for risk assessments. Other recent trends in the evolution of human developmental neurotoxicology research are notable. At the same time that the endpoints considered important to society were broadened to include sub-clinical as well as clinical impairments within a particular health domain, the range of health domains of interest as possible targets of neurotoxicant exposures has also broadened. For a variety of reasons, neurotoxicological studies traditionally focused on cognitive morbidities, defined rather narrowly and often consisting solely of IQ, as the critical endpoints. Beginning in the 1980s, however, attention was drawn in the general pediatric literature to what was called “the new morbidity,” referring to behavioral disorders and maladaptive psychosocial function. A concern with such disorders is increasingly reflected in neurotoxicological studies, with diagnoses such as juvenile delinquency, attention deficit hyperactivity disorder, and autism spectrum disorder serving as the endpoints of interest. Another trend is increasing sophistication of the methods used to address the critical issue of confounding bias by characterizing more accurately and comprehensively the panoply of factors that, apart from the chemical exposure of interest, can affect a child’s health. Determining how a chemical exposure fits into the complex web of influences on child development, its effects perhaps being exacerbated by some of these other influences and mitigated by others, also

Preface

v

became an important part of the evaluation of study hypotheses. These trends resulted in studies that have become increasingly complex and multidisciplinary, requiring, at least, the collaboration of toxicologists, developmental psychologists, epidemiologists, biostatisticians, and analytical chemists. The contributions of psychiatrists, sociologists and cultural anthropologists will be increasingly important as the range of endpoints of interest continues to broaden and the child development models used to evaluate chemical effects become richer in depth and complexity. The purpose of this book is to describe the state-of-the-art in the design, conduct, and interpretation of human developmental neurotoxicology studies. Authors were asked to do more, however, than to describe current knowledge and methods in their fields of research. They were asked to address the question, “How can we do better?” and encouraged to identify the advances that need to be made in order to allow investigators to clarify the “known unknowns” and to identify the “unknown unknowns” that will be the foci of future research. Chapters in section one focus on specific environmental chemical exposures, including mercury, PCBs, lead, and solvents. Although the first three are among the chemicals that have been most intensively studied, many knowledge gaps remain, continuing to inspire debate and to render the risk assessment process contentious. Because of the current world-wide concern about mercury toxicity, two chapters are devoted to this metal. Although there is some overlap in the topics covered, the chapters provide complementary perspectives on the conduct and interpretation of mercury studies as well as on their public health implications. The chapters in section two focus on the developmental neurotoxicities associated with intentional exposures to chemicals or chemical mixtures. Some of these are medications administered therapeutically, such as anti-epileptic and chemotherapeutic drugs, while others are drugs used recreationally, including tobacco, alcohol, and cocaine. The potential adverse effects of exposures to recreational drugs generally receive much more attention than do the potential adverse effects of exposures to therapeutic drugs, perhaps due, at least in part, to differences in the risk-benefit calculus appropriate to these two classes of exposures. Exposure to therapeutic drugs occurs as a result of a decision that the avoidance of a health risk associated with a medical condition outweighs the risks associated with their use, which might be substantial (i.e., “side effects”). In contrast, use of recreational drugs, particularly during pregnancy, would not be expected to provide any health benefit to the fetus or child, which could make the risk of even subtle “side effects” on a child enough to tip the balance in favor of avoiding the exposure. Chapters in section three focus on critical issues in the assessment of exposure and outcome. Separate chapters focus on the special considerations germane to the assessment of exposures to accumulative chemicals, such as many metals, and to the assessment of exposures to chemicals with relatively short biological residence times, such as organophosphate pesticides. Another chapter focuses on special issues that pertain to characterizing voluntary exposures, such as marijuana and cocaine. The chapters focusing on outcome assessment address considerations in assembling, modifying, and validating a battery of tests, the emerging role for neuroimaging modalities in assessing neurotoxicity, and the special challenges faced in studying the contributions of childhood neurotoxicant exposures to the development of adult neurologic disease. Chapters in section four provide perspectives on several aspects of the analysis and interpretation of developmental neurotoxicity studies. These include the strategies used to identify potential confounding variables to include in regression models and the potential utility of analytic strategies such as structural equation modeling in characterizing the relationships among exposures, outcomes, confounders, and mediators. Another chapter

vi

Preface

shows how our understanding of exposure-outcome associations can be enriched by taking into account the broader social environment within a child that is developing and building multi-level models of complex inter-relationships. Two other chapters address issues pertaining to study interpretation, one focusing on how consideration of the experimental animal literature on a chemical’s developmental neurotoxicity can inform both the choice of methods used in human studies, and the inferences drawn about behavioral mechanisms and, ultimately, about causality. A chapter that focuses on the effects of chemical exposures on thyroid hormone signaling pathways illustrates the daunting distance we have to go to bridge the yawning chasm that separates the observations made in epidemiologic studies of developmental neurotoxicity and an understanding of the biological mechanisms generating those observations. The chapters in section five place the contributions of developmental neurotoxicity research in a broader context. Two chapters provide the perspectives of groups who could be characterized as “consumers” of such studies, namely those who use the data clinically to manage patients exposed to neurotoxic chemicals, and those who use the data as the basis for formulating public policy. They help us to understand ways in which this research can be designed, conducted, analyzed in ways that will make the data more useful to those who apply the findings. The final chapters alert us to the hazards that can arise in doing research that threatens industrial interests or that enriches investigators who place themselves in positions of conflict of interest. These caveats remind us of the quite profound effects that developmental neurotoxicity research have on people’s health and livelihoods, and the special responsibilities we assume in undertaking such work. David C. Bellinger

Contents Preface : : : : iii Contributors : : : : xv SECTION ONE: ENVIRONMENTAL CHEMICALS 1. Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Bernard Weiss Mercury as an Environmental Poison : : : : 1 Anticipation of the Current Debate : : : : 3 Minamata and Iraq : : : : 3 The Current Scene : : : : 6 Other Complications : : : : 8 The Risk Equation and Effect Modification : : : : 9 Inorganic Mercury : : : : 11 Delayed Toxicity : : : : 13 Postnatal Exposures in Humans : : : : 16 Unresolved Issues : : : : 18 References : : : : 19 2. Developmental Neurotoxicity Associated with Dietary Exposure to Methylmercury from Seafood and Freshwater Fish . . . . . . . . . . 25 Philippe Grandjean, Sylvaine Cordier, and Tord Kjellstro¨m Introduction : : : : 25 Prospective Study Designs and Settings : : : : 26 Cross-Sectional Studies : : : : 29 Exposure Assessment : : : : 32 Outcome Variables : : : : 34 Confounding Variables : : : : 36 Public Health Relevance : : : : 38 References : : : : 40

vii

viii

Contents

3. Effects of Developmental PCB Exposure on Neuropsychological Function in Epidemiological Studies: Issues and Research Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Deborah C. Rice Identification of Types of Behavioral Deficits Produced by PCB Exposure : : : : 44 Choice of Congeners and Tissue Compartment as Determinants of Exposure : : : : 49 Determination of Effects of Multiple Chemicals in Addition to PCBs : : : : 51 Contribution of Postnatal Effects to Neurotoxicity : : : : 54 Determination of the Relationship between Exposure and Effect : : : : 57 Thyroid Hormones as a Potential Mediator of Neurotoxic Effects : : : : 58 Summary and Overall Conclusions : : : : 58 References : : : : 60 4. Lead Neurotoxicity in Children: Knowledge Gaps and Research Needs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 David C. Bellinger Introduction : : : : 67 Selection of Exposure Biomarkers and Sources of Misclassification Errors : : : : 68 Selection of Critical Outcomes : : : : 71 Functional Form of the Concentration–Response/ Concentration–Effect Relationship : : : : 73 Reversibility : : : : 74 Effect Modification : : : : 75 Conclusion : : : : 77 References : : : : 77 5. Effects of Organic Solvents on Reproductive Outcome and Offspring Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 Christine Till, Gideon Koren, and Joanne Rovet Basic Principles : : : : 83 Organic Solvents in the Central Nervous System : : : : 87 Fetal Vulnerability to Organic Solvents : : : : 88 Reproductive and Developmental Effects in Animals Following Prenatal Exposure to Organic Solvents : : : : 89 Reproductive and Developmental Effects Following Occupational Exposure to Organic Solvents : : : : 90 Organic Solvent Inhalant Abuse : : : : 96 Problems Studying Reproductive Hazards in Human Studies : : : : 97

Contents

ix

Summary and Recommendations : : : : 98 References : : : : 99 SECTION TWO: MEDICINAL AND RECREATIONAL SUBSTANCE USE 6. The Structural and Functional Teratology of Antiepileptic Medications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 Jane Adams, Jennifer Anne Lantz Gavin, and Patricia A. Janulewicz Fundamentals of Teratology and Neurobehavioral Teratology : : : : 103 Research in the Early Years: The Role of Epilepsy as a Disease Contributing to Adverse Outcomes of Pregnancy : : : : 106 The Effects of Polytherapy Versus Monotherapy Treatments : : : : 107 The Teratogenic Effects of Monotherapy with Phenytoin, Phenobarbital, Carbamazepine, and Valproic Acid : : : : 116 Individualized Predictors of Susceptibility to Adverse Outcome that Can be Assessed During Pregnancy or at Birth : : : : 124 Future Research Needs : : : : 125 References : : : : 126 7. Chemotherapy Agents for Treatment of Acute Lymphoblastic Leukemia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 Christine Mrakotsky and Deborah P. Waber Introduction : : : : 131 Pathophysiology and Prevalence of ALL : : : : 132 Evolution of Treatment Protocols : : : : 132 Chemotherapy Agents: Efficacy and Toxicity : : : : 133 Neuropsychological Sequelae of Treatment for ALL with Chemotherapy Only : : : : 135 Methodological Considerations, Limitations, Outstanding Issues and Data Needs : : : : 139 Conclusion : : : : 142 References : : : : 143 8. Prenatal Tobacco and Postnatal Environmental Tobacco Smoke Exposure and Children’s Cognitive and Behavioral Functioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . Michael Weitzman, Todd A. Florin, and Megan Kavanaugh Introduction : : : : 149 Characteristics of Women Associated with Maternal Smoking : : : : 150 Potential Pathway for Adverse Effects : : : : 151

149

x

Contents

Neurocognitive and Behavioral Outcomes Associated with Maternal Smoking : : : : 153 Research Implications : : : : 158 Summary : : : : 159 References : : : : 160 9. Lessons Learned: Fetal Alcohol Spectrum Disorders . . . . . . . . . . 169 Christie L. McGee, Susanna L. Fryer, and Sarah N. Mattson Introduction : : : : 169 Methodology : : : : 170 Review of Neuropsychological Literature : : : : 173 Long-Term Consequences : : : : 181 Conclusions : : : : 185 References : : : : 186 10. Neurodevelopmental Sequelae of Prenatal Cocaine Exposure . . . . 193 Linda C. Mayes and Marjukka Pajulo Models of Neurobehavioral Teratology : : : : 193 Prenatal Cocaine Exposure : : : : 197 Human Model of Prenatal Exposure : : : : 208 The Interaction of Prenatal Cocaine Exposure and a Drug-Dependent Environment : : : : 211 Summary : : : : 216 References : : : : 218 SECTION THREE: EXPOSURE AND OUTCOME ASSESSMENTS 11. Immutable Elements—Variable Effects: Exposure Assessment for Neurotoxic Metals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Alan H. Stern Introduction : : : : 231 General Considerations for Metals : : : : 231 General Approaches to Exposure Assessment for Developmental Neurotoxic Metals : : : : 233 Application of Pharmacokinetic Models : : : : 237 Mercury as a Case Study : : : : 239 References : : : : 248 12. Studying the Relation Between Pesticide Exposure and Human Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 Dana B. Barr, Asa Bradman, Natalie Freeman, Robin M. Whyatt, Richard Y. Wang, Luke Naeher, and Brenda Eskenazi Introduction : : : : 253 Toxicology : : : : 256 Human Studies on the Neurotoxic Effects of Pesticides in Children : : : : 258

Contents

xi

Factors Affecting Fetal, Infant, and Early Childhood Exposure : : : : 266 Assessing Exposure to Neurotoxic Pesticides : : : : 269 Recommendations for Future Studies : : : : 275 References : : : : 276 13. Assessing In Utero Exposure to Cannabis and Cocaine . . . . . . . . 287 Gale A. Richardson, Marilyn A. Huestis, and Nancy L. Day Introduction : : : : 287 Prevalence : : : : 287 Self-Report Methods : : : : 288 Biological Measures : : : : 291 Comparisons of Self-Report and Biological Markers : : : : 296 Conclusions and Recommendations : : : : 297 References : : : : 298 14. Neurobehavioral Test Batteries for Children Diane S. Rohlman and W. Kent Anger Identifying the Problem : : : : 303 Planning the Research Study : : : : 304 Methods Development : : : : 309 Quality Control : : : : 311 An Example: Developing a Test Battery for Preschool Children : : : : 312 Analyzing the Results : : : : 315 Conclusions : : : : 316 References : : : : 316

. . . . . . . . . . . . . . . . 303

15. Role of Neuroimaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321 Patricia A. Janulewicz, Roberta F. White, and Carole Palumbo Neuroimaging Techniques : : : : 322 Imaging in Developmental Neurotoxicology Research : : : : 325 Faroese Imaging Study : : : : 327 Discussion : : : : 332 References : : : : 332 16. Early Life Environmental Exposures and Neurologic Outcomes in Adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 Marc G. Weisskopf, Robert O. Wright, and Howard Hu Introduction : : : : 341 Critical Periods in Brain Development : : : : 342 Are Early Life Exposures Related to Adult Neurologic Outcomes? : : : : 344 Difficulties in Studying the Effect of Early Life Exposures on Adult Neurologic Outcomes : : : : 347 How Can the Effect of Early Life Exposures on Adult Neurologic Outcomes Best be Studied? : : : : 351

xii

Contents

Summary : : : : 353 References : : : : 353 SECTION FOUR: ANALYSIS AND INTERPRETATION 17. The Critical Concept of Control in Human Neurobehavioral Toxicology Studies: We Can, and Must, Do Better . . . . . . . . . . . . 361 Paul Stewart Introduction : : : : 361 Additional (Theoretical) Criteria for Covariates : : : : 372 References : : : : 376 18. Environmental Risk Assessment with Structural Equation Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Esben Budtz-Jørgensen Summary : : : : 379 Introduction : : : : 379 The Faroese Mercury Study : : : : 380 The Standard Approach to Effect Estimation : : : : 381 Structural Equation Models : : : : 382 Estimation of Exposure Measurement Error : : : : 383 Effect Estimation in Structural Equation Models : : : : 385 The Benchmark Approach to Setting Safety Standards : : : : 387 Standard Application of the Benchmark Approach : : : : 388 Sophisticated Benchmark Analysis : : : : 389 Discussion : : : : 390 References : : : : 391 19. Incorporating the Social-Ecological Perspective into Studies of Developmental Neurotoxicity . . . . . . . . . . . . . . . . . . . . . . . . . . 393 Virgina A. Rauh, Frederica Perera, and Jennifer F. Culhane Introduction : : : : 393 The Confluence of Social and Chemical Risks : : : : 394 Interactions Between Social and Chemical Risks : : : : 402 References : : : : 407 20. Assessing the Neurobehavioral Effects of Early Toxicant Exposure: A Perspective from Animal Research . . . . . . . . . . . . . 415 Barbara J. Strupp and Stephane A. Beaudin Introduction : : : : 415 Considerations in Task Selection and Design : : : : 416 Some Cautionary Notes : : : : 433 Summary : : : : 436 References : : : : 438

Contents

xiii

21. The Developmental Neurotoxicology of Chemicals Disrupting Thyroid Hormone Signaling . . . . . . . . . . . . . . . . . . . . 447 R. Thomas Zoeller The EDSTAC Final Recommendation for Screening Anti-thyroid Activities : : : : 447 The HPT Negative Feedback System : : : : 449 Mechanism of Thyroid Hormone Action : : : : 449 Thyroid Hormone Exerts Time- and Dose-Dependent Effects on Brain Development : : : : 451 Polychlorinated Biphenyls (PCBs) : : : : 452 Conclusion : : : : 454 References : : : : 455

SECTION FIVE: DEVELOPMENTAL NEUROTOXICITY: CLINICAL PRACTICE, RISK ASSESSMENT ETHICS 22. Public Health Perspectives on Developmental Neurotoxicology . . 463 Jennifer S. Mammen and Lynn R. Goldman Public Health Burden of Developmental Disabilities/Learning Disabilities in Children : : : : 463 Assessment of Risk of Potential Developmental Neurotoxic Agents : : : : 464 Tracking Developmental Disabilities : : : : 476 Policy Issues : : : : 476 Conclusions : : : : 479 References : : : : 479 23. Research to Clinical Practice: A Pediatric Environmental Health Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483 April A. Harper and Michael W. Shannon Introduction : : : : 483 The Role of Scientific Research in Clinical Practice : : : : 484 Specific Research Needs : : : : 485 Conclusion : : : : 493 References : : : : 494 24. Evaluating Neurotoxic Effects: Epidemiological, Epistemic, and Economic Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499 Herbert L. Needleman Introduction : : : : 499 The Growth of Understanding of Childhood Lead Toxicity : : : : 500 The Conundrum of Causality : : : : 500 Science, Universities, and the Market Place : : : : 504 Summary : : : : 506 References : : : : 506

xiv

Contents

25. Knowledge, Norms, and the Politics of Risk: Ethical Issues in Policy-Relevant Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 Virginia Ashby Sharpe Introduction : : : : 509 The Normative Context of Policy-Relevant Science : : : : 510 New Mandates in Policy-Relevant Science : : : : 514 Corporate Construction of Risk and Conflicts of Interest : : : : 516 Conflict-of-Interest Disclosure and the Right-to-Know : : : : 517 Community-Based Research : : : : 519 Conclusion : : : : 520 References : : : : 521 Index : : : : 525

Contributors

Jane Adams Department of Psychology, University of Massachusetts Boston, Boston, Massachusetts, U.S.A. W. Kent Anger Center for Research on Occupational and Environmental Toxicology, Oregon Health and Science University, Portland, Oregon, U.S.A. Dana B. Barr National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia, U.S.A. Stephane A. Beaudin New York, U.S.A.

Division of Nutritional Sciences, Cornell University, Ithaca,

David C. Bellinger Department of Neurology, Children’s Hospital Boston, Harvard Medical School and Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A. Asa Bradman Center for Children’s Environmental Health, School of Public Health, University of California, Berkeley, California, U.S.A. Esben Budtz-Jørgensen Department of Biostatistics, University of Copenhagen, Copenhagen and Institute of Public Health, University of Southern Denmark, Odense, Denmark and Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts, U.S.A. Sylvaine Cordier Rennes, France

National Institute for Health and Medical Research (INSERM),

Jennifer F. Culhane Department of Obstetrics and Gynecology, Drexel University College of Medicine, Philadelphia, Pennsylvania, U.S.A. Nancy L. Day Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, U.S.A. Brenda Eskenazi Center for Children’s Environmental Health, School of Public Health, University of California, Berkeley, California, U.S.A. xv

xvi

Contributors

Todd A. Florin American Academy of Pediatrics, Center for Child Health Research and Department of Pediatrics, The University of Rochester School of Medicine and Dentistry, Rochester and Department of Pediatrics, New York University School of Medicine, New York, New York, U.S.A. Natalie Freeman Center for Environmental and Human Toxicology, College of Veterinary Medicine and College of Public Health and Health Professions, University of Florida, Gainesville, Florida, U.S.A. Susanna L. Fryer Center for Behavioral Teratology, Department of Psychology, San Diego State University, San Diego, California, U.S.A. Jennifer Anne Lantz Gavin Department of Psychology, University of Massachusetts Boston, Boston, Massachusetts, U.S.A. Lynn R. Goldman Environmental Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, U.S.A. Philippe Grandjean Department of Environmental Medicine, University of Southern Denmark, Odense, Denmark and Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A. April A. Harper Division of General Pediatrics, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A. Howard Hu Department of Environmental Health, Harvard School of Public Health and The Channing Laboratory, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, U.S.A. Marilyn A. Huestis Chemistry and Drug Metabolism, Intramural Research Program, National Institute on Drug Abuse, National Institutes of Health, Baltimore, Maryland, U.S.A. Patricia A. Janulewicz Department of Environmental Health, Boston University School of Public Health and Department of Psychology, University of Massachusetts, Boston, Massachusetts, U.S.A. Megan Kavanaugh American Academy of Pediatrics, Center for Child Health Research and Department of Pediatrics, The University of Rochester School of Medicine and Dentistry, Rochester and Department of Pediatrics, New York University School of Medicine, New York, New York, U.S.A. Tord Kjellstro¨m National Institute of Public Health, Stockholm, Sweden and Australian National University, Canberra, Australian Capital Territory, Australia and Wellington School of Medicine and Health Sciences, Wellington, New Zealand Gideon Koren Division of Clinical Pharmacology and Toxicology, The Hospital for Sick Children, Toronto, Ontario, Canada

Contributors

xvii

Jennifer S. Mammen Environmental Health Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland, U.S.A. Sarah N. Mattson Center for Behavioral Teratology, Department of Psychology, San Diego State University, San Diego, California, U.S.A. Linda C. Mayes

Yale Child Study Center, New Haven, Connecticut, U.S.A.

Christie L. McGee Center for Behavioral Teratology, Department of Psychology, San Diego State University, San Diego, California, U.S.A. Christine Mrakotsky Department of Psychiatry, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A. Luke Naeher U.S.A.

College of Public Health, University of Georgia, Athens, Georgia,

Herbert L. Needleman Department of Psychiatry, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A. Marjukka Pajulo Department of Child Psychiatry, University of Tampere, Tampere, Finland and Yale Child Study Center, New Haven, Connecticut, U.S.A. Carole Palumbo Department of Neurology, Boston University School of Medicine, Aphasia Research Center, VA Boston Healthcare System, Boston, Massachusetts, U.S.A. Frederica Perera Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, U.S.A. Virginia A. Rauh Heilbrum Department for Population and Family Health, Mailman School of Public Health, Columbia University, New York, New York, U.S.A. Deborah C. Rice Environmental and Occupational Health Program, Maine Center for Disease Control and Prevention, Augusta and Center for Integrative and Applied Toxicology, University of Southern Maine, Portland, Maine, U.S.A. Gale A. Richardson Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, U.S.A. Diane S. Rohlman Center for Research on Occupational and Environmental Toxicology, Oregon Health and Science University, Portland, Oregon, U.S.A. Joanne Rovet Ontario, Canada

Department of Pediatrics, The Hospital for Sick Children, Toronto,

Michael W. Shannon Division of Emergency Medicine, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A. Virginia Ashby Sharpe Center for Clinical Bioethics, Georgetown University Medical Center, Washington, D.C., U.S.A.

xviii

Contributors

Alan H. Stern Division of Environmental and Occupational Health, University of Medicine and Dentistry of New Jersey, School of Public Health, Piscataway, New Jersey, U.S.A. Paul Stewart Department of Psychology, State University of New York at Oswego, Oswego, New York, U.S.A. Barbara J. Strupp Division of Nutritional Sciences and Department of Psychology, Cornell University, Ithaca, New York, U.S.A. Christine Till

Toronto Rehabilitation Institute, Toronto, Ontario, Canada

Deborah P. Waber Department of Psychiatry, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A. Richard Y. Wang National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia, U.S.A. Bernard Weiss Department of Environmental Medicine, University of Rochester, School of Medicine and Dentistry, Rochester, New York, U.S.A. Marc G. Weisskopf Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A. Michael Weitzman American Academy of Pediatrics, Center for Child Health Research and Department of Pediatrics, The University of Rochester School of Medicine and Dentistry, Rochester and Department of Pediatrics, New York University School of Medicine, New York, New York, U.S.A. Roberta F. White Department of Environmental Health, Boston University School of Public Health, Boston, Massachusetts, U.S.A. Robin M. Whyatt Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, U.S.A. Robert O. Wright Department of Pediatrics, Children’s Hospital and The Channing Laboratory, Brigham and Women’s Hospital, Harvard Medical School and Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A. R. Thomas Zoeller Biology Department, University of Massachusetts-Amherst, Amherst, Massachusetts, U.S.A.

SECTION ONE: ENVIRONMENTAL CHEMICALS

1

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma Bernard Weiss Department of Environmental Medicine, University of Rochester, School of Medicine and Dentistry, Rochester, New York, U.S.A.

MERCURY AS AN ENVIRONMENTAL POISON The New York Times of August 26, 1991, carried a headline on its first page that reverberates with debates that are even more strident today: “Despite Era of Concern for the Earth, Mercury is Re-emerging as a Peril.” It pointed to the deepening concern over contamination by mercury of aquatic life and how it had aroused governmental agencies charged with environmental protection and public health to take preventive and regulatory action. It described the origins of the catastrophe in Minamata, Japan, a fishing village whose inhabitants suffered an epidemic of methylmercury poisoning from contaminated fish. It told of the conflicts faced by sport and subsistence fishermen about the safety of consuming their catch, and of the confusion and wariness on the part of fish consumers. The intensity of the debate has not waned. Instead, it has even widened in scope to include forms of mercury other than the methyl form found in aquatic species. It now includes the vaccine preservative Thimerosal, indicted as a risk factor for autism because it is 50% ethylmercury which, like methylmercury, is a potent neurotoxicant. It also includes dental amalgams, an argument carried on for 150 years but becoming more forceful because of the availability of alternatives. Advocacy groups and individuals have raised alarms because chewing releases mercury vapor from fillings that then is inhaled. Why is mercury the focus of such intense discussion at this time? It is not just the evidence of its potency as a toxicant, which Hunter (1), in his treatise on occupational medicine, described in his review of the “ancient metals.” It more likely is a symptom of our distrust of the information made available to us in the realm of food and chemical safety. We have been deceived or misled about chemical hazards on so many occasions at the same time we’ve received reassurances that our concerns are unfounded that we raise a shield of skepticism. We often have seen how commercial or political interests distort the truth, and have witnessed so many disastrous episodes that it requires a suspension of hardened cynicism to believe in reassurances. Methylmercury is the principal subject addressed here, but in the environment it is entangled with other forms of mercury and other neurotoxic entities. Of all mercury species, methylmercury, however, is the one with the broadest implications for public 1

2

Weiss

health. It is found primarily in a major food source, fish, so it affects populations worldwide. It became a model for behavioral teratology as a definitive example of an environmental neurotoxicant, and it contains lessons useful to the study of other agents with the potential to interfere with neurobehavioral development (2). Methylmercury also illustrates the problems posed to regulatory agencies and their need to translate into exposure standards questions such as the shape of dose-response functions and the inherent limitations of epidemiology. In most instances, the translation is a challenge because most of the toxicity information takes the form of high-dose experiments in animals and, except for mass poisonings, low-level environmental exposures in humans, requiring a succession of tentative extrapolations to harmonize these disparities. Like other agents discussed in this volume, methylmercury exemplifies how an appropriate definition of low-level exposure is a constantly moving target. From some perspectives, discussions of neurotoxicity attributable to environmental chemicals involve a marginal problem compared, say, to the serious disabilities of the fetal alcohol syndrome (FAS). Comparability is not the issue, however. Even FAS can be seen as a minor problem compared to the lives erased by guns, child abuse, and auto accidents. It is the inexorable accumulation of many different risks that is so appalling. Exposure to environmental toxicants, although incurring far less dramatic effects than FAS, is more widely distributed in the population, and their effects, because they are so insidious, are easily overlooked. A small effect on IQ, such as an effect size of 0.33 (one-fifth of a standard deviation or 3 IQ points, producing a mean score of 97 rather than 100) assigns 3.2% of the population to the category “retarded” rather than 2.3%. Figure 1 depicts such a condition. It is the kind of situation that animates how we view the neurobehavioral risks of lead and PCBs as well. That is, their adverse effects are

Proportion

0.03

0.02

0.01

0.00

Proportion

0.03

0.02

0.01

0.00 40

60

80

100 IQ

120

140

160

Figure 1 Upper chart shows an IQ distribution with a mean of 100 and SD of 15. The dark area represents the 2.3% of the population below 70. The light area represents those with IQs below 100 and above 70. The lower chart depicts an IQ distribution with a mean of 97. Here, 3.2% of the population falls below 70. IQ of 100 is shown on both charts as the vertical line.

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

3

expressed primarily as shifts in the distribution of population scores rather than, as with clinical assessments, deficits in individuals. This is the core issue for methylmercury, too.

ANTICIPATION OF THE CURRENT DEBATE In 1968, we held the first of over 20 Rochester Conferences on Environmental Toxicity (3). By that time, Swedish scientists had alerted the world that the catastrophe of methylmercury poisoning at Minamata, Japan, represented not a local problem but rather one with global dimensions. They had demonstrated that environmental discharges of all forms of mercury (metallic, inorganic divalent, phenylmercury) can all be converted by natural processes in aquatic settings to the highly toxic methylmercury form (4). They had been alerted to the existence of a methylmercury problem in Sweden’s waters by a decline in sea bird populations, in species such as ospreys, that depended on fish for sustenance. Their findings recapitulated the questions that had been posed by Rachel Carson (5) about pesticides when she observed diminished bird populations. By analyzing the mercury content of bird feathers in museums and comparing them to contemporary levels, they demonstrated the sharp rises coincident with the introduction of organic mercury fungicides into Swedish agriculture. (Feathers, like growing hair, reflect methylmercury concentrations in the blood). They also showed how the process of bioconcentration could magnify toxic exposures: pike in Swedish waterways bore flesh levels of methylmercury as much as 3000 times greater than the ambient water concentration due to increases at successive trophic levels as predators consumed contaminated prey. At the 1968 meeting, Robert Risebrough, of the University of California, Berkeley, asked the following question: Why should mercury be a problem in the northern countries rather than in the more temperate areas?

Fredrik Berglund of the National Institute of Public Health, Sweden, replied: I think one reason that this problem does not exist in the United States with mercury is that the levels are not known.I feel, personally, that the problem also exists here as it does in other parts of the world but it is not recognized.

Recognition soon came, with the discovery, in 1970, of methylmercury contamination in the Great Lakes, stemming mainly from the discharge of metallic mercury by chlor-alkali plants. These plants use large pools of liquid mercury as electrodes to convert brine into chlorine and sodium hydroxide. The current methylmercury agenda, with its emphasis on prenatal exposure, was set the following year with a mass chemical poisoning in Iraq (6). This history parallels the concurrent development of neurobehavioral toxicology as a science.

MINAMATA AND IRAQ Two mass chemical disasters awakened us to the threats posed to developing brains by methylmercury. One occurred in Japan, the other in Iraq. Progessive stages in the evolution of neurobehavioral toxicology can be traced by how those two mass poisonings helped spawn a discipline that pulled toxicology from its traditional reliance on lethality and pathology towards criteria such as IQ scores and performance on behavioral tests.

4

Weiss

What the Japanese later called Minamata Disease arose from methylmercury contamination of Minamata Bay, which adjoins a fishing village in the southernmost Japanese island of Kyushu. A change in catalyst use in the production of acetaldehyde by the Chisso factory on the shores of the bay resulted in the discharge of mercury into the bay, where it contaminated the fish and shellfish consumed by the villagers, especially fishermen and their families. The tragedy of Minamata has been told many times and in many ways, but perhaps none as effectively as in the photo essay by Eugene and Aileen Smith (7). It depicted with one unnerving episode the price Japan had paid for industrialization carried out with little concern for its environmental consequences. Minamata disease, after a lengthy period of puzzlement about its etiology, was eventually discovered to be due to methylmercury exposure. Its signs, in adults, had been known for decades because of accidents. They are listed in Table 1. The victims of that tragedy extend beyond the group whose manifestations were easily identified. A much larger group most surely bears scars that are not as easily or directly ascribed to methylmercury because their legacy came in the form of functional deficits that required close study with appropriate tests and comparisons at a time when quantitative measures of exposure were not available. The drama of that episode, however, has tended to blur an important environmental lesson. Like some other environmental contaminants, methylmercury delivered its toxicological message insidiously. Figure 2 charts the rising incidence of Minamata Disease over a three-years period. Cases appeared sporadically. Dr. Hajime Hosokawa, a physician employed by Chisso, courageously proposed that a link existed between the factory and the continuing appearance of cases. His hypothesis was dismissed by Chisso management, but he based his argument on an almost intuitive epidemiology and, perhaps, drew from a famous experiment with cats described by Eto et al. (8). By 1958, almost no cat could be found in the Minamata district. Even as far back as 1952, they had begun to display convulsions, ataxia, and what the inhabitants termed “dancing disease.” Because of this history, a cat experiment was designed to test the conclusion, declared by a study group from Kumamoto University, that organic mercury was responsible for the outbreak. Ten cats were fed food mixed with industrial waste produced in the Chisso plant. The results were not made public, but Eto et al. (8) found the specimens and conducted pathological and chemical assays. Typical lesions of methylmercury poisoning were observed in the central nervous system tissue, and were associated with markedly elevated mercury levels (9). Table 1

Signs and Symptoms of Methylmercury Poisoning in Adults

Sensory Paresthesia (numbness and tingling) Pain in limbs Visual disturbances (field constriction) Hearing disturbances Asterognosis (discrimination by touch) Motor Disturbances of gait Weakness, leg unsteadiness, falling Thick, slurred speed (dysarthria) Tremor Other Headaches, rashes, “mental disturbance”

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

5

Onset of Minamata Disease by Month 60 Number of Cases

50 40 30 20

"Mystery Dlsease" report by Dr. Hosokawa

10 0 12/53

6/54

12/ 54

6/ 55

cumulative

12 / 55

6/ 56

12 / 56

new cases

Data courtesy of H. Satoh and C. Watanabe

Figure 2 Accumulation of cases of Minamata disease between December 1953 and December 1956.

Figure 2 reflects a property of methylmercury toxicity that helped obscure the connection between fish and shellfish consumption and Minamata disease: the period of silent toxicity between exposure and onset of effect. Although 95% of ingested methylmercury is absorbed, delayed toxicity, or displacement in time between exposure and its consequences, makes it difficult to discern a relationship, a gap that can deceive investigators about etiology or lead to disasters that could have been averted or diminished if acted on earlier. This property of methylmercury is discussed later. Minamata also provided the first suggestions of fetal vulnerability. Pregnant women who evinced no signs of poisoning themselves gave birth to offspring with severe neurological deficits who later received the diagnosis of Fetal or Congenital Minamata Disease. These prenatally-exposed children were characterized by severe neurological impairments that included ataxia, neuromotor disability, seizures, hearing loss, microcephaly and cognitive impairment (10). Some of these individuals, now in their 50s, are maintained in a special unit attached to the Minamata Disease Research Institute, which was formed and is supported by the Japanese government. In a bizarre repetition of Minamata about 10 years later, another Japanese town, Niigata, also experienced an epidemic of methylmercury poisoning (11). Again, it arose from a factory that manufactured acetaldehyde and that, in the process, discharged inorganic and methylmercury into a waterway, in this instance, the Agano river. The authors report that 690 adults have been certified as victims of methylmercury poisoning. They also report that even many of those exposed prenatally, whose mothers’ hair levels may have exceeded 50 ppm, a level now considered excessive, show little evidence of adverse effects. These measurements have not been verified by current analytical methods, however. The second mass poisoning, and the one best documented, occurred in Iraq in the winter of 1971–72. A severe drought in the summer of 1971 impelled the Iraqi government to purchase 80,000 tons of seed wheat for planting in 1972. It also requested that the wheat be treated with a methylmercury fungicide, a request made in error. Hungry farmers in rural Iraq, rather than planting the wheat, baked it into bread. During a three-month period that winter, as many as 5000 victims may have died with as many as 50,000 suffering severe illness.

6

Weiss

Because Thomas W. Clarkson at the University of Rochester had recently reported on a sulfhydryl-binding resin that accelerated the clearance of methylmercury from the body, he was contacted by the Iraqi Ministry of Health to secure the resin. The ultimate outcome of that request was establishment of a laboratory in Baghdad and a survey of the exposed population in the countryside (12). It led to the use of hair as a biomarker of exposure history and, of supreme importance, a clear demonstration of the exquisite sensitivity of the fetal brain. Children whose mothers had been pregnant during the outbreak were examined at about 30 months of age to assess developmental milestones such as walking, a difficult task because of undocumented birth dates coupled with the need to conduct examinations in the field. Nevertheless, those data suggested that maternal body burdens in the range of those seen in fish-consuming populations might pose a risk to neurobehavioral development (13) and eventually stimulated several prospective studies in such populations. Interpretation of the results of those studies provides the basis for the current debates about the neurodevelopmental risks of methylmercury. Minamata and Iraq aroused enormous anxieties about their implications for child development (14). Both episodes identified the human fetal brain as the most vulnerable target of methylmercury neurotoxicity, an expectation confirmed by studies in animal models (15,16). Such models have provided the main basis for exploring the mechanisms underlying the pathological features observed in human brains. These were described by Choi et al. (17), who examined the brains of two full term newborn human infants in Iraq who were exposed to methylmercury early in pregnancy due to maternal ingestion of methylmercury–contaminated bread. The brains contained high levels of mercury. Disturbed development was marked by incomplete or abnormal migration of neurons to the cerebellar and cerebral cortices, and deranged cortical organization. Numerous heterotopic neurons were seen in the white matter of cerebrum and cerebellum. The characteristic laminar pattern of the cerebral cortex was distorted in many regions. Such a pattern of damage is unlike the focal neuronal damage observed in poisoned adults and children exposed postnatally. Studies in mice by Sager et al. (18) and Rodier et al. (19) indicate that mitosis is blocked by methylmercury exposure during metaphase. The mechanism appears to be disruption of microtubules in the mitotic spindle. Mitotic arrest can prove devastating to the developing brain, especially before birth, when the bulk of neurons are formed. Choi (20), also in mice, noted abnormal patterns of neuronal migration and cortical cytoarchitecture, another feature of methylmercury neuropathology.

THE CURRENT SCENE In practice, the only source of methylmercury exposure is food from aquatic and marine environments. Because these represent crucial food sources, and a dominant one for protein in some populations, methylmercury ascended to a problem of global dimensions. Although several smaller epidemiological studies have been undertaken (21–27), two large, longitudinal cohort studies occupy the center of the risk debate. One, undertaken by investigators from the University of Rochester, was situated in the Republic of the Seychelles, an island group in the Indian Ocean. Another, conducted by investigators from the University of Southern Denmark, in Odense, was situated in the Faroe Islands in the North Atlantic. Both research groups chose to study the relationship between prenatal exposure to methylmercury and a variety of neurodevelopmental measures. In the Seychelles project, methylmercury concentration in maternal hair was chosen as the biomarker, a choice

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

7

governed by data demonstrating a close correlation between blood levels and levels in growing hair and by ease of storage for transportation and later assays. Because scalp hair grows about 1.1 cm per month, a 12 cm hair strand can recapitulate a one-year exposure history, a vital factor in tracing the outbreak of poisoning in Iraq (6). McDowell et al. (28) point out another advantage of hair assays; namely, because they are noninvasive, they elicit much better response rates from population subgroups such as children that usually resist blood sampling. The hair:blood ratio roughly averages 250:1, but with substantial variation (29,30) and higher ratios in children. The question of how hair levels relate to brain levels will be discussed later. Based on the autopsy material from the Seychelles (31), maternal hair and infant brain levels are closely correlated. The Faroes investigators selected cord blood, which reflects exposure during the later stages of pregnancy, as their main index of exposure, but additional analyses showed a high correlation with hair levels. Maternal hair levels averaged 6.9 ppm in the Seychelles and 4.27 in the Faroes. In the U.S., recent NHANES data showed a geometric mean of 0.12 ppm in children, and 0.20 ppm in women (32). Frequent fish consumers, compared with non-consumers, showed geometric mean hair mercury levels for women of 0.38 versus 0.11 ppm and levels for children of 0.16 versus 0.08 ppm. In the Seychelles, mothers reported eating a mean of 12 fish meals weekly. In the Faroes, adults reported mean daily consumption of 72 grams of fish, 12 grams of whale muscle, and 7 grams of whale blubber. The differences in consumption patterns are significant; the methylmercury source in the Seychelles is ocean fish, while in the Faroes it is primarily pilot whales. The differences create an interpretive dilemma, as discussed later, because these whales are heavily contaminated with other neurotoxicants such as PCBs and a variety of other organohalogens. The Seychelles cohort, consisting of 779 mother-infant pairs enrolled in 1989–1990, has so far been assessed at 6.5, 19, 29, 66, and 107 months of age by a variety of neuropsychological tests appropriate for the cohort’s stage of development (33). For example, at 19 months of age, the main assay consisted of the Bayley Scales of Infant Development. At 107 months of age, the investigators were able to deploy a broad variety of instruments ranging from those assessing motor function (e.g., grooved pegboard) to those assessing cognitive function (e.g., Boston Naming Test) to those assessing conduct (e.g., Child Behavior Checklist). From all of these assessments, based on multiple regression analyses that included potential confounders, only a handful of scattered endpoints revealed any correlation with mercury exposure indices. And, of these, some showed a positive relationship between exposure level and performance, but the relationship depended on age. The Faroes cohort was also assessed with a battery of tests that included neurophysiological (e.g., visual and auditory evoked potentials) as well as neuropsychological (e.g., motor function, cognitive function, subjective state) instruments. The study was initiated with the collection of maternal hair and cord blood levels from 1022 singleton births. At about 7 years of age, 917 of the children were recruited to undergo testing (34). Multiple regression analyses, adjusted for confounders, comprised the main statistical technique. Neither the neurophysiological endpoints nor clinical examinations showed a relationship with methylmercury exposure. In contrast to the Seychelles results, analyses of the neuropsychological indices demonstrated significant negative associations with prenatal methylmercury measures in all five domains chosen for assessment: motor, attention, visuospatial, language, and memory. The different conclusions derived from the Seychelles and Faroes studies remain puzzling but, also, contentious. A National Academy of Sciences committee, charged with

8

Weiss

establishing a risk assessment of methylmercury in seafood, chose the Faroes data as its basis because it offered a clear outcome (35). Specifically, it relied upon the Boston Naming Test as the index measure. It then applied Benchmark Dose (BMD) modeling to derive an exposure level associated with a specified risk value. For such calculations, a Benchmark Response (BMR) is chosen; for example, a value of 5%, which means an increased risk of an abnormal response of 5%. A BMD is then calculated by fitting a mathematical model to the dose-response function, and, on the basis of that function, deriving the dose of the agent that would be predicted to produce, for a BMR of 5%, an excess risk of that magnitude. Because of statistical uncertainty about the central estimate BMD, a lower 95% confidence limit (BMDL) for the BMD is also computed. The committee determined a BMDL of 58 ppb in cord blood (equal approximately to 12 ppm in hair) as the appropriate exposure level from which to derive risk values. The value of 58 ppb was extrapolated to an equivalent intake of 1.1 mg/kg of methylmercury per day. In setting exposure standards, it is common to divide a value such as the BMDL by one or more uncertainty factors to provide a margin of protection. Dividing that value by uncertainty factors summing to about 10 yields a Reference Dose (RfD) (equivalent to an acceptable daily intake) of 0.1 mg/kg per day. The EPA defines a RfD as an exposure level that is without adverse effects over a 70 years lifespan. This RfD translates into a weekly consumption of one 198 g (7 oz) can of tuna for an adult. BMD models have also been applied to the Seychelles data set (36). As the authors note, such models can prove useful even when a statistically significant dose-related trend is absent. On the basis of an increase of 0.1 in the probability of an adverse response (the BMR), maternal hair BMDLs calculated from the Seychelles data ranged from about 19 to about 30 ppm. These values are not markedly different from the Faroes BMDL, nor from BMDLs based upon a New Zealand cohort (23,24) or the data from Iraq (37). The Faroes BMDL values can roughly be translated into brain levels. In nonhuman primates, the blood-brain ratio is approximately 3–10:1 (38–41). In the Seychelles cohort, comparing maternal blood with infant brains secured at autopsy yielded ratios between 5.1 and 6.7 (31). Based on these values, we would expect brain levels at the BMDL to roughly approximate 300 ppb. Infant brains from the Seychelles were examined by Lapham et al. (42). They found no evidence of neuropathology, assessed by criteria similar to those of Choi et al. (17); these brains all exhibited levels of mercury less than 300 ppb, however, while discernible anatomical abnormalities (by conventional pathological techniques) are seen in experimental animals at levels of 1000 ppb and above (43).

OTHER COMPLICATIONS The underlying question for the risk assessment of methylmercury is converting exposure magnitude into risk, which is then translated into the form of the dose-response function. Response takes the form of scores on one or more neuropsychological tests. Dose takes the form of methylmercury in blood or hair. Methylmercury is a biomarker of seafood consumption. (The term seafood is used generically to refer to fish, crustaceans, sea mammals, and other forms of aquatic life eaten by humans and coming from both marine and freshwater sources). In general, those who consume more of these foods exhibit higher levels of methylmercury. Investigators pursuing the question of methylmercury risks, like toxicologists in general, traditionally assume a monotonic dose-response function; that is, that higher exposure levels correspond to greater adverse effects. The dose-response relationship for any particular agent and endpoint may assume a different shape due either

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

9

to the nature of the biological response or because of interactions or confounding with other factors (cf., 44). Instead, it might assume a U-shaped function, as suggested in some data from the Seychelles project (45), especially in populations where fish rather than other aquatic species comprise major portions of the diet. Adverse developmental outcomes, that is, could be associated with both low and high levels of methylmercury, and optimal outcomes might be associated with intermediate values, as described below. U-shaped functions are common in nutrition. For example, manganese is required by neonates for normal skeletal and inner ear development and as components of critical enzymes (46). Elevated manganese exposures in neonatal rats, however, can also induce neurochemical and behavioral abnormalities (47). Vitamins share the same kind of U-shaped dose-response properties and, like manganese, reflect different endpoints in the two arms of the U. In the methylmercury example, lower levels might reflect diminished intake of important nutrients such polyunsaturated long chain fatty acids (PUFAs). The n-3 (omega-3) fatty acids, such as docosahexanoic acid (DHA), are abundant in fish. Higher levels of methylmercury would overcome such beneficial properties and induce toxicity. This phenomenon may underlie the finding of an inverse relationship between maternal hair mercury levels and scores of male offspring at 66 and 107 months of age on the Conners Rating Scales, which are used as measures of ADHD (45). These results do not contradict the recognized threats to brain development posed by higher levels of methylmercury exposure; they merely suggest that the neurotoxic risks associated with low levels of exposure may be concealed by the benefits of fish consumption for which methylmercury serves as an exposure marker. Choline levels are also high in fish. Zeisel (48) reviewed the evidence suggesting an essential role for choline in human neurodevelopment and cognitive functioning. Studies in rats have demonstrated that supplementation with choline during embryonic development or immediately following birth can result in improved memory performance, which is maintained throughout life (49). Whether supplementation or deprivation later in postnatal development could affect complex cognitive, sensory, motor or behavioral functions is unclear. Common developmental disorders including ADHD, dyslexia, dyspraxia and autistic spectrum disorders (ASD) may also involve functional deficiencies or imbalances in n-3 and n-6 fatty acids (50). But the Seychellois diet is high in these fatty acids, and the levels of PUFAs in blood in Seychellois women is six times higher than in U.S. women. THE RISK EQUATION AND EFFECT MODIFICATION Few epidemiological studies examining exposure to environmental neurotoxicants can without ambiguity dissociate themselves from the multitude of other toxic risks encountered during development nor, in some cases, such as pesticides, from the question of potential public health benefits. For some environmental contaminants, such as dioxin and lead, no benefits are attached to exposure. For pesticides, it could be argued that exposure must be weighed against the benefits of pest control and efficient agricultural production (51). For seafood, the dilemma is how to construct a risk assessment equation across several dimensions. We are faced with three axioms: 1. All fish, crustaceans, and sea mammals contain methylmercury in variable concentrations. 2. Fish, crustaceans, and sea mammals also contain nutrients, again in variable concentrations, such as n-3 fatty acids, choline, and iodide, that are essential for brain development.

10

Weiss

3. Many fish, crustaceans, and sea mammals are also contaminated with a variety of toxic contaminants other than methylmercury; for example, cadmium, PCBs, polybrominated diethyl ethers (PBDEs) dioxins, DDT and its metabolites, other organochlorine pesticides, and a variety of polycyclic aromatic hydrocarbons such as benzo[a]pyrene. Disentangling the individual contributions of such contaminants, all of which can exert adverse effects on brain function and development, is a daunting, if not impossible challenge, especially because of the potential for additive or synergistic combinations. The mixture problem, a legacy of the extent to which we have thoughtlessly contaminated the environment, is hardly confined to seafood, and pervades every facet of current toxicology. Achieving an appropriate balance of risks and benefits for seafood presents an unusual dilemma that needs further exploration. The third axiom above noted that other common contaminants, as discussed in other chapters in this book, offer hazards of their own. Methylmercury is found primarily in muscle tissue. Organohalogens are concentrated in fatty tissue. Fish that are low in methylmercury and high in n-3 fatty acids, such as salmon and herring, because of their high fat content, also tend to contain higher concentrations of the organohalogens. The consumer, as a result, is compelled to choose between toxicants in selecting different varieties of seafood. Farm-raised salmon, for example, may be contaminated by levels of PCBs exceeding those recommended by agencies such as the U.S. EPA (52). These authors also noted wide regional differences in the degree of contamination. Fish raised in farms in Scotland and the Faroes contained the highest concentrations of PCBs, while those raised in Chile contained the least. The high lipid solubility of PCBs and other organohalogens presents a special problem for brain development. The fetal and neonatal brain is 60% lipid, and it consumes a high proportion (70%) of the energy drawn from the mother. Its energy consumption is one reason it is so vulnerable. Fetal and infant brain development are highly dependent on the availability of two long-chain polyunsaturated fatty acids, docosahexanocic acid (DHA) and arachidonic acid (AA). Seafoods provide these nutrients in relative abundance compared to terrestrial sources (53). These authors argue, in fact, that the evolution of the hominid brain depended on the availability of these fatty acids, which were provided in fish consumed by littoral communities. The organohalogen connection, for which PCB levels provide an exposure index, creates interpretive complications. Fangstrom et al. (54), in their analyses of PCBs in pregnant Faroese women, observed high concentrations of PCBs in serum, “possibly the highest so far reported in a population” and attributed these values primarily to the consumption of pilot whale blubber. Some neurotoxic effects of PCB exposure were uncovered in the 7 years assessment (34). The later stratified analyses conducted by Grandjean et al. (55) suggested that neurotoxicity stemming from PCB exposure may be detectable only at the higher levels of methylmercury, which suggests an additive or more complex model. Whether joint PCB and methylmercury exposure can be disentangled by statistical methods is unclear because we do not understand the mechanisms of interaction, or, indeed, if interaction is the proper term for this situation (56). A study of sport fish consumers residing in a community on the shore of Lake Ontario, designed primarily to explore the influence of prenatal PCB exposure on neurobehavioral development, has also unearthed a complex PCB-methylmercury relationship (57). It is, in essence, a mirror-image of Grandjean et al. (55). In this cohort, in offspring exhibiting the higher PCB levels, methylmercury concentrations in maternal

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

11

hair provided a significant predictor of performance on the McCarthy Scales of Children’s Abilities. In a somewhat parallel analysis of PCB-methylmercury interactions, Grandjean et al. (58) reported that breast feeding in the Faroes is associated with diminished growth, an outcome attributed to both methylmercury and PCBs. Put another way, the benefits of breast feeding for neuropsychological development and immune function in a community such as the Faroes have to be balanced against the risk of growth retardation. These PCB-methylmercury interactions can be viewed from the perspective of effect modification as well as from a mechanistic perspective. Bellinger (59) defines the term to describe a phenomenon in which the effects of one variable, such as dose, depend upon the levels of another variable, such as socioeconomic status. Bellinger pointed to the latter as a model, because of his finding that an effect of lead on IQ scores could be seen in children from low SES but not in high SES families. Analogously, in the Ontario and Faroe cohorts, an effect of PCBs (or methylmercury) in combination with methylmercury (or PCBs) depended on the level of the other agent.

INORGANIC MERCURY The risk assessment challenge of methylmercury in seafood is also compounded by simultaneous exposure to inorganic mercury from sources other than food. In the form of mercury vapor (Hg8), elemental mercury is a potent neurotoxicant. The complications for methylmercury risk assessment due to concurrent exposure to the vapor species arise from two sources. One is the potential developmental neurotoxicity of mercury vapor itself. The other is the possibility that the vapor and organic species share common mechanisms of toxicity. Because assays of mercury in blood often do not distinguish inorganic from organic species, much of the data in the literature cannot be used to compare their individual contributions to selected endpoints. Although mercury vapor is a venerable and potent adult neurotoxicant, its hazards for nervous system development are virtually unknown. The fetus may be exposed to both the methyl and vapor forms. While methylmercury is consumed in the diet, exposures to the vapor form may occur in the workplace, as a result of certain religious practices employing metallic mercury (60), through the use of mercury-containing cosmetics (61), or, the most common source, by mercury vapor emitted by dental amalgam restorations and inhaled by pregnant women. Both the organic and vapor forms of mercury are then transmitted to the fetus. Chewing is known to produce vapor, which is then inhaled by the mother and carried to the fetus. Information about the joint effects of these two mercury species is scant, but, on the basis of mechanistic considerations and sketchy animal data, warrants careful consideration. Berlin (62) argues forcefully that inorganic mercury could pose a threat to fetal brain development. He notes, based on in vitro experiments, that toxic effects are discernible at concentrations of 1–10 ng/g in tissue, which lie in the range measured in infants and fetuses from amalgam-bearing mothers. Further, he argues, anatomical and behavioral disorders are seen in rats and in monkeys exposed developmentally at fetal brain concentrations of 10–200 ng/g, which fall below the concentrations seen to produce such effects with methylmercury. In addition, he points to epidemiological studies indicating that occupational exposure to mercury vapor is associated with elevated risks of mental retardation and infants labelled as small for gestational age.

12

Weiss

Inorganic mercury levels in fetal tissue are significantly correlated with the number of maternal amalgam surfaces (63). And Bjornberg et al. (64) report that inorganic Hg in cord blood increased significantly with increasing number of maternal dental amalgam fillings. Not only are billions of individuals world-wide exposed to the emission of mercury vapor from amalgam restorations, but some pregnant women suffer an increase in dental problems such as caries and gingivitis, and become candidates for dental restorations (65). In some communities, dental care, including restorations, is promoted as part of prenatal care. Like methylmercury, inhaled mercury vapor is transported from mother to fetus, across the placenta, in much the same manner in which it crosses from the blood into the brain (66,67). Furthermore, we also know that maternal vapor exposure in humans is correlated with mercury blood levels in the fetus (68). Ionic mercury neither enters the brain nor penetrates the fetus. One source of concern about coexposure arises from morphological findings in primates exposed to inhaled mercury vapor (summarized in 69). Prenatal exposure resulted in brain growth retardation, sulcal abnormalities and increases in the number of heterotopic neurons in the cerebrum. In particular, the subpially localized heterotopic neurons with their disoriented apical dendrites indicated arrested cell migration. Similar changes of subpially localized cells have been noted in humans after exposure to methylmercury (17). Consequently, both empirical and mechanistic reasons suggest that combined exposure to methylmercury and inhaled mercury vapor might together produce effects more severe than those seen from exposure to either agent alone. These are based on the following considerations: (1) both agents produce prenatal damage, (2) both produce behavioral changes in animals at levels that are not overtly toxic, (3) arrested neuronal migration is detectable after both agents, and (4) both agents result in the release in the brain of the same toxic species of mercury, namely, mercuric mercury. Further, as with methylmercury, infants and children seem significantly more vulnerable than adults to mercury vapor toxicity. The syndrome of acrodynia (“painful limbs”), also known as “Pink Disease” because of the erythema and prolonged episodes of irritability and crying, began to make its appearance in medical journals early in this century (70). Its etiology was ascribed mainly to infectious diseases. Only in 1947 was it discovered to be a manifestation of mercury toxicity, when Warkany and Hubbard (71), at the University of Cincinnati, detected high levels of mercury in the urine of children with Pink Disease. The sources consisted of teething powders, which contained calomel (mercurous chloride), and some anti-helminthics. When calomel was withdrawn from teething powders in England, Pink Disease reports fell sharply except for instances probably arising from other exposure sources. Although inorganic mercury compounds were responsible for the early outbreaks, almost all of the contemporary reports of Pink Disease in the medical literature point to mercury vapor as the exposure source (72). Despite the fact that inhaled mercury vapor distributes to the fetal brain in humans (63) and monkeys (73), no investigations of subclinical developmental changes in children, like those conducted for in utero exposure to lead and methylmercury (74), have been published. A study of pregnant women in Sweden (75) found that 28% of the mercury in blood during early pregnancy proved to be inorganic while the rest was comprised of methylmercury. The inorganic component was mostly contributed by dental amalgams. And, as noted earlier, Bjornberg et al. (64), also showed a close relationship between number of maternal amalgams and inorganic mercury in cord blood, but a lower proportion of total mercury, perhaps because of changes in dental care. These values

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

13

should be viewed against the figures derived from the 1999–2000 NHANES survey, which indicated that about 8% of women had blood mercury concentrations higher than the U.S. EPA recommended RfD of 5.8 mg/L (76), which assumes that almost all the mercury in blood is contributed by the methyl form. In a more recent assessment, Mahaffey et al. (32), also based on the 1999–2000 NHANES survey, used measures of both total and inorganic mercury in blood to calculate the organic (methyl) component. They noted that, as the organic proportion rose with an increasing number of fish meals, the inorganic proportion dwindled, but that is to be expected if the vapor exposure remains fairly constant. This argument is not a valid dismissal of the potential contribution of mercury vapor to mercury-associated developmental neurotoxicity. In fact, it could be argued that, as organic mercury intake increases, the potential neurotoxicity of the vapor component becomes increasingly important. In fact, if animal data provide any guidance, the combined exposure may even be synergistic (77). One complication arising from these possibilities is that measures of cord blood mercury often measure only total mercury (27,30,78). Although total mercury is closely related to dietary intake of methylmercury, it is conceivable that the inorganic component may itself be correlated with later neuropsychological measures. Hair levels of mercury do not reflect vapor exposure, so that for most purposes it may be the preferred index. Speciation is also crucial for measuring transfer of mercury to the infant from breast milk. Oskarsson et al. (79) measured mercury concentrations in breast milk, blood, and hair samples, collected six weeks after delivery, from 30 Swedish women. Milk samples showed an average of 51% of total mercury in the form of inorganic mercury. In blood, similar to the findings of Vahter et al. (75), an average of only 26% was present in the inorganic form. Both total and inorganic mercury concentrations in blood and milk were significantly correlated with the number of amalgam restorations. Estimated methylmercury intake from fish and concentrations in breast milk were not related. The potential confounding of different mercury species both in the measurement of endpoints and in biomarker assays warrants precautions such as measuring mercury in maternal and infant urine; urinary mercury provides an index of vapor exposure and is correlated with the number of amalgam restorations.

DELAYED TOXICITY Latent, or delayed, toxicity appies to the phenomenon in which detectable expressions of adverse effects are displaced in time from an acute exposure or after exposure has ceased (80). It represents the core issue, not just for methylmercury, but for all neurodevelopmental toxicants. Cognitive difficulties, retarded language acquisition, and attention deficit disorders, for example, are identified only at a stage of maturation when the organism is expected to be functioning appropriately for that stage. The main challenge for developmental neurotoxicology is how to connect such manifestations to exposures occurring earlier in life. One impediment to securing such an anchor is the documented but rather limited plasticity of the developing brain. The prevailing doctrine used to be that young brains can recover even from the type of severe damage that disables adult brains. One source of this belief was the experiments of Margaret Kennard, who, in the 1930s and 1940s, compared the effects of brain lesions in infant monkeys with presumably similar lesions in adult monkeys and found far less functional disturbance in the infants than in the adults.

14

Weiss

Examined more carefully, however, the doctrine lacked sound support. Kennard, in fact, later reported that deficits in motor function not observed during infancy began to emerge as the lesioned monkeys matured (81). Similar findings were reported after pyramidal tract lesions. In rat studies, aberrant social behaviors in rats subjected to amygdala lesions during infancy appeared only when they reached puberty. The classical experiments of Goldman (82) provided compelling evidence for the phenomenon that has been called, “growing into” the lesion. Monkeys subjected to removal of dorsolateral cortex when infants did not differ from controls on a delayed alternation task at 15 months of age. At 2 years of age, the lesioned monkeys committed significantly more errors than the control monkeys even though they had practiced the task 9 mo earlier. One way to account for such observations is to presume that the behavioral deficits appear only at the time when the damaged region would typically begin to subserve the function being tested. Another way to interpret the phenomenon is to view it as the structure becoming “committed” to a specific range of functions as the organism matures. Prenatal exposure to relatively modest MeHg concentrations (e.g., !10 ppm in maternal hair) may influence only those higher order cognitive functions that develop with maturity. As a result, deficits may not begin to appear until the brain has developed to a stage at which it has “grown into” the lesions caused by earlier exposure. Johnson and Almli (83) describe the phenomenon as follows: “One of the most perplexing results found when brain damage is sustained during infancy is that many behavioral deficits do not become manifest until after considerable time has elapsed.Goldman.has suggested that delayed effects following brain damage sustained at an early age may be related to the degree of functional maturity of the brain region-behavior relationship under study.With maturation, as that brain region would normally become committed to the behavior in question, deficits in behavior of the brain-damaged animal become manifest.” The temporal patterning sequence seen by Gogtay et al. (84) in their MRI mapping study of cortical development reveals that higher-order association cortices mature only after lower-order somatosensory and visual cortices, the functions of which they integrate, are developed. Changes are taking place even into the late teens, so that earlier damage might not emerge, if the “growing into the lesion” phenomenon holds in this instance, until that stage and with tests based on complex cognitive function. The experimental data are supported by observations in humans. Lenneberg (85), on the basis of his experience with brain-injured children offered the following comment: “One may say that the child with a perinatal cerebral injury only gradually ‘grows into his symptoms,’ and that both lesions and symptoms have their own ramified consequences, often affecting distant structures years after the primary injury.” These venerable lesion studies are reviewed here to illustrate the principle that the brain is a dynamic entity and that it changes through the lifespan in often dramatic ways that are reflected in behavior. This principle needs to guide why the conclusions we have reached on the basis of our current information about methylmercury should also remain in a dynamic state, so to speak. Our appreciation of how the consequences of exposure may vary through the lifespan remains fairly primitive. Two animal studies are illuminating examples of this principle. Spyker (15) administered 4 or 8 mg/kg methylmercury to pregnant mice on gestational day 7, 9, or 12. She then maintained them for a lifetime. Most of the offspring displayed no overt signs of toxicity at weaning. At about 30 days of age, the treated offspring showed deviant behavior in open-field testing and in swimming behavior. As they matured, and especially as they approached senescence, they began to exhibit a multitude of disabilities: obesity,

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

15

kyphosis, muscular atrophy, difficulty in righting themselves, and an assortment of less obvious behavioral deficits. They also died sooner than control mice. Two studies in rats, Newland and Rasmussen (86) and Newland et al. (87), also demonstrated interactions with aging following developmental exposure to methylmercury. As they aged, the treated subjects began to display deviations from control performance on complex behavioral tasks. These were not grossly aberrant performances; the deficits occurred earlier during senescence (86), or were expressed as slower acquisition (87). Evidence from nonhuman primates also indicates that effects of low exposures to MeHg during postnatal development may not emerge until later in life. Such delayed neurotoxicity, appearing years after cessation of exposure, was suggested in reports from Minamata, and the publications cited above provide further evidence for such a phenomenon. These primate studies provide even more impressive evidence for delayed toxicity. In one study, Rice (88) exposed one group of monkeys to methylmercury beginning during gestation and continuing until 4 years of age). In tests of auditory function conducted at 11 and 19 years of age, the monkeys exhibited deficits in pure-tone thresholds, especially at the later age. Another group of monkeys received low daily doses of methylmercury from birth to 7 years of age. After a six-years hiatus, at 13 years of age, they began to display mild signs of somatosensory dysfunction in their home cages (40) consisting mainly of impaired dexterity and clumsiness in handling items of food. These observations were validated by tests of vibration sensitivity (89). The monkeys dosed during gestation and until 4 years of age also showed somatosensory deficits (90). Figure 3 depicts the situation graphically and compares it to the Minamata outbreak. There, latency periods as long as 15 years have been reported and described as “Minamata disease of late onset” (91).

Delayed onset in primates* start dosing

end dosing

onset effects

50µgHg /kg /day

0

3

6 9 years of age

12

15

Late onset of Minamata cases** End of acute Minamata outbreak

‘60

‘63

Reports of “new cases”

66 69 1960 - 1975

‘72

‘75

Figure 3 The late onset of methylmercury poisoning in nonhuman primates and in humans after exposure to methylmercury in Minamata.*Based on Rice (1996) in primates and **Igata (1991) in humans.

16

Weiss

These primate experiments, although revealing sensory and motor deficits resulting from developmental exposures, failed to demonstrate the kind of cognitive performance deficits assayed by certain forms of schedule-controlled operant behavior (92,93). Newland et al. (94), relying on another form of operant behavior, did demonstrate deficits associated with prenatal exposure. In this situation, the exposed animals showed retarded adaptation to shifts in reinforcement contingencies. Different monkey species as well as different dosing schedules may underlie the different findings.

POSTNATAL EXPOSURES IN HUMANS The primary question addressed by human developmental studies up to now is whether prenatal exposure to methylmercury is associated with adverse effects on behavioral function during childhood. The long-term impact of postnatal dietary exposure has largely been overlooked. Exposure doesn’t cease at birth, however. In populations in which fish comprises a significant part of the diet, it is consumed over a lifetime. We remain largely ignorant of the joint contribution to neurobehavioral outcomes of prenatal and postnatal exposure. The two largest epidemiological studies, one in the Faroe Islands (34), the other in the Seychelles (33), have provided data on pre-adolescent ages. A recent reanalysis of the Seychelles data (95) suggests the possibility that postnatal exposure may incur subtle adverse effects that only become apparent as children approach adolescence. Nonlinear multiple regression analyses of scores on the Child Behavior Checklist at 107 months of age showed a significant adverse association, on the Thought Problems subscale, with contemporary hair levels of methylmercury. Figure 4 shows a partial residual plot for depicting the association (pZ0.002). No effects are seen below approximately 8 ppm, but there is an increasing trend (adverse effect) above that level. Thought Problems

Smoothed Postnatal MeHg

0.4 0.3 0.2 0.1 0.0

0

5

10 15 Postnatal MeHg

20

25

Figure 4 Plot showing the relationship between scores on the Thought Problems subscale of the Child Behavior Checklist and postnatal methylmercury exposure as measured by hair levels at 107 months of age. The analysis was based on the application of semiparametric additive models. Above about 10 ppm the data points are fewer and the confidence intervals widen. The rug plot (vertical marks) along the bottom illustrates the distribution of postnatal exposure in this cohort. Source: From Ref. 95.

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

17

These results, indicating effects of postnatal exposure, to some extent parallel those of Murata et al. (96) with the Faroe Islands cohort on brainstem auditory evoked responses at 14 years of age. When tested at age 7 years, (34,97), the cohort exhibited delayed latencies associated with elevated prenatal exposures based on cord blood. The newer data, based on age 14 years, showed increased delays associated, as well, with current exposures as measured by the children’s hair levels of mercury. Subtle adverse consequences emerging during later childhood and adolescence, as also seen in experimental animals, should come as no surprise. Adolescence, like the perinatal period, is a tumultuous stage of development during which profound changes are occurring in brain function (84). Adams and colleagues (98) noted that “in addition to the dramatic neuroendocrinological and physical changes associated with adolescence.it is clear that the brain of the adolescent also undergoes striking transformations.Several brain regions undergo prominent remodeling.These regions include the prefrontal cortex.These data suggest that the adolescent period of brain development should be a time of particular vulnerability to insult.However, there has been surprisingly little investigation in either humans or animals of the vulnerability of the adolescent brain to developmental perturbation. Moreover, adverse impacts on neurocognitive and behavioral functions during such periods are likely to have protracted consequences for latter success.” This question deserves to be urgently pursued in methylmercury research. These recent data are especially relevant because, in fish-consuming populations, prenatal and postnatal exposures are entwined. However, prenatal and postnatal exposures seem to result in different patterns of brain damage, and, as a result, tend to produce different phenotypes. Prenatal exposure disrupts cell migration and differentiation throughout the developing brain in mice (19) and in humans (20). Such extensive damage would be expected to induce widespread impairment manifested as deficits in cognitive function, sensory function, motor function, information processing and language, memory, attention, and social communication. The syndrome of what Japanese investigators labeled Congenital Minamata Disease, which was marked overtly by neurological signs, included such global effects. In contrast, postnatal exposure, at least in adults, appears to be predominantly focal with most damage centered in areas deep in the sulci such as the calcarine fissure and the folds of the cerebellum in both nonhuman primates (99) and humans (9). It is this pattern of damage that accounts for the signs and symptoms associated with methylmercury poisoning (Table 1) before the sensitivity of the developing brain was appreciated. Aging The emphasis on early development has tended to obscure a larger issue: the consequences of lifetime exposure. The overwhelmingly predominant source of methylmercury is fish and other aquatic species. Fish consumption takes place through a lifetime and, during advanced age, the brain, as seen in animal studies, may begin to reflect the consequences of cumulative exposure. In fact, in its investigation of the catastrophe at Minamata, Japan, the Kumamoto University committee made the following statement (100): “.the problem about the relationship between a small amount of methylmercury pollution for a long period and its accumulation in the brain still remains obscure.Subclinical Minamata disease was sometimes revealed by detectable symptoms during the aging process.The subclinical Minamata disease.could be called a delayed type of Minamata disease in aged people.” A more quantitative assessment of this finding can be found in Kinjo et al. (101). They compared over one thousand Minamata disease patients aged 40 or over with

18

Weiss

controls matched by age and sex. A questionnaire interview surveyed subjective complaints and activities of daily living (ADL). The patients had significantly higher rates of all complaints than the controls, particularly in the categories of sensory disturbances such as paresthesias and motor difficulties such as tremor and weakness. Differences between patients and controls increased with age, and ADL disability in MD patients was aggravated by aging. The observations in Minamata emphasize how crucial it is, to more fully understand the risks stemming from developmental exposures, to study outcomes not only during the relatively early phases of the life cycle, but throughout the lifetime (87,102). The risks are ambiguous because early, low-level exposures may produce undetected, latent (103), or “silent” damage (104,105) that will be manifested only when functional capacities are challenged by other conditions, such as aging (106,107), drugs (108,109), or complex behavioral situations (86,87). The effects ascribed to aging underscore the principle that neurotoxicity may be manifested differently at different stages of the life cycle (88,92). Certainly, older humans handle drugs and toxicants differently from younger adults who handle them differently from children and neonates. Furthermore, the age question deserves a greater emphasis than it has received in the past. It is important to ask how such exposures might affect “aging” processes themselves (110). The pioneering experiments of Spyker (15) and Spyker et al. (111), the primate studies of Rice (90,92), and the aging models offered by Weiss and Simon (107) and Weiss et al. (106) show how these issues become closely intertwined when addressing the long-term consequences of exposures to toxic agents during early development.

UNRESOLVED ISSUES Despite the flood of research unleashed by the disasters in Minamata and Iraq, perplexing questions about the risks to neurobehavioral development posed by methylmercury persist. One stimulus for such questions is the source of exposure. Fish is an essential part of the diet for many populations, so exposure is inescapable. Even in populations where other protein sources are readily available, such as ours, fish remains an important food. Whatever risks are associated with fish consumption need to be weighed against its nutritional properties, some of which were discussed earlier. Moreover, methylmercury is hardly the only contaminant to be considered in evaluating the potential risks of fish consumption. Our grasp of how such risks combine is relatively primitive (56). Second, we must recognize that the selection of endpoints for assessing methylmercury neurotoxicity reveals incompatabilities between epidemiological and laboratory research. The human studies, in part because of their reliance on standardized tests—perhaps inescapably—have featured cognitive function. Animal studies indicate deficits in motor and sensory function to be the clearest outcomes of exposure, in parallel with poisonings in adult humans. The appropriate endpoints need to be pursued more intensively in developmentally-exposed humans. Finally, the issue that has been least adequately addressed is how to formulate risks based on lifetime exposure. We are especially deficient in our understanding of the combined effects of prenatal and postnatal exposure. Development is a continuous process, although sometimes treated as though it is separated into a series of individual journeys. Delayed or latent effects, and the contribution of aging, with its accompanying decline in compensatory processes, have not been incorporated into the risk equation.

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

19

REFERENCES 1. Hunter D. The Diseases of Occupations. Boston: Little, Brown, 1969. 2. Weiss B. Perspectives on methylmercury as a global health hazard. Neurotoxicology 1995; 16:577–578. 3. Miller MW, Berg GG, eds. Chemical Fallout. Springfield, IL: CC Thomas, 1969. 4. Jensen S, Jernelov A. Biological methylation of mercury in aquatic organisms. Nature 1969; 223:753–754. 5. Carson R. Silent Spring. Boston: Houghton Mifflin, 1962. 6. Bakir F, Damluji SF, Amin-Zaki L, et al. Methylmercury poisoning in Iraq. Science 1973; 181:230–241. 7. Smith WE, Smith AM. Minamata. Austin, TX: Rinehart and Winston, 1975. 8. Eto K, Yasutake A, Nakano A, Akagi H, Tokunaga H, Kojima T. Reappraisal of the historic 1959 cat experiment in Minamata by the Chisso Factory. Tohoku J Exp Med 2001; 194:197–203. 9. Eto K. Minamata disease. Neuropathology 2000; 20:S14–S19. 10. Harada Y. Congenital Minamata disease. In: Tsubaki R, Irukayama K, eds. Minamata disease: Methyl mercury poisoning in Minamata and Niigata. Tokyo, Japan: Kodansha, 1977:209–239. 11. Saito H, Sekikawa T, Taguchi J, et al. Prenatal and postnatal methyl mercury exposure in Niigata, Japan: adult outcomes. Seychelles Med Dent J 2004; 7:138–146. 12. Myers GJ, Davidson PW, Cox C, Shamlaye C, Cernichiari E, Clarkson TW. Twenty-seven years studying the human neurotoxicity of methylmercury exposure. Environ Res 2000; 83:275–285. 13. Cox C, Clarkson TW, Marsh DO, Amin-Zaki L, Tikriti S, Myers GG. Dose-response analysis of infants prenatally exposed to methyl mercury: an application of a single compartment model to single-strand hair analysis. Environ Res 1989; 49:318–332. 14. Davidson PW, Myers GJ, Weiss B. Mercury exposure and child development outcomes. Pediatrics 2004; 113:1023–1029. 15. Spyker JM. Behavioral teratology and toxicology. In: Weiss B, Laties VG, eds. Behavioral Toxicology. New York: Plenum Press, 1975:311–349. 16. Watanabe C, Satoh H. Evolution of our understanding of methylmercury as a health threat. Environ Health Perspect 1996; 104:367–379. 17. Choi BH, Lapham LW, Amin-Zaki L, Saleem T. Abnormal neuronal migration, deranged cerebral cortical organization. and diffuse white matter astrocytosis of fetal human brain: a major effect of methylmercury poisoning in utero. J Neuropathol Exp Neurol 1978; 37:719–733. 18. Sager PR, Aschner M, Rodier PM. Persistent, differential alterations in developing cerebellar cortex of male and female mice after methylmercury exposure. Brain Res 1984; 314:1–11. 19. Rodier PM, Aschner M, Sager PR. Mitotic arrest in the developing CNS after prenatal exposure to methylmercury. Neurobehav Toxicol Teratol 1984; 6:379–385. 20. Choi BH. The effects of methylmercury on the developing brain. Prog Neurobiol 1989; 32:447–470. 21. Cordier S, Garel M, Mandereau L, Morcel H, Doineau P, Gosme-Seguret S, et al. Neurodevelopmental investigations among methylmercury-exposed children in French Guiana. Environ Res 2002; 89:1–11. 22. Dolbec J, Mergler D, Passos CJS, de Morais SS, Lebel J. Methylmercury exposure affects motor performance of a riverine population of the Tapajos river, Brazilian Amazon. Int Arch Occup Environ Health 2000; 73:195–203. 23. Kjellstrom T, Kennedy P, Wallis S, Mantell C. Physical and mental development of children with prenatal exposure to mercury from fish. Stage 1. Preliminary tests at age 4. Solna, Sweden: National Swedish Environmental Board Report No. 3080, 1986.

20

Weiss

24. Kjellstrom T, Kennedy P, Wallis S, et al. Physical and mental development of children with prenatal exposure to mercury from fish. Stage 2. Interviews and psychological tests at age 6.Solna, Sweden: National Swedish Environmental Board Report No. 3642, 1989. 25. Lebel J, Mergler D, Branches F, et al. Neurotoxic effects of low-level methylmercury contamination in the Amazonian Basin. Environ Res 1998; 79:20–32. 26. McKeown-Eyssen GE, Ruedy J, Neims A. Methyl mercury exposure in northern Quebec. II. Neurologic findings in children. Am J Epidemiol 1983; 118:470–479. 27. Ramirez GB, Pagulayan O, Akagi H, et al. Tagum study II: Follow-up study at two years of age after prenatal exposure to mercury. Pediatrics 2003; 111:e289–e295. 28. McDowell MA, Dillon CF, Osterloh J, et al. Hair mercury levels in U.S. children and women of childbearing age: reference range data from NHANES 1999–2000. Environ Health Perspect 2004; 112:1165–1171. 29. Rice DC. The US EPA reference dose for methylmercury: sources of uncertainty. Environ Res 2004; 95:406–413. 30. Stern AH, Smith AE. An assessment of the cord blood:maternal blood methylmercury ratio: implications for risk assessment. Environ Health Perspect 2003; 111:1465–1670. 31. Cernichiari E, Brewer R, Myers GJ, et al. Monitoring methylmercury during pregnancy: maternal hair predicts fetal brain exposure. Neurotoxicology 1995; 16:705–710. 32. Mahaffey KR, Clickner RP, Bodurow CC. Blood organic mercury and dietary mercury intake: national health and nutrition examination survey, 1999 and 2000. Environ Health Perspect 2004; 112:562–570. 33. Myers GJ, Davidson PW, Cox C, et al. Prenatal methylmercury exposure from ocean fish consumption in the Seychelles child development study. Lancet 2003; 361:1686–1692. 34. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-years-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 35. National Research Council. Committee on the Toxicological Effects of Methylmercury. Toxicological Effects of Methylmercury. Washington, DC: National Academy Press, 2000. 36. Crump KS, Van Landingham C, Shamlaye C, et al. Benchmark concentrations for methylmercury obtained from the Seychelles Child Development Study. Environ Health Perspect 2000; 108:257–263. 37. U.S. EPA. Mercury Report to Congress, Volume VI: Characterization of Human Health and Wildlife Risks from Anthropogenic Mercury Emissions in the United States. EPA452/R-97-001f. Washington, DC:U.S. Environmental Protection Agency, 1997. 38. Vahter M, Mottet NK, Friberg L, Lind B, Shen DD, Burbacher T. Speciation of mercury in the primate blood and brain following long-term exposure to methyl mercury. Toxicol Appl Pharmacol 1994; 124:221–229. 39. Evans HL, Garman RH, Weiss B. Methylmercury: exposure duration and regional distribution as determinants of neurtoxicity in nonhuman primates. Toxicol Appl Pharmacol 1977; 41:15–33. 40. Rice DC. Brain and tissue levels of mercury after chronic methylmercury exposure in the monkey. Toxicol Environ Health 1989; 27:189–198. 41. Stinson CH, Shen DM, Burbacher TM, Mohamed MK, Mottet NK. Kinetics of methyl mercury in blood and brain during chronic exposure in the monkey Macaca fascicularis. Pharmacol Toxicol 1989; 65:223–230. 42. Lapham LW, Cernichiari E, Cox C, et al. An analysis of autopsy brain tissue from infants prenatally exposed to methymercury. Neurotoxicology 1995; 16:689–704. 43. Burbacher TM, Rodier PM, Weiss B. Methylmercury developmental neurotoxicity: a comparison of effects in humans and animals. Neurotoxicol Teratol 1990; 2:191–202. 44. Calabrese EJ, Baldwin LA. Hormesis: the dose-response revolution. Ann Rev Pharmacol Toxicol 2003; 43:175–197. 45. Davidson PW, Myers GJ, Cox C, et al. Effects of prenatal and postnatal methylmercury exposure from fish consumption on neurodevelopment: outcomes at 66 months of age in the Seychelles Child Development Study. JAMA 1998; 280:701–707.

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

21

46. Hurley LS. The roles of trace elements in foetal and neonatal development. Philos Trans R Soc Lond B Biol Sci 1981; 294:145–152. 47. Tran TT, Chowanadisai W, Crinella FM, Chicz-DeMet A, Lonnerdal B. Effect of high dietary manganese intake of neonatal rats on tissue mineral accumulation, striatal dopamine levels, and neurodevelopmental status. Neurotoxicology 2002; 23:635–643. 48. Zeisel SH. Choline: needed for normal development of memory. J Am Coll Nutr 2000; 19:528S–531S. 49. Meck WH, Williams CL. Simultaneous temporal processing is sensitive to prenatal choline availability in mature and aged rats. Neuroreport 1997; 8:3045–3051. 50. Richardson AJ. Clinical trials of fatty acid treatment in ADHD, dyslexia, dyspraxia and the autistic spectrum. Prostaglandins Leukot Essent Fatty Acids 2004; 70:383–390. 51. Weiss B, Amler S, Amler RW. Pesticides. Pediatrics 2004; 113:1030–1036. 52. Hites RA, Foran JA, Carpenter DO, Hamilton MC, Knuth BA, Schwager SJ. Global assessment of organic contaminants in farmed salmon. Science 2004; 303:226–229. 53. Broadhurst CL, Wang Y, Crawford MA, Cunnane SC, Parkington JE, Schmidt WF. Brainspecific lipids from marine, lacustrine, or terrestrial food resources: potential impact on early African Homo sapiens. Comp Biochem Physiol B Biochem Mol Biol 2002; 131:653–673. 54. Fangstrom B, Athanasiadou M, Grandjean P, Weihe P, Bergman A. Hydroxylated PCB metabolites and PCBs in serum from pregnant Faroese women. Environ Health Perspect 2002; 110:895–899. 55. Grandjean P, Weihe P, Burse VW, et al. Neurobehavioral deficits associated with PCB in 7-years-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001; 23:305–317. 56. Newland MC. Neurobehavioral toxicity of methylmercury and PCBs: Effects-profile and sensitive populations. Environ Toxicol Pharmacol 2002; 12:119–128. 57. Stewart PW, Reihman J, Lonky EI, Darvill TJ, Pagano J. Cognitive development in preschool children prenatally exposed to PCBs and MeHg. Neurotoxicol Teratol 2003; 25:11–22. 58. Grandjean P, Budtz-Jorgensen E, Steuerwald U, et al. Attenuated growth of breast-fed children exposed to increased concentrations of methylmercury and polychlorinated biphenyls. FASEB J 2003; 17:699–701. 59. Bellinger DC. Effect modification in epidemiologic studies of low-level neurotoxicant exposures and health outcomes. Neurotoxicol Teratol 2000; 22:133–140. 60. Riley DM, Newby CA, Leal-Almeraz TO, Thomas VM. Assessing elemental mercury vapor exposure from cultural and religious practices. Environ Health Perspect 2001; 109:779–784. 61. U.S. Centers for Disease Control. Mercury poisoning associated with beauty cream–Texas, New Mexico, and California, 1995–1996. Morbidity Mortality Weekly 1996; 45:400–403. 62. Berlin M. Mercury in dental amalgam: a risk analysis. Seychelles Med Dent J 2004; 7:154–158. 63. Drasch G, Schupp I, Hofl H, Reinke R, Roider G. Mercury burden of human fetal and infant tissues. Eur J Pediat 1994; 153:607–610. 64. Bjornberg KA, Vahter M, Petersson-Grawe K, et al. Methyl mercury and inorganic mercury in Swedish pregnant women and in cord blood: influence of fish consumption. Environ Health Perspect 2003; 111:637–641. 65. Salvolini E, Di Giorgio R, Curatola A, Mazzanti L, Fratto G. Biochemical modifications of human whole saliva induced by pregnancy. Br J Obstet Gynaecol 1998; 105:656–660. 66. Clarkson TW. The pharmacology of mercury compounds. Ann Rev Pharmacol 1972; 12:375–406. 67. Clarkson TW, Friberg L, Hursh JB, Nylander M. The prediction of mercury intake from amalgams. In: Clarkson TW, Friberg L, Nordberg FG, Sager PR, eds. Biological Monitoring of Toxic Meals. New York: Plenum Press, 1988:246–260. 68. Clarkson TW. The three modern faces of mercury. Environ Health Perspect 2002; 110:11–23. 69. Goering PL, Galloway WD, Clarkson TW, Lorscheider FL, Berlin M, Rowland AS. Toxicity assessment of mercury vapor from dental amalgams. Fundam Appl Toxicol 1992; 19:319–329.

22

Weiss

70. Weiss B, Clarkson TW. Mercury toxicity in children. In: Chemical and Radiation Hazards to Children. Ross Pediatric Conferences, 1982:52–58. 71. Warkany J, Hubbard DM. Adverse mercurial reactions in the form of acrodynia and related conditions. Am J Dis Child 1951; 81:335–373. 72. Curtis HA, Ferguson SD, Kell RL, Samuel AH. Mercury as a health hazard. Arch Dis Child 1987; 62:293–295. 73. Warfvinge K. Mercury distribution in the neonatal and adult cerebellum after mercury vapor exposure of pregnant squirrel monkeys. Environ Res 2000; 83:93–101. 74. World Heath Organization. Environmental Health Criteria 101, Methylmercury. Geneva, Switzerland: World Health Organization, 1990. 75. Vahter M, Akesson A, Lind B, Bjors U, Schutz A, Berglund M. Longitudinal study of methylmercury and inorganic mercury in blood and urine of pregnant and lactating women, as well as in umbilical cord blood. Environ Res 2000; 84:186–194. 76. Schober SE, Sinks TH, Jones RL, et al. Blood mercury levels in US children and women of childbearing age, 1999–2000. JAMA 2003; 289:1667–1674. 77. Fredriksson A, Dencker L, Archer T, Danielsson BR. Prenatal coexposure to metallic mercury vapour and methylmercury produce interactive behavioural changes in adult rats. Neurotoxicol Teratol 1996; 18:129–134. 78. Bilrha H, Roy R, Moreau B, Belles-Isles M, Dewailly E, Ayotte P. In vitro activation of cord blood mononuclear cells and cytokine production in a remote coastal population exposed to organochlorines and methylmercury. Environ Health Perspect 2003; 111:1952–1957. 79. Oskarsson A, Schultz A, Skerfving S, Hallen IP, Ohlin B, Lagerkvist BJ. Total and inorganic mercury in breast milk in relation to fish consumption and amalgam in lactating women. Arch Environ Health 1996; 51:234–241. 80. Weiss B, Reuhl K. Delayed neurotoxicity: a silent toxicity. In: Chang LW, ed. Principles of Neurotoxicology. New York: Marcel Dekker, 1994:765–784. 81. Kennard MA. Cortical reorganization of motor function. Studies on series of monkeys of various ages from infancy to maturity. Arch Neurol Psychiat 1942; 48:227–240. 82. Goldman PS. An alternative to deelopmental plasticity: heterology of CNS structures in infants and and adults. In: Stein DG, Rosen JJ, Butters N, eds. Plasticity and Recovery of Function in the Central Nervous System. New York: Academic Press, 1974:149–174. 83. Johnson D, Almli CR. Age, brain damage, and performance. In: Finger S, ed. Recovery from brain damage. Research and theory. New York: Plenum Press, 1978:115–134. 84. Gogtay N, Giedd JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis AC, et al. Dynamic mapping of human cortical development during childhood through early adulthood. Proc Natl Acad Sci USA 2004; 10121:8174–8179. 85. Lenneberg EH. The effect of age on the outcome of cortical nervous system disease in children. In: Isaacson RI, ed. The Neuropsychology of Development. New York: Wiley, 1968. 86. Newland MC, Rasmussen EB. Aging unmasks adverse effects of gestational exposure to methylmercury in rats. Neurotoxicol Teratol 2000; 22:819–828. 87. Newland MC, Reile PA, Langston JL. Gestational exposure to methylmercury retards choice in transition in aging rats. Neurotoxicol Teratol 2004; 26:179–194. 88. Rice DC. Age-related increase in auditory impairment in monkeys exposed in utero plus postnatally to methylmercury. Toxicol Sci 1998; 44:191–196. 89. Rice DC, Gilbert SG. Effects of developmental methylmercury exposure or lifetime lead exposure on vibration sensitivity function in monkeys. Toxicol Appl Pharmacol 1995; 134:161–169. 90. Rice DC. Evidence for delayed neurotoxicity produced by methylmercury. Neurotoxicology 1996; 17:583–596. 91. Igata A. Epidemiological and clinical features of Minamata Disease. In: Suzuki T, Imura N, Clarkson TW, eds. Advances in Mercury Toxicology. New York: Plenum Press, 1991:439–457. 92. Rice DC. Effects of pre- plus postnatal exposure to methylmercury in the monkey on fixed interval and discrimination reversal performance. Neurotoxicology 1992; 13:443–452.

Methylmercury: A Model Neurotoxicant and Risk Assessment Dilemma

23

93. Gilbert SG, Burbacher TM, Rice DC. Effects of in utero methylmercury exposure on a spatial delayed alternation task in monkeys. Toxicol Appl Pharmacol 1993; 123:130–136. 94. Newland MC, Yezhou S, Logdberg B, Berlin M. Prolonged behavioral effects of in utero exposure to lead or methyl mercury–reduced sensitivity to changes in reinforcement contingencies during behavioral transitions and in steady state. Tox Appl Pharmacol 1994; 126:6–15. 95. Myers GJ, Davidson PW, Shamlaye CF, et al. The Seychelles Child Development Study of methylmercury from fish consumption: analysis of subscales from the Child Behavior Checklist at age 107 months in the main cohort. Seychelles Med Dent J 2004; 7:107–114. 96. Murata K, Weihe P, Budtz-Jorgensen E, Jorgensen PJ, Grandjean P. Delayed brainstem auditory evoked potential latencies in 14-years-old children exposed to methylmercury. J Pediatr 2004; 144:177–183. 97. Murata K, Weihe P, Renzoni A, et al. Delayed evoked potentials in children exposed to methylmercury from seafood. Neurotoxicol Teratol 1999; 21:343–348. 98. Adams J, Barone S, Jr., LaMantia A, et al. Workshop to identify critical windows of exposure for children’s health: neurobehavioral work group summary. Environ Health Perspect 2000; 108:535–544. 99. Garman RH, Weiss B, Evans HL. Alkylmercurial encephalopathy in the monkey (Saimiri sciureus and Macaca arctoides): a histopathologic and autoradiographic study. Acta Neuropathologica 1975; 32:61–74. 100. Research Committee on Minamata Disease. Kumanoto, Japan: Kumamoto University, 1973. 101. Kinjo Y, Higashi H, Nakano A, Sakamoto M, Sakai R. Profile of subjective complaints and activities of daily living among current patients with Minamata disease after 3 decades. Environ Res 1993; 63:241–251. 102. Berlin CM, Kacew S. Environmental chemicals in human milk. In: Kacew S, Lambert GH, eds. Environmental Toxicology and Pharmacology of Human Development. Washington, DC: Taylor & Francis, 1997:67–93. 103. Isaacson RL. The myth of recovery from early brain damage. In: Ellis NR, ed. Aberrant Development in Infancy: Human and Animal Studies. Hillsdale, NJ: Erlbaum, 1975:1–25. 104. Grant CA. Pathology of experimental methylmercury intoxication: Some problems of exposure and response. In: Miller MW, Clarkson TW, eds. Mercury, Mercurials, and Mercaptans. Springfield, IL: Charles C. Thomas, 1973:294–312. 105. Reuhl KR, Chang LW. Effects of methylmercury on the development of the nervous system. Neurotoxicology 1979; 1:21–55. 106. Weiss B, Clarkson TW, Simon W. Silent latency periods in methylmercury poisoning and in neurodegenerative disease. Environ Health Perspect 2002; 110:851–854. 107. Weiss B, Simon W. Quantitative Perspectives on the long-term toxicity of methylmercury and similar poisons. In: Weiss B, Laties VG, eds. Behavioral Toxicology. New York: Plenum, 1975:429–435. 108. Eccles CU, Annau Z. Prenatal methyl mercury exposure, II: Alterations in learning and psychotropic drug sensitivity in adult offspring. Neurobehav Toxicol Teratol 1982; 4:377–382. 109. Rasmussen EB, Newland MC. Developmental exposure to methylmercury alters behavioral sensitivity to D-amphetamine and pentobarbital in adult rats. Neurotoxicol Teratol 2001; 23:45–55. 110. Williams JR, Spencer PR, Stahl SM, et al. Interactions of aging and environmental agents: the toxicological perspective. In: Baker SR, Rogul M et al, eds. Environmental Toxicity and the Aging Processes. New York: Liss, 1987:81–135. 111. Spyker JM, Sparber SB, Goldberg AM. Subtle consequences of methylmercury exposure: behavioral deviations in offspring of treated mothers. Science 1972; 177:621–623.

2 Developmental Neurotoxicity Associated with Dietary Exposure to Methylmercury from Seafood and Freshwater Fish Philippe Grandjean Department of Environmental Medicine, University of Southern Denmark, Odense, Denmark and Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A.

Sylvaine Cordier National Institute for Health and Medical Research (INSERM), Rennes, France

Tord Kjellstro¨m National Institute of Public Health, Stockholm, Sweden and Australian National University, Canberra, Australian Capital Territory, Australia and Wellington School of Medicine and Health Sciences, Wellington, New Zealand

INTRODUCTION Methylmercury is a ubiquitous contaminant of seafood and freshwater fish (1). Human exposures to this toxicant have increased over time, due to anthropogenic mercury pollution, and because modern fishing technology allows catching large predatory species that accumulate methylmercury. Dramatic reminders of the neurotoxic potential of methylmercury occurred in the poisoning incidents in Minamata and Niigata, Japan, in the 1950s and 1960s and the subsequent contamination of bread in Iraq in 1971 from methylmercury-treated seed grains (1,2). Recent assessments of methylmercury neurotoxicity have been published by national and international bodies, e.g., the U.S. National Research Council (3), the U.S. Environmental Protection Agency (4), and the Joint FAO/WHO Expert Committee on Food Additives (5). While agreeing that the developing brain is the main target for methylmercury, the reports suggest that two of the largest prospective studies of developmental neurotoxicity have reached “different” conclusions. Indeed, one study seems to disagree with several others, but the neurotoxicology literature includes many examples of studies that are ‘non-positive’ without offering convincing evidence that neurotoxic effects were absent (6). 25

26

Grandjean et al.

Like other scientific inquiries, epidemiological studies will always render tentative knowledge, since no scientific process can provide absolute proof. However, the observational studies carried out so far to explore the low-level neurotoxicity of methylmercury offer strong evidence of the public health impact of exposures that are common in many populations that depend on marine food or freshwater fish. The present chapter will review the main issues in regard to the current risk assessment of methylmercury.

PROSPECTIVE STUDY DESIGNS AND SETTINGS In the absence of clinical controlled trials, prospective epidemiological studies provide evidence of the greatest importance. When choosing a setting, the researchers must try to maximize the statistical power and validity of the study by securing a wide exposure interval, a high participation rate, also at follow-up, feasibility of applying sensitive outcome measures, and limited impact of potential confounders. Three major prospective cohort studies on adverse effects of prenatal methylmercury exposure have been published, each conducted in a population with a high average intake of seafood. The first study was carried out in New Zealand. A group of 11,000 new mothers was initially screened by a questionnaire survey at the time their child was born in 1978; of these, about 1000 had consumed at least 3 fish meals per week during pregnancy. Their hair mercury levels were analyzed (7). Seventy-three mothers ‘one of whom had twins’, had hair mercury levels above 6 mg/g and were defined as the “high-exposure group”. High mercury exposure from seafood was mainly due to consumption of sharks, which in New Zealand usually contain methylmercury concentrations of 1–5 mg/g (7). At age 4 years, 31 highexposure children and 31 reference children with lower exposure were matched for potential confounders, i.e., mother’s ethnic group, age, child’s birthplace and birth date. The highexposure group showed lower scores on the Denver Developmental Screening test (7). A follow-up of the original cohort was carried out at age 6 years, now with three control groups with lower prenatal mercury exposure (8). During pregnancy, mothers in two of these control groups had high fish consumption and average hair mercury concentrations of 3–6 mg/g and 0–3 mg/g, respectively. Matching parameters were maternal ethnic group, age, smoking habits, residence, and sex of the child. At this time, 61 of 74 high exposure children were available for examinations (8). Lead levels in cord blood and garden soil were tested to assess potential confounding, but there was no association between lead and methylmercury exposure. Results of the psychological performance tests correlated well. Stepwise robust multiple regression analysis showed that the full (and performance) Wechsler Intelligence Scale for Children (WISC-R), the McCarthy Scales of Children’s Abilities (perceptual and motoric), and the Test of Oral Language Development were most strongly associated with the maternal hair mercury concentration (8). The proportion of the variance in test results due to hair mercury above 6 mg/g was about 2%, which was similar to the influence of social class and home language, two of the main confounders accounted for in the analysis. The robust regression analysis reduced the impact of one extreme outlier (with maternal hair-mercury above 80 mg/g). A reanalysis of the full database of this study (9) replicated the association between high maternal mercury exposure and reduced test performance, but the statistical significance was very much influenced by the extreme outlier. When this subject was excluded some tests became statistically significant. Later on, a study was initiated in the Faroe Islands in the North Atlantic (10). The first birth cohort consisted of 1,022 children born during a 21-month period in 1986–1987.

Methylmercury from Seafood and Freshwater Fish

27

The Faroes constitute a prosperous fishing community where excess exposure to methylmercury is widely prevalent because of the traditional habit of eating meat from the pilot whale. Ingestion of whale blubber causes exposure to lipophilic contaminants, notably polychlorinated biphenyls (PCBs). Assessment of prenatal methylmercury exposure was based on mercury concentrations in cord blood and in maternal hair; they spanned a range of about 1000-fold. Lead levels in cord blood were low. All cohort members were invited for the first detailed examination at school age (7 years), and a total of 917 of eligible children (90.3%) completed the examinations that lasted about 5 hours. The physical examination included a sensory function assessment and a functional neurological examination with emphasis on motor coordination and perceptual-motor performance. Main emphasis was placed on detailed neurophysiological and neuropsychological function tests that had been selected as sensitive indicators of abnormalities thought to be caused by methylmercury. A repeat examination was carried out at age 14 years, again with a high participation rate, and the clinical test battery was very similar to the one previously applied, with proper adjustment for the participants now being adolescents. The main finding at the 7-year follow-up in the Faroes was that decrements in attention, language, verbal memory, and, to a lesser extent, in motor speed and visuospatial function, were associated with prenatal methylmercury exposure; the cord– blood mercury concentration was the best risk indicator (10). These findings were robust in the full Faroes data set in analyses controlled for age, sex and confounders, and they persisted after exclusion of high-exposure subjects. Support for these findings were seen in some of the neurological tests, but particularly in delays in brainstem auditory evoked potentials (11). Likewise, prenatal methylmercury exposure is associated with a decrease in the normal heart rate variability, and this indication of poorer nervous system control of heart function is reflected in a tendency of increased blood pressure (12). Data on the 14-year follow-up are currently being processed, but two recent publications address the neurophysiological outcomes that are unlikely to be affected by socioeconomic confounders (13,14). Mercury-associated delays in brainstem auditory evoked potentials remained at 14 years, as did the decreased heart rate variability. These two functions were associated, perhaps due to their joint involvement of brainstem nuclei, but the association became weaker when adjusted for mercury exposure. These findings suggest that brainstem toxicity may be an important component of developmental methylmercury neurotoxicity. The brainstem auditory evoked potentials also showed a delay in peak V, i.e., the signal elicited by the transmission of the electrical signal to the midbrain, but this delay was associated only with the current methylmercury exposure (Fig. 1) (13). The average hair-mercury concentration in the adolescents examined was about 1 mg/g, i.e., close to the current Reference Dose (RfD) established by the U.S. EPA. This observation suggests that the vulnerability of the brain may extend into the teenage period, and that even exposures similar to the RfD may cause adverse effects. Two birth cohorts were formed in the Seychelles, both involving about 800 children, i.e., about 50% of all children born during the year (15). For exposure assessment, a hair sample was obtained from the mother, about six months after birth. The hair segment that represented the pregnancy period was identified from the assumption that hair grows 1.1 cm per month. The authors have noted that the first pilot study was not as wellcontrolled as the main or longitudinal study: there were fewer covariates, medical records were not reviewed as carefully, and there was less information on socio-economic status. A subset of 217 children from the pilot cohort was evaluated at 66 months (16). Maternal hair mercury was negatively associated with the McCarthy General Cognitive Index and Perceptual Performance subscale, and the Preschool Language Scale Total

28

Grandjean et al.

BAEP III – V 20 Hz (ms)

1.94 1.90 1.86 1.82 1.78 1.74 0.01

0.10 1.00 Child Hair Mercury Concentration (mg/g)

10.00

Figure 1 Latency of peak V of the brainstem auditory evoked potentials recorded in 859 Faroese children at 14 years adjusted for sex and age (27). The association is estimated in a generalized additive model analysis, using the current hair-mercury concentration as an indicator of current exposure. The broken lines indicate the point-wise 95% confidence interval for the doseresponse relationship. Each vertical line above the horizontal axis represents one observation at the exposure level indicated. The average exposure is similar to the U.S. EPA Reference dose (1 mg/g).

Language and Auditory Comprehension subscale. When statistically determined outliers and points considered to be influential were removed from the analyses, statistical significance of the association remained only for auditory comprehension. The main Seychelles study included evaluation of the children at 6.5, 19, 29 and 66 months of age, and again at 8 years. No association with maternal hair mercury was found for most endpoints in these children (16,17). At 29 months there was an association between mercury exposure and decreased activity level in boys only, who also showed a possible mercury-associated delay in age for walking, but the latter was not significant when adjusted for confounders. The authors cautioned that subtle neurological and neurobehavioral effects are more likely to be detected in older rather than younger children. The most detailed examination was carried out at age 8 years using tests that were similar to those applied in New Zealand and the Faroes. In calculating possible effects of prenatal methylmercury exposure, the regression equations included adjustment for postnatal exposure. No association between deficits and maternal hair-mercury concentrations was evident (17). Despite the apparent differences between the three studies of mercury-exposed populations, they may not necessarily be in disagreement. For example, the confidence intervals for the Faroes and Seychelles studies overlap, and the findings therefore do not significantly differ (18). Further, some differences would be anticipated, because these two studies used different methods for assessment of exposures and outcomes (Table 1), and due to different epidemiological settings. The New Zealand study population may be most similar to continental Europe and North America, even though about half of the high-exposure mothers were of Pacific Island descent. Their diet is high in ocean fish, and very few of them smoke, but otherwise their exposures in daily life would not be very different from other New Zealand women or, indeed, women from other Western societies.

Methylmercury from Seafood and Freshwater Fish Table 1

29

Main Differences Between the Prospective Studies of Methylmercury-Exposed Children

Attribute Source of exposure Exposure assessment Concomitant exposures Language Socioeconomic setting Family setting Outcome tests Clinical examiners

New Zealand

Faroes

Shark and other large Whale, ocean fish and ocean fish shellfish Maternal hair Cord blood and maternal hair Lead in house paint PCBs (whale blubber) and air English (and Pacific Faroese (and Danish) languages) Industrialized Industrialized Western Scandinavian Urban, mixed cultures Traditional Omnibus Domain-related and neurophysiological Clinical specialists Clinical specialists

Seychelles Ocean fish Maternal hair Pesticide use in tropics Creole (English and French) Middle-income developing Mainly matriarchal Omnibus and domainrelated Nurse/student

The neurotoxicity of methylmercury should not be judged from a single observational study but needs to be evaluated on the basis of the overall strength of the total evidence available. It is fortunate that three prospective projects have been carried out under different circumstances, and that supplementary information is available from cross-sectional studies. Their strengths and weaknesses need to be considered jointly. Following a review of the cross-sectional studies, we will consider the main issues in this assessment.

CROSS-SECTIONAL STUDIES Due to logistical difficulties in carrying out prospective studies in many exposed communities, most studies were cross-sectional (Table 2). In addition, some research teams have recently embarked on potential prospective studies, but results published so far warrant consideration only as cross-sectional. Again, researchers have chosen populations that include representation of high-level exposures to methylmercury. However, due to the remote setting, e.g., in the Arctic or the Amazon basin, less sophisticated parameters had to be chosen, also taking into account the possible differences in culture, language, and school education. Some of studies are particularly important, because they focus on the impact of methylmercury from freshwater fish. Thus, in the Amazonian region, fish contamination has been increased by local mercury pollution from gold-mining activities. In the Arctic, the traditional diet includes sea mammals and other species high in the food chain, thereby causing an increased risk of high intake of biomagnified methylmercury. The groups of children examined have ranged from tens to a few hundred. Selection bias, especially when studying older children, is possible, and usually no information was available about children who were unavailable for the study. Most likely such selection would result in healthier children being examined, thereby potentially obscuring exposurerelated effects. The studies also differ in regard to the exposure intervals covered. Exposure levels range from an average of about 5 mg/g maternal hair (Cree Indians in Northern Que´bec,

234 children (12–30 mo)

131 infant–mother pairs

21/36 children exposed (3–15 years) 15 non-exposed

149/153 children (7 yr)

354/420 children (7–12 yr)

Peru (20)

Ecuador (21)

Madeira (11)

Brazil (22)

Population (age range)

Canada (19)

Country (ref)

mgZ11.0 mg/g mgZ11.6 mg/g

Child hair Maternal hair (rZ0.80)

Freshwater fish (gold mining area)

Neurological exam Neuropsychological tests: Finger tapping, Hand–eye coordination test Continuous performance test WISC-R: digit spans block designs Stanford–Binet Bead memory test Neurophysiological tests: Brainstem auditory evoked potentials Pattern-reversal visual evoked potentials Neuropsychological tests: Finger-tapping Santa ana dexterity test WISC-R digit span Stanford-Binet tests: Copying Recall Bead memory

mgZ3.8 mg/g mgZ9.6 mg/g

Child hair Maternal hair

Audiological tests Auditory brainstem evoked potentials (10 children exposed)

maZ16.2 mg/L

Child’s blood at time of exam

Freshwater fish (gold mining area)

Marine fish

Growth parameters Developmental milestones

mgZ7 mg/g [1–29]

Growth parameters Neurological exam

Outcomes measured

maZ6 mg/g [0–24]

Level of exposure

Maternal hair during pregnancy

Maternal hair during pregnancy

Exposure biomarker

Marine fish

Freshwater fish

Main source of exposure

pZ0.003 pZ0.02 ns

ns pZ0.001 ns

Correlation I–III (and I–V) interpeak latency with maternal Hg Correlation N145 withmaternal Hg

ns

ns

ns Association with I-III interpeak latency (left ear only)

ns

Abnormalities of tendon reflexes (boys only, no dose-response)

Results

Table 2 Description of Cross-Sectional Studies on the Association Between Prenatal Exposure to Methylmercury and Neurodevelopmental Effects 30 Grandjean et al.

ns (but highly significant when merged with the Madeira data)

Abbreviations: ma, arithmetic mean; mg, geometric mean.

median Z1.63 mg/g median Z1.65 mg/g

Several neurophysiological tests: Brainstem auditory evoked potentials

Child hair Maternal hair

327 children (7 yr)

Akita, Japan (26)

Seafood

Correlation hand-eye coordination test with maternal Hg (pZ0.01)

Neurological exam Neuropsychological tests: Finger tapping Hand-eye coordination test Continuous performance test WISC-R: digit spans block designs Stanford-Binet Bead memory test Neurophysiological tests: Brainstem auditory evoked potentials Pattern-reversal visual evoked potentials

mgZ5.5 mg/g mgZ15.5 mg/g

Marine mam- Child hair mals (arctic Maternal hair traditional food)

43 children (7–12 yr)

Greenland (25)

ns pZ0.006

ns p!0.001 ns

ns

Increased tendon reflexes,

Neurological exam (Amiel-Tison) Neuropsychological tests Finger tapping Stanford-Binet test Block Copying Bead memory McCarthy Digit forward Leg coordination

mgZ10.2 mg/g mgZ12.7 mg/g

Child hair Maternal hair (high exposure area) (rZ0.30)

Freshwater fish (gold mining)

French 248/290 children Guiana (24) neurological ex. (6mo–6 yr) 206/243 children neuropsy. tests (5–12 yr)

pZ0.03

Neurological optimality scale

mgZ20.4 mg/L mgZ4.1 mg/g

Cord blood Maternal hair

182 infants (2 wk)

Maternal seafood diet

Faroe Islands (23)

Methylmercury from Seafood and Freshwater Fish 31

32

Grandjean et al.

Faroes, and two groups in the Amazon Basin) up to 13 mg/g in French Guiana, and even higher in Northern Greenland. Despite the differences in cultural settings and other limitations, these studies show a number of concordant findings when taking into account the different protocols and research strategies. One of the new prospective studies is the Faroese Cohort 2 that included 182 singleton term births. These children were first examined by the Neurological Optimality Score (NOS) at age two weeks (adjusted for gestational age). Detailed information was obtained on exposures both to methylmercury and to lipophilic pollutants. The NOS showed significant decreases at higher cord–blood mercury concentrations, while PCB was not important (23). A small study carried out in the northernmost Greenland included data on prenatal exposure levels on about half of the subjects (25). While anticipating the results from a newly initiated prospective study in Japan (27), a cross-sectional study has analyzed preserved umbilical cords as a measure of the children’s prenatal exposure levels (26). Because the developing brain is considered the main target of methylmercury toxicity, evaluation of these studies must be based on the assumption that exposures measured at the time of the postnatal examination represent causative exposures at the time of the greatest vulnerability of the nervous system. Postnatal exposure may of course add to the overall toxic impact, thereby complicating the evaluation if exposures have not been constant. Overall, these studies tend to confirm that attention, motor coordination and speed, and visuospatial function are sensitive targets of methylmercury toxicity. Language and verbal memory may also be vulnerable, but were not evaluated.

EXPOSURE ASSESSMENT A crucial element in the risk assessment process is a correct measure of exposure in terms of relevant period and accurate level. Because the exposure is not a matter of design, the validity of the exposure assessment depends on the degree to which the exposure parameters reflect the “true” exposure. Exposure biomarkers should be considered only proxy variables, which are always imprecise to some extent. This issue is important, because exposure misclassification is likely to be nondifferential and will therefore cause underestimation of the true effect of the exposure. In the prospective studies, samples for mercury analysis have included maternal hair, cord blood, and cord tissue collected at parturition. In the Seychelles (16), maternal hair was collected at a delay of 6 months after delivery. Some studies have also used maternal dietary questionnaires to obtain information on the origin and approximate magnitude of the methylmercury exposure. Cross-sectional studies, as opposed to cohort studies, rely on surrogate measures of fetal exposure, because the children were enrolled several years after birth. In the populations studied, it seems reasonable to assume that diet has changed little over time due to the preponderant role of fish among available resources. In addition, fish contamination by mercury has most likely remained fairly stable in the years following birth. Such stability was evidenced by the study in Peru, in which peak and average concentrations of mercury in maternal hair during pregnancy are very similar and in French Guiana where hair mercury levels measured five years apart in different villages were remarkably constant. Most studies in Table 2 used maternal hair collected at time of child’s examination. In two studies, the child’s exposure (blood or hair measurement) was used instead due

Methylmercury from Seafood and Freshwater Fish

33

to incomplete data on maternal exposure levels, and because a good correlation was documented between the child’s and the mother’s exposure levels (22). This association is to be expected when older children are examined in communities where they are usually sharing the meals with their parents at home. If the child’s current exposure does not provide a correct ranking between individuals, it could introduce exposure misclassification and thus underestimate the risk. However, if the child’s exposure is lower than the maternal level, an exaggeration of the risk may ensue. The length of maternal hair analyzed is also important for assessment of mercury exposures during pregnancy. Although hair growth rates are known to vary, a 90 mm hair sample obtained at parturition or shortly thereafter would be thought to represent the average mercury exposure during the whole pregnancy. In New Zealand, monthly (10 mm) exposure levels varied by a factor of 2, and on average the peak monthly exposure was about 50% higher than the 9-month average (8). In the Faroes, similar variations were recorded, with coefficients of variation mostly being below 25% (28). Comparison of hair from the Faroes and the Seychelles showed similar short-term variations (29). Until recently, the degree of imprecision has been assumed to be reflected by laboratory imprecisions, although these low levels of imprecision (usually about 5% or less) could not explain why associations between mercury concentrations in hair and blood often show wide scattering. Mercury analyses should always be supported by detailed quality control procedures (e.g., blind comparison with other experienced laboratories and use of certified samples) as was first done in the New Zealand study (8). Although frequently used for feasibility reasons, the maternal hair mercury concentration is likely to be a rather imprecise measure, particularly in regard to fetal exposure. Among sources of variability are hair growth rate, hair type, hair color, external contamination, and leaching due to permanent hair treatments (30). As a contributory reason for differences in study outcomes, temporal variability in exposure has been suggested. The concern is that ‘bolus’ exposures might be more toxic than steady exposures at an average level. This hypothesis is difficult to test, but experimental evidence (31) suggests that fetal brain concentrations change more slowly than maternal blood concentrations. In addition, exposure variability is likely to introduce error in the exposure assessment, and such misclassification would be likely to cause an underestimation of the dose–response relationship. In agreement with this prediction, exclusion of Faroese cohort subjects with variable exposures during gestation tended to increase the associations between the mercury exposure and the deficits (28). In international comparisons, three main types of hair structure are recognized (i.e., African, Caucasian, and Oriental), but good data for calibration with blood concentrations exists only for the latter two hair types. Thus, for the African population in the Seychelles, translation of hair-mercury results to blood concentrations and intake levels must currently be based on data mainly from Caucasian populations. Recent studies have now documented that the coefficient of variation for the hair-mercury imprecision is over 50%, i.e., twice the level found for the blood concentration (32). The overall effect of such non-differential imprecision is that the regression coefficients decrease, the P-values increase, and adjustment for confounders with better precision cause additional bias toward the null hypothesis (33). When calculating an exposure level from the hair mercury concentration, an average hair-to-blood ratio of 250 is generally used (4). This ratio is in accordance with recent evidence on Caucasian and Oriental hair (30), but is known to vary considerably between individuals. The 95th percentile differs from the median by a factor between 2 and 3. It also depends on the concentration level, and it changes with age (32).

34

Grandjean et al.

OUTCOME VARIABLES The validity of outcome variables depends on their sensitivity to the exposure under study and the specificity in this regard, i.e., lack of sensitivity to the influence of other factors, including confounders. The choice of effect parameters must at the same time be feasible and appropriate for the age of the children, and for the setting of the study. Tests that depend only minimally on cooperation of the subject have the advantage of being less likely to be affected by motivation. The more advanced neuropsychological tests are only possible when a child has reached school age. However, such tests may be of uncertain validity, if they have not previously been applied in the same culture. In addition, many tests require special skills of the examiner. All of these issues need to be considered when evaluating the study findings. Most studies employed a battery of neurobehavioral tests, and some appeared to be more sensitive to mercury neurotoxicity than others. Simple comparison of regression coefficients may provide suggestions for the most sensitive parameter, at least within the confines of a particular study. To facilitate such comparisons, the regression coefficient may be expressed as a proportion of the standard deviation of the test result, or as a delay in mental development calculated from the regression coefficient for age. Alternatively, benchmark dose levels may be used as a basis for comparison. Thus, the most sensitive neurological, neuropsychological, and neurophysiological effects parameters all exhibit benchmark dose levels of 5–10 mg/g hair. Despite the great variability of the study settings and the outcome variables, this observation suggests a substantial degree of concordance and that the combined evidence therefore is quite convincing in regard to the dose– response relationship.

Neurological Tests Many of the studies included a neurological exam (11,19,20,23,24). Unfortunately protocols of examination differ between studies: in the Canadian and French Guiana studies, evaluation of sensory functions, cranial signs, muscle tone, stretch reflexes and coordination were conducted in young children. Both studies reported abnormal tendon reflexes in association with maternal hair concentration. However, contrary to the expectation, these signs were mild and isolated, and the reproducibility of the assessment in the French Guiana study was reported to be poor (24). In the study in Peru (20), essential details of the neurological evaluation (items, age of the children examined) were not presented precluding informed expertise of their results. In this study, frequencies of abnormal signs were not associated with maternal mercury levels. In Madeira and the first Faroes cohort, the neurological examination put emphasis on motor coordination and perceptual motor performance (10,11); children who failed the most difficult of the 19 functional neurological tasks tended to have slightly higher exposures than children who performed well. In summary, the clinical neurological tests provide limited evidence linking lowdose methylmercury exposures to detectable abnormalities. It is very likely however that this absence of positive findings partly reflects the lack of sensitivity of this type of evaluation in this range of exposure levels. Thus, performance on a clinical test is rated by the examiner, thereby introducing a potential subjective aspect, and scoring is usually a simple pass-fail or pass-questionable-fail, thereby limiting the sensitivity.

Methylmercury from Seafood and Freshwater Fish

35

Developmental Tests Among the prospective studies, the New Zealand children were examined at 4 years of age using the Denver Developmental Screen Test (DDST). It consists of four major function sectors: gross motor, fine motor, language, and personal–social, where possible scores are abnormal, questionable, or normal. The prevalence for developmental delay in children was 52% for children of high mercury mothers and 17% for those of mothers of the reference group; the delays most frequently affected fine motor and language sectors. The Bayley Scales of Infant Development (BSID) was used in the Seychelles, but no mercuryassociated deficits were reported. Developmental tests may be useful for such studies in small children, they may be less dependent on differences in culture than are tests appropriate for older children, but they may be of limited sensitivity to subtle changes.

Neuropsychological Tests While likely to be more sensitive in revealing early neurotoxic changes, neuropsychological test results may also be affected if the administration is not standardized, and they may show examiner-dependence. Further, they may be sensitive to details in the test situation, such as the use of an interpreter, changes in temperature, and other aspects that may be important when a test is used for the first time in a particular culture. In New Zealand, two blinded psychologists tested the same children with a shortened version of the standard tests used and documented a remarkable agreement (8). In the Faroes and several other studies, tests were administered to all children by the same examiner, thus limiting the possible impact of examiner-related differences. Traditionally, studies in this field have included standard intelligence test batteries, because of the wealth of information available on such tests and the implications of lowlevel performance. The New Zealand study applied the Wechsler Intelligence Scale for Children (WISC) and McCarthy Scale of Children’s Abilities at age 6 years. Likewise, the Seychelles examinations included the McCarthy Scales. These intelligence tests may not be the most appropriate and sensitive for effects of methylmercury. Still, significant effects were found on the WISC and McCarthy in the New Zealand study. The New Zealand study also used the “Test of Oral Language Development” (a standard test widely used in New Zealand school programs), and this test appeared to be the most affected by methylmercury exposure. In the Faroes, the approach taken was to emphasize tests that reflected functional domains, e.g., attention, motor speed, and verbal memory. The functions chosen were those that were most likely to be affected by developmental methylmercury exposure, as judged from location of neuropathological lesions in poisoning cases and as illustrated by studies of other developmental neurotoxicants, especially lead. In the Faroes, the Boston Naming test appeared to be the most sensitive outcome. Similar tests were included when the Seychelles cohort was examined at age 8 years. However, application of the Boston Naming test in a different culture may not be unproblematic. For example, the sequence of stimuli may not necessarily represent an increasing degree of difficulty, e.g., if pictures of an igloo or an acorn are not as familiar to children in a tropical developing country as they are to northern children from the Faroes, or the United States, where the test was developed. In addition, each stimulus was meant to have one correct answer, but since Seychellois is a mixed language, then some stimuli in the Boston Naming test are known to have as many as three correct answers. Thus, even if the same test material and instructions (after translation) are used, the validity of the tests will necessarily differ.

36

Grandjean et al.

A variety of neuropsychological tests, including WISC subtests were used under different circumstances in the cross-sectional studies in Madeira, Greenland, Brazil and French Guiana. In the first two and in an Indian group studied in Brazil, all tests were administered with the use of an interpreter, thereby making the test results less reliable. Likewise, the Continuous Performance test used in the Seychelles was not supervised by an examiner, thereby allowing for possible untoward variability that would be less likely if the examiner was present. Results on such computer-assisted tests will also be affected by the prior computer experience of the child. Due to the type of populations studied, the researchers attempted to avoid culturebiased and language-dependent tests, thereby precluding evaluation of important domains. The functions evaluated focused on motor speed and motor coordination, visuospatial organization, attention, and short-term memory. Several tests were common to these studies and also overlap with the prospective studies, e.g., finger tapping, and Stanford-Binet Bead Memory. At increased exposure levels, reduced scores were evidenced on the Santa-Ana dexterity test in Brazil and the McCarthy leg coordination test in French Guiana. In these two studies, scores on the Stanford-Binet Copying test (that measures visuospatial organization) were negatively associated with mercury exposure with very similar regression coefficients in the two studies. Several types of errors occurred in these tests, and the French Guiana study pointed more specifically at rotation errors among younger children (5–6 years old). Such errors would suggest possible insult in the parietal lobes of the brain resulting in developmental delay in the learning to place objects in space (34). Whether or not this type of test would appear sensitive in other populations living in other cultural environments needs to be established. Neurophysiological Testing As an objective evaluation of brain dysfunction that is probably less sensitive to confounding, neurophysiological tests have been applied in several studies. Their applicability requires advanced instrumentation and depends on skilled examiners. An outcome that has previously been found to be sensitive to lead exposure is brainstem auditory evoked potentials. They are recorded using surface electrodes placed on the skull while the child listens to a stimulus in one ear. The transmission of the electrical signals within the brain is then recorded as peaks that represent the acoustic nerve, an intermediate connection in the pons, and the midbrain. The latency of peak III was significantly increased at higher intrauterine exposure to mercury. Parallel associations were found in 7-year-old children in the Faroes, in Madeira, and in Japan, and these observations were then replicated in the Faroese cohort when examined at 14 years. A smaller study from Ecuador also reported delays in peak III at higher exposure levels. In addition, prolonged latencies of peak V among the 14 years-old were linked to the current mercury exposure only and therefore suggested on effect of postnatal mercury exposure (Fig. 1) and that this effects is different from those caused by mercury exposure during fetal development.

CONFOUNDING VARIABLES Three major reasons for confounding have been noted as to why a mercury effect might have been overestimated: (a) association of mercury intake with exposure to other neurotoxic pollutant(s); (b) other types of residual confounding; and (c) inadequate adjustment for multiple comparisons (35). The best protection against confounding problems is to select a study setting where such concerns are unlikely and, if relevant, may be adjusted for

Methylmercury from Seafood and Freshwater Fish

37

appropriately. Thus, a homogeneous society with limited differences in socioeconomic and cultural factors should be chosen. The existence of residual confounding can never be fully excluded, but there is little reason to invoke “phantom” covariates to explain away a biologically plausible association between methylmercury exposure and neurobehavioral deficits. In addition, most attention is usually paid to confounders that affect the outcomes in the same direction as the exposure under study, but confounders may also have the opposite effect of attenuating the apparent impact of the exposure. Thus the potential for overestimation of a toxic effect should not be raised without paying equal attention to the risk of underestimation. Standard multiple regression methods are often used for controlling for confounding effects. Even in the absence of confounding, adjustment for such established predictors as sex, age, and maternal intelligence should be included to obtain a more precise estimate of the mercury effect. In general, the prudent approach is usually to include all covariates that may be potential confounders. However, in situations were the exposure is measured with some degree of imprecision (as is the case here), this approach may result in biased estimates when covariates are less imprecise. Inclusion of such covariates, which are associated with the exposure but without any explanatory power in regard to the effect, will then increase the underestimation of the effect of the exposure of interest (33). As a main concern in regard to confounding, socioeconomic conditions vary substantially between the study settings. Although mercury neurotoxicity was reported in almost all studies, differences within each study could be important. New Zealand and the Faroes represent relatively wealthy, industrialized populations, where socioeconomic differences are thought to be limited, but most of the other studies were carried out in developing countries, where basic sanitary problems are common, and where nutritional deficiencies may occur, both of which may be difficult to adjust for. In the Seychelles, stunting still occurs in a small percentage of children (36). In New Zealand, ethnic differences appeared to play a role, but the analysis was based on matching of the children in the different exposure groups for ethnicity. An additional factor of possible interest is consanguinity, which is more frequent in island populations and other isolated communities. However, to cause confounding, the degree of consanguinity would have to be associated with mercury exposure and at the same time result in neurobehavioral deficits. Both assumptions seem unlikely, but documentation from an individual study would be a major undertaking. The family structure and home environment are documented as important determinants of childhood development. Within the populations studied, circumstances may vary. For example, only about 25% of the births in the Seychelles are nuptial, while an additional 50% are recognized by a father, but about 25% of children have no known father (37). Accordingly, children of the Seychelles cohort were said to be accompanied by a “care-giver,” often a relative, with whom the child was living (18). The variable family structure, which may be difficult to adjust for in statistical analyses, contrasts with the more uniform circumstances of most other studies with a traditional and stable family structure. In the New Zealand study, low social class and non-English home language reduced the score in some tests (not surprising) and more than 6 months of breastfeeding increased the score in some tests (8). These variables were accounted for in the multiple regression analysis. Among other known developmental neurotoxicants, none is as prevalent as ethanol. Most studies reported that maternal alcohol use during pregnancy was minimal, but in some cases it may be difficult to assess because of the importance of home-brewed beverages, e.g., in the Seychelles (38).

38

Grandjean et al.

In particular the Faroese are exposed to PCBs, in this case from eating whale blubber. Detailed analyses of the Faroes data failed to show any important impact of PCB exposure on the neurotoxicity outcomes (39,40). Inclusion of PCB exposure in a structural equation model attenuated the mercury effect somewhat; mercury remained statistically significant, but PCB was far from significant (41). In New Zealand and the Seychelles, the ocean fish consumed is unlikely to be contaminated by PCB, and the same would be the case with freshwater fish in the Amazon Basin. A reverse effect may occur if subjects who do not eat mercury-containing fish instead consume fruits and vegetables that contain pesticides. Such exposures are more likely in tropical developing countries. The neurotoxicity of many pesticides could then potentially cause neurodevelopmental effects in children with low-level mercury exposures, thus blurring the dose-response relationship. Although pesticide use might be a cofactor in the Seychelles, no information is available to evaluate this possibility. Certain essential nutrients in fish and seafood may provide beneficial effects on brain development, thereby possibly counteracting adverse effects of the contaminants. This possibility has often been mentioned in regard to ocean fish (17). Perhaps, if ocean fish contains higher concentrations of essential n-3 fatty acids than do freshwater fish, then this difference could perhaps explain why the mercury dose–response relationship appears to be steeper in populations that rely on river fish. Chemical analyses of ocean fish from New Zealand show that selenium concentrations do not depend on fish size, while mercury concentrations increased linearly (42). Although selenium has been considered to potentially provide protection against mercury effects, cord–blood selenium concentrations in the Faroes did not impact on mercury-associated deficits. It seems unlikely that essential nutrients would counteract all aspects of mercury neurotoxicity.

PUBLIC HEALTH RELEVANCE Public health authorities have recently begun to issue dietary advisories in regard to the most contaminated types of freshwater fish and seafood. However, some discussions on methylmercury toxicity have erroneously pictured the situation as a conflict between a negative and a positive study. This misleading characteristic may be related to disagreements between regulatory agencies and has been exploited by vested interest groups. In epidemiology, the term “non-positive” is often used in regard to studies that were unable to detect a particular effect. Also, no matter how positive a study is, all observational studies have weaknesses, and a prudent judgment should be based on the total amount of evidence available, no on single studies, whether positive or not. The present chapter has demonstrated that the scientific evidence on methylmercury neurotoxicity is fairly consistent, and that adverse effects may even occur at low-level exposures. There is no dispute about the very serious prenatal effects that occurred in Minamata at maternal hair-mercury concentrations in the range of 10–100 mg/g (1–3). It would seem intuitively logical that less severe effects may occur at the exposure ranges found in the more recent studies. Because of the global significance of methylmercury contamination of food, these scientific findings need to be expressed in terms that may facilitate an evaluation of their public health significance. The Faroes study showed that each doubling in prenatal mercury exposure corresponded to a delay of one or two months in mental development at age 7 years (10). Because rapid development occurs at that age, such delays may be important. Also, even small shifts in a measure of central tendency may be associated with large changes in the tails of the distribution. Such developmental delays are likely to be permanent, at least

Methylmercury from Seafood and Freshwater Fish

39

in part, but the long-term implications are unknown. The experience with lead neurotoxicity suggests that such effects are likely to remain and that they may even become more apparent with time. A shift in IQ levels was documented in the New Zealand study (8). The average WISC-R full-scale IQ for the study population (nZ237) was 93. In the group with maternal hair mercury above 6 mg/g (nZ61) the average was 90. The average exposure in the latter group would be about 4-fold higher than in the study population as a whole. Another way of presenting these shifts in IQ is to estimate the increased number of subjects with very low IQ as methylmercury exposure increases. In New Zealand, an IQ below 70 (Zmental retardation) was twice as common (increase from 5 to 10%) in the highest hair mercury group (O10 mg/g) compared to the group with hair mercury below 6 mg/g (42). For the IQ range 71–85, the increase was from 20 to 25%, but due to the small number of children, this result was not statistically significant. Another approach was used in the Faroes in the absence of formal IQ tests. The regression coefficients were expressed as a proportion of the standard deviation of the test results (22). The most sensitive outcome parameters show a decrement of about 10% of the standard deviation at each doubling of the prenatal methylmercury exposure level. Had an IQ scale been used, with a standard deviation of 15 IQ points, a doubling in the exposure could have caused a deficit of about 1.5 IQ points. More detailed calculations based on the three WISC tests administered are in agreement with this conclusion (43). These findings are also in good agreement with the New Zealand data, and recent calculations of associated diminished lifetime economic productivity suggest that societal expenses may be very substantial (44). Each of these estimations is associated with some degree of uncertainty. Some scientific uncertainties are bound to remain, although new prospective cohort studies on methylmercury neurotoxicity are starting to provide new evidence, e.g., from ongoing research in Japan. However, the documentation is not going to expand substantially or otherwise provide much clearer guidance for regulatory agencies. It should also be recognized that the question as to whether to base decisions either on proof of harm or on precaution cannot be settled from epidemiological evidence. The experience with lead research (6) has amply illustrated that apparent disagreement is likely to occur between studies carried out by different methods in different settings. We therefore should not anticipate complete coherence among all available evidence. Accordingly, decisions on preventive efforts should be justified by the scientific database at large, taking into account its various uncertainties and inconsistencies. The potential costs and other societal consequences of policy decisions—including decisions to do nothing—deserve fair consideration. However, these issues should be addressed in parallel to and separate from the discussion of toxicological and epidemiological concerns. Otherwise, the erroneous impression will be generated that disagreements on preventive measures are solely due to uncertainties in epidemiologic evidence.

ACKNOWLEDGMENTS This work was supported by the U.S. National Institute of Environmental Health Sciences (ES09797 and ES11681) and the Danish Medical Research Council. The contents of this paper are solely the responsibility of the authors and do not represent the official views of the NIEHS, NIH or any other funding agency.

40

Grandjean et al.

REFERENCES 1. United Nations Environment Programme. Global Mercury Assessment. Geneva, 2002. 2. Tsubaki T, Irukayama K. Minamata Disease: Methylmercury Poisoning in Minamata and Niigata, Japan. Amsterdam: Elsevier Scientific Publishing Company, 1977. 3. National Research Council Toxicological Effects of Methylmercury. Washington: National Academy Press, 2000. 4. U.S. Environmental Protection Agency, Office of Science and Technology, Office of Water (2001). Water quality criterion for the protection of human health: methylmercury, Final. EPA823-R-01-001. Washington, 2001. http://www.epa.gov/waterscience/criteria/methylmercury/ document.html. (Accessed 19 December, 2003). 5. JECFA (Joint FAO/WHO Expert Committee on Food Additives). Sixty-first meeting, Rome, 10–19 June 2003. Summary and conclusions. ftp://ftp.fao.org/es/esn/jecfa/jecfa61sc.pdf. (Accessed 19 December, 2003). 6. Needleman HL, Bellinger D. Studies of lead exposure and the developing central nervous system: a reply to Kaufman. Arch Clin Neuropsychol 2001; 16:359–374. 7. Kjellstro¨m T, Kennedy P, Wallis S, Mantell C. Physical and mental development of children with prenatal exposure to mercury from fish. Stage 1: Preliminary tests at age 4. (Report 3080) Stockholm, National Swedish Environmental Protection Board, 1986. 8. Kjellstro¨m T, Kennedy P, Wallis S, et-al. Physical and mental development of children with prenatal exposure to mercury from fish. Stage 2, interviews and psychological tests at age 6. (Report 3642) Stockholm, National Swedish Environmental Protection Board, 1989. 9. Crump KS, Kjellstrom T, Shipp AM, Silvers A, Stewart A. Influence of prenatal mercury exposure upon scholastic and psychological test performance: benchmark analysis of a New Zealand cohort. Risk Anal 1998; 18:701–713. 10. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 11. Murata K, Weihe P, Renzoni A, et al. Delayed evoked potentials in children exposed to methylmercury from seafood. Neurotoxicol Teratol 1999; 21:343–348. 12. Sørensen N, Murata K, Budtz-Jorgensen E, Weihe P, Grandjean P. Prenatal methylmercury exposure as a cardiovascular risk factor at seven years of age. Epidemiology 1999; 10:370–375. 13. Murata K, Weihe P, Budtz-Jørgensen E, Jørgensen PJ, Grandjean P. Delayed brainstem auditory evoked potential latencies in 14-year-old children exposed to methylmercury. J Pediatr 2004; 144:177–183. 14. Grandjean P, Murata K, Budtz-Jorgensen E, Weihe P. Cardiac autonomic activity in methylmercury neurotoxicity: 14-year follow-up of a Faroese birth cohort. J Pediatr 2004; 144:169–176. 15. Shamlaye CF, Marsh DO, Myers GJ, et al. The Seychelles child development study on neurodevelopmental outcomes in children following in utero exposure to methylmercury from a maternal fish diet: background and demographics. Neurotoxicology 1995; 16:597–612. 16. Myers GJ, Davidson PW, Cox C, et al. Summary of the Seychelles child development study on the relationship of fetal methylmercury exposure to neurodevelopment. Neurotoxicology 1995; 16:711–716. 17. Myers GJ, Davidson PW, Cox C, et al. Prenatal methylmercury exposure from ocean fish consumption in the Seychelles child development study. Lancet 2003; 361:1686–1692. 18. Keiding N, Budtz-Jørgensen E, Grandjean P. Prenatal methylmercury exposure in the Seychelles [letter]. Lancet 2003; 362:664–665. 19. McKeown-Eyssen G, Ruedy J, Neims A. Methylmercury exposure in northern Quebec. II. Neurologic findings in children. Am J Epidemiol 1983; 118:470–479. 20. Marsh DO, Turner MD, Smith JC, Perez VMH, Allen P, Richdale N. Fetal MeHg study in a Peruvian fish eating population. Neurotoxicology 1995; 16:717–726. 21. Counter SA, Buchanan LH, Laurell G, Ortega F. Blood mercury and auditory neuro-sensory responses in children and adults in the Nambija gold mining area of Ecuador. Neurotoxicology 1998; 19:185–196.

Methylmercury from Seafood and Freshwater Fish

41

22. Grandjean P, White R, Nielsen A, Cleary D, de Oliveira-Santos E. Methylmercury neurotoxicity in Amazonian children downstream from gold mining. Environ Health Perspect 1999; 107:587–591. 23. Steuerwald U, Weihe P, Jørgensen PJ, et al. Maternal seafood diet, methylmercury exposure, and neonatal neurologic function. J Pediatr 2000; 136:599–605. 24. Cordier S, Garel M, Mandereau L, et al. Neurodevelopmental investigations among methylmercury-exposed children in French Guiana. Environ Res 2002; 89:1–11. 25. Weihe P, Hansen JC, Murata K, et al. Neurobehavioral performance of Inuit children with increased prenatal exposure to methylmercury. Int J Circumpolar Health 2002; 61:41–49. 26. Murata K, Sakamoto M, Nakai K, et al. Effects of methylmercury on neurodevelopment in Japanese children in relation to the Madeiran study. Int Arch Occup Environ Health 2004; 77:571–579. 27. Nakai K, Suzuki K, Oka T, et al. The Tohoku Study of Child Development: a Cohort Study of the effects of perinatal exposures to methylmercury and environmentally persistent organic pollutants on neurobehavioral development in Japanese children. Tohoku J Exp Med 2004; 202:227–237. 28. Grandjean P, White RF, Weihe P, Jørgensen PJ. Neurotoxic risk caused by stable and variable exposure to methylmercury from seafood. Ambul Pediatr 2003; 3:18–23. 29. Lanzirotti A, Jones KW, Clarkson TW, Grandjean P. Human health risks from methyl mercury in fish. Science Highlights—National Synchroton Light Source Activity Report. Upton, NY: Brookhaven National Laboratory, 2002 pp. 97–99. 30. Grandjean P, Jørgensen PJ, Weihe P. Validity of mercury exposure biomarkers. In: Wilson SH, Suk WA, eds. Biomarkers of Environmentally Associated Disease. Boca Raton, FL: CRC Press/Lewis Publishers, 2002:235–247. 31. Lewandowski TA, Pierce CH, Pingree SD, Hong S, Faustman EM. Methylmercury distribution in the pregnant rat and embryo during early midbrain organogenesis. Teratology 2002; 66:235–241. 32. Budtz-Jørgensen E, Grandjean P, Jørgensen PJ, Weihe P, Keiding N. Association between mercury concentrations in blood and hair in methylmercury-exposed subjects at different ages. Environ Res 2004; 95:385–393. 33. Budtz-Jørgensen E, Keiding N, Grandjean P, Weihe P, White RF. Consequences of exposure measurement error for confounder identification in environmental epidemiology. Stat Med 2003; 22:3089–3100. 34. Sullivan, K. Neurodevelopmental aspects of methylmercury exposure: neuropsychological consequences and cultural issues. PhD Thesis in Behavioral Neuroscience, Boston University School of Medicine, 1999. 35. NIEHS. Workshop organized by Committee on Environmental and Natural Resources (CENR), Office of Science and Technology Policy (OSTP), The White House: Scientific Issues Relevant to Assessment of Health Effects from Exposure to Methylmercury, November 18–20, 1998. (Accessed December 19, 2003). Available from: http://ntp.niehs.nih.gov/main_pages/ PUBS/MethMercWkshpRpt.html. (Accessed 10 January, 2006). 36. WRI (World Resources Institute) Earth Trends: The Environmental Information Portal, 2003. Variable: Children’s Health: Stunting in Children Under 5. Washington. http://earthtrends.wri. org/text/POP/variables/387.htm. (Accessed 5 March, 2004). 37. MISD (Management & Information Systems Division), Statistics & Database Administration Section (SDAS) of the Ministry of Information Technology & Communication of the Republic of Seychelles, 2003. Statistical Bulletin Quarterly: Population and Vital Statistics 2002. Victoria, Mahe: table 3 (Registered Live Births By Year, Month of Registration, Sex and Status, 1998–2002). 38. Perdrix J, Bovet P, Larue D, Yersin B, Burnand B, Paccaud F. Patterns of alcohol consumption in the Seychelles Islands (Indian Ocean). Alcohol Alcohol 1999; 34:773–785. 39. Budtz-Jørgensen E, Keiding N, Grandjean P, White RF, Weihe P. Methylmercury neurotoxicity independent of PCB exposure [letter]. Environ Health Perspect 1999; 107:A236–A237.

42

Grandjean et al.

40. Grandjean P, Weihe P, Burse VW, et al. Neurobehavioral deficits associated with PCB in 7-year-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001; 23:305–317. 41. Budtz-Jørgensen E, Keiding N, Grandjean P, Weihe P. Estimation of health effects of prenatal methylmercury exposure using structural equation models. Environ Health 2002; 1:22. 42. Kjellstro¨m T. Methyl-mercury exposure and intellectual development in vulnerable groups in New Zealand. Proceedings of the US–Japan Workshop, Nov. 2000. Minamata, Japan, National Institute for Minamata Disease, 2000. 43. Budtz-Jørgensen E, Debes F, Weihe P, Grandjean P. Adverse mercury effects in 7-year-old children expressed as loss in “IQ”, 2005. Available from: www.chef-project.dk. 44. Trasande L, Landrigan PJ, Schechter C. Public health and economic consequences of methyl mercury toxicity to the developing brain. Environ Health Perspect 2005; 113:590–596.

3 Effects of Developmental PCB Exposure on Neuropsychological Function in Epidemiological Studies: Issues and Research Needs Deborah C. Rice Environmental and Occupational Health Program, Maine Center for Disease Control and Prevention, Augusta and Center for Integrative and Applied Toxicology, University of Southern Maine, Portland, Maine, U.S.A.

Attention focused on the adverse consequences of PCB exposure on the developing fetus following poisoning episodes in Taiwan (1–4) and Japan (5). Ingestion of rice oil, contaminated with PCBs, dibenzofurans and dibenzodioxins, by women before or during pregnancy resulted in infants with hyperpigmentation, mental retardation, and other overt effects at high exposures, and cognitive impairment and behavior problems at lower exposures. It was clear from these unfortunate incidents that the fetus was much more sensitive to these chemicals than were adults. In the 1980s, investigators in Michigan studied the consequences of developmental exposure to PCBs in women who did or did not consume Lake Michigan fish (6). During the same time period, the effects of in utero exposure were examined in a North Carolina cohort whose exposure was through the general food supply (7). Both studies documented adverse effects during infancy associated with maternal or fetal PCB body burden. The Michigan study found effects on IQ, attention, memory, and reading ability that persisted at least through 11 yr of age (8,9). In the 1990s, several longitudinal prospective studies were implemented to examine the effects of PCBs and other contaminants on neuropsychological development of children. These studies were able to take advantage of more sensitive and accurate chemical analyses of PCBs and related chemicals than were available in earlier studies, as well as to use information from previous studies to choose or develop behavioral outcome measures. A study in Oswego, New York, examined offspring of mothers who did or did not consume Lake Ontario fish (10), whereas studies in Germany (11) and the Netherlands (12) assessed the effects of exposure through the general food supply. All three studies 43

44

Rice

reported deficits associated with developmental PCB exposure. A study in the Faroe Islands, designed to study the effects of methylmercury, also examined the effects of in utero exposure to PCBs (13). A study in the Inuit population in Arctic Canada is in the early stages (14). The epidemiological literature on the effects of PCBs on behavior in children was reviewed most recently and extensively by Schantz et al. (15), as well as in earlier publications (16,17). Congruence between studies in the types of deficits identified is good in most cases; complete agreement in every detail is not to be expected in observational studies such as these (18). This chapter is not intended to be a review of the literature, but rather a discussion of issues that require further study for optimal protection of public health.

IDENTIFICATION OF TYPES OF BEHAVIORAL DEFICITS PRODUCED BY PCB EXPOSURE The evidence that PCBs are neurotoxic is compelling. Studies in monkeys (19–23) and rats (24–32) documented deficits in a variety of neuropsychological functions as a consequence of developmental exposure to PCBs. PCBs also produce effects on neurotransmitter systems and brain metabolic processes (33–39). Comparison of exposure among epidemiological studies is not straightforward for several reasons. The early studies (Michigan, North Carolina) used older, less sensitive analytical methodology that did not determine concentrations of individual congeners. Newer studies analyzed different congeners and different numbers of congeners. In addition, PCBs were assayed in different tissues (cord blood, maternal blood, breast milk, cord tissue), with some studies measuring levels in more than one tissue. Nonetheless, a comparison was performed across studies based on normalization of data to concentrations of congener 153 in maternal serum (40). Results from the five studies designed to assess the effects of PCBs were not dissimilar, with a good deal of overlap in the range of estimated concentrations. North Carolina had the lowest median exposure (80 ng/g lipid) and Germany the highest (140 ng/g lipid). However, the accuracy of the estimates for the Michigan and North Carolina studies are less certain than those from the more recent studies. The population in the Faroe Island study, designed to assess the effects of methylmercury, had much higher PCB exposures than other studies (median 450 ng/g lipid). There is a reasonable degree of consistency in both experimental approach and findings with respect to measures of early psychomotor and cognitive development (Table 1). All the studies designed to assess the effects of PCBs—Michigan, North Carolina, Dutch, German, and Oswego—measured IQ between 3 and 5 yr of age (41–46). The American studies used the McCarthy Scales, whereas the European studies used the Kaufman Assessment Battery for Children (K-ABC). All studies except the North Carolina study reported adverse effects on cognitive performance. The Dutch study was designed to assess the contribution of prenatal exposure and exposure through breastfeeding to any observed effects. Therefore, the cohort was recruited so that half the mothers breast-fed and half did not, with half of each group recruited from Rotterdam or Gro¨ningen. Differential effects on IQ were observed in breast- and formulafed infants; effects were detected in the full cohort and formula-fed infants, but not the breast-fed group. The North Carolina, Dutch, and Oswego studies assessed neurological status during early infancy, and all found associations between prenatal exposure and infant

Vigilance task (55)

np

K-ABC (42)

McCarthy attenuated (61) Tower of London, 9 yr; Rey Complex figure test, negative (53) Vigilance task (54) simple reaction time (53) Reynell language development scale (42)

Negative McCarthy (3–5 yr) (41) np

McCarthy (45,46)

Attention/response inibition/processing speed Language

np

np

Vigilance task (119) freedom from distractibility, WISC-R (8)

Word comprehension, Woodcock Reading mastery test; reading comprehension, verbal comprehension, WISC-R (8)

Cognitive effects in later WISC-R, 11 yr (8) childhood

np

McCarthy attenuated (44) np

K-ABC attenuated (66) np

PDI (51)

MDI (50,92)

Negative (49)

Cognitive effects 4–7 yr np

K-ABC (43)

np

np

(49)

np

np

MDI (11)

np

(Continued)

McCarthy (44)

(52)

Negative (11)

np

Precchtl (12)

Eaters and noneaters of Lake Ontario fish NBAS (94)

Oswego

NBAS (based on fish consumption) (48) NBAS (91)

General population

Germany

Infant neurological status Fagan test of recognition memory Bayley scales of infant development Cognitive effects 3–4 yr

The Netherlands

General population General population, half breast-fed, half not

North Carolina

Eaters and non-eaters of Lake Michgan fish

Michigan

Summary of Identified Associations Between PCB Exposure and Adverse Neuropsychological Effects

Study population

Table 1

PCB Exposure on Neuropsychological Function 45

Vocabulary and information scores, WISC-R (8) np Rating scale (9)

Abbreviations: np, not performed; negative, no effect of PCBs.

Social behavior Activity

Memory

Michigan

np np

np

North Carolina Auditory Verbal Learning Task negative (53) CBCL (59) Rating scale, play behavior (54,60)

The Netherlands

np np

np

Germany

Table 1 Summary of Identified Associations Between PCB Exposure and Adverse Neuropsychological Effects (Continued)

np np

np

Oswego

46 Rice

PCB Exposure on Neuropsychological Function

47

neurological status (7,12,47). The Oswego study found poorer autonomic function and failure to habituate (a measure of learning) associated with increased PCB exposure. The North Carolina and Dutch studies reported hypotonicity associated with increased PCBs; the North Carolina study reported increased abnormal reflexes, for which the Oswego study found a trend. In the Michigan study, maternal fish consumption was associated with motoric immaturity, poorer lability of states, abnormal reflexes, and a greater degree of startle (48). Cord serum levels were not predictive in the Micigan study. However, there were only a limited number of cord blood samples, particularly from highexposure infants. Four of the studies assessed performance on the Bayley Scales of Infant Development, with only the Michigan study failing to find an effect (11,49–51). Three studies used the Fagan test of visual recognition memory: Michigan (49), Oswego (52), and German (11). Effects were observed in the Michigan and Oswego studies, but not the German, possibly as a consequence of poor experimental control (11). Language development was assessed in two studies. The Dutch study found adverse effects of in utero exposure to PCBs in formula-fed infants using the Reynell Language Development Scale (42). The Michigan study found negative effects on verbal IQ, verbal comprehension, and reading ability at 11 yr of age (8). Effects were not observed in the Dutch study on the Auditory Verbal Learning Test (53), which is not a test of language function per se, but assesses short-and longer-term memory for word recall. Performance on a vigilance task was assessed in three studies at various ages between 3.5 and 11 yr. The Dutch (54), Oswego (55), and Michigan studies (56) all found effects related to PCB exposure: in fact, all three studies found increased errors of commission (failure of response inhibition). The Michigan investigators also found a decreased number correct responses and increased errors of commission at four years of age on the Sternberg Memory paradigm, a computerized test of working memory that allows responding to digits not on a sample list (57). The Oswego study also reported an interaction between the size of the corpus collosum (a major brain fiber tract subserving interhemispheric communication) and increased PCB body burden on errors of commission (55). The Dutch investigators found increased response latencies and more variable latencies on a simple reaction time task related to increased PCB exposure at nine years of age (53). This task is similar to the vigilance task, but does not provide the opportunity to respond on incorrect stimuli (errors of commission). No information was provided on failure to respond to the target. The effects observed on the vigilance task are indicative of deficits in “executive function,” the ability to organize behavior in time and space. The Dutch investigators found deficits in spatial executive function on the Tower of London in their cohort at nine years of age (53). No effects of PCB exposure were found on the Rey Complex Fig. Task, which assesses visuospatial organization and memory. The Michigan investigators reported increased perseverative errors on the Wisconsin Card Sort test at 11 yr, as well as deficits in attention on the Digit Cancellation task (57). The Michigan study also reported slower reaction time on a mental rotation task (57), which in the absence of effects on accuracy in this task or reaction time on other tasks, may be considered to be a consequence of a deficit in response organization, again indicative of a problem with executive function. However, no effects were observed in this cohort on the Stroop Color-Word Test, another test of executive function. Observed effects were associated with prenatal and not postnatal exposure. Both the Dutch (54) and the Michigan (58) studies found changes in activity. The Michigan study reported a decrease in activity at 4 yr as assessed by the rating of three adults. The Dutch study reported increased activity as assessed by a parents’ rating scale, but a decrease in free play behavior, at 3.5 yr of age.

48

Rice

The Dutch study also assessed problem behavior using the Child Behavior Check List (CBCL) (59) at 3.5 yr. They reported effects on the internalizing, withdrawn/ depressed and aggressive scales, but not oppositional or overactive scales. Scores on the anxious scale were borderline. Sexually dimorphic play behavior was examined in the Dutch cohort at 7.5 yr using the Pre-School Activity Inventory to test the hypothesis that PCBs and dioxins exert effects on behavior via endocrine disruption (60). Prenatal PCB concentrations were associated with less masculinized play behavior in boys and more masculinized behavior in girls, whereas higher prenatal dioxin levels were associated with more feminized play behavior in both boys and girls. An important issue that has been examined in the Dutch and Michigan studies is the possibility of effect modification: specifically, the salutary effects of an advantaged home environment in reducing the adverse effects of PCBs on children’s development. In the Dutch study, effects were observed on the K-ABC at 3.5 yr in formula-fed but not breast-fed infants (42). These findings were originally attributed to the nutritional benefits of breastfeeding. However, they are more likely the result of the fact that mothers with a higher socio-economic status and education breast-feed their babies longer. The Dutch study found negative effects at 6.5 yr of age on the McCarthy Scales in less- but not moreadvantaged children (61). Analyses revealed that it was because formula-fed infants were from less advantaged homes, and not the formula-versus-breast-fed dichotomy per se, that accounted for the difference in performance on the McCarthy Scales (61). Similarly, the Michigan investigators found that effects on IQ at four and 11 yr were present in infants breast-fed fewer than 6 wk, but not those breast-fed more than 6 wk (62). As in the Dutch study, the Michigan investigators reported that the lack of association between PCB exposure and IQ at 4 or 11 yr of age in breast-fed babies could be accounted for by statistically by quality of parental intellectual input, and that adverse effects were strongest in children of less verbally competent mothers (63,64). The findings in these studies are consistent with studies in lead-exposed children, in which adverse effects were greater in less advantaged children (65). It appears that high-quality parental care may ameliorate, or at least attenuate, the effects of neurotoxic agents. The German investigators studied the effects of home environment and PCB exposure independently. They reported that HOME score was positively associated with mental and motor development on the Bayley Scales at 30 mo and on the K-ABC at 42 mo, whereas increasing milk PCB concentrations were associated with poorer performance, when each variable was adjusted for the other (43). At 72 mo, effects of the HOME score were still a predictor of positive outcome on the K-ABC in this relatively advantaged population (66). The effects of PCBs were no longer significant based on 70 children (less than half the original cohort), although the direction was still negative. The Faroe Islands study, which was designed to study the effects of in utero methylmercury exposure, did not include early development assessment, nor was IQ measured. Instead, a number of domain-specific functions were assessed at seven years of age (67). Only limited effects of PCBs were found prior to control for methylmercury exposure (13) despite the high PCB concentrations (40). A negative association was found between cord tissue PCB levels and performance on the Boston Naming Test, a test of language development. This is consistent with the effects on language in the Dutch and Michigan studies. Effects were also found on a continuous performance test. No effects of PCBs were found on any endpoint in the Faroe Islands study after controlling for methylmercury exposure, although there was some indication of effects on several endpoints in children in the highest tertile with respect to methylmercury. Conclusions. It appears, based on information presently available, that the effects of PCB on IQ, identified at 3–4 yr of age, attenuate or disappear as the child reaches school

PCB Exposure on Neuropsychological Function

49

age in studies begun in the 1990s. It is unknown whether effects will reappear, as was the case for lead in the Cincinnati study (68,69). It also seems, however, that children from less advantaged homes may still be vulnerable to effects of PCBs on IQ at school age. Investigators of the various studies have assessed measures other than IQ in a somewhat individualistic manner, and have identified behavior domains affected by PCBs that require more detailed investigation to characterize effects. One area is that of executive functions: attention, temporal and spatial organization of behavior, and response inhibition. Deficits on spatial tasks have been observed in rats (25,26,29,70), and failure of response inhibition and poor temporal organization of behavior have been observed as a consequence of PCB exposure in monkeys (19,20). Adverse effects in these domains have been observed in the modern PCB studies, but require fuller characterization. Perseverative behavior was observed at 11 yr of age in the Michigan cohort (56). Perseverative behavior as a consequence of PCB exposure in animals was observed in monkeys (21). It is important to assess perseverative behavior in the newer studies in which exposures are lower than those observed in the Michigan study. Another area that requires exploration is social- and antisocial- behavior. Poorer outcome on the CBCL suggests that these children may be at risk for increased criminality in the future, as was the case for lead (71–73). The effects of PCBs and dioxins on masculinization or feminization of play behavior suggests that the potential behavioral consequences of effects on sex hormones should be pursued. Effects of contaminants on social interactions and behavioral consequences of perturbation of sex hormones have been little studied, yet are vitally important to the functioning of the individual in society.

CHOICE OF CONGENERS AND TISSUE COMPARTMENT AS DETERMINANTS OF EXPOSURE PCBs are a mixture of as many as 209 congeners with varying degrees and patterns of toxicity. The fingerprint of PCB congeners in human tissue following environmental exposure is reasonably consistent between populations (74), although there may be differences resulting from sources and recency of exposure (75). It is vitally important to understand the relative potency and pattern of deficits produced by different congeners or congener classes on nervous system function and development to provide the best advice for protection of public health. It is also important to understand in which tissue(s) determination of PCB concentrations provides the best predictor of neurotoxicity. The more recent studies had the opportunity to perform congener-specific analysis using sensitive analytical methodology. Investigators used different strategies regarding which PCB congeners and tissues to analyze. The Oswego study measured 68 congeners or congener pairs in cord blood, with no analysis of maternal blood (10). Breast milk was analyzed from a small subset of women at varying times during the first six months following delivery, thereby essentially obviating the possibility of assessing any effects of postnatal exposure via breast milk. The most highly chlorinated PCB congeners (C 7–9) were predictive of fish consumption, whereas the lower chlorinated homologs (C 1–3 or 4–6) were not. In addition, for most measures the most highly chlorinated congeners were a better predictor of performance than total PCBs or lower chlorinated homologs during infancy (10,47,52) and early childhood (44). The Oswego study did not measure dioxins, and only measured two of the less potent dioxin-like PCB congeners (105 and 118). However, because body burdens of these two congeners may represent a substantial percentage of dioxin-like congeners, they can represent a large proportion of the total toxic equivalency quotient (TEQ) (76).

50

Rice

Only the Dutch study focused on discriminating effects of total PCBs from those of dioxins and dioxin-like PCBs on neuropsychological function. Dioxin-like compounds are those that activate the aryl hydrocarbon (Ah) receptor as a mechanism of action for producing toxic effects. The relative potency compared to 2,3,7,8-tetrachlorinated dibenzo-p-dioxin (TCDD, the most potent dioxin) has been determined by an international committee for the dozen dioxin-like PCB congeners as well as other chemicals that act through Ah receptor activation (77). Individual toxic equivalency factors (TEFs) can be combined into an overall TEQ to estimate the “dioxin-like” potency of a mixture. The Dutch investigators measured, in both maternal and cord blood, the four congeners that typically are found at the highest concentrations in human tissue (congeners 118, 138, 153, and 180) (51). They also measured 17 dioxins and furans, 6 coplanar or mono-ortho coplanar (dioxin-like) PCB congeners, and 20 ortho-substituted congeners in breast milk shortly after birth from the half of the mothers who breast-fed their infants. PCBs and other lipid-soluble chemicals are at higher concentrations in milk than blood, so that sampling of breast milk allowed analysis of more congeners with greater accuracy. The dioxins, furans, and dioxin-like congeners were used to calculate dioxin TEQs separately, or as a total TEQ. This provided the opportunity to determine the association between performance and concentrations of dioxins, dioxin-like- and non-dioxin-like PCBs in breast-fed infants, as well as the sum of the four congeners in maternal and cord blood and breast milk in the full cohort. In general, maternal blood PCB levels were the best predictors of performance measures on the Bayley Scales during infancy (51), and the K-ABC at 3.5 yr of age (42). The sum of PCBs in cord blood was predictive of five outcomes, whereas maternal blood PCB levels were predictive of nine outcomes. This may well be due to the higher levels in maternal compared to cord blood, allowing more accurate measurement of PCB concentrations. PCB levels in milk and milk TEQ were both predictive of only one measure in early assessments: neurological status during infancy (12,51). Similarly, planar, mono-ortho, or dioxin TEQ were not predictive of free play behavior, performance on a vigilance task, or activity according to a parents’ questionnaire at 3.5 yr (54). Total TEQ, dioxin-TEQ, and planar PCB TEQ were each predictive of a more adverse score on the internalizing scale of the CBCL, whereas the sum of the four congeners measured in maternal or cord blood were not (59). All measures were predictive of adverse scores on the withdrawal/depressed scale. In a subsequent study on play behavior, all four measures were predictive of differential behavior in boys and girls: dioxin TEQ in milk predicted more feminized play behavior in both sexes, whereas the sum of the four PCB congeners in cord or maternal blood predicted less masculinized behavior in boys and the sum of the four PCB congeners in milk predicted more masculinized behavior in girls (60). Performance on the Tower of London was assessed at nine years in half the cohort (from Rotterdam but not Gro¨ningen) (53). Higher levels of the four congeners measured in maternal plasma were associated with longer and more variable response times on a simple reaction time task, and lower scores on the Tower of London. Other exposure variables (e.g., cord blood or TEQ) were not compared in this publication. The German study only measured three congeners (153, 138, and 180) in cord plasma. They also measured these congeners in breast milk at 2 and 4 wk postpartum, which were combined to yield a single value. Adverse effects were associated with milk PCB levels on early infant psychomotor and memory tests (11) and performance on the K-ABC at 3.5 yr (43). The authors considered milk PCB concentrations to be a more accurate measure of prenatal exposure than cord blood, since levels are much higher and therefore presumably analyzed more accurately. This contrasts with the Dutch study, however, in which cord (but particularly maternal) blood PCB concentrations predicted

PCB Exposure on Neuropsychological Function

51

cognitive performance at 3.5 yr, but milk levels, taken on the half of the cohort that breast fed, did not. The Faroe Islands study measured five congeners in cord tissue in half the cohort only (118, 138, 153, 170, and 180), but used only 138, 153, and 180 as exposure markers (13), precluding exploration of issues related to optimal markers of exposure. This issue has been little explored in animals. In a series of studies in rats with dioxin (TCDD) and five PCB congeners with or without dioxin-like properties (28, 118, 153, 77, 95), Schantz and colleagues (25,26,70) found all to be neurotoxic, but with different patterns of impairment on the two tasks examined. Rice and colleagues (78–82), on the other hand, reported minimal neurotoxicity following developmental exposure to congener 126, whereas effects were observed in other organ systems. Conclusions. The issue of which congeners are producing toxicity is an important one for public health. It appears from the Dutch study that TEQ in breast milk did not predict performance on most outcome measures early in life, but then neither did non-dioxin-like congeners in breast milk. TEQ was predictive of outcome on non-cognitive endpoints at 3.5 yr of age, and on the masculine/feminine dimension of play behavior at 6.5 yr. The Oswego study measured a reasonable number of congeners, but did not include most dioxin-like PCBs, including the most potent congeners, 126 and 169; therefore, there is not the opportunity to determine whether TEQ is predictive of performance. In the Oswego study, the more highly chlorinated PCBs best predicted performance, but that may be because these congeners are more reliably analyzed than lower-chlorinated congeners because of lack of interference from other chemicals. Both the German and Faroe Islands studies used only three congeners as markers of exposure. The question of whether some congeners are more toxic than others (although this may be assumed), and which those may be, remains largely unaddressed in epidemioloigcal studies. Although only measuring a few congeners may provide a reasonable indication of PCB body burden, such a strategy runs the risk of failing to adequately monitor the congeners or congener classes associated with toxicity. The Oswego study found that total PCBs, even based on 68 congeners, was often not predictive of outcome, whereas highly chlorinated PCBs were. Similarly, the Dutch investigators found that dioxin TEQ predicted some effects that the four congeners used as a surrogate for total PCBs did not. It is also unclear which tissue may be the best to use for PCB determination. The Dutch study suggests that maternal blood in preference to cord blood or milk is a better predictor, at least through early childhood, whereas the German study indicates that milk may be preferable to cord blood. The Oswego study measured only cord blood but multiple congeners, and found effects on numerous endpoints. With current knowledge, it is not possible to recommend one tissue over another; the best strategy is to measure all three compartments.

DETERMINATION OF EFFECTS OF MULTIPLE CHEMICALS IN ADDITION TO PCBS A reality that must be addressed in the evaluation of effects of PCBs on neuropsychological function in children is the fact that all of us are exposed to multiple contaminants, from conception (or pre-conception) onward. Exposures to other chemicals can serve as confounders or effect modifiers. In some instances, the degree of collinearity may be sufficiently high that effects of PCBs cannot be differentiated from those of other chemicals.

52

Rice

No study can measure every chemical believed to possibly be relevant when the study is planned, and the list is constantly expanding. (For example, there is increasing concern about the effects of brominated flame retardants, which are bioaccumulated, bioconcentrated, persistent, and may have many of the same adverse health effects as PCBs. These chemicals were not of concern when the studies under discussion were planned and implemented.) Investigators are forced to choose which co-contaminants may be the most relevant for their study population. Different decisions concerning which and how many chemicals to measure have been made in various studies, making comparison between studies sometimes difficult. Lead is a known neurotoxicant, and exposure is ubiquitous in industrialized countries. All investigators have measured blood lead concentrations in the children. Since lead was not the focus of the study, however, exposure was not necessarily measured as accurately as other contaminants. For example, the Oswego study obtained cord blood lead concentrations from the full cohort, but only a single lead measurement from a subset of the cohort during early childhood. The Michigan study measured blood lead levels in the children at four and 11 yr of age (8). Other studies only measured cord blood, including the German (43), Inuit (14) and Faroe Island (67) studies. The Dutch study measured blood lead and cadmium concentrations of the children at 18 mo of age (83). Blood lead concentrations during early childhood, not cord blood lead, is the best predictor of performance at school age (84,85). Since the correlation between lead and PCBs is presumably low in all recent studies, it is unlikely to be a confounder of the effects of PCBs. However, controlling for lead, particularly childhood lead levels, may have strengthened or revealed effects of PCBs in some studies. Measuring lead concentrations at appropriate ages would have the ancillary benefit of allowing exploration of the effects of lead on the endpoints under study, as was done in additional analyses of the Michigan study (57). In addition, lead and PCBs have similar effects on synaptic plasticity in brain slices (86), suggesting the possibility of interactive effects. Methylmercury is another known developmental neurotoxicant that may be associated with adverse effects in PCB studies. Co-exposure could be particularly important in studies in fish-eating populations. The Michigan study was criticized for not measuring body burdens of methylmercury (or mercury) (87), since Lake Michigan fish are known to be contaminated with methylmercury. The Oswego study, designed as a replicate of the Michigan study, did measure methylmercury in maternal hair sectioned to correspond with the first and second half of pregnancy (52). The Faroe Islands study was designed to study the effects of methylmercury, not PCBs, and analyzed both maternal hair and cord blood for total mercury (67). (In contrast, PCB concentrations were measured in cord tissue in half the cohort.) The Inuit study also measured mercury in cord and maternal blood, as well as hair sectioned to represent each of the three trimesters of pregnancy (14), in recognition of relatively high exposure to methylmercury in the Arctic. Neither the Dutch nor German studies measured methylmercury concentrations. There has been considerable discussion concerning whether blood or hair mercury is a better measure of exposure: the National Academy of Sciences report on the health effects of methylmercury expressed the opinion that both were good (88). However, the Faroe Islands study found cord blood to be a better predictor than maternal hair, at least as assessed by p-values (67). PCB and methylmercury levels are unlikely to be so highly correlated in either fish-eater or general population studies that effects cannot be differentiated. In the Oswego study, which recruited consumers of Lake Ontario fish, the two contaminants were not highly correlated (rZC0.11) (44), probably because fish is the only source of methylmercury exposure, whereas there are of many dietary sources of PCB exposure. In addition, PCB and methylmercury levels are not necessarily highly correlated even

PCB Exposure on Neuropsychological Function

53

in fish. Methylmercury concentrations in fish in Lake Ontario are relatively low, and methylmercury body burdens are also low in the cohort. Methylmercury has not been found to be a predictor of performance on most endpoints, although a mercury-PCB interaction was identified on the McCarthy Scales at 3.5 yr of age (44). A possible influence of methylmercury on the effects of PCBs was reported in the Faroe Islands study, in which PCB and mercury body burdens are both relatively high, with a correlation of about 0.4 based on wet-weight PCB concentrations (89). There was no statistical evidence for an interaction between methylmercury and PCBs when regression analyses for effects of methylmercury were performed separately according to PCB tertiles (89). However, there was some evidence that the effects of PCBs on several measures were only present in children in the highest tertile respect to methylmercury (67). Methylmercury exposure was not measured in the North Carolina, German, or Dutch studies. Because PCB exposure in these populations was through the general food supply, the correlation between PCBs and methylmercury may have been low. However, as with lead, failure to measure methylmercury in these studies is a missed opportunity to assess potential interactions of PCBs and another important neurotoxicant. Of perhaps greatest importance for interpretation of PCB studies is the inclusion (or not) of the measurement of other lipophilic chemicals. The source of PCBs in humans is consumption of fatty animal products, which is also the source of other lipophilic contaminants. These chemicals therefore may be highly correlated within individuals. Of perhaps the most concern is dioxins and the related furans. Dioxins affect behavior in animal models (27,90). It would seem important, therefore, to measure dioxins (and furans) in addition to dioxin-like PCB congeners. The only study to do so thus far has been the Dutch study, in which dioxin TEQ and the sum of PCBs in breast milk, and maternal blood or cord blood were each used as exposure measures (see section on choice of congeners). Differential effects were related to total PCBs, dioxin, and/or PCB TEQ on a number of outcome measures. Pesticides are another class of neurotoxic chemicals that may interact with PCBs in producing adverse health consequences. DDE was not found to be associated with effects in the Oswego study on infant neurological status or performance on the McCarthy Scales at 36 or 54 mo (44). On the Fagan test of infant recognition memory, DDE was significantly associated with performance in the univariate analysis, but not after controlling for PCBs (52). The North Carolina study assessed effects of prenatal exposure to PCBs and DDE without controlling for the other chemical. They reported that both chemicals were associated with adverse outcome on the Brazelton Neonatal Behavioral Assessment Scales (91), whereas on the Bayley Scales, PCBs predicted poorer performance at 6 and 12 mo and DDE predicted better performance at six months (92). The correlation between PCB and DDE in breast milk was 0.23 (93), so that failure to include the alternate chemical in the statistical analyses ignored a potential confounder. The Faroe Islands study is less clear with regard to possible effects of DDE. In a paper reporting the effects of PCBs (13), tissue concentrations of both PCBs and DDE were reported. Univariate analyses of the association of DDE and performance were apparently performed, and the authors stated that “[b]ecause of.collinearity, the regression coefficients of [DDE] were very similar to those for.PCB (data not shown).” Results were not provided. DDE was not included as a covariate in the PCB analyses. One is left with the conclusion that both PCBs and DDE were significant predictors of outcome. It may not have been possible to disentangle the effects of DDE and PCBs, but the paper does not adequately report the findings of the study. In contrast, in a subsequent

54

Rice

cohort study in the Faroe Islands, methylmercury was a significant predictor of poorer infant neurological status, whereas PCBs and DDE were not (94). The Oswego study also measured mirex and hexachlorobenzene, which were not found to be predictive of performance. The Inuit study measured a total of 11 pesticides, including DDT and DDE. The Inuit study failed to find an association between neurological status in early infancy and multiple pesticides, including DDT/DDE, whereas adverse effects were associated with PCBs (95). The Michigan study measured seven organochlorine pesticides, which were unrelated to performance, as were PBBs (9). Other studies have not measured pesticide body burdens, so whether pesticide exposure is producing effects in these cohorts, including possibly interacting with PCBs, is unknown. Conclusions. From the results of studies to date, several inferences may be made. Based on the Dutch study, it appears that the effects of non-dioxin-like PCBs may often be distinguishable from those of dioxins and dioxin-like PCBs as determined by TEQ. However, since this is the only study to examine this issue, definite conclusions cannot be drawn. It also appears that, at the low body burdens of methylmercury present in the Oswego study, there is not usually an interaction between methylmercury and PCBs. However, there may be an interaction when body burdens of both contaminants are high, as in the Faroe Islands cohort. Body burdens of lead have generally been low in recent studies, and not associated with effects. The Oswego study is the only recent study to publish an assessment of the effects of multiple pesticides on neuropsychological function, and no effects related to pesticide body burden have been found. It appears that the contribution of effects from other chemicals has not compromised the ability to interpret the effects of PCBs in these cohorts. It is nonetheless important to continue to measure multiple contaminants in any new studies, and to continue to include measured contaminants in analyses as potential confounders or co-contributors to neurotoxicity. In addition, the relative contribution of PCBs and dioxins to neurotoxicity, and the types of deficits produced by each, is an important issue that requires much more research.

CONTRIBUTION OF POSTNATAL EFFECTS TO NEUROTOXICITY Since PCBs are lipophilic, there are much higher concentrations in breast milk relative to maternal or cord blood. Babies who breast feed for even a few weeks receive much larger amounts of PCBs via nursing than they received in utero (96,97), and blood PCB levels during childhood are largely reflective of exposure through breast milk (98,99). Breast milk PCB concentrations decrease rapidly over several weeks as the infant nurses; the best way for a woman to decrease her own store of PCBs is to breast feed (76,91,100,101). Since developmental processes in the nervous system are ongoing well after birth (102), it seems reasonable that postnatal PCB exposure should have an effect. It was therefore somewhat puzzling that early studies reported effects associated with prenatal but not postnatal exposure on infant and early childhood neuropsychological development (7,8,45,46,49,50,58,92,103). The Dutch study is best suited to examine the contribution of postnatal exposure to behavioral outcome: since it was designed specifically to assess the effects of exposure through breastfeeding versus in utero exposure, half the cohort was breast-fed for at least 6 wk and half was not breast-fed at all. Breast-fed infants performed better on neurological assessments during infancy (51), but this may be the result of the more advantaged social environment of the infants that were breast-fed rather than breastfeeding per se (61). Postnatal TEQ, but not measures of prenatal exposure, was associated with poorer performance on the Bayley PDI at seven months, negating the positive effects of

PCB Exposure on Neuropsychological Function

55

breastfeeding at higher exposures (51). Prenatal exposure was negatively associated with performance on the Bayley PDI at three months, the neurological optimality score at 18 mo, and outcome on a neurological examination at two weeks (12,51). Effects on early neuropsychological function and cognitive performance at 3.5 or 6.5 yr were reported to be related to prenatal exposure but not postnatal exposure (42,61). Significant associations were observed between the concurrent blood level of the child and increased reaction time on a vigilance task, more hyperactive behavior, and poorer attention at 3.5 yr in the breast-fed group (54). PCB blood levels of breast-fed infants were considerably higher than formula-fed infants, and are presumably largely the result of breastfeeding (97,98). An interesting strategy for the determination of pre-versus postnatal effects was adopted by the Dutch investigators in an assessment of performance on domain-specific tasks at 9 yr (53). They divided the Rotterdam half of the cohort into six groups: formula-fed, low or high prenatal exposure as assessed by maternal blood levels; breast-fed for less than 16 wk, low or high prenatal exposure; and breast-fed for more than 16 wk, low or high prenatal exposure. On the Tower of London, there was evidence of prenatal effects (formula-fed high vs. low) as well as a postnatal effect (breast-fed low vs. formula-fed low). There was a marginal effect of breast-fed long versus breast-fed short, and other comparisons were as would be expected (breast-fed short or long vs. formula-fed). On a simple reaction-time task, only prenatal exposure was predictive of performance. The German study measured three congeners in breast milk and cord blood, and recorded weeks of breastfeeding. They reported a negative association between prenatal exposure, as measured by breast milk but not cord blood PCB concentrations, and performance on the Bayley at 30 mo and the K-ABC at 42 mo (43). They also reported a negative effect of postnatal exposure on cognitive performance on the K-ABC at 42 mo of age, measured either by the child’s concurrent blood PCB concentration or breast milk concentration X weeks of breastfeeding. Potential effects of postnatal exposure were apparently not assessed before 42 mo. Based on 70 children (fewer than half the original cohort), effects of neither milk PCB levels (maternal exposure) nor the child’s concurrent blood PCB concentration (postnatal exposure) were significantly associated with the mental processing composite scale of the K-ABC at 72 mo of age, although the trend was negative for both (66). In the Michigan study, prenatal exposure as measured by umbilical cord blood was associated with poorer performance during infancy and at four years of age (45,46,49,62,103). In contrast, exposure via breastfeeding as determined by breast milk PCB concentration X weeks of breastfeeding was not predictive of cognitive performance. However, decreased activity as rated by three adults was associated with postnatal exposure through breastfeeding at four years of age (9). Measures of IQ, attention, and memory were associated with a composite marker of prenatal exposure including cord and maternal blood and breast milk PCB concentrations at 4 and 11 yr (8,57). In a reexamination of the effects at 4 and 11 yr (63), effects were observed on nine of 21 outcome measures in infants breast-fed fewer than six weeks, and on two different measures in infants breast-fed more than six weeks. These latter effects (reaction time on a mental rotation task and errors on the Seashore Rhythm test) may be chance findings, or may reflect an influence of postnatal exposure. In the Michigan study, prenatal exposure was measured by cord blood and/or breast milk PCB concentrations. One breast milk sample was collected within 4.5 mo postpartum, with a median of one month (57). Postnatal exposure was calculated by the measured concentration in breast milk X weeks of breastfeeding. The presumably large variability in time of collection would result in significant exposure misclassification, given that PCB milk concentrations would change dramatically over the first several

56

Rice

months of breastfeeding. Control for prenatal exposure was included in the postnatal analyses, and the child’s blood PCB concentration at four years was used as a control variable in the prenatal analyses in the re-analyses (63). There was no information concerning how many blood samples were above the limit of detection with the old methodology available at the time the study was performed. However, the mean and standard deviations appeared similar to those of cord blood, in which 70% of samples were below the limit of detection. It is therefore likely that the control for concurrent body burden on prenatal effects was relatively poor. In contrast, prenatal exposure was measured by a composite of breast milk, maternal serum, and cord serum levels at 4 and 11 yr, and so was likely a more accurate indicator of prenatal exposure than a child’s blood PCB concentration at 4 yr was an indicator of postnatal exposure (see comments below on over-controlling). No effects were attributed to postnatal exposure as measured by the child’s concurrent blood PCB level, after controlling for prenatal exposure. Conclusions. Several design issues are relevant to the ability of studies performed to date to identify effects of postnatal exposure. Two of the three modern studies (Dutch, German) measured PCB levels in milk taken shortly after birth. The Dutch study collected only one sample, whereas the German study collected milk at two and four weeks postpartum, and averaged them to derive a measure of PCB milk concentration. Both studies multiplied a PCB concentration X weeks of breastfeeding to estimate exposure through breastfeeding. Neither study collected sequential samples beginning within a day or so of birth and ascertained the PCB “wash-out” curve, which is presumably exponential and with an unknown half-time (either in terms of central tendency or on an individual basis). Multiplying one value X weeks of exposure would result in exposure misclassification, which may be substantial if there was a large variability in the length of time of breastfeeding (such that babies stopped exposure at different points on a rather steep wash-out curve). Exposure misclassification of course biases results toward not finding an effect even if there is one. The North Carolina study considered the decline in breast milk concentrations over the duration of breast feeding in the estimation of exposure through breast milk. They sampled milk at birth and several times thereafter. The average percent decrease in milk PCB levels across time based on all women, with linear extrapolation between time points, was used in the estimation of total PCB intake (93). Lack of sufficient data from individual women precluded modeling wash-out curves for individuals (Gladen, personal communication). While not an optimum strategy, this approach is an improvement over other studies in that it incorporated at least an approximation of the decrease in breast milk levels over time. The North Carolina study did not find any effect of PCBs on the McCarthy Scales at 3–5 yr. Further testing was not performed, so that the possibility that postnatal exposure may have predicted effects later in childhood could not be explored. The Dutch and German studies measured blood PCB concentration of the child concurrent with testing at 3.5 yr. Both studies reported effects associated with the child’s concurrent PCB body burden, which is largely the result of exposure through breast milk (97,98). However, this measure too is an imprecise indicator of the child’s exposure during infancy. It is not known whether the child’s body burden at the time of testing (reflecting ongoing effects of exposure to PCBs) or accurate determination of early postnatal exposure would have been the best predictor of performance. The second issue that might result in an underestimation of the effects of breastfeeding is the fact that the effects of breastfeeding were generally assessed controlling for in utero exposure, using blood and/or breast milk concentrations. First, in utero and postnatal exposure via breast milk will be highly correlated, even given that mothers breast feed for different lengths of time, since breast milk is highest following

PCB Exposure on Neuropsychological Function

57

delivery and decreases markedly over a few weeks. Therefore, the early and highest exposure shortly after birth is essentially eliminated from statistical analysis by this procedure, leaving the inaccurate single concentration X weeks of feeding as the measure of postnatal exposure. This analytic strategy is based on the assumption that the primary contribution to any observed deficits will always be in utero exposure. Although it is established that in utero exposure may produce adverse effects (from infants not breast-fed) the strategy of controlling for in utero exposure may represent overcontrolling. The most recent analysis by the Dutch investigators avoided this dilemma by dichotomizing maternal PCB concentration and duration of breastfeeding, eliminating the temptation to control for prenatal exposure when examining the effects of breastfeeding (60). The importance of the contribution of PCBs from breastfeeding to performance is still very much an open question. We do not know how important exposure to PCBs through breastfeeding may be, nor do we understand the interaction of PCB exposure with the benefits of breastfeeding on neuropsychological development. We also do not have much understanding of what kinds of behaviors may be affected by postnatal exposure to PCBs. Although the Dutch study did not find effects of postnatal exposure on cognition at 3.5 yr, the German study did. The Dutch study did find an effect on executive function associated with postnatal exposure at nine years of age, however. Effects on the child’s concurrent blood levels were also associated with other endpoints in the Dutch study. Moreover, the Oswego study did not provide the opportunity to examine postnatal effects, so that the emphasis on prenatal exposure as the important (or only) contributor to adverse outcome may be misleading.

DETERMINATION OF THE RELATIONSHIP BETWEEN EXPOSURE AND EFFECT There is no information published to date on the shape of the relationship between exposure of the infant (or child) to PCBs and performance on any measure. It is unknown whether the relationship is best fit by a linear model, which would suggest that there is no threshold within the range of body burdens studied, or whether it is sublinear (shallower slope at lower body burdens), suggesting that there is a threshold of body burden below which there does not appear to be an adverse effect. This is of vital importance to our understanding of the potential burden for society that ubiquitous exposure to PCBs may produce. Investigators of these studies performed analyses to determine whether there was a statistical association between exposure and performance on one or more measures, which is the standard form of analysis. In contrast to PCBs, the shape of the exposureeffect relationship, and whether there may be a threshold, has been studied for the neurotoxicants lead and methylmercury, both by individual investigators (104–106) and government agencies (88,107–109). For those neurotoxicants, there is evidence that the relationship may be supralinear: i.e. a relatively steeper slope, and therefore greater relative effect, at lower body burdens than higher. Some limited information may be gleaned from graphic representation of data in publications. For example, in the Dutch study, performance on the K-ABC was represented graphically in five categories (42). For all scores (overall cognitive, sequential processing, and simultaneous processing), all four of the higher groups performed more poorly than the lowest group (sum PCB in maternal plasma !1.5 ppb). Information is not available concerning performance in the lowest category, so no conclusions may be drawn concerning possible threshold. Similarly, there appears to be a more-or-less monotonic

58

Rice

relationship in the Oswego study between increased PCB cord blood levels (those with values below the limit of detection, compared to other groups,) for performance on the McCarthy General Cognitive Index, Perceptual and Quantitative Scales (44). In the German study, the second quintile with respect to PCB milk levels appears unimpaired relative to the lowest quintile on the Bayley Scales at 30 or 42 mo (43), suggestive of a threshold. Such ad hoc divination by the reader of course is thoroughly inadequate in determining the association between body burden and effect. The Michigan investigators have performed a Bench Mark Dose (BMD) analysis of four outcomes measured at 11 yr of age (110). This analysis is useful for risk assessment, in that it calculates an exposure (body burden) associated with a defined risk, which can then serve as the starting point for deriving an acceptable intake level. The issue of the shape of the exposure-effect relationship was not explored, however: a linear relationship was apparently assumed. The failure on the part of most investigators to determine the relationship between exposure and effect represents a serious gap in the information required to formulate public health policy.

THYROID HORMONES AS A POTENTIAL MEDIATOR OF NEUROTOXIC EFFECTS Developmental PCB exposure affects circulating thyroid levels in both mother and offspring (111), and effects on thyroid function have been postulated as a mechanism for PCB neurotoxicity. Schantz et al. (112) reported no correlation between degree of thyroid suppression in blood and neurotoxicity following exposure to various congeners in rats. Moreover, the pattern of behavioral effects was different for different congeners, suggesting that even if thyroid suppression is contributing to the behavioral effects, it cannot be the only mechanism. The Dutch study is the only environmental epidemiological study to measure levels of circulating thyroid hormone. They found decreased concentrations of T3 and/or T4 in mothers before and after delivery, and decreased infant T3 and increased TSH at both two weeks and three months of age (113). Effects were associated with TEQ in breast milk. It is unknown whether the effects on circulating thyroid hormones were correlated with performance on any measure. Whereas it is clear that thyroid hormones are important for neurological development (111,114), the degree to which suppression of thyroid hormones by PCBs mediates neurotoxicity, if at all, at environmental exposures is unknown. However, as pointed out by Zoeller (115), the relationship between relevant events in the brain and circulating levels of thyroid hormones in the blood may be weak at best, so that elucidation of this issue is probably best pursued in experimental studies.

SUMMARY AND OVERALL CONCLUSIONS 1. There is good congruence between studies demonstrating that prenatal exposure to PCBs is associated with adverse effects on early neuromotor development and measures of IQ in early childhood. In the Michigan study, effects on IQ were still present at 11 yr of age. In newer studies, effects on IQ attenuated at 4–6 yr of age. However, there is evidence that effects on IQ continue: specifically, that

PCB Exposure on Neuropsychological Function

2.

3.

4.

5.

6.

7.

59

effects are present in children with less advantaged but not more advantaged home environments. In addition, the Dutch investigators reported effects on executive function at nine years of age. Effects have been observed on behavioral domains in addition to cognition. Deficits in attention and failure of response inhibition have been observed in several studies. Differences in activity level, play behavior, and a clinical scale of problem behavior were documented in the Dutch study. Most non-cognitive effects have been studied in early childhood, and whether effects will still be observed as the children age is unknown. However, the Dutch investigators reported differences in sexually dimorphic play behavior at 6.5 yr of age. Measurement of IQ in the studies discussed here should continue, and newer studies (such as the Inuit) should include measures of IQ. It is not yet established that effects on IQ disappear with age. It is also important to assess additional functional domains: attention, impulsivity, language processing, and social and sexually dimorphic behaviors. There is currently some evidence for effects on these domains, but effects need to be fully characterized. This may be especially important since some of the effects associated with PCB exposure are similar to those produced by lead: deficits in attention (116) and impulse control (117), perseveration on the Wisconsin Card Sort Test (118), and abnormal scores on the CBCL (72). Deficits in these functional domains may contribute to increased criminality, suggesting that similar consequences may be associated with PCB exposure as the children get older. The shape of the relationship between exposure (body burden) and effects is unknown. It is not determined whether there is evidence for a threshold within the range of body burdens of individuals in these studies, or conversely, whether effects are actually relatively greater at lower body burdens. This issue is of vital importance for public health policy decisions. The relative potency of specific congeners or congener classes to produce neurotoxic effects is unknown. The epidemiological studies have been unelucidating in this regard, and only a few congeners have been explored in animal models. It might be fruitful to model dose-effect functions for individual congeners or congener classes from current studies that measured a sufficient number of congeners (Dutch, Oswego). The contribution of the effects of other chemicals in addition to PCBs is not well characterized. There is some evidence, particularly for the Faroe Islands study, that PCBs and methylmercury may interact to produce neurotoxicity. There does not appear to be an adverse effect of DDE or other pesticides assessed in the Oswego study. Results from the Inuit study, which measured a number of pesticides, are yet to be published. The issue of the contribution of dioxin, and dioxin-like PCBs, is of considerable importance, and has only been examined in the Dutch study. Results suggest that for some types of outcomes, TEQ, including only planar PCB TEQ, may be predictive of outcome, whereas total PCB concentration is not. The relative importance of postnatal exposure to observed effects is inadequately examined. There have not been accurate data collected on postnatal exposure during infancy. Even so, the two modern studies that have determined PCB blood level of the child have reported effects related to concurrent body burden. In addition, the Dutch investigators observed effects associated with duration of breast feeding at nine years of age. This issue is obviously important with respect to advice regarding breastfeeding.

60

Rice

In summary, there is compelling evidence that developmental exposure to PCBs produces behavioral impairment on a number of domains. However, there are significant gaps in our understanding concerning the effects of PCBs. One of the most important from a public health perspective is the lack of information on the shape of the exposure-effect relationship, including whether there is evidence for a threshold. Another critical area from a public health perspective is the characterization of the effects of postnatal exposure, and at what concentrations adverse effects of PCB exposure potentially negates the positive effects of breastfeeding on neuropsychological function.

REFERENCES 1. Guo YL, Lai TJ, Chen SJ, Hsu CC. Gender-related decrease in Raven’s progressive matrices scores in children prenatally exposed to polychlorinated biphenyls and related contaminants. Bull Environ Contam Toxicol 1995; 55:8–13. 2. Chen YJ, Hsu CC. Effects of prenatal exposure to PCBs on the neurological function of children: a neuropsychological and neurophysiological study. Dev Med Child Neurol 1994; 36:312–320. 3. Chen YC, Yu ML, Rogan WJ, Gladen BC, Hsu CC. A 6-year follow-up of behavior and activity disorders in the Taiwan Yu-cheng children. Am J Public Health 1994; 84:415–421. 4. Chen YCJ, Guo YL, Hsu CC, Rogan WJ. Cognitive development of Yu-Cheng (Oil Disease) children prenatally exposed to heat-degraded PCBs. JAMA 1992; 268:3213–3218. 5. Harada M. Intrauterine poisoning. Bull Inst Constit Med 1976; 25:38–61. 6. Fein G, Jacobson J, Jacobson S, Schwartz P, Dowler J. Prenatal exposure to polychlorinated biphenyls: effects on birth size and gestational age. J Pediatr 1984; 105:315–320. 7. Rogan WJ, Gladen BC, McKinney JD, et al. Neonatal effects of transplacental exposure to PCBs and DDE. J Pediatr 1986; 109:335–341. 8. Jacobson JL, Jacobson SW. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. N Engl J Med 1996; 335:783–789. 9. Jacobson JL, Jacobson SW. A 4-year follow-up study of children born to consumers of great lake fish. J Great Lakes Res 1993; 19:776–783. 10. Stewart P, Darvill T, Lonky E, Reihman J, Pagano J, Bush B. Assessment of prenatal exposure to PCBs from maternal consumption of Great Lakes fish: an analysis of PCB pattern and concentration. Environ Res 1999; 80:S87–S96. 11. Winneke G, Bucholski A, Heinzow B, et al. Developmental neurotoxicity of polychlorinated biphenyls (PCBs): cognitive and psychomotor functions in 7-month old children. Toxicol Lett 1998;102–103: 423–428. 12. Huisman M, Koopman-Esseboom C, Fidler V, et al. Perinatal exposure to polychlorinated biphenyls and dioxins and its effect on neonatal neurological development. Early Hum Dev 1995; 41:111–127. 13. Grandjean P, Weihe P, Burse VW, et al. Neurobehavioral deficits associated with PCB in 7-year-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001; 23:305–317. 14. Muckle G, Ayotte P, Dewailly E, Jacobson SW, Jacobson JL. Prenatal exposure of the Northern Quebec Inuit infants to environmental contaminants. Environ Health Perspect 2001; 109:1291–1299. 15. Schantz SL, Widholm JJ, Rice DC. Effects of PCB Exposure on neuropsychological function in children. Environ Health Perspect 2003; 111:357–376. 16. Schantz SL. Developmental neurotoxicity of PCBs in humans: what do we know and where do we go from here? Neurotoxicol Teratol 1996; 18:217–228. 17. Rice DC. Relationship between measures of exposure to PCBs/dioxins and behavioral effects in recent developmental studies. Hum Ecol Risk Assess 2001; 7:1059–1077.

PCB Exposure on Neuropsychological Function

61

18. Rice DC. PCBs and behavioral impairment: are there lessons we can learn from lead? Open Peer Commentary Neurotoxicol Teratol 1996; 18:229–232. 19. Rice DC. Effects of postnatal exposure of monkeys to a PCB mixture on spatial discrimination reversal and DRL performance. Neurotoxicol Teratol 1998; 20:391–400. 20. Rice DC. Effect of postnatal exposure to a PCB mixture in monkeys on multiple fixed interval-fixed ratio performance. Neurotoxicol Teratol 1997; 19:429–434. 21. Rice DC, Hayward S. Effects of postnatal exposure to a PCB mixture in monkeys on nonspatial discrimination reversal and delayed alternation performance. Neurotoxicology 1997; 18:479–494. 22. Schantz SL, Levin ED, Bowman RE, Heironimus MP, Laughlin NK. Effects of perinatal PCB exposure on discrimination-reversal learning in monkeys. Neurotoxicol Teratol 1989; 11:243–250. 23. Levin ED, Schantz SL, Bowman RE. Delayed spatial alternation deficits resulting from perinatal PCB exposure in monkeys. Arch Toxicol 1988; 62:267–273. 24. Widholm JJ, Clarkson GB, Strupp BJ, Crofton KM, Seegal RF, Schantz SL. Spatial reversal learning in aroclor 1254-exposed rats: sex-specific deficits in associative ability and inhibitory control. Toxicol Appl Pharmacol 2001; 174:188–198. 25. Schantz SL, Seo BW, Moshtaghian J, Peterson RE, Moore RW. Effects of gestational and lactational exposure to TCDD or coplanar PCBs on spatial learning. Neurotoxicol Teratol 1996; 18:305–313. 26. Schantz SL, Seo BW, Wong PW, Pessah IN. Long-term effects of developmental exposure to 2,2 0 ,3,5 0 ,6-pentachlorobiphenyl (PCB 95) on locomotor activity, spatial learning and memory and brain ryanodine binding. Neurotoxicology 1997; 18:457–468. 27. Schantz SL, Widholm JJ. Cognitive effects of endocrine-disrupting chemicals in animals. Environ Health Perspect 2001; 109:1197–1206. 28. Lasky RE, Widholm JJ, Crofton KM, Schantz SL. Perinatal exposure to Aroclor 1254 impairs distortion product otoacoustic emissions (DPOAEs) in rats. Toxicol Sci 2002; 68:458–464. 29. Roegge CS, Seo BW, Crofton KM, Schantz SL. Gestational-lactational exposure to Aroclor 1254 impairs radial-arm maze performance in male rats. Toxicol Sci 2000; 57:121–130. 30. Goldey ES, Crofton KM. Thyroxine replacement attenuates hypothyroxinemia, hearing loss, and motor deficits following developmental exposure to Aroclor 1254 in rats. Toxicol Sci 1998; 45:94–105. 31. Herr DW, Goldey ES, Krofton KM. Developmental exposure to Aroclor 1254 produces lowfrequency alteration in adult rat brainstem auditory evoked responses. Fundam Appl Toxicol 1996; 33:120–128. 32. Goldey ES, Kehn LS, Lau C, Rehnberg GL, Crofton KM. Developmental exposure to polychlorinated biphenyls (Aroclor 1254) reduces circulating thyroid hormone concentrations and causes hearing deficits in rats. Toxicol Appl Pharmacol 1995; 135:77–88. 33. Wong PW, Garcia EF, Pessah IN. Ortho-substituted PCB 95 alters intracellular calcium signaling and causes cellular acidification in PC12 cells by an immunophilin-dependent mechanism. J Neurochem 2001; 76:450–463. 34. Wong PW, Joy RM, Albertson TE, Schantz SL, Pessah IN. Ortho-substituted 2,2 0 ,3,5 0 ,6-pentachlorobiphenyl (PCB 95) alters rat hippocampal ryanodine receptors and neuroplasticity in vitro: evidence for altered hippocampal function. Neurotoxicology 1997; 18:443–456. 35. Kodavanti PR, Tilson HA. Neurochemical effects of environmental chemicals: in vitro and in vivo correlations on second messenger pathways. Ann NY Acad Sci 2000; 919:97–105. 36. Kodavanti PR, Tilson HA. Structure–activity relationships of potentially neurotoxic PCB congeners in the rat. Neurotoxicology 1997; 18:425–441. 37. Richardson JR, Miller GW. Acute exposure to Aroclor 1016 or 1260 differentially affects dopamine transporter and vesicular monoamine transporter 2 levels. Toxicol Lett 2004; 148:29–40.

62

Rice

38. Seegal RF, Bush B, Shain W. Lightly chlorinated ortho-substituted PCB congeners decrease dopamine in nonhuman primate brain and in tissue culture. Toxicol Appl Pharmacol 1990; 106:134–144. 39. Seegal RF, Okoniewski RJ, Brosch KO, Bemis JC. Polychlorinated biphenyls alter extraneuronal but not tissue dopamine concentrations in adult rat striatum: an in vivo microdialysis study. Environ Health Perspect 2002; 110:1113–1117. 40. Longnecker MP, Wolff MS, Gladen BC, et al. Comparison of polychlorinated biphenyl levels across studies of human neurodevelopment. Environ Health Perspect 2003; 111:65–70. 41. Gladen B, Rogan W. Effects of perinatal polychlorinated biphenyls and dichlorodiphenyl dichloroethane on later development. J Pediatr 1991; 119:58–63. 42. Patandin S, Lanting CI, Mulder PG, Boersma ER, Sauer PJ, Weisglas-Kuperus N. Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in Dutch children at 42 months of age. J Pediatr 1999; 134:7–9. 43. Walkowiak J, Wiener J-A, Fastabend A, et al. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on psychodevelopment in early childhood. Lancet 2001; 358:1602–1607. 44. Stewart PW, Reihman J, Lonky EI, Darvill TJ, Pagano J. Cognitive development in preschool children prenatally exposed to PCBs and MeHg. Neurotoxicol Teratol 2003; 25:11–22. 45. Jacobson JL, Jacobson SW, Humphrey HEB. Effects of in utero exposure to polychlorinated biphenyls and related contaminants on cognitive functioning in young children. J Pediatr 1990; 116:38–45. 46. Jacobson JL, Jacobson SW, Padgett R, Brumitt G, Billings R. Effects of prenatal PCB exposure on cognitive processing efficiency and sustained attention. Dev Psychol 1992; 28:297–306. 47. Stewart P, Reihman J, Lonky E, Darvill T, Pagano J. Prenatal PCB exposure and neonatal behavioral assessment scale (NBAS) performance. Neurotoxicol Teratol 2000; 22:21–29. 48. Jacobson JL, Jacobson SW, Fein GG, Schwartz PM, Dowler JK. Prenatal exposure to an environmental toxin: a test of the multiple effects model. Dev Psychol 1984; 20:523–532. 49. Jacobson SW, Jacobson JL, Fein GG. Environmental toxins and infant development. In: Fitzgerald HE, Lester BM, Yogman MW, eds. Theory and Research in Behavioral Pediatrics. New York: Plenum Press, 1986:96–146. 50. Rogan WJ, Gladen BC. PCBs, DDE, and child development at 18 and 24 months. Ann Epidemiol 1991; 1:407–413. 51. Koopman-Esseboom C, Weisglas-Kuperus N, de Ridder MAJ, Van der Paauw CG, Tuinstra LGMTh, Sauer PJJ. Effects of polychlorinated biphenyl/dioxin exposure and feeding type on infants’ mental and psychomotor development. Pediatrics 1996; 97:700–706. 52. Darvill T, Lonky E, Reihman J, Stewart P, Pagano J. Prenatal exposure to PCBs and infant performance on the Fagan test of infant intelligence. Neurotoxicology 2000; 21:1029–1038. 53. Vreugdenhil HJ, Mulder PG, Emmen HH, Weisglas-Kuperus N. Effects of perinatal exposure to PCBs on neuropsychological functions in the Rotterdam cohort at 9 years of age. Neuropsychology 2004; 18:185–193. 54. Patandin S, Veenstra J, Mulder PGH, Sewnaik A, Sauer PJJ, Weisglas-Kuperus N. Attention and activity in 42-month-old Dutch children with environmental exposure to polychlorinated biphenyls and dioxins. Effects of Environmental Exposure to Polychlorinated Biphenyls and Dioxins on Growth and Development in Young Children. Ph.D. Thesis. Erasmus University, Rotterdam, 1999: 123–142 55. Stewart P, Fitzgerald B, Reihman J, et al. Prenatal PCB exposure, the corpus collosum, and response inhibition. Environ Health Perspect published 2003. 56. Jacobson JL, Jacobson SW. Specific effects of prenatal PCB exposure on attention and behavior in school-age children. XX Int Neurotoxicol Meet, Little Rock AR, Feb. 2003 9–12. 57. Jacobson JL, Jacobson SW. Prenatal exposure to polychlorinated biphenyls and attention at school age. J Pediatr 2003; 143:780–788. 58. Jacobson JL, Jacobson SW, Humphrey HEB. Effects of exposure to PCBs and related compounds on growth and activity in children. Neurotoxicol Teratol 1990; 12:319–326.

PCB Exposure on Neuropsychological Function

63

59. Patantin S, Koot HM, Sauer PJJ, Weisglas-Kuperus N. Problem behavior in Dutch preschool children in relation to background polychlorinated biphenyl and dioxin exposure. Effects of Environmental Exposure to Polycholorinate Biphenyls and Dioxins on Growth and Development in Young Children Ph.D. Thesis. Erasmus University, Rotterdam, 1999: 143–156. 60. Vreugdenhil HJ, Slijper FM, Mulder PG, Weisglas-Kuperus N. Effects of pernatal exposure to PCBs and dioxins on play behavior in Dutch children at school age. Environ Health Perspect 2002; 110:A593–A598. 61. Vreugdenhil HJI, Lanting CI, Mulder PGH, Boersma ER, Weisglas-Kuperus N. Effects of prenatal PCB and dioxin background exposure on cognitive and motor abilities in Dutch children at school age. J Pediatr 2002; 140:48–56. 62. Jacobson SW, Jacobson JL. Breastfeeding and intelligence. Lancet 1992; 339:926. 63. Jacobson JL, Jacobson SW. Breastfeeding and gender as moderators of teratogenic effects on cognitive development. Neurotoxicol Teratol 2002; 24:349–358. 64. Jacobson SW, Chiodo LM, Jacobson JL. Breastfeeding effects on intelligence quotient in 4- and 11-year-old children. Pediatrics 1999; 103:e71–e77. 65. Bellinger DC. Effect modification in epidemiologic studies of low-level neurotoxicant exposures and health outcomes. Neurotoxicol Teratol 2000; 22:133–140. 66. Winneke G, Kraemer U, Walkowiak J, et al. Delay of neurobehavioral development following pre- and postnatal PCB exposure: persistent or reversible? The second PCB workshop— recent advances in the environmental toxicology and health effects of PCBs. Brno, Czech Republic, May 2002. Hedelberg: Springer-Verlag, 2002. Abstract published in NaunynSchmiedeberg’s Archives of Pharmacolgy (2002), Volume 365 supplement. 67. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 68. Dietrich KN, Succop PA, Berger OG, Hammond PB, Bornschein RL. Lead exposure and the cognitive development of urban preschool children: the Cincinnati Lead Study cohort at age 4 years. Neurotoxicol Teratol 1991; 13:203–211. 69. Dietrich KN, Succop PA, Bornschein RL, et al. Lead exposure and neurobehavioral development in later infancy. Environ Health Prespect 1990; 89:13–19. 70. Schantz SL, Moshtaghian J, Ness DK. Spatial learning deficits in adult rats exposed to orthosubstituted PCB congeners during gestation and lactation. Fundam Appl Toxicol 1995; 26:117–126. 71. Needleman HL, McFarland C, Ness RB, Fienberg SE, Tobin MJ. Bone lead levels in adjudicated delinquents. A case control study. Neurotoxicol Teratol 2002; 24:711–717. 72. Needleman HL, Riess JA, Tobin MJ, Biesecker GE, Greenhouse JB. Bone lead levels and delinquent behavior. J Am Med Assoc 1996; 275:363–369. 73. Dietrich KN, Ris MD, Succop PA, Berger OG, Bornschein RL. Early exposure to lead and juvenile delinquency. Neurotoxicol Teratol 2001; 23:511–518. 74. Gladen BC, Doucet J, Hansen LG. Assessing human polychlorinated biphenyl contamination for epidemiologic studies: lessons from patterns of congener concentrations in Canadians in 1992. Environ Health Perspect 2003; 111:437–443. 75. Gerstenberger SL, Dellinger JA, Hansen LG. Concentrations and frequencies of polychlorinated biphenyl congeners in a native American population that consumes great lakes fish. J Toxicol Clin Toxicol 2000; 38:729–746. 76. Hong CS, Xiao J, Casey AC, Bush B, Fitzgerald EF, Hwang SA. Mono-ortho- and non-orthosubstituted polychlorinated biphenyls in human milk from Mohawk and control women: effects of maternal factors and previous lactation. Arch Environ Contam Toxicol 1994; 27:431–437. 77. Van den Berg M, Birnbaum L, Bosveld AT, et al. Toxic equivalency factors (TEFs) for PCBs, PCDDs, PCDFs for humans and wildlife. Environ Health Perspect 1998; 106:775–792. 78. Bushnell PJ, Rice DC. Behavioral assessments of learning and attention in rats exposed perinatally to 3,3 0 ,4,4 0 ,5-pentachlorobiphenyl (PCB 126). Neurotoxicol Teratol 1999; 21:381–392.

64

Rice

79. Crofton KM, Rice DC. Low-frequency hearing loss following perinatal exposure to 3,3 0 ,4,4 0 ,5-pentachlorobiphenyl (PCB 126) in rats. Neurotoxicol Teratol 1999; 21:299–301. 80. Geller AM, Bushnell PJ, Rice DC. Behavioral and electrophysiological estimates of visual thresholds in awake rats treated with 3,3 0 ,4,4 0 ,5-pentachlorobiphenyl (PCB 126). Neurotoxicol Teratol 2000; 22:521–531. 81. Rice DC, Hayward S. Effects of exposure to 3,3 0 ,4,4 0 ,5-pentachlorobiphenyl (PCB 126) throughout gestation and lactation on behavior (concurrent random interval-random interval and progressive ratio performance) in rats. Neurotoxicol Teratol 1999; 21:679–687. 82. Rice DC. Effect of exposure to 3,3 0 ,4,4 0 ,5-pentachlorobiphenyl (PCB 126) throughout gestation and lactation on development and spatial delayed alternation performance in rats. Neurotoxicol Teratol 1999; 21:59–69. 83. Weisglas-Kuperus N, Patandin S, Berbers GA, et al. Immunologic effects of background exposure to polychlorinated biphenyls and dioxins in Dutch preschool children. Environ Health Perspect 2000; 108:1203–1207. 84. Bellinger DC, Stiles KM, Needleman HL. Low-level lead exposure, intelligence and academic achievement: a long-term follow-up study. Pediatrics 1992; 90:855–861. 85. Dietrich KN, Berger OG, Succop PA, Hammond PB, Bornschein RL. The developmental consequences of low to moderate prenatal and postnatal lead exposure: intellectual attainment in the Cincinnati Lead Study Cohort following school entry. Neurotoxicol Teratol 1993; 15:37–44. 86. Carpenter DO, Hussain RJ, Berger DF, Lombardo JP, Park H-Y. Electrophysiologic and behavioral effects of perinatal and acute exposure of rats to lead and polychlorinated biphenyls. Environ Health Perspect 2002; 110:377–386. 87. NIEHS (National Institute of Environmental Health Sciences). Scientific issues relevant to assessment of health effects from exposure to methylmercury. Workshop organized by Committee on Environmental and Natural Resources (CENR) Office of Science and Technology Policy (OSTP), The White House, Raleigh, NC, November 18–20, 1998. 88. NRC (National Research Council). Committee on the Toxicological effects of Methylmercury, Board on Environmental Studies and Toxicology, Toxicological effects of methylmercury. Committee on the Toxicological Effects of Methylmercury, Board on Environmental Studies and Toxicology, Commission on Life Sciences, National Research Council. Washington, DC: National Academy Press, 2000. 89. Budtz-Jørgensen E, Keiding N, Grandjean P, White RF, Weihe P. Methylmercury neurotoxicity independent of PCB exposure. Environ Health Perspect 1999; 107:A236–A237. 90. Schantz SL, Ferguson SA, Bowman RE. Effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on behavior of monkeys in peer groups. Neurotoxicol Teratol 1992; 14:433–446. 91. Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated biphenyls (PCBs) and dichlorodiphenyl dichloroethylene (DDE) in human milk: effects of maternal factors and previous lactation. Am J Public Health 1986; 76:172–177. 92. Gladen BC, Rogan WJ, Hardy P, Thullen J, Tingelstad J, Tully M. Development after exposure to polychlorinated biphenyls and dichlorodiphenyl dichloroethene transplacentally and through human milk. J Pediatr 1988; 113:991–995. 93. Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated biphenyls (PCBs) and dichlorodiphenyl dichloroethene (DDE) in human milk: effects on growth, morbidity, and duration of lactation. Am J Public Health 1987; 77:1294–1297. 94. Steuerwald U, Weihe P, Jørgensen PJ, et al. Maternal seafood diet, methylmercury exposure, and neonatal neurologic function. J Pediatr 2000; 136:599–605. 95. Muckle G, XXI International Neurotoxicology Meeting. Honolulu, HI, Feb 9–12, 2004. 96. Jacobson JL, Humphrey HEB, Jacobson SW, Schantz SL, Mullin MD, Welch R. Determinants of polychlorinated biphenyls (PCBs), polybrominated biphenyls (PBBs), and dichlorodiphenyl trichloroethane (DDT) levels in the sera of young children. Am J Public Health 1989; 79:1401–1404.

PCB Exposure on Neuropsychological Function

65

97. Patandin S, Dagnelie PC, Mulder PG, et al. Dietary exposure to polychlorinated biphenyls and dioxins from infancy until adulthood: A comparison between breastfeeding, toddler, and longterm exposure. Environ Health Perspect 1999; 107:45–51. 98. Patanin S, Weisglas-Kuperus N, de Ridder MAJ, et al. Plasma polychlorinated biphenyl levels in Dutch preschool children either breast-fed or formula-fed during infancy. Am J Public Health 1997; 87:1711–1714. 99. Lanting CI, Fidler V, Huisman M, Boersma ER. Determinants of polychlorinated biphenyl levels in plasma from 42-month-old children. Arch Environ Contam Toxicol 1998; 35:135–139. 100. Schecter A, Ryan JJ, Papke O. Decrease in levels and body burden of dioxins, dibenzofurans, PCBs, DDE, and HCB in blood and milk in a mother nursing twins over a thirty-eight month period. Chemosphere 1998; 37:1807–1816. 101. Abraham K, Papke O, Gross A, et al. Time course of PCDD/PCDF/PDB concentrations in breastfeeding mothers and their infants. Chemosphere 1998; 37:1731–1741. 102. Rice DC, Barone S. Critical periods of vulnerability for the developing nervous system: evidence from humans and animal models. Environ Health Perspect 2000; 3:511–533. 103. Jacobson JL, Jacobson SW. New methodologies for assessing the effects of prenatal toxic exposure on cognitive functioning in humans. In: Evans MS, ed. Toxic Contaminants and Ecosystem Health, a Great Lakes Focus. New York: John Wiley & Sons, Inc., 1988:373–388. 104. Davidson PW, Kost J, Myers GJ, Cox C, Clarkson TW, Shamlaye CF. Methylmercury and neurodevelopment: reanalysis of the Seychelles Child Development Study outcomes at 66 months of age. JAMA 2001; 285:1291–1293. 105. Canfield RL, Henderson CR, Cory-Slechta DA, Cox C, Jusko TA, Lanphear BP. Intellectual impairment in children with blood lead concentrations below 10 micrograms per deciliter. N Engl J Med 2003; 348:1517–1526. 106. Bellinger DC, Needleman HL. Intellectual impairment and blood lead levels. N Engl J Med 2003; 349:500–502. 107. Budtz-Jørgensen E, Keiding N, Grandjean P. Benchmark modeling of the Faroese methylmercury data. Final Report to U.S. EPA. Research Report 99/5. Copenhagen: Department of Biostatistics, University of Copenhagen, 1999. 108. Budtz-Jørgensen E, Grandjean P, Keiding N, et al. Benchmark dose calculations of methylmercury-associated neurobehavioral deficits. Toxicol Lett 2000;112–113: 193–199. 109. Schwartz J. Low-level lead exposure and children’s IQ: a meta-analysis and search for a threshold. Environ Res 1994; 65:42–55. 110. Jacobson JL, Janisse J, Banerjee M, Jester J, Jacobson SW, Ager JW. A benchmark dose analysis of prenatal exposure to polychlorinated biphenyls. Environ Health Perspect 2002; 110:393–398. 111. Zoeller RT, Dowling ALS, Herzig CTA, Iannacone EA, Gauger KJ, Bansal R. Thyroid hormone, brain development, and the environment. Environ Health Perspect 2002; 110:355–361. 112. Schantz SL, Seo BW, Moshtaghian J, Amin S. Developmental exposure to polychlorinated biphenyls or dioxin: do changes in thyroid function mediate effects on spatial learning? Am Zool 1997; 37:399–408. 113. Koopman-Esseboom C, Huisman M, Weisglas-Kuperus N, et al. PCB and dioxin levels in plasma and human milk of 418 Dutch women and their infants. Predictive value of PCB congener levels in maternal plasma for fetal and infant’s exposure to PCBs and dioxins. Chemosphere 1994; 28:1721–1732. 114. Howdeshell KL. A model of the development of the brain as a construct of the thyroid system. Environ Health Perspect 2002; 110:337–348. 115. Zoeller RT. Polychlorinated biphenyls as disrupters of thyroid hormone action. In: Robertson LW, Hansen LG, eds. PCBs: Recent Advances in Environmental Toxicology and Health Effects. University of Kentucky Press, Lexington, Kentucky, 2001:265–271.

66

Rice

116. Winneke G, Brockhaus A, Collet W, Kraemer V. Modulation of lead-induced performance deficit in children by varying signal rate in a serial choice reation task. Neurotoxicol Teratol 1989; 11:587–592. 117. Needleman HL, Gunnoe C, Leviton A, et al. Deficits in psychologic and classroom performance of children with elevated dentine lead levels. N Engl J Med 1979; 300:689–695. 118. Stiles KM, Bellinger DC. Neuropsychological correlates of low-level lead exposure in school-age children: a prospective study. Neurotoxicol Teratol 1993; 15:27–35. 119. Jacobson, JL, Jacobson, SW, Prospective studies of exposure to an environmental contaminant: the challenge of hypothesis testing in a multivariate correlational context. Psychol Schools 2004; 46: 625–637.

4 Lead Neurotoxicity in Children: Knowledge Gaps and Research Needs David C. Bellinger Department of Neurology, Children’s Hospital Boston, Harvard Medical School and Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A.

INTRODUCTION It was in Australia in the 1890s, two millennia after lead was recognized as an adult neurotoxicant, that childhood lead poisoning was first described (1). For the next halfcentury, it was conceptualized largely as a clinical intoxication that was simply “present” or “absent.” Moreover, the intoxication was thought to have no lasting impact if a child did not become encephalopathic, which tends to occur at blood lead levels greater than 100 mg/dL. The observations of Byers and Lord (2) posed the first major challenge to this view. In this case series, many of the lead poisoned children who presented without signs of encephalopathy nevertheless suffered significant, enduring behavioral and intellectual problems that interfered substantially with their well-being. This suggested that lead poisoning is better characterized as a continuum than as a dichotomy, with subtler forms of neurological deficit occurring among children with lead exposures that do not produce frank toxicity. Progress was slow, however, in identifying milestones on this continuum and characterizing their relationships to lead dose. Textbooks of pediatrics published as recently as the late 1960s identified a blood lead level of 60 mg/dL as the upper limit of the normal range (3). In light of today’s knowledge, this is remarkable. It is somewhat less so, however, if considered in the context of blood lead screening data from that era, which revealed that among young children residing in areas of large, older cities such as Chicago, Philadelphia, and New York, the mean blood lead level differed by only a little more than a factor of two from the levels at which some children become encephalopathic. For instance, in Baltimore in the 1950s, 90% of low-income 7- to 60-month-old children attending well-baby clinics and hospital outpatient departments had a blood lead level above 30 mg/dL and 26% had a level above 60 (4). Other screening data from the 1960s suggested that 20% to 45% of children had a level greater than 40 mg/dL (5). The margin of safety was clearly dangerously slim. It appears that until the final decades of the 20th century, public health recommendations rested, to a large extent, on the erroneous assumption that the blood lead level that was average within the general population of children was also the level that was “physiologically normal” and thus posed no hazard. 67

68

Bellinger

Although population blood lead levels have fallen as a result of the numerous restrictions placed on the uses of lead, as recently as the late 1970s the mean level of U.S. preschool children was 15 mg/dL and 88% had a level greater than 10 mg/dL (6). By the early 1990s, the mean had declined to 4 mg/dL, with 9% of children having a level greater than 10 mg/dL (7). The decline has continued. The current mean among young children, based on NHANES, 1999–2002 data, is 1.9 mg/dL, with 1.6% having a level greater than 10 mg/dL (8). This still represents approximately 310,000 children, and disparities still exist, with poor and non-Hispanic black children suffering the greatest exposures. The economic benefit of the presumed increase in IQ scores associated with the blood lead decline between 1976 and 1999 was estimated to be $110 to $319 billion for each year’s cohort of 2-year-old children (9). The dramatic decline in children’s blood lead levels has prompted some to conclude that because the mean level is now so close to 0 mg/dL, the steps taken in recent decades to abate key sources and pathways of exposure have brought us to the threshold of eliminating lead neurotoxicity in children. This is, however, an artifact of the measurement units used, by convention, to express blood lead level. If blood lead were measured in mg/L instead of mg/dL, the current mean level would be 20 mg/L; in ng/L, it would be 20,000. What was the blood lead level of humans likely to have been before they began to disturb the natural distribution of lead in the earth’s crust by mining and other activities? Two estimates are available. One, based on the relationship between blood and bone lead levels in contemporary animals and humans, is 0.016 mg/d (10). The other, based on a pharmacokinetic model, is 0.06 to 0.12 mg/dL (11). Therefore, the current mean of 2 mg/dL among U.S. children, although a 90% reduction from the mean during the late 1970s, could be as much as two orders of magnitude greater than children’s “natural” blood lead level. One can hope that most of the population distribution of blood lead levels now falls below the level at which lead produces demonstrable neurotoxicities in children, but it is possible that this goal still lies far ahead. The goal of this chapter is to identify some major lacunae in our understanding of the neurotoxicities associated with childhood lead poisoning and the impediments to filling them. The issues discussed bear on the selection of exposure biomarkers; the selection of critical outcomes; the functional form of the concentration–response/effect relationship; the reversibility of effects; and possible bases for individual differences in vulnerability.

SELECTION OF EXPOSURE BIOMARKERS AND SOURCES OF MISCLASSIFICATION ERRORS Given that the central nervous system is considered the critical target organ for childhood lead poisoning, it would be most helpful to be able to measure, in vivo, the concentration of lead at the cellular site(s) of action in the brain. Because such measurements are not currently feasible, investigators must rely on measurements of lead in more readily accessible but peripheral tissues, most commonly blood and bone. The relationship between brain lead and lead in each of these surrogate tissues is poorly understood, although the pharmacokinetics clearly differ among these compartments. In both rodents (12) and non-human primates (13), brain lead level falls much more slowly than blood lead level following chelation with succimer and, in the rodent, in non-chelated animals after cessation of exposure. These observations suggest that using blood lead as an index of lead in the brain will result in exposure misclassification, although the magnitude of this bias in any specific

Lead Neurotoxicity in Children

69

setting will be difficult to characterize. The most likely direction, however, would be underestimation of the amount of lead in the brain. As an exposure biomarker, blood lead level has other limitations. Only about 5% of an individual’s total body lead burden resides in blood. Furthermore, blood consists of several sub-compartments. More than 90% of lead in whole blood is bound to red cell proteins such as hemoglobin, with the balance in plasma. From a toxicological perspective, this unbound fraction is likely to be the most important sub-compartment of blood lead because of the ease with which it diffuses into soft tissues. The concentration of lead is plasma is much lower than in whole blood, however. For example, in a group of pregnant women with blood lead levels below 10 mg/dL, plasma lead levels were less than 0.3% of the whole blood lead level (14). The use of whole blood lead as a surrogate for plasma lead could be justified if the ratio of whole blood lead:plasma lead were well characterized, but this is not so. At least some studies suggest that it varies several-fold among individuals with the same blood lead level (14–17). Moreover, the ability of red cells to bind lead is limited, so the ratio of blood lead: plasma lead would be expected to be non-linear. Thus, interpreting whole blood lead level as a proxy for plasma lead level, which, itself, is a proxy for brain lead level, will likely result in some exposure misclassification. The greater relative abundance of lead in whole blood makes its measurement much easier (and cheaper) than the measurement of lead in plasma. With the number of toxicokinetic steps separating brain lead level from the exposure biomarkers usually measured, it is remarkable that we have made as much progress as we have in characterizing the concentration-response/effect relationships. Another limitation in the use of blood lead as the exposure biomarker is that its residence time in blood is closely linked to red cell lifetime, with a half-time on the order of 30 days, although it can be somewhat longer in individuals with higher total body burdens because of re-equilibration of lead stored in deeper and shallower pools. In nonchelated children, the time for blood lead to decline to a value less than 10 mg/dL was linearly related to baseline blood lead level. In children with starting levels of 25–29 mg/dL, the mean duration was 24 months, compared to 9 months in children with starting levels of 10–14 mg/dL (18). A single blood lead measurement might therefore provide limited information about an individual’s lead exposure history, a difficulty frequently cited regarding the interpretation of cross-sectional studies, in which blood lead is often measured only once, and sometimes only well after the period when children’s blood lead levels peak (18–30 mo). If it is exposures to lead in the early postnatal years that are most detrimental to children’s development (19), categorizing a child’s exposure status based on the blood lead level that is contemporaneous with the measurement of neurodevelopment at school-age could result in exposure misclassification. Unless intraindividual stability of serial blood lead levels is very high within a study cohort, misclassification would probably be non-differential, resulting in bias towards the null. This concern must be qualified by recent data from some longitudinal studies indicating that concurrent blood lead level, even at ages well beyond 18 to 30 mo, is sometimes the strongest predictor of late outcomes (20). Age-related changes in vulnerability, and the reasons why it might differ across studies, remain uncertain. It might be that among children with chronically elevated exposure, but not in children with relatively low lifetime exposure, blood lead level measured at school-age is a reasonably good marker of cumulative exposure. That concurrent blood lead level is a stronger predictor of schoolage outcomes than is blood lead level in the early postnatal years does not necessarily imply greater vulnerability of the brain to ongoing than to past exposure. The development of X-ray-fluorescence (XRF) methods for measuring lead in mineralized tissues engendered the hope that they would allow better characterization and

70

Bellinger

reconstruction of exposure history. Such tissues are a long-term lead storage site, with a half-life measured in decades, containing approximately 90% of the total body lead burden in adults and 70% in children. Thus, bone lead is an index with a long exposure averaging time (21). Moreover, some data suggest that bone lead, which can be mobilized by many physiologic and pathophysiologic processes, is a major contributor to plasma lead. XRF methods have proven useful in studying individuals with occupational lead exposure (22,23), those living in highly polluted environments (24), and those who grew up when community lead exposures were relatively high (25). In a relatively highly exposed cohort of pregnant women in Mexico City, higher bone lead levels at one month post-partum were associated with reduced birth weight (26), less infant weight gain (27), smaller head circumference and birth length (28), and slower infant development (29). Among children living near a large lead smelter in Yugoslavia, IQ at age 10–12 yr was more strongly associated, inversely, with tibia lead level than with blood lead level (30). Current XRF methods for measuring bone lead levels have limitations, however. Temporal features of exposure history cannot easily be discerned. Some progress has been made toward this goal by examining the spatial distribution of lead in teeth in relation to the relative abundance of stable lead isotopes (31), but the specialized technologies needed to carry out these analyses are unlikely ever to be widely available, and the unpredictability of tooth exfoliation makes this tissue difficult to collect unless the study design involves contact with (and the cooperation of) participants at the appropriate ages. Current XRF methods might not be sufficiently sensitive for studies of the health effects of low-dose community exposures. The bone lead levels of a large percentage of subjects might be below the detection limit, for example, 80% in a case-control study of bone lead levels and juvenile delinquency (32). Efforts continue to modify the instrumentation or measurement protocols to reduce the detection limit. Lead appears to be deposited at sites of most active calcification (33). In children, this is trabecular bone, in which the rate of fractional resorption in early childhood is high. Depending on the amount of a child’s ongoing exposure, lead might not permanently remain in bone, making bone lead level an inaccurate index of lifetime lead exposure. Finally, it is difficult to compare the performance of different laboratories using XRF methods to measure bone lead because of the absence of standard reference materials. A major research need is the development and validation of biomarkers of critical dose that, compared to blood lead or bone lead, are fewer toxicokinetic steps removed from the sites of lead’s actions in the brain. One promising front in the effort to deduce the contents of the “black box” separating external dose and clinical disease is the measurement of processes and products that potentially mediate the association between them. For example, magnetic resonance spectroscopy (MRS) has been used in small case series to measure the ratio of N-acetylaspartate (NAA) to creatine, which is a marker of neuronal and axonal damage, and thus an early biological effect rather than a biomarker of exposure. In children, higher lead exposures are associated with lower NAA:creatine ratios in the frontal gray matter and, to a lesser extent, in frontal white matter (34,35). Similarly, an adult who had higher bone and blood lead levels than did his monozygotic twin had both greater neuropsychological deficits and lower NAA:creatine ratios in the hippocampus, frontal lobe, and midbrain (36). While much remains uncertain about the interpretation of MRS, the use of this and other biochemical imaging methods, in combination with more conventional structural and functional imaging methods, might bring us closer to understanding the mechanisms of lead neurotoxicity. Imprecision in exposure classification can result in estimates of critical dose that are insufficiently protective (37). All other things being equal, exposure misclassification will result in the greatest reduction of statistical power in studies that focus on the lower end of

Lead Neurotoxicity in Children

71

the exposure range. Exposure-related outcome differences would be expected to be smaller at lower than at higher blood lead levels (excepting a supralinear functional form; see “Functional Form of the Concentration–Response/Concentration–Effect Relationship”), so the size of the error introduced by misclassification will be greater in relation to mean blood lead level, representing a greater proportion of the coefficient of variation.

SELECTION OF CRITICAL OUTCOMES The functions measured as indicators of lead’s developmental neurotoxicity have, overwhelmingly, been neurocognitive deficits. Because of the intense controversy that attended proposals to impose significant restrictions on lead’s uses, many study publications focused entirely on analyses of the association between children’s lead levels and their scores on apical tests, such as IQ, that integrate performance over diverse domains. This is understandable in light of the risk assessor’s need to base exposure standards on endpoints for which consensus is firm with regard to individual and societal importance and for which deficits can be readily valued (i.e., monetized). Because individuals with the same IQ can present with markedly different patterns of strengths and weaknesses, however, IQ is perhaps the least useful piece of data collected in a clinical evaluation. A low score suggests the magnitude of a child’s struggles, but relatively little about their nature and, thus, about strategies for amelioration. In most studies, assessment batteries include, in addition to IQ, age-appropriate standardized tests of specific aspects of neurodevelopmental function, administered in hopes that the pattern of test scores will help to identify the neuropsychological mechanisms that underlie the inverse associations between lead and apical test scores. Lead-associated deficits on diverse domain-focused tests have been reported, involving attention, language, executive functions, visual-motor integration, and fine motor skills. Some have suggested that the findings provide a sufficient basis for inferring a “behavioral signature” for lead neurotoxicity (38,39). Any particular study has tended to identify leadassociated deficits in only a subset of these domains, however, and not all children within a study sample display all of the deficits noted at the group level. This is true even among children who suffer clinical lead poisoning (J. Rosen, personal communication). The limited consistency across studies in the domains identified as lead-sensitive provides relatively weak evidence for a true behavioral signature that can, with good sensitivity and specificity, identify the child whose impairments are related to lead and not to some other chemical, biological, or psychosocial factor (40). Some argue that behavioral signatures, in this strict diagnostic sense, have been identified only for Rett Syndrome, Prader–Willi Syndrome, and Lesch–Nyhan Syndrome (41). Furthermore, the deficits that have been proposed as contributing to a behavioral signature for lead are, themselves, vague, referring to broad abstract constructs (e.g., attention) rather than to specific processes. The processes within these broad domains that are lead-sensitive would have to be specified and defined operationally before any purported behavioral signature would be useful. Some of the difficulties encountered in drawing strong inferences about a behavioral signature, and weaker inferences about the more common relative deficits of lead-exposed children, stem from psychometric limitations of the tests that are used to assess different domains. These limitations have potentially serious implications for any effort to use batteries of such tests to draw inferences about relative deficits, a first step in determining whether a behavioral signature can be identified. The key data needed to identify a relative

72

Bellinger

deficit is a lead-associated decrement on the test used to assess one domain but not on the test used to assess another domain (in neuropsychological terms, a “double dissociation”). To support the inference that this does, in fact, reflect a relative deficit, it is necessary, furthermore, to demonstrate that the tests used to assess different domains are equivalent in their discriminating power in the population that generated the study sample. This is rarely, if ever, done. If it is not, a relative deficit is only one of the hypotheses that could explain the findings (42). In this regard, a promising recent development is the use of integrated assessment batteries assembled to be interpretable in terms of current models in cognitive neuroscience. Complex constructs such as “attention” are not monolithic and unidimensional. Mirsky and Duncan’s (43) model distinguishes five dimensions of attention (encode, sustain, focus-execute, shift, and stability), so its administration enables an investigator to determine whether only some of these dimensions are sensitive to lead (39,44). Another theory-driven battery applied in lead studies (45) is the Cambridge Neuropsychological Automated Battery (CANTAB) (46). The CANTAB assesses functions considered to be sensitive to deficits in the prefrontal cortex and hippocampus (visual attention and memory, working memory, planning). It also includes some tasks that are homologues of tasks used in animal studies (e.g., delayed match-to-sample, delayed alternation, intra- and extra-dimensional shift), making it easier to draw explicit links between the human epidemiological and experimental animal literatures on lead (47). A potential drawback of applying such models is that the basic research underlying them is ongoing, and the investigator interested in applying them to lead must commit to one of sometimes several competing models of complex constructs. As Cory-Slechta (48) cautioned, with respect to attention, “.it may be. differences in frameworks defining attention coupled with a lack of operational definitions for the different response classes comprising attention that have limited any substantive advancement of our understanding of the role of lead in attention and how this might contribute to the reported deficits in learning associated with lead.” It has long been apparent that the spectrum of brain-based disorders of childhood associated with increased lead exposure is considerably broader than solely neurocognitive deficits. Since Byers and Lord’s (2) early report, in which the educational failure of many of the lead-poisoned children was attributed to explosive temper and aggressive behaviors, serious behavioral pathologies have been included among the signs and sequelae of lead poisoning. Although many studies of lead neurotoxicity have included assessments of children’s psychiatric and psychosocial status, the tools employed were relatively crude, often limited to parent- or teacher-completed screening questionnaires such as the Child Behavior Checklist (or Teacher Report Form), the Rutter Scales, or the Connors’ Rating Scales (49–55). No cohort study of lead has incorporated both sophisticated exposure assessment and psychiatric assessment tools linked to diagnostic criteria specified in the Diagnostic and Statistical Manual of Mental Disorders. Lead has been hypothesized to increase the risk of a variety of psychiatric disorders. Low-level lead exposure has repeatedly been linked to behavioral dimensions such as distractibility and inattentiveness that are central to diagnostic classification of ADHD, but the evidence linking it to impairments so severe that they take a child across the diagnostic threshold is limited. Other pathologies associated with lead exposure in studies employing a variety of designs (ecologic, case-control, prospective cohort) include anti-social behavior, juvenile delinquency, and criminality (56–59). Despite the greater research attention devoted to externalizing disorders, lead exposure has been linked, as well, to an increased risk of internalizing behavior disorders, including anxiety and depression (52), and even autism (60–62). Whether increased

Lead Neurotoxicity in Children

73

exposure is etiologic or secondary to behavioral abnormalities associated with autism is uncertain, however. Case studies suggest an association between exposure and affective or schizophreniform psychosis (63,64), and a recent case-control study involving schizophrenic adults and matched controls found an increased risk of the disease among individuals whose mothers had higher levels of amino levulinic acid in archived serum samples collected in the second trimester of pregnancy (65). Replication of this intriguing finding using a more specific lead biomarker is needed, however.

FUNCTIONAL FORM OF THE CONCENTRATION–RESPONSE/ CONCENTRATION–EFFECT RELATIONSHIP Consensus opinion regarding the “lowest observed adverse effect level” for lead neurotoxicity in children has, over recent decades, moved inexorably toward lower and lower levels. Ironically, this is, at least in part, a result of the extraordinary success of measures implemented to reduce population exposures to lead. Historically, the blood lead distribution of the control group included in a study reflected the distribution in the contemporaneous source population. In effect, however, this constrained the research hypotheses that could be addressed. For instance, a study conducted in the early 1970s compared the development of children with blood lead levels of 58 to 137 mg/dL to the development of a control group for whom the mean blood lead level was 38 mg/dL (20 to 55 mg/dL) (66). The two groups had developmental problems of comparable severity, leading the investigator to conclude that the problems of the “exposed” group did not reflect lead toxicity, but, like those of the control group, poor home environments. In retrospect, a more likely explanation is that the development of children in both groups had been compromised by lead toxicity, with exposure misclassification producing a false negative result. In the late 1970s, when, as noted earlier, nearly 9 of 10 U.S. preschool children had a blood lead level greater than 10 mg/dL, it was difficult to assemble a control group suitable for evaluating the risk associated with levels of 10 to 20 mg/dL. As a result, the concentration-response/effect relationship estimated in any study was conditional on the lower range of the blood lead distribution of the study sample, and any deficits observed among more highly exposed children allowed inferences only about relative, rather than absolute, toxicity, complicating inter-study comparisons of quantitative results. It has become clear over the past decade that the current CDC screening guideline of 10 mg/dL has no greater biological significance than did the former screening guidelines of 40, 30, or 25 mg/dL. In the Boston prospective study, blood lead level at 2 yr of age was inversely associated with IQ at age 10, although for 90% of the children that level was less than 13 mg/dL (67). In a cross-sectional study of 6-yr-old German children, 95% of whom had a blood lead level less than 9 mg/dL, higher levels were associated with lower verbal intelligence and more false-positive responses on a CPT test (68). In a sample of Taiwanese school children, blood lead level (mean 5.5 mg/dL, SD 1.9) was inversely related to academic achievement (69). In 6 to 16 yr old children in the NHANES III survey, concentration-related deficits in reading and arithmetic scores were found even when analyses were restricted to children with concurrent blood lead levels below 2.5 mg/dL (38). Essentially linear relationships extending down to 1 mg/dL were found between concurrent blood lead level and IQ, naming, visual-motor integration, and attention (39). The most significant challenge to the view that 10 mg/dL can be regarded as “safe” was posed by Canfield et al. (20) They applied semi-parametric models with penalized

74

Bellinger

splines to their data, essentially allowing the data to reveal the functional form that best describes them. It demonstrated that the IQ decline per mg/dL increase in blood lead was greater below 10 mg/dL than it was above 10 mg/dL. The estimated slope of the IQ decline per mg/dL among children for whom the maximum blood lead level measured was less than 10 mg/dL was 0.74 points per mg/dL and 0.13 points per mg/dL among children for whom the maximum blood lead levels exceeded 10 mg/dL. A similar supra-linear relationship was found in a re-analysis of the Boston prospective study (70). Despite the clear, and serious, policy implications, close attention has not always been paid to identifying the functional form that best fits the data. Frequently a log transformation is applied to blood lead level to normalize the distribution. If a linear model is fitted to the log-transformed exposure data, this, in effect, reflects an assumption that the concentration-effect relationship, on the measured scale, is nonlinear, with the slope decreasing in magnitude with increasing concentration, i.e., a greater effect of lead at lower than at higher blood lead levels. Non-linear dose–effect relationships are not uncommon in toxicology (71,72), although many of these are claimed to be examples of hormesis, with low doses of a toxicant being associated with a beneficial effect rather than an adverse effect. Moreover, a biological mechanism for supralinearity for lead has not been identified. Perhaps the predominant mechanism at very low blood lead levels is rapidly saturated, and that a different process that is less rapidly saturated becomes predominant at blood lead levels greater than 10 mg/dL. This ad hoc explanation is more descriptive than explanatory, however, and the specific processes that would produce this result have not yet been identified. An important caveat regarding efforts to specify the functional form of the dose– effect relationship is that the accuracy that can be achieved is constrained by the extent to which the biomarker of lead concentration does, in fact, reflect the concentration at the critical target organ, the brain. To the extent that it does not, aspects of the functional form that best describes a given set of observations might be artifactual.

REVERSIBILITY The absence of a clear operational definition of “reversibility” is a major impediment to drawing inferences about the natural history of any adverse effect associated with an accumulative neurotoxicant such as lead. It cannot be simply that a performance deficit remains detectable after external exposure has ended because it could reflect ongoing toxicity due to lead remaining at the critical target organ or lead deposited at the organ post-exposure as the result of redistribution of lead among body pools. As noted earlier, brain lead level can remain elevated long after blood lead level falls. A true test of reversibility requires that every lead atom has been cleared from the body. This being unattainable, investigators must exploit opportunities that permit only weaker tests of hypotheses about reversibility. These include assessing the persistence of deficits previously associated with early lead biomarkers and evaluating performance changes associated with natural experiments that involve assessing changes following events (including time), such as a change in external exposure or chelation that perturb the equilibrium of lead among different body pools. The data are somewhat inconsistent with regard to the reversibility of developmental adversities associated with low-level prenatal exposures to lead. In some studies, such as the Port Pirie cohort study, such associations were never consistently observed. In others, associations seen in infancy attenuated by preschool age (67,73). In yet others,

Lead Neurotoxicity in Children

75

associations between prenatal lead levels and various late outcomes were reported as late as middle adolescence (57,74,75). Results are more consistent in showing that higher postnatal lead biomarkers are associated with persistent neurocognitive deficits (44,67,74–78). In only a relatively few instances, however, have the longitudinal nature of the data been fully addressed in the analyses (79). Epidemiological studies have been unable to answer the question of whether the persistent deficits are the effects of exposure during some circumscribed critical period, cumulative exposure, or ongoing exposure. In experimental nonhuman primate studies, in which the temporal characteristics of exposure can be manipulated, neurocognitive deficits associated with higher developmental exposures persist for years following cessation of exposure (80). This suggests that ongoing external exposure is not necessary to maintain the deficits, although, as noted previously, it is not possible to exclude a possible role for ongoing endogenous exposures attributable to the redistribution of lead stores among different compartments. Studies of the late outcomes of children treated for lead poisoning support the hypothesis that the deficits are persistent and, possibly, permanent. The Treatment of Lead-Poisoned Children trial (TLC) enrolled children with pre-treatment blood lead levels of 20 to 44 mg/dL (NZ780) and randomly assigned them to receive either a placebo or therapeutic chelation using succimer. At 36-months post-treatment, no significant differences in cognition or behavior were noted between the succimer and placebo groups (81). Current blood lead level was significantly associated with cognitive performance at baseline, 36-months post-treatment, and at 7 yr of age, however, and the regression coefficients were similar to those estimated in observational studies (i.e., w3 point IQ decline per 10 mg/dL increase in blood lead) (82). Preliminary data suggest, however, that succimer does produce beneficial effects on cognition when administered to lead-exposed rats (83) and juvenile (but not adult) primates (84). The reasons for the apparent discrepancy between these findings and those of the TLC trial are not clear.

EFFECT MODIFICATION Plots of children’s values on a lead biomarker and their scores on a neurobehavioral test invariably reveal tremendous scatter, with the overall pattern better described as vaguely ovoid than as a straight line. In other words, the test scores observed at a given value of a lead biomarker are not coincident, but define a distribution, with the Y-value predicted by the model representing the best estimate of the central tendency. The smaller the standard deviation of the distribution, the more precise the prediction (i.e., less residual variance). The amount of residual variance will generally be greater in the presence of: (1) errors (imprecision) in characterizing concentration and outcome, (2) incomplete characterization of outcome variance attributable to factors other than lead, and (3) true biological variability in response that is not captured by the model. Insofar as issues pertaining to biomarker measurements were discussed previously, this section focuses on the second and third possibilities. The R2 values of the “final” regression models reported rarely exceed 50%, and are often considerably lower. Some of this can be attributed to the fact that the goal of model construction is usually to insure that important sources of confounding bias are not overlooked rather than to maximize prediction of the outcome. Much of the methodological discussion pertaining to lead studies has focused on the likelihood that, in a particular study, residual confounding by unmeasured or poorly measured factors biased the estimate of the partial coefficient for lead. One convention that developed in the

76

Bellinger

1980s was the obligatory inclusion of certain potential confounders as a necessary (but not sufficient) condition for study credibility. The most important of these were maternal IQ and quality of the home environment (the age-appropriate version of the Home Observation for Measurement of the Environment). Adjustments are also generally made for a variety of variables identified as potential confounders, but usually on an ad hoc, study-by-study basis insofar as confounding is a characteristic of a specific dataset rather than an inherent characteristic of an association (85). Nevertheless, it is clear that variance in outcome scores that is potentially informative is left to fester in the model error term. What additional factors might be measured and modeled in an effort to reduce the size of this error term? By and large, studies of developmental exposure to lead (and other neurotoxicants) have included in the pool of candidate confounders individual- and family-level factors (e.g., birth weight, birth order, family social class). The ecology of a developing child is considerably broader and more complex than this, however. In some respects, it resembles an onion, with the family representing an inner layer, the neighborhood and community the next outermost layer, and the culture representing the outermost layer. A focus solely on individual and family-level factors reflects an implicit assumption that any influences on child development of more distal aspects of the ecological context are mediated entirely by more proximal factors. Therefore, the argument goes, precise and valid measurements of proximal factors obviate the need to measure distal factors. Recent epidemiologic studies demonstrate the inadequacy of this assumption insofar as distal factors frequently remain significant predictors of health status even after adjustment for individual and family-level factors (86). In studies of lead neurotoxicity, this type of multi-level modeling could integrate into a single analysis data on characteristics that cross the boundaries between layers of influence (41). For example, in a study of the risk among African-American women of delivering a moderately lowbirth weight baby, Rauh and colleagues (87) found that the risk was significantly associated with indicators of community socioeconomic conditions, even after adjustments for individual-level risk factors. The biomarker model developed by the Committee on Biological Markers of the National Research Council (88) incorporates “susceptibility” factors that influence the rate of transition, or likelihood of transition, along the continuum from external dose to clinical disease. Progress has been slow in identifying factors that either exacerbate or mitigate the risk of lead neurotoxicity. Nutritional status is one candidate. Adjusting for the severity of environmental lead contamination, iron-deficient children had higher blood lead levels than iron-replete children (89). One interpretation of these data is that children experiencing the same external dose can experience different internal doses. In another study of iron status, a decline in blood lead level was associated with improved cognitive performance in iron-sufficient but not in iron-deficient children (90). Among the possible explanations for this finding is that iron deficiency contributes to pharmacodynamic variability, increasing the toxicity of a given lead dose. Finally, in a third study, the intellectual deficit associated with an increased blood lead level was greater among undernourished children than well-nourished children (91). Other studies suggest that the cognitive deficits associated with lead exposure are greater, and the prognosis more dire, among children of lower socioeconomic status (92–94). These epidemiological observations are consistent with those in an experimental rodent model (95). On a spatial learning task, the performance of lead-exposed rats reared in an enriched environment exceeded the performance of control rats reared in standard cages, and did not differ significantly from that of control rats reared in the enriched environment. Moreover, environmentally-enriched lead-exposed rats showed significant

Lead Neurotoxicity in Children

77

recovery of N-methyl-D-aspartate subunit 1 mRNA and brain-derived neurotrophic factor in the hippocampus. Genetic polymorphisms that influence lead toxicokinetics and/or toxicodynamics have been identified, although most of the work has focused on adults who were occupationally exposed to lead. Compared to workers with the wild type allele of amino levulinic acid dehydratase, workers with the variant allele had a higher mean blood lead level (96), greater lead-associated renal dysfunction (97,98), and an increased risk of amyotrophic lateral sclerosis (96). The slope of the association between floor dust lead and blood lead is steeper among children with the less common variant of the vitamin D receptor (Fox 1 or B) than among children with the wild-type allele (99). In adults, these same alleles are associated with higher blood lead levels (100) and increased blood pressure (101). Greater lead-associated reductions in renal function have been observed in adults with a variant allele of nitric acid synthetase (eNOS Asp) (97). Adults with variants of the hemochromatosis gene (C282Y and/or H63D) have higher patella lead levels (102). Only one polymorphism has been shown to modify lead neurotoxicity. Lead workers with the apolipoprotein E4 allele showed greater lead-associated decreases in neurobehavioral function than did workers with the E1, E2, or E3 alleles (23). Clearly this work has only begun, and no conclusions can be drawn at present about the likelihood that any of these polymorphisms modify lead neurotoxicity in children.

CONCLUSION The history of research on lead neurotoxicity in children is considerably more extensive than is that of research on any other nervous system toxicant, yet important issues clearly remain unresolved. It does seem to be, as the truism goes, that as we learn more, we are stimulated to ask new questions. But it is important to recognize the changes that have occurred over time in the scope of the questions asked about lead. We are no longer asking, “Does low-level lead exposure affect children’s neurodevelopment?” In fact, the phrase, “low-level” is increasingly being recognized as a term that has little biological content and is interpretable only in a specific historical context. Now the questions of interest focus on the shape of the dose-effect relationship at blood lead levels below 10 mg/dL, the identification of biological and socio-environmental factors that modulate an individual child’s risk, expansion of the neurological endpoints being explored as possible targets of lead toxicity, and so on. Fortunately, public health interventions implemented as a result of the answers investigators provided to the more basic questions have substantially decreased the risk of lead toxicity borne by today’s children. Nevertheless, the job is by no means finished. More than 300,000 children, among whom poor and minority children are over-represented, still have blood lead levels that exceed the CDC action guideline. And we know that 10 mg/dL is not a “bright line” separating toxicity from safety, but only a risk management tool. If the dose-effect relationship is supralinear in the range !10 mg/dL, the foundation of the risk calculus will need to be reconsidered, as will the appropriate public health response.

REFERENCES 1. Lin-Fu JS. Lead poisoning and undue lead exposure in children: history and current status. In: Needleman HL, ed. Low Level Lead Exposure: The Clinical Implications of Current Research. New York: Raven Press, 1980:5–16.

78

Bellinger

2. Byers RK, Lord EE. Late effects of lead poisoning on mental development. Am J Dis Child 1943; 66:471–494. 3. Lin-Fu JS. Modern history of lead poisoning: a century of discovery and rediscovery. In: Needleman HL, ed. Human Lead Exposure. Boca Raton: CRC Press, 1992:23–43. 4. Bradley JE, Powell AE, Niermann W, McGrady JR, Kaplan E. The incidence of abnormal levels of lead in a metropolitan pediatric clinic. J Pediatr 1956; 49:1–6. 5. Lin-Fu JS. Undue absorption of lead among children- a new look at an old problem. N Engl J Med 1972; 286:702–710. 6. Mahaffey KR, Annest JL, Roberts J, Murphy RS. National estimates of blood lead levels: United States, 1976–1980. Association with selected demographic and socioeconomic factors. N Engl J Med 1983; 307:573–579. 7. U.S. Centers for Disease Control. Update: blood lead levels—United States, 1991–1994. Morbid Mortal Wkly Rep 1997; 46:141–146. 8. U.S. Centers for Disease Control. Blood lead levels—United States, 1999–2002. Morbid Mortal Wkly Rep 2005; 54:513–516. 9. Grosse SD, Matte TD, Schwartz J, Jackson RJ. Economic gains resulting from the reduction in children’s exposure to lead in the United States. Environ Health Perspect 2002; 110:563–569. 10. Flegal AR, Smith DR. Lead levels in preindustrial humans. N Engl J Med 1992; 326:1293–1294. 11. Mushak P. New directions in the toxicokinetics of human lead exposure. Neurotoxicology 1993; 14:29–42. 12. Stangle DR, Strawderman MS, Smith D, Kuypers M, Strupp BJ. Reductions in blood lead overestimate reductions in brain lead after repeated succimer regiments in a rodent model of childhood lead exposure. Environ Health Perspect 2004; 112:302–308. 13. Cremin JD, Luck ML, Laughlin NK, Smith DR. Efficacy of succimer chelation for reducing brain lead in a primate model of human lead exposure. Toxicol Appl Pharmacol 1999; 161:283–293. 14. Smith D, Hernandez-Avila M, Tellez-Rojo MM, Mercado A, Hu H. The relationship between lead in plasma and whole blood in women. Environ Health Perspect 2002; 110:263–268. 15. Hernandez-Avila M, Smith D, Meneses F, Sanin LH, Hu H. The influence of bone and blood lead on plasma lead levels in environmentally-exposed adults. Environ Health Perspect 1998; 106:473–477. 16. Bergdahl IA, Vahter M, Counter SA, et al. Lead in plasma and whole blood from leadexposed children. Environ Res 1999; 80:25–33. 17. Chuang HY, Schwartz J, Gonzalez-Cossio T, et al. Interrelations of lead levels in bone, venous blood, and umbilical cord blood with exogenous lead exposure through maternal plasma lead in peripartum women. Environ Health Perspect 2001; 109:527–532. 18. Roberts JR, Reigart JR, Ebeling M, Hulsey TC. Time required for blood lead levels to decline in nonchelated children. J Toxicol Clin Toxicol 2001; 39:153–160. 19. Pocock S, Smith M, Baghurst PA. Environmental lead and children’s intelligence: a systematic review of the epidemiological evidence. BMJ 1994; 309:1189–1197. 20. Canfield RC, Henderson CR, Cory-Slechta DA, Cox C, Jusko TA, Lanphear BP. Intellectual impairment in children with blood lead concentrations below 10 mg per deciliter. N Engl J Med 2003; 348:1517–1526. 21. Hu H, Rabinwitz M, Smith D. Bone lead as a biological marker in epidemiologic studies of chronic toxicity: conceptual paradigms. Environ Health Perspect 1998; 106:1–8. 22. Schwartz BS, Stewart WF, Bolla KI, et al. Past adult lead exposure is associated with longitudinal decline in cognitive function. Neurology 2000; 55:1144–1150. 23. Stewart WF, Schwartz BS, Simon D, Kelsey K, Todd AC. ApoE genotype, past adult lead exposure, and neurobehavioral function. Environ Health Perspect 2002; 110:501–505. 24. McNeill FE, Stokes L, Brito JA, Chettle DR, Kaye WE. 109Cd K X ray fluorescence measurements of tibial lead content in young adults exposed to lead in early childhood. Occup Environ Med 2000; 57:465–471.

Lead Neurotoxicity in Children

79

25. Korrick SA, Hunter DJ, Rotnitsky A, Hu H, Speizer FE. Lead and hypertension in a sample of middle-aged women. Am J Public Health 1999; 89:330–335. 26. Gonzalez-Cossio T, Peterson KE, Sanin L, et al. Decrease in birth weight in relation to maternal bone lead burden. Pediatrics 1997; 100:856–862. 27. Sanin LH, Gonzalez-Cossio T, Romieu I, et al. Effect of maternal lead burden on infant weight and weight gain at one month of age among breastfed infants. Pediatrics 2001; 107:1016–1023. 28. Hernandez-Avila M, Peterson KE, Gonzalez-Cossio T, et al. Effect of maternal bone lead on length and head circumference at birth. Arch Environ Health 2002; 57:482–488. 29. Gomaa A, Hu H, Bellinger D, et al. Maternal bone lead as an independent risk factor for fetal neurotoxicity: a prospective study. Pediatrics 2002; 110:110–118. 30. Wasserman GA, Factor-Litvak P, Liu X, et al. The relationship between blood lead, bone lead, and child intelligence. Child Neuropsychol 2003; 9:22–34. 31. Gulson BL. Tooth analyses of sources and intensity of lead exposure in children. Environ Health Perspect 1996; 104:306–312. 32. Needleman HL, McFarland C, Ness RB, Fineberg SE, Tobin MJ. Bone lead levels in adjudicated delinquents: a case-control study. Neurotoxicol Teratol 2002; 24:711–717. 33. Aufderheide AC, Wittmers LE. Selected aspects of the spatial distribution of lead in bone. Neurotoxicology 1992; 13:809–820. 34. Trope I, Lopez-Villegas D, Lenkinski RE. Magnetic resonance imaging and spectroscopy of regional brain structure in a 10-yr-old boy with elevated blood lead levels. Pediatrics 1998; 101:E7. 35. Trope I, Lopez-Villegas D, Cecil KM, Lenkinski RE. Exposure to lead appears to selectively alter metabolism of cortical gray matter. Pediatrics 2001; 107:1437–1442. 36. Weisskopf MG, Hu H, Mulkern R, et al. Cognitive deficits and magnetic resonance spectroscopy in adult monozygotic twins with lead poisoning. Environ Health Perspect 2004; 112:620–625. 37. Grandjean P, Budtz-Jorgensen E, Keiding N, Weihe P. Underestimation of risk due to exposure misclassification. Int J Occup Med Environ Health 2004; 17:131–136. 38. Lanphear BP, Dietrich KN, Auinger P, Cox C. Cognitive deficits associated with blood lead concentrations !10mg/dL in US children and adolescents. Public Health Rep 2000; 115:521–529. 39. Chiodo LM, Jacobson SW, Jacobson JL. Neurodevelopmental effects of postnatal lead exposure at very low levels. Neurotoxicol Teratol 2004; 26:359–371. 40. Bellinger DC. Interpreting the literature on lead and child development: the neglected role of the “experimental system.” Neurotoxicol Teratol 1995; 17:201–212. 41. Bellinger DC. Future directions for neurobehavioral studies of environmental neurotoxicants. Neurotoxicology 2001; 22:645–656. 42. Strauss ME, Allred LJ. Methodological issues in detecting specific long-term consequences of perinatal drug exposure. Neurobehav Toxicol Teratol 1986; 8:369–373. 43. Mirsky AF, Duncan CC. A nosology of disorders of attention. In: Wasserstein J, Wolf LE, eds. Adult attention deficit disorder: Brain mechanisms and life outcomes. Ann NY Acad Sci 2001;931:17–32 44. Bellinger D, Hu H, Titlebaum L, Needleman HL. Attentional correlates of dentin and bone lead levels in adolescents. Arch Environ Health 1994; 49:98–105. 45. Canfield RL, Gendle MH, Cory-Slechta DA. Impaired neuropsychological functioning in lead-exposed children. Dev Neuropsychol 2004; 26:513–540. 46. Luciana M. Dopamine-opiate modulations of reward-seeking behavior: implications for the functional assessment of prefrontal development. In: Nelson CA, Luciana M, eds. Handbook of Developmental Cognitive Neuroscience. Cambridge, MA: The MIT Press, 2003:647–662. 47. Rice DC. Behavioral effects of lead: commonalities between experimental and epidemiologic data. Environ Health Perspect 1996; 104:337–351. 48. Cory-Slechta DA. Lead-induced impairments in complex cognitive function: Offerings from experimental studies. Child Neuropsychol 2003; 9:54–75.

80

Bellinger

49. Needleman HL, Gunnoe C, Leviton A, et al. Deficits in psychologic and classroom performance of children with elevated dentine lead levels. N Engl J Med 1979; 300:689–695. 50. Yule W, Urbanowicz M-A, Lansdown R, Millar I. Teachers’ ratings of children’s behaviour in relation to blood lead levels. Brit J Dev Psych 1984; 2:295–305. 51. Thomsen GO, Raab G, Hepburn WS, Hunter R, Fulton M, Laxen DP. Blood-lead levels and children’s behaviour: results from the Edinburgh lead study. J Child Psychol Psychiatry 1989; 30:515–528. 52. Bellinger D, Leviton A, Allred E, Rabinowitz M. Pre- and postnatal lead exposure and behavior problems in school-aged children. Environ Res 1994; 66:12–30. 53. Wasserman G, Staghezza-Jaramillo B, Shrout P, Popovac D, Graziano J. The effect of lead exposure on behavior problems in preschool children. Am J Public Health 1998; 88:481–486. 54. Burns J, Baghurst P, Sawyer M, McMichael A, Tong S. Lifetime low-level exposure to environmental lead and children’s emotional and behavioral development at ages 11–13 year: The Port Pirie Cohort Study. Am J Epidemiol 1999; 149:740. 55. Needleman HL, Reiss JA, Tobin MJ, Biesecker GE, Greenhouse JB. Bone lead levels and delinquent behavior. JAMA 1996; 275:363–369. 56. Denno DW. Biology and Violence. New York: Cambridge University Press, 1990. 57. Dietrich KN, Ris MD, Succop PA, Berger OG, Bornschein RL. Early exposure to lead and juvenile delinquency. Neurotoxicol Teratol 2001; 23:511–518. 58. Stretesky PB, Lynch MJ. The relationship between lead exposure and homicide. Arch Pediatr Adolesc Med 2001; 155:579–582. 59. Nevin R. How lead exposure relates to temporal changes in, I.Q., violent crime, and unwed pregnancy. Environ Res 2000; 83:1–22. 60. Accardo P, Whitman B, Caul J, Rolfe U. Autism and plumbism. Clin Pediatr 1988; 27:41–44. 61. Cohen DJ, Johnson WT, Caparulo BK. Pica and elevated blood lead level in autistic and atypical children. Am J Dis Child 1976; 130:47–48. 62. Eppright TD, Sanfacon JA, Horwitz EA. Attention deficit hyperactivity disorder, infantile autism, and elevated blood-lead: a possible relationship. Missouri Med 1996; 93:136–138. 63. Schottenfeld R, Cullen M. Organic affective illness associated with lead intoxication. Am J Psychiatry 1984; 141:1423–1426. 64. McCracken J. Lead intoxication psychosis in an adolescent. J Am Acad Child Psychiatry 1987; 26:274–276. 65. Opler MGA, Brown AS, Graziano JH, et al. Prenatal lead exposure, d-aminolevulinic acid, and schizophrenia. Environ Health Perspect 2004; 112:548–552. 66. Kotok D. Development of children with elevated blood lead levels: a controlled study. J Pediatr 1972; 80:57–61. 67. Bellinger DC, Stiles KM, Needleman HL. Low-level lead exposure, intelligence, and academic achievement: a long-term follow-up study. Pediatrics 1992; 90:855–861. 68. Walkowiak J, Altmann L, Kramer U, et al. Cognitive and sensorimotor functions in 6-year-old children in relation to lead and mercury levels: Adjustment for intelligence and contrast sensitivity in computerized testing. Neurotoxicol Teratol 1998; 20:511–521. 69. Wang CL, Chuang HY, Ho CK, Yang CY, Tsai JL, Wu TN. Relationship between blood lead concentration and learning achievement among primary school children in Taiwan. Environ Res 2002; 89:12–18. 70. Bellinger DC, Needleman HL. Intellectual impairment and blood lead levels. N Engl J Med 2003; 349:500. 71. Calabrese EJ, Baldwin LA. The frequency of U-shaped dose responses in the toxicological literature. Toxicol Sci 2001; 62:330–338. 72. Calabrese EJ, Baldwin LA. The hormetic dose–response model is more common than the threshold model in toxicology. Toxicol Sci 2003; 71:246–250. 73. Bellinger D, Sloman J, Leviton A, Rabinowitz M, Needleman HL, Waternaux C. Low-level lead exposure and children’s cognitive function in the preschool years. Pediatrics 1991; 87:219–227.

Lead Neurotoxicity in Children

81

74. Wasserman GA, Liu X, Popovac D, et al. The Yugoslavia Prospective Lead Study: contributions of prenatal and postnatal lead exposure to early intelligence. Neurotoxicol Teratol 2000; 81:811–818. 75. Ris MD, Dietrich KN, Succop PA, Berger OG, Bornschein RL. Early exposure to lead and neuropsychological outcome in adolescence. J Int Neuropsychol Soc 2004; 10:261–270. 76. Needleman HL, Schell A, Bellinger D, Leviton A, Allred EA. Long-term effects of childhood exposure to lead at low dose: an 11-year follow-up report. N Engl J Med 1990; 322:83–88. 77. Tong S, Bagurst P, McMichael A, Sawyer M, Mudge J. Lifetime exposure to environmental lead and children’s intelligence at 11–13 years: the Port Pirie Cohort Study. BMJ 1996; 312:1569–1575. 78. Fergusson DM, Horwood LJ, Lynskey MT. Early dentine lead levels and educational outcomes at 18 years. J Child Psychol Psychiatry 1997; 38:471–478. 79. Coscia JM, Ris MD, Succop P, Dietrich KN. Cognitive development of lead exposed children from ages 6 to 15: an application of growth curve analysis. Child Neuropsychol 2003; 9:10–21. 80. Rice DC. Behavioral impairment produced by developmental lead exposure: Evidence from primate research. In: Needleman HL, ed. Human Lead Exposure. Boca Raton, FL: CRC Press, 1992:137–152. 81. Rogan WJ, Dietrich KN, Ware JH, et al. The effect of chelation therapy with succimer on neuropsychological development in children exposed to lead. N Engl J Med 2001; 344:1421–1426. 82. Dietrich KN, Ware JH, Salganik M, et al. Effect of chelation therapy on the neuropsychological and behavioral development of lead-exposed children after school entry. Pediatrics 2004; 114:19–26. 83. Stangle D, Strawderman M, Smith D, Strupp BJ. Succimer chelation significantly ameliorates the lasting cognitive and affective dysfunction produced by early lead exposure in a rodent model. Toxicol Sci 2003; 72:128 (abstract). 84. Laughlin NK, Smith DR. Succimer chelation therapy alleviates some lead-induced cognitive impairments in as non-human primate model of childhood lead exposure. Neurotoxicol Teratol 2001; 22:142 (abstract). 85. Bellinger DC. Effect modification in epidemiologic studies of low-level neurotoxicant exposures and health outcomes. Neurotoxicol Teratol 2000; 22:133–140. 86. Diez Roux AV. The study of group-level factors in epidemiology: rethinking variables, study designs, and analytical approaches. Epidemiol Rev 2004; 26:104–111. 87. Rauh VA, Andrews HF, Garfinkel RS. The contribution of maternal age to racial disparities in birthweight: a multilevel perspective. Am J Public Health 2001; 91:1815–1824. 88. Committee on Biological Markers of the National Research Council. Biological markers in environmental health. Environ Health Perspect 1987; 74:3–9. 89. Bradman A, Eskenazi B, Sutton P, Athanasoulis M, Goldman L. Iron deficiency associated with higher blood lead in children living in contaminated environments. Environ Health Perspect 2001; 109:1079–1084. 90. Ruff HA, Markowitz ME, Bijur PE, Rosen JF. Relationships among blood lead levels, iron deficiency, and cognitive development in two-year-old children. Environ Health Perspect 1996; 104:180–185. 91. Gardner JM, Walker SP, Chang SM, Vutchkov M, Lalor GC. Undernutrition and elevated blood lead levels: effects on psychomotor development among Jamaican children. Public Health Nutr 1998; 1:177–179. 92. Winneke G, Kramer U. Neuropsychological effects of lead in children: interactions with social background variables. Neuropsychobiology 1984; 11:195–204. 93. Bellinger DC, Leviton A, Waternaux C, Needleman HL, Rabinowitz M. Low-level lead exposure, social class, and infant development. Neurotoxicol Teratol 1988; 10:497–503. 94. Bellinger D, Leviton A, Sloman J. Antecedents and correlates of improved cognitive performance in children exposed in utero to low levels of lead. Environ Health Perspect 1990; 89:5–11.

82

Bellinger

95. Guilarte TR, Toscano CD, McGlothan JL, Weaver SA. Environmental enrichment reverses cognitive and molecular deficits induced by developmental lead exposure. Ann Neurol 2003; 53:50–56. 96. Kamel F, Umbach DM, Lehman TA. Amyotropic lateral sclerosis, lead, and genetic susceptibility: Polymorphisms in the d-aminolevulinic acid dehydratase and vitamin D receptor genes. Environ Health Perspect 2003; 111:1335–1339. 97. Weaver VM, Schwartz BS, Ahn K-D, et al. Associations of renal function with polymorphisms in the d-aminolevulinic acid dehydratase, vitamin D receptor, and nitric oxide synthase genes in Korean lead workers. Environ Health Perspect 2003; 111:1613–1619. 98. Wu M-T, Kelsey K, Schwartz J, Sparrow D, Weiss S, Hu H. A d-aminolevulinic acid dehydratase (ALAD) polymorphism may modify the relationship of low-level lead exposure to uricemia and renal function: the Normative Aging Study. Environ Health Perspect 2003; 111:335–340. 99. Haynes EN, Kalkwarf HJ, Hornung R, Wenstrup R, Dietrich K, Lanphear BP. Vitamin D receptor fok1 polymorphism and blood lead concentration in children. Environ Health Perspect 2003; 111:1665–1669. 100. Schwartz B. Associations of blood lead, dimercaptosuccinic acid-chelatable lead, and tibia lead with polymorphisms in the vitamin D receptor and d-aminolevulinic acid dehydratase genes. Environ Health Perspect 2000; 108:949–954. 101. Lee BK, Lee GS, Stewart WF, et al. Associations of blood pressure and hypertension with lead dose measures and polymorphisms in the vitamin D receptor and delta-aminolevulinic acid dehydratase genes. Environ Health Perspect 2001; 109:383–389. 102. Wright RO, Silverman EK, Schwartz J, et al. Association between hemochromatosis genotype and lead exposure among elderly men: the Normative Aging Study. Environ Health Perspect 2004; 112:746–750.

5 Effects of Organic Solvents on Reproductive Outcome and Offspring Development Christine Till Toronto Rehabilitation Institute, Toronto, Ontario, Canada

Gideon Koren Division of Clinical Pharmacology and Toxicology, The Hospital for Sick Children, Toronto, Ontario, Canada

Joanne Rovet Department of Pediatrics, The Hospital for Sick Children, Toronto, Ontario, Canada

Many women continue to work throughout pregnancy, making the link between women’s workplace exposure to organic solvents and reproductive hazards an important public health issue. The effect of this exposure on human offspring development is important to the discussion of developmental neurotoxicology since exposure to organic solvents is widespread and, over the last three decades, increasing evidence has accrued to suggest that many solvents can be potent teratogens. However, current biological understanding of the effects of organic solvents on reproductive outcome remain far from definitive and confined to a small number of organic solvents. This chapter reviews the evidence pertaining to the effects of organic solvent exposure on reproductive outcomes and offspring development with a focus on occupational exposures to industrial solvents. Evidence from animal experiments is included in order to establish whether organic solvents interfere with reproduction. Epidemiological evidence and clinical case studies are reviewed for increased risk of spontaneous abortion, birth defects, neurobehavioral deficits, and visual defects in offspring of women exposed to organic solvents during pregnancy. Methodological issues that need considering when interpreting human and animal studies are also discussed.

BASIC PRINCIPLES Classes of Organic Solvents Organic solvents comprise a large and structurally diverse group of industrially important chemical compounds used in industry for the purpose of extracting, dissolving, or 83

84

Till et al.

suspending non-water soluble materials. There are thousands of industrial solvents, and these can be grouped according to their chemical composition. The major classes of organic solvents include aromatic and aliphatic hydrocarbons, halogenated compounds, aldehydes, ketones, amines, esters, ethers, glycols, and alcohols (Table 1). Complex solvents such as kerosene, rubber solvent, varnish, painters’ naphtha, and mineral spirits contain a mixture of components determined by the petroleum sources, processing methods, and mixing ratios. This section reviews the uses and some basic principles of organic solvents focusing on the properties related to reproductive outcome.

Sources of Exposure to Organic Solvents Organic solvents represent one of the most common kinds of industrial chemicals. They are used in a variety of industrial applications including the manufacture of paints, glues, dyes, plastics, and metals; clothing and textile industries; and in the preparation of processed foods and pharmaceuticals. Many female-dominated occupations such as laboratory technicians, hairdressers, photo developers, and dry cleaning workers involve the use of organic solvents. Exposure to organic solvents can also occur through household product use in spray adhesives, paint removers, floor and tile cleaners, and varnishes. Pregnant women are often exposed to organic solvents because they are commonly engaged in home renovations for preparation of additional family members. Exposure can also occur more broadly from environmental contaminants or through substance abuse. Organic solvents have been reported as contaminants in drinking water (1) and as air pollution (2). One route of organic solvent exposure that is on the rise today is deliberate exposure in pursuit of an intoxicating solvent “high” (3,4). According to a national survey of high school students in the United States, approximately 17% of adolescents have sniffed inhalants at least once in their lives (5).

Properties of Organic Solvents Most organic solvents are liquids that boil in the range of 758C to 2208C and share common physical characteristics including high volatility, which refers to its tendency to evaporate into a gas or vapor, and solvency, which refers to its ability to pass through intact skin. As a general rule, the higher the volatility of a solvent, the more it will be present in the breathing zone, and hence, the higher the exposure. The high volatility of most solvents under conditions of standard use and the efficient transfer of vapors into the bloodstream indicates that inhalation is the major route of exposure. With respiratory exposure, vapors enter the alveoli of the lungs and diffuse readily across a large surface area of respiratory membranes from high concentration in the lung’s inspired air to low concentrations in the blood and tissues. Dermal absorption represents a second major route of exposure for many solvents as skin contact is very common due to the casual nature of solvent use. Solvents that are both lipid and water soluble pass through intact skin easily and lead to a defatting of skin or skin irritation. Many solvents are very lipid soluble and can therefore cross the blood-brain barrier and accumulate in lipid-rich areas of the body, such as depot fats and myelin or brain white matter. Because of their lipid solubility, organic solvents enter the bloodstream with ease, although the rate of membrane transport also depends upon the condition of the individual. For example, between 40% to 80% of the inhaled dose is absorbed when the individual is at rest, and this amount increases with exercise and pregnancy because blood flow to the lung and alveolar ventilation are elevated. Therefore, the unique pharmacokinetics

Effects of Organic Solvents on Reproductive Outcome Table 1

Common Classes of Organic Solvents and Exemplars

Structural class

Exemplars

Aliphatic Methane, ethane, propane, butane, pentane, hydrocarbons isobutane, petroleum ether, hexane, naphtha, gasoline, kerosene Aromatic Benzene, methylbenzene (toluene), hydrocarbons dimethylbenzene (xylene), trimethylbenzenes, tetramethylbenzene (durene), ethylbenzene, propylbenzene, isopropylbenzene (cumene), butylbenzene, vinylbenzene (styrene) Chlorinated Monochloromethane (methyl chloride), hydrocarbons Dichloromethane (methylene chloride), Monochloroethane (chloroform), 1,1,1trichloroethane (methyl chloroform), 1,1,2-trichloroethane (TCE; vinyl trichloride), vinyl chloride, trichloroethane (trichloroethylene), tetrachloromethane (carbon tetrachloride), 1,1,2,2tetrachloroethane (acetylene tetrachloride), tetrachloroethylene (perchloroethyelene), dichlorodifluoromethane (Freon 12) Esters Methyl formate, ethyl acetate, isopropyl acetate, methyl acetate, isoamyl acetate Ketones Dimethyl ketone (acetone), methyl ethyl ketone (MEK; butanone), methyl n-propyl ketone (MPK; 2-pentanone), methyl isopropyl ketone (MIK; 3-methyl-2-butanone), methyl-n-butyl ketone (MBK; 2-hexanone), methyl n-amyl ketone (MAK; 2-heptanone), ethyl n-butyl ketone (3-heptanone) Aldehydes Acetaldehyde, formaldehyde, (formalin), glutaraldehyde, fluoroacetaldehyde

Ethers

Glycols

Alcohols

85

Dimethyl ether, diethyl ether (ether), diisopropyl ether, dibutyl ether, divinyl ether, methyl propyl ether, tetrahydrofuran, dioxane, isopropyl ether Ethylene glycol, propylene glycol, triethylene glycol, butyl cellusolve (2-butoxyethanol), cellusolve (2-ethoxyethanol), methyl cellusolve (2-methoxyethanol) Methanol, ethanol, phenol, isopropanol, Propanol, butanol

Principle source/use Coatings and insecticides, dry cleaning fluid, spot, remover, excellent degreasers and solvents for paints and epoxies Solvents for acrylic, paints, adhesives, cements, glues, veneers, plastics; used in printing

Used as degreasers, dry cleaning agents, paint, strippers; found in coatings, resins, tars, aerosol propellant, solvents for adhesives, waxes, glues, oils, etc.

Fumigant insecticides Nail polish remover (acetone), solvents for dyes, resins, adhesives, lacquers, oils, varnish; used in manufacture of plastics, explosives, cosmetics, medicines

Used in laboratories, plastic and resin industries, in synthesis of sealants, cosmetics, disinfectants, wood preservatives Gasoline engine primer; ingredients in dyes, resins, waxes, paints, gums, lacquers, fuels and cellulose nitrate Used as lubricants; ingredients in cosmetics and coatings; to lower freezing point of liquids

Used for cleaning, paint removal (Continued)

86

Till et al.

Table 1 Common Classes of Organic Solvents and Exemplars (Continued) Structural class Alkyl nitrites Others

Exemplars Propyl, isopropyl nitrite, butyl, isobutyl nitrite, amyl, isoamyl nitrite Freon, turpentine, dimethylformamide (DMF), carbon disulfide, pyridine, amides, amines

Principle source/use Jet fuel, room deodorizer

inherent to the pregnant woman raise concern about potential adverse effects of the exposure. Because solvents constitute a heterogeneous group of chemicals, there are many potential metabolic breakdown pathways. However, in most instances there is involvement of the P450 enzyme system and the glutathione pathways. These pathways catalyze oxidative and conjugation reactions to form water-soluble substances, which can be excreted in urine. For most solvents, metabolism occurs primarily in liver. However, it is generally recognized that other organs such as kidney and lung can also exhibit a capacity for biotransformation.

Monitoring Organic Solvents in the Workplace The estimation of the dose of organic solvents usually involves measuring either the solvent or its metabolite in urine or blood. Biological monitoring is useful in determining solvent exposure occurring through dermal absorption, whereas hygiene monitoring in the workplace (i.e., measurement of airborne concentrations and/or dermal contact) only approximates inhalation exposure and is dependent upon the employer’s ability or willingness to collect the exposure data. Furthermore, the efficacy of biological monitoring depends on the properties of the solvent. For example, solvents with half-lives in the range of several hours (6) are excreted rapidly, and because levels of metabolite may fluctuate from hour to hour, they may not be detected reliably by biological monitoring. Therefore, the timing of sample collection relative to the termination of exposure is critical for reliable estimates of uptake. Hygiene monitoring in the workplace also only approximates inhalation exposure because the concentration of solvents is dependent on several solvent- and human-related factors. Solvent-related factors include: (1) the level and duration of exposure, (2) specific physiochemical features of each solvent, (3) the use of protective equipment, and (4) synergistic effects of simultaneous exposure to solvent mixtures (i.e., reduced rate of metabolism of one solvent in the presence of other chemicals) (6). With respect to humanrelated factors, physiological factors such as body build, ethnic and genetic differences, percentage body fat, and blood and tissue solubility affect the pharmacokinetic profile of organic solvents (7). Toxicity of organic solvents also varies according to levels of physical exercise (associated with increased pulmonary capillary blood flow), diurnal metabolic cycles, and alcohol use (8). Regarding the latter, Baker et al. (6) reported that ethanol ingestion decreases the metabolic clearance rate of xylene by about one half-life due to the metabolic interaction between xylene and alcohol, which are both degraded by the same enzyme (viz., aldehyde dehydrogenase). As a result of the ethanol-induced metabolic inhibition, concurrent alcohol and organic solvent exposure slows clearance of the solvent, and so may be expected to prolong internal exposure to the neurotoxin.

Effects of Organic Solvents on Reproductive Outcome

87

Thus, because hygeine monitoring provides only an estimate of exposure for an individual and the fetus, it must be combined with adequate health data, as mentioned above, to determine risk. Transfer of Organic Solvents to the Fetus During pregnancy, lipid-soluble and low-molecular-weight solvents can easily traverse the placental barrier into the amniotic fluid where they are ingested and readily absorbed by the conceptus at all stages of gestation (9–11). Ghantous and Danielsson (10) used autoradiography and liquid scintillation methods to study the placental distribution of radiolabeled toluene, xylene, and benzene and their metabolites in pregnant mice. The distributions of these solvents were observed at regular intervals from 0 to 24 hr following a 10-min inhalation period. All solvents reached a high concentration in dams’ lipid-rich tissues (e.g., brain and fat) and well perfused organs (e.g., liver and kidney) immediately after inhalation and in every tissue, except fat, were eliminated in one hour. Peak levels were reached between 30 min and one hour after inhalation and then were all rapidly eliminated except for the benzene metabolite, which was retained in maternal liver and kidney. For all solvents, volatile radioactivitiy was observed immediately and up to one hour after inhalation in placenta and fetuses regardless of stage of gestation. Although concentrations of toluene, xylene, and benzene were much lower in fetal tissues than in maternal tissues, those in most fetal tissues were evenly distributed in early gestation. The one exception was the fetal liver, which reached high levels in late gestation. Thus, these aromatic hydrocarbons may be transported to the fetus paraplacentally through fetal membranes and amniotic fluid, with the possibility of fetal swallowing or dermal absorption.

ORGANIC SOLVENTS IN THE CENTRAL NERVOUS SYSTEM Organic solvents can act as central nervous system (CNS) depressants and cause general anesthetic effects by inhibiting brain and spinal cord activity. In the extreme, high level exposure to organic solvents can render the individual unconscious and even result in death. With repeated low-level chronic exposure (or high-level acute exposure), these compounds can accumulate in the lipid membranes of nerve cells and, in some cases, disrupt normal excitability of the nerve tissues and adversely affect normal nerve impulse conduction (12). Because these compounds tend to accumulate in lipid-rich areas of the body, such as myelin or brain white matter, there is a high potential for neurotoxicity. These effects would also be expected to predominate in the developing fetal CNS when myelin is forming. As a rule, the degree of CNS depressant activity increases with the carbon chain length. This increased toxicity is most evident when larger functional groups are added to small organic compounds, since the increase in molecular size disproportionately decreases the water solubility and increases the lipophilicity. However, this generalization is relevant only for chemicals with a five- or six-carbon chain length (e.g., pentane, hexane) while molecular size beyond this point results in a decrease in vapor pressure, thereby changing the pattern of inhalation dramatically. Another property that increases the degree of CNS depressant activity is the number of unsaturated bonds (organic structures where hydrogens have been deleted, forming one or more double or triple bonds). Typically, unsaturated chemicals (e.g., hexene) are more potent CNS depressants than their saturated (single-bond) counterparts (e.g., hexane).

88

Till et al.

Similarly, CNS depressant properties of organic compounds are enhanced by increasing the degree of halogenation [e.g., chlorine (Cl), bromine (Br)] and, to a lesser extent, adding alcoholic (–OH) functional groups. For example, while methane and ethane have no significant anaesthetic properties, their corresponding alcohol analogs (methanol and ethanol) are potent CNS depressants. Similarly, while methylene chloride (i.e., CH2Cl2) has appreciable anaesthetic properties, chloroform (i.e., CHCl3) and carbon tetrachloride (i.e., CHCl4) are even more potent compared to methylene chloride. Table 2 shows the relative CNS depressant and irritant potency of organic compounds. The approximate rankings shown in Table 2 rely on basic comparisons among the unsubstituted chemical analogs and become less applicable in broader comparisons among the larger, more complex and multisubstituted compounds.

FETAL VULNERABILITY TO ORGANIC SOLVENTS The fetus is uniquely vulnerable to toxin exposures for a number of reasons: First, because of the incomplete development of the blood-brain barrier, the fetus has greater exposure to organic solvents than adults. Second, accumulation of organic solvents may be further exacerbated because of the relatively high lipid content of the developing brain compared to the rest of the body. Third, the mechanisms for solvent metabolism, in particular P-450, are functioning only marginally in the fetus. At birth, human and animal offspring have only 20 to 40% of the adult concentration of cytochrome P450 and do not reach adult levels of P450 enzymes until one year of age (13). This reduced metabolic capacity of the young favors the increase of solvent absorption compared to adults and makes them more susceptible to neurotoxic insult. Also, the fetus is more susceptible to neurotoxic insult than adults because of its rapid nervous system growth and development. For example, during organogenesis, systems such as the visual system develop and establish essential neural connections. If the timing of exposure falls within critical periods of development, vital connections between nerve cells may fail to form and, as a consequence, there is high risk that the developing nervous system will be permanently damaged. Despite these characteristics that make the fetus more vulnerable to toxic substances than adults, most occupational threshold values are determined based on the adult worker, not the fetus. This therefore calls into question the adequacy of workplace standards for pregnant women. Table 2

Relative CNS Depressant and Irritant Potency of Selected Organic Solvent Classes

Potency High

Low Source: From Ref. 12.

CNS depressant potential Halogen-substituted compounds Ethers Esters Organic acids Alcohols Alkenes Alkanes

Membrane and tissue irritant potential Amines Organic acids AldehydesZketones Alcohols Alkanes

Effects of Organic Solvents on Reproductive Outcome

89

REPRODUCTIVE AND DEVELOPMENTAL EFFECTS IN ANIMALS FOLLOWING PRENATAL EXPOSURE TO ORGANIC SOLVENTS Animal studies have reported that a variety of organic solvents cross the placenta and result in increased embryotoxicity as reflected by lower conception rates, smaller litter sizes, and increased rates of miscarriage and structural abnormalities. Malformations following prenatal solvent exposure in animals include hydrocephaly, exencephaly, skeletal defects, and cardiovascular abnormalities (9,14–17). Also described are delayed maturation of the cerebral cortex (9) and changes to the lipid class composition of the cerebral cortex (18). Studies of behavioral toxicity in the offspring of animals following maternal exposure to organic solvents have revealed significant dose-dependent effects such as delayed reflex and motor development (16), altered rates of behavioral habituation to novel environments (19), deficits in spatial learning and memory (20), and increases in spontaneous activity (16). In one study by Nelson, Taylor, Selzer, and Horning (21), offspring of SpragueDawley rats, which were exposed to 900 parts per million (ppm) perchloroethylene (PCE), showed alteration in growth as well as neurobehavioral changes, including increased activity in open field tests (21). Neurochemical analyses of the animal brains revealed significant reductions in levels of acetylcholine and dopamine, particularly among pups exposed on gestation days 7 to 13. However, as with the human epidemiological studies, findings in animal studies are often conflicting and many of the positive associations between solvent exposure and teratogenicity depend on the specific solvent, dose, route of administration, and the indirect effects of maternal toxicity. The most convincing evidence for the developmental toxicity of solvents per se comes from studies that controlled for maternal toxicity and used exposure conditions similar to what might be encountered in the workplace. One such study reported rib abnormalities in mice exposed to 400 ppm of toluene for 7 hr per day on gestational days 7 to 16, but no effect with lower exposures of 200 ppm (22). In another study that showed no differences in maternal body weight, rabbits exposed to medium and high levels of diethylene glycol dimethyl ether showed prenatal mortality and malformations reflecting rib fusion and clubbing of limbs (14). While research on animals has convincingly demonstrated that a variety of solvents readily cross the placenta and have the potential to disrupt neurodevelopment, even in the absence of overt maternal toxicity, it is still difficult to make generalizations from animal to human data for several reasons. First, most animal studies use higher dose levels than are accepted for humans, and these are provided via a variety of administration routes. While solvent absorption in humans occurs primarily through inhalation or dermal absorption, animals are often administered solvents via gavage or intravenous injection. Second, as a result of developmental differences between animals and humans at the time of birth, the teratogenic impact of certain chemicals may be less in those animals whose critical period of brain development is postnatal. A third limitation is that specific mixtures of solvents, which occur commonly in many occupational settings, are usually not tested in animals. Fourth, there is debate about how to extrapolate the findings to humans because teratogenic activity varies among species. For example, mice metabolize PCE more rapidly than rats, whereas humans metabolize PCE less rapidly than rats. Also, the brain-body weight fraction and the cardiac output to the brain are greater in humans than in rodents. Therefore, studies only in animals are not adequate for assessing potential reproductive and developmental risks associated with human occupational exposure to organic solvents during pregnancy.

90

Till et al.

REPRODUCTIVE AND DEVELOPMENTAL EFFECTS FOLLOWING OCCUPATIONAL EXPOSURE TO ORGANIC SOLVENTS The current scientific literature on reproductive risks associated with solvent exposure focuses mainly on the risks to the fetus during pregnancy. This focus makes it seem as if women are more susceptible to risk than men. However, this perspective is a misconception because some substances may be more harmful for men (23–25), while others may be more harmful for women. Thus, the present focus on reproductive risks of women’s workplace exposures does not imply a distinction between reproductive disorders to men or women. Risk of Infertility Fertility can be measured by a number of indices, which include: fecundability ratios (FR), or the probability of conceiving in a given menstrual cycle; and time to pregnancy (TTP) adequacy of menstrual and ovarian function. Studies of fertility indicate disruption of the many biological mechanisms responsible for conception (or the organs responsible for them) increases the time to achieve a pregnancy. This is particularly important for lowdose exposures that do not necessarily result in frank infertility. One of the earliest epidemiological studies carried out in Denmark sought to examine the link between occupational solvent exposures and reduced fecundability (26). Occupational exposures and TTP data were collected for more than 4000 pregnant couples. For women, the exposures that were significantly associated with infertility and/ or delayed conception included anesthetic gases, dry cleaning chemicals, metals (such as those for welding, lead, cadmium, mercury), plastic manufacturing, heat, and noise. Several recent studies have examined fecundity and exposure. For example, PlengeBonig and Karmaus reported a significantly reduced fecundability ratio [FRZ0.47, 95% confidence interval (CI) 0.29 to 0.77] among women, but not men, employed in the printing industry and exposed to toluene (27). In a retrospective study of TTP in female wood workers exposed to different exposure levels of formaldehyde, Taskinen et al. assessed the fecundability density ratio (FDR) (28), which represents the average incidence densities of pregnancies for exposed women compared to pregnancies of employed unexposed women. High exposure to formaldehyde was found to be significantly associated with delayed conception (FDRZ0.64, 95% CI 0.43 to 0.92), while FDR was still lower if the women did not use gloves (FDRZ0.51, 95% CI 0.28 to 0.92). No association was found between exposure to phenols, dusts, or organic solvents and prolonged TTP. In a study of women exposed to organic solvents, Sallmen et al. observed reduced fertility (29) reflecting decreased FRs among women exposed to organic solvents in shoe factories (FRZ0.28, 95% CI 0.11 to 0.71), dry cleaning shops (FRZ0.44, 95% CI 0.28 to 0.86), and the metal industry (FRZ0.58, 95% CI 0.34 to 0.98). Several studies have focused on the effects of occupational exposures to organic solvents on biological processes that may affect fertility, such as menstrual cycle length and variability. Women with high exposure to toluene (30) and dry cleaning workers exposed to perchloroethylene (31) were reported to have increased dysmenorrhea compared to controls. In contrast, a study of more than 1500 plastics industry workers found no association nor a dose-response relationship between styrene exposure and menstrual disorders. Rather, menstrual disorders were linked to smoking, the woman’s age, chronic disease, and nulliparity (32). Further studies with detailed exposure assessments and controlling for other work-related factors are needed to determine whether dysmenorrhea is related to exposure to other specific solvents.

Effects of Organic Solvents on Reproductive Outcome

91

Risk of Spontaneous Abortion Spontaneous abortion (SAB) has been extensively studied in relation to organic solvent exposure using cross-sectional, case-control, and historical cohort designs. Although this research has reported a significantly increased risk of SAB in certain occupational settings, there appears to be discrepancies among occupations (Table 3). For example, increased risk of SAB was associated with solvent exposure among women employed in leather and shoe manufacturing (33), the graphics industry (34), laundry and dry cleaning (35), laboratories (36,37), painting (38), and pharmaceutical factories (37), whereas other studies reported that laboratory workers (33,39) and workers in the microelectronics industry (40) did not have increased SAB rates. These discrepancies may reflect the different levels of exposure between occupational groups, as well as differences with respect to types of solvents and solvent mixtures used. With regard to type of organic solvent, an increased risk of SAB has been associated with maternal exposure to toluene, styrene, trichloroethylene, methylene chloride, and aliphatic hydrocarbons. Significant excesses of SAB have also been reported among women exposed to (1) styrene and polystyrene during the production of plastics; (2) tetrachloroethylene in dry cleaning work; and (3) toluene in shoe and audio speaker manufacturing (40). Among laboratory workers, whose exposure usually involves more than one agent, the risk of SAB was most consistent among women exposed to aliphatic hydrocarbons (34,36). Similarly, a study of women employed in the pharmaceutical industry found an increased risk of SAB following exposure to methylene chloride (41). In the latter study, Taskinen et al. used logistic regression to show that risk increased in relation to the number of solvents used with women who used four or more solvents having an increase in the SAB rate (statistically insignificant) compared to those exposed to one to three solvents (41). Taken together, these studies show a fairly consistent association between maternal solvent exposure and evidence of SAB. However, it is difficult to interpret the consistency in results across studies because it is unclear how well the studies were able to estimate retrospectively the actual exposure during pregnancy.

Risk of Congenital Malformations Several case-control studies have indicated that offspring of women exposed to organic solvents during pregnancy are more likely to have birth defects. Although analysis of the specific types of defects did not reveal any particular trend, the wide range of anomalies associated with solvent exposure is not surprising given the diversity in chemicals that are used. Holmberg and Nurminen (42) in a case-control study examined the association between occupational factors during pregnancy and congenital defects of the CNS, using data from the Finnish Registry of Congenital Malformations. Personal interviews, which were conducted on 120 cases and their referents revealed that women exposed to solvents during the first trimester of pregnancy were 4 times more likely to have a child with a CNS anomaly than women who were not exposed during pregnancy. Other studies have reported increased cardiovascular abnormalities (43), limb reduction defects (24), renalurinary or gastrointestinal defects (44), and oral cleft malformations (45–47) in the progeny of solvent-exposed women. Hemminki and colleagues (48) conducted a large case-control study in which they prospectively investigated the association between maternal occupation and children

92

Till et al.

Table 3 Studies on the Relationship Between Maternal Occupational Exposure to Organic Solvents and Spontaneous Abortion Study population Laboratory work

No. of exposed pregnancies; cases/controls

Relative risk

95% C.I. or p value

Solvent exposure

576

1.3

0.9–9

Multiple

299

0.7

0.3–1.14

n/a

38/57

1.0

0.6–1.7

Multiple

23/21

2.3

1.1–4.3

44/130

2.3a

p!.06

46/357

1.7a

1.0–2.9

5/38

2.9

1.0–8.8

3/27

1.10a

0.33–3.73

839

3.09a

p!.01

Graphics

7/3

5.2

1.3–20.8

Shoe factory

5/2

9.3

1.0–84.7

Dry cleaning

247/680

3.6a

p!.05

TCE

2.7

0.7–11.2

TCE

low exposure high exposure Pharmaceutical work Factory workers (rubber, plastics, & machines) Painters

Microelectronics industry Leather manufacturing

4/5

a

References Axelsson et al. 1984 (39) McDonald et al. 1988 (33) Taskinen et al. 1994 (36)

Methylene chloride Solvents

Taskinen et al. 1986 (41) Heidam, 1984 (38)

Solvents, phthalic acid Aliphatic hydrocarbons n/a

Heidam, 1984 (38)

Aliphatic hydrocarbons Toluene

Lipscomb et al. 1991 (40) McDonald et al. 1988 (33) Lindbohm et al. 1990 (34) Lindbohm et al. 1990 (34) Kyyronen et al. 1989 (35) Lindbohm et al. 1990 (34)

Value indicates adjusted odds ratio.

born with major malformations. All cases and their referents from 1967 to 1977 were studied for a total of 3300 pairs. Risk factors of malformations were considered as possible confounding factors, and were controlled using multivariate analyses. Results showed an increased risk of CNS, oral cleft, and musculoskeletal malformations in the

Effects of Organic Solvents on Reproductive Outcome

93

offspring of women in industrial trades, many of whom used solvents. However, no specific chemical was identified as the causative agent because the majority of workers were exposed to more than one chemical, or chemicals were not specified. This drawback makes it difficult to draw any conclusions about the risk of congenital malformations of specific occupations or solvents. In a case-control study from Montreal (44), certain specific chemicals were systematically studied as risk factors of congenital defects. This study was restricted to women who had worked at least 30 hr during the first trimester of pregnancy. Cases included 301 women who gave birth to a child with major congenital defects or who had their pregnancy terminated because of a major defect detected early in pregnancy while controls were women who gave birth to a normal child in the same hospital. Each case was matched to a control on gravidity, education, maternal age and the date of delivery. The exposure assessment was based on interviews immediately after delivery. Results of a matched pair analysis revealed a significantly higher overall frequency of chemical exposure in cases than referents (63,47). Analysis of the chemical categories revealed a significant association between aromatic solvents and malformations (ORZ2.3, pZ0.04), which was most evident in the urinary tract group (9:0). This difference was mainly restricted to the group exposed to toluene. Among laboratory workers, some studies have shown increased frequencies of congenital malformations (34,43,49) while others have found no such association (39,50). Inconsistencies also exist among studies of women working in the leather industry. For example, Garcia and Fletcher (47) reported an increased risk of oral clefts among offspring of leather workers, whereas McDonald et al. (33) reported an increased risk of stillbirth, but not malformations, among these workers. Differences in the composition of the study populations and the intensity of exposure may partly explain the inconsistent results. A recent study conducted in Toronto by Khattak and colleagues (51) investigated the risk of major fetal malformations following maternal occupational exposure to organic solvents during pregnancy. The offspring of 125 women who had been exposed to various types of organic solvents for at least the first trimester of pregnancy and 125 offspring of women not exposed were systematically studied. Major birth defects were found in 13 progeny in the exposed group but only 1 child in the control group [relative risk (RR)Z13.0; 95% CI: 1.8 to 99.5]. Furthermore, malformations were more likely to occur among women who reported symptoms associated with their exposure compared with those who were asymptomatic. Although this study avoided many of the methodological shortcomings of retrospective studies, its small sample size and heterogeneity of exposure types has made it difficult to draw conclusions about the reproductive hazards of specific solvents or occupations. McMartin et al. (52) used meta-analysis to summarize the risk for spontaneous abortion and major malformations from maternal occupational exposure to solvents during pregnancy. This method is advantageous because it is more objective than a general literature review and it also involves pooling research results from various studies, thereby increasing the statistical power of the analysis. Results of the McMartin study found a trend towards an increased risk of spontaneous abortions (ORZ1.25; 95% CI: 0.99–1.58) and a significantly increased risk of major malformations (OR was 1.64; 95% CI: 1.16–2.30) following maternal exposure to solvents. Although the risk for spontaneous abortion was small, these authors concluded that the addition of one study of similar effect size would have rendered this trend statistically significant. Hence, the need for further research in this area is warranted.

94

Till et al.

Neurodevelopmental Effects To date, little is known about the long-term effects of organic solvent exposure during pregnancy on offspring neurodevelopment. This represents a major gap in knowledge because the disabling effects of a teratogenic substance often do not emerge at birth but rather, when the potentially vulnerable cognitive or behavioral system has matured or compensatory mechanisms are no longer adequate (53). The lack of studies examining neurodevelopmental effects is of concern because lower doses of organic solvents, such as those commonly found in the workplace, may induce more subtle manifestations of toxicity in the absence of physical malformations in offspring and direct effects on the mother. To illustrate this point further, one only needs to compare the effects of fetal alcohol syndrome (FAS) and fetal alcohol effect (FAE). Since both groups show a consistent pattern of neuropsychological deficits, but only the FAS group has facial dysmorphology (54), these data therefore suggest that degree of deficit is independent of the presence of physical features associated with prenatal alcohol exposure. One prospective study by Eskenazi, Gaylord, Bracken and Brown (55) examined the neurobehavioral development of 41 3.5-year old children whose mothers were occupationally exposed to solvents during pregnancy. Results showed no significant differences from controls in (1) cognitive status measured by the McCarthy Scales of General Abilities or (2) and child’s behavior and personality according to maternal ratings on questionnaires. However, the authors acknowledged that exposure levels may have been too low to produce noticeable neurobehavioral deficits. Since these researchers grouped exposure as a dichotomous variable (i.e., yes or no) but did not report the dose nor exposure duration, it is possible that their dichotomous stratification approach was not sufficiently sensitive to detect a true effect of exposure. Moreover, the general measures used in this study and mother’s report of developmental milestones, behavior, and personality may have been insensitive to detect any subtle and specific differences that may have existed between the solvent-exposed and control group at the specific age tested. The authors of the study did note that “subtle effects of low-level solvent exposure may not be detectable four years after the exposure because of the plasticity of the developing brain to compensate for early insult”. Alternatively, neurobehavioral effects may not emerge until a later age when more complicated cognitive skills develop, such as reading and arithmetic. In another study involving a large cohort of women hairdressers from the Netherlands, Kersemaekers, Roeleveld, Zielhuis, and Gabreels (56) compared the neurodevelopment of offspring of 9000 hairdressers born between 1986 and 1993 to 9000 age-matched children of clothing sales clerks, who served as the reference group. All women completed a questionnaire about their reproductive history, the child’s developmental milestones, and whether the child had seizures during fever. Results indicated that children of hairdressers born between 1986 and 1988 had significant delays in speaking first words and first sentences, whereas those born between 1991 and 1993 did not. For both study periods, seizures during fever were reported more often in the children of hairdressers than clothing salespersons. The findings therefore signify that there may be specific effects on child language development among offspring of hairdressers in the earlier (1986 to 1988) compared to the later period (1991 to 1993). Although not directly tested, the authors speculated that this difference reflected changes in the chemicals used by the hairdressers during the different time periods. Work from our lab may be more revealing. We examined cognitive and behavioral functioning prospectively in 33 3 to 7 year old children born to women who were occupationally exposed to organic solvents during pregnancy. We compared these

Effects of Organic Solvents on Reproductive Outcome

95

children to age-matched unexposed children (57,58). All children were ascertained through an antenatal counseling service in Toronto, Canada, for potential exposure to teratogens, and the reference group was deemed to be exposed to a nonharmful substance (e.g., acetaminophen). The children were tested with a comprehensive battery of neuropsychological tests that included subtests from the Developmental Neuropsychology Assessment (NEPSY), a visual continuous performance test (CPT), as well as parent-rated measures of children’s behavior. Exposure information was documented by the counseling service at the time of pregnancy via a structured questionnaire, and we estimated exposure intensity from this information. We found that while the performance of the exposed group was within the normal range overall, comparison with controls revealed that the exposed group performed more poorly on indices of expressive language (p!0.01), receptive language (p!0.01), and graphomotor ability (p!0.001). They performed comparably to controls on measures of attention and visuospatial and fine-motor abilities. As predicted by a dose-response model, negative associations were found between exposure levels and scores on the language and graphomotor tests reflecting poorer performance with increasing exposure level. A more recent study from our group (59) reported additional evidence of an association between weak language and visuomotor abilities in 32 children aged 3 to 9 years whose mothers were occupationally exposed to organic solvents during pregnancy, controlling for educational level and maternal IQ. These convergent findings provide confirmatory evidence that occupational exposure to organic solvents during pregnancy may be associated with selective deficits in the progeny. However, the highly heterogeneous solvent exposures in both studies and problems in estimating dose made it difficult to draw any conclusions about specific effects of chemicals.

Effects on Visual Functioning Specific visual functions have proven to be useful markers for CNS dysfunction in the research on adults with occupational solvent exposure. Studies have shown acquired color vision loss in adult workers after chronic, low-level exposure to styrene (60–63), toluene (64), n-hexane (65), and mixtures of solvents (66,67), as well as reduced contrast detection (68,69). Similar effects on vision, including reduced contrast sensitivity and degeneration of retinal ganglion cells, have been reported in macaque monkeys exposed to carbon disulfide (70,71). These well-documented effects on the mature visual system led our group to explore the impact of gestational exposure to organic solvents on the developing visual system (57). We used clinical measures to assess color vision and visual acuity in 32 of the children described above whose mothers were occupationally exposed to organic solvents during pregnancy and the same matched controls. Our results revealed that relative to controls, the solvent-exposed group showed mild to severe impairments in red-green (p!.05) and blue-yellow (p!.005) color discrimination as well as reduced visual acuity (p!.05). The increased incidence of blue-yellow color discrimination deficits in the exposed group is of particular interest because defects in this color axis are not inherited as are defects in the red-green range. Furthermore, this study also reported a nonsignificant trend reflecting clinical red-green color vision loss in 3 of the 32 exposed children (2 males and 1 female) versus no controls. It is interesting to note that these results were found despite the study’s exclusion of children with a family history of a heritable retinal disease or congenital color vision loss. Thus, our findings suggested that maternal occupational exposure to organic solvents during pregnancy is associated with increased risk of visual abnormalities in the progeny.

96

Till et al.

A larger study conducted by our group (72) prospectively evaluated the impact of prenatal solvent exposure on selective aspects of infant visual functioning (viz., contrast sensitivity, visual acuity, color vision) using visual evoked potential (VEP) techniques. The sample consisted of 21 infants born to women who were occupationally exposed to solvents during pregnancy compared with 27 non-exposed age-matched infants. Mothers and their infants were recruited from an antenatal counseling service, and exposure was estimated using questionnaire data obtained during pregnancy. Testers were masked to exposure status. Results showed a significant reduction in contrast sensitivity at low spatial frequencies for solvent-exposed infants compared to controls. Moreover, infants with high level exposure had significantly lower acuity than those with low exposure, suggesting evidence of a dose-response relation. With respect to color vision, the incidence of abnormal waveforms was significantly higher in the exposed group for the red-green chromatic response, but not the blue-yellow and achromatic response, signifying abnormalities in long- and medium-wavelength neural mechanisms. Overall, the findings suggest that exposure to organic solvents during pregnancy, at levels encountered in the workplace, is associated with selective visual deficits in offspring. In adults, tests of visual functioning are frequently added to standardized batteries of neurobehavioral tests since they are simple, noninvasive, and easy to use in the workplace as a screen for specific CNS impairments. Findings in children now suggest an increased risk of visual deficits following low-level gestational exposure to organic solvents. Prenatal exposure to other neurotoxic agents, including mercury or methylmercury (73,74), cocaine (75), and alcohol (76–79), has also been associated with visual disturbances. These findings in children suggest that the addition of visual tests to batteries may provide important information for detecting alterations to the developing nervous system.

ORGANIC SOLVENT INHALANT ABUSE Inhalant abuse, also known as “sniffing” or “huffing,” refers to the intentional inhalation of volatile chemicals and gases to achieve a desired intoxication (5). The most commonly abused volatile solvents include toluene and 1,1,1-trichloroethylane, through the sniffing of gasoline, glue, and spray paints. A sense of euphoria similar to alcohol-like intoxication is generally achieved with toluene exposures of at least 500 ppm, whereas confusion, auditory and visual hallucinations, and inhibition and incoordination occur at exposure levels of 600–800 ppm (80). The popularity of toluene as a recreational psychotropic agent is related to its relative ease of accessibility, low cost, and misperceived lack of addictive qualities (80). Of growing concern is the potential teratogenicity of this substance abuse, especially as many inhalant abusers are of child-bearing age. This popularity of volatile inhalants among adolescents raises concerns about potential unwanted pregnancies. Teratogenic effects have been reported in children born to mothers who abused solvents during pregnancy (3,81). In 1979, Toutant and Lippmann (81) proposed the term fetal solvent syndrome to describe a neonate born to a mother with a 14-year addiction to toluene and a 3-year history of alcohol intake estimated at six beers per week. This child presented with low birth weight, short length, and facial features similar to that of FAS. Seen in particular were craniofacial anomalies including microcephaly, a flat nasal bridge, a hypoplastic mandible, short palpebral fissures, mildly low-set ears, and a sloping forehead. Since this first report, more than 100 children have been described in the literature as exhibiting fetal solvent syndrome as a result of maternal solvent abuse during pregnancy (5).

Effects of Organic Solvents on Reproductive Outcome

97

In 1985, Hersh et al. (3) published a report describing clinical findings among three children aged 3 to 4 year who were exposed to toluene in utero. The clinical findings were microcephaly, CNS dysfunction, attentional deficits, developmental delay including language deficits, and variable growth deficiency. Phenotypic similarities included a small midface and narrow bifrontal diameter; short palpebral fissures, deep-set eyes, and low-set ears; micrognathia; and mild limb abnormalities, including blunted fingertips and small fingernails. These clinical characteristics are comparable to those described in children with FAS. In fact, of 18 toluene-exposed infants examined by Pearson et al. (4), 83% had features similar to FAS. While these similar teratogenic effects may signify a common mechanism for toluene and alcohol, this has not yet been established. The concomitant exposure to both toluene and alcohol may also potentiate the adverse effects of alcohol through ethanol-induced metabolic inhibition. Hersh et al. noted that gasoline inhalation during pregnancy results in a different facial phenotype and more significant neurodevelopmental impairments, including spastic quadriparesis, than prenatal exposure to toluene (3). Such differences may reflect the presence of lead and the minimal aromatics found in gasoline when the phase-down of lead as a gasoline additive was still incomplete. Our group conducted a neuropsychological examination as well as tests of vision on a 33-month old boy and a 40-month old girl, both of whom were exposed prenatally to toluene from maternal glue-sniffing during pregnancy. The children, who were biological siblings, separated in age by 13 months, were adopted as infants into separate families. Cognitive and motor abilities were formally assessed using the Mullen Scales of Early Learning for the boy and the Wechsler Primary and Preschool Scale of Intelligence (WPPSI-R) for the girl. Surprisingly, the children’s profiles were notably different. The boy’s results revealed age appropriate levels of cognitive and motor abilities, which were all within the average to high average range, whereas the girl displayed below average abilities overall with stronger verbal than non-verbal skills. Interestingly, however, both children demonstrated specific difficulties in the low to below average range on measures of graphomotor control, consistent with previous findings on offspring of women exposed occupationally to organic solvents during pregnancy (58). Clinical and electrophysiological techniques also used to assess selective visual functions in the two children showed they both had difficulties with color discrimination, while the girl also showed reduced contrast sensitivity and acuity. In addition, refraction results were normal for the boy and indicated astigmatism for the girl.

PROBLEMS STUDYING REPRODUCTIVE HAZARDS IN HUMAN STUDIES Several methodological issues make the assessment and monitoring of solvent effects in human offspring especially challenging. One consideration is that the mechanisms by which many solvents exert their toxicity are complex and vary among solvents and individuals. Factors to be accounted for in quantifying the exposure must include environmental conditions (e.g., ventilation, temperature), individual characteristics (e.g., body weight, physical exercise, genetic factors in susceptibility), and effect-modifying factors (e.g., smoking, alcohol consumption, drug intake, exposure to other chemicals). Moreover, the potential for toxicological interaction increases with use of solvent mixtures because of the possible unpredictable additivity, synergism, or potentiation of effects. Because these factors are difficult to control, most studies cannot confirm exposure levels or identify precisely the chemicals in multi-exposure occupations. Studies that combine

98

Till et al.

women with low and high exposures or classify exposure by job title may fail to detect true effects of the exposure. Assessment of exposure is further complicated by the fact that organic solvents are rapidly metabolized and no biological assay is available to evaluate fetal exposure. Finally, exposure to more than one chemical may also produce unknown byproducts that may be more toxic than the individual parent compound. Another problem is the retrospective nature of data collection, which makes confirmation of diagnoses and exposures difficult. For example, Lindbohm et al. (34) assessed recall bias in workers exposed to organic solvents during pregnancy and showed that 19% (9/47) of women did not report exposure to a specific solvent within a 24 month follow-up period. Similar findings by Eskenazi et al. (55) and Till et al. (82) indicated that many women exposed to solvents during pregnancy were inaccurate when asked to recall their exposure at a follow-up interview conducted 3 to 6 year after pregnancy. For example, findings by Till et al. indicated low percentage agreement between the two occasions for reports of duration of exposure (41%), protective barrier use (48%), and symptomatology (41%). When reports were not in perfect agreement, women at time of follow-up tended to report longer durations of exposure, increased use of protective barriers, and more symptoms than during pregnancy. Such variation may reflect response biases in the data collected at time of pregnancy or recall biases at time of follow-up, and are of concern because they can severely affect estimates of human teratological risk. Finally, factors such as small sample size and selection bias make it difficult to draw any causal associations between occupational exposure and pregnancy outcome. Clearly, all of these limitations attest to the need for good documentation of exposure during pregnancy, prospective designs, and strict methodological criteria.

SUMMARY AND RECOMMENDATIONS A large body of evidence has been accumulated on the effects of organic solvent exposure on the worker and developing fetus. As summarized in this review, this information indicates that adverse reproductive and developmental outcomes are commonly found following workplace exposure to organic solvents. Epidemiological studies suggest an increased risk of reduced fecundability, spontaneous abortions, and congenital malformations, especially among workers in laboratories and the pharmaceutical, graphics, and painting industries. Certain solvents, including aromatic, aliphatic, and chlorinated hydrocarbons have been consistently associated with increased risk of spontaneous abortions. It is important to note that many of the adverse effects were observed in mothers who were not symptomatic and exposed to organic solvents near or below the standard occupational exposure levels. With respect to neurodevelopmental outcomes of the offspring, some studies suggest an association between solvent exposure and specific neurocognitive deficits, particularly reduced expressive and receptive language abilities and poor performance on tasks of visuospatial and graphomotor abilities. Recent findings have also suggested an increased risk of color vision deficits and reduced acuity in children born to women exposed to organic solvents. Unfortunately, there are still few long-term studies on the behavioral effects of solvent exposure during pregnancy, and of the studies conducted to date, many are complicated by methodological weaknesses. Therefore, despite the large number of women exposed daily to solvents in the workplace and the presumed increased vulnerability of the fetus to the toxic insult, knowledge of neurobehavioral effects of solvent exposure remains sparse. This significant gap makes estimation of risk for particular occupations or solvent exposures virtually impossible. Positive findings encourage further studies with improved

Effects of Organic Solvents on Reproductive Outcome

99

study design and methods, particularly improved assessment of exposure and a diverse range of end points, including measures of visual functioning. In conclusion, the protection of pregnant women and their unborn children against occupational reproductive hazards represents a major challenge to our society. Not only are thousands of new chemicals developed each year, they are typically introduced without adequate testing for reproductive and developmental toxicity. Therefore, women are right to suspect reproductive health hazards as important issues in their workplaces and their lives. Until new policies are established to protect women from reproductive hazards in the workplace, women should be advised to minimize exposure to organic solvents during pregnancy. This exposure reduction must occur early in pregnancy to prevent effects during critical periods of organogenesis and it should extend throughout pregnancy to protect the developing fetus.

ACKNOWLEDGMENTS This work was supported by a grant from the Workplace Safety and Insurance Board of Ontario and by a studentship to Christine Till from the Canadian Institutes of Health Research. Gideon Koren is holder of the Research Leadership for Better Pharmacotherapy during Pregnancy and Lactation (Hospital for Sick Children) and the Ivey Chair in Molecular Toxicology, University of Western Ontario.

REFERENCES 1. Witkowski KM, Johnson NE. Organic-solvent water pollution and low birth weight in Michigan. Soc Biol 1992; 39:45–54. 2. Marshall EG, Gensburg LJ, Deres DA, Geary NS, Cayo MR. Maternal residential exposure to hazardous wastes and risk of central nervous system and musculoskeletal birth defects. Arch Environ Health 1997; 52:416–425. 3. Hersh JH, Podruch PE, Rogers G, Weisskopf B. Toluene embryopathy. J Pediatr 1985; 106:922–927. 4. Pearson M, Hoyme H, Seaver L, Rimsza M. Toluene embryopathy: delineation of the phenotype and comparison with fetal alcohol syndrome. Pediatrics 1994; 93:211–215. 5. Jones HE, Balster RL. Inhalant abuse in pregnancy. Obstet Gynecol Clin North Am 1998; 25:153–167. 6. Baker EL, Jr., Smith TJ, Landrigan PJ. The neurotoxicity of industrial solvents: a review of the literature. Am J Ind Med 1985; 8:207–217. 7. Sato A, Endoh K, Kaneko T, Johanson G. A simulation study of physiological factors affecting pharmacokinetic behaviour of organic solvent vapours. Br J Ind Med 1991; 48:342–347. 8. Cherry N. Neurobehavioural effects of solvents: the role of alcohol. Environ Res 1993; 62:155–158. 9. Stoltenburg-Didinger G, Altenkirch H, Wagner M. Neurotoxicity of organic solvent mixtures: embryotoxicity and fetotoxicity. Neurotoxicol Teratol 1990; 12:585–589. 10. Ghantous H, Danielsson BR. Placental transfer and distribution of toluene, xylene and benzene, and their metabolites during gestation in mice. Biol Res Pregnancy Perinatol 1986; 7:98–105. 11. Ghantous H, Danielsson BR, Dencker L, Gorczak J, Vesterberg O. Trichloroacetic acid accumulates in murine amniotic fluid after tri- and tetrachloroethylene inhalation. Acta Pharmacol Toxicol (Copenh) 1986; 58:105–114. 12. Teaf CM. Properties and effects of organic solvents. In: Williams PP, James RC, Roberts SM, eds. Principles of Toxicology, Environmental and Industrial Applications. New York: John Wiley & Sons, Inc., 2000:199–205.

100

Till et al.

13. Mason JM. Teratogens. In: Klaassen CD, Amdur MO, Doull J, eds. Cassarett and Doull’s Toxicology: The Basic Sciences of Poisons. New York: MacMillan Co., 1986. 14. Schwetz BA, Price CJ, George JD, Kimmel CA, Morrissey RE, Marr MC. The developmental toxicity of diethylene and triethylene glycol dimethyl ethers in rabbits. Fundam Appl Toxicol 1992; 19:238–245. 15. Narotsky MG, Kavlock RJ. A multidisciplinary approach to toxicological screening: II. Developmental toxicity. J Toxicol Environ Health 1995; 45:145–171. 16. Kishi R, Chen BQ, Katakura Y, Ikeda T, Miyake H. Effect of prenatal exposure to styrene on the neurobehavioral development, activity, motor coordination, and learning behavior of rats. Neurotoxicol Teratol 1995; 17:121–130. 17. Thiel R, Chahoud I. Postnatal development and behaviour of Wister rats after prenatal toluene exposure. Arch Toxicol 1997; 71:258–265. 18. Kyrklund T, Alling C, Haglid K, Kjellstand P. Chronic exposure to trichloroethylene: lipid and acyl group composition in gerbil cerebral cortex and hippocampus. Neurotoxicol 1983; 4:35–42. 19. Bornschein RL, Hastings L, Manson JM. Behavioral toxicity in the offspring of rats following maternal exposure to dichloromethane. Toxicol Appl Pharmacol 1980; 52:29–37. 20. von Euler G, Ogren SO, Li XM, Fuxe K, Gustafsson JA. Persistent effects of subchronic toluene exposure on spatial learning and memory, dopamine-mediated locomotor activity and dopamine D2 agonist binding in the rat. Toxicol 1993; 77:223–232. 21. Nelson BK, Taylor BJ, Setzer JV, Hornung RW. Behavioral teratology of perchloroethylene in rats. J Environ Pathol Toxicol 1980; 3:233–250. 22. Courtney KD, Andrews JE, Springer J, et al. A perinatal study of toluene in CD-1 mice. Fundam Appl Toxicol 1986; 6:145–154. 23. Lindbohm ML. Effects of parental exposure to solvents on pregnancy outcome. J Occup Environ Med 1995; 37:908–914. 24. Schwartz DA, Newsum LA, Heifetz RM. Parental occupation and birth outcome in an agricultural community. Scand J Work Environ Health 1986; 12:51–54. 25. Welch LS, Schrader SM, Clapp DE, Halperin WE, Turner TW, Hornung RW. Seme quality in workers exposed to ethylene glycol ethers on shipyard painters: II. Male reproduction. Am J Ind Med 1988; 14:509–526. 26. Rachootin P, Olsen J. The risk of infertility and delayed conception associated with exposures in the Danish workplace. J Occup Med 1983; 25:394–402. 27. Plenge-Bonig A, Karmaus W. Exposure to toluene in the printing industry is associated with subfecundity in women but not in men. Occup Environ Med 1999; 56:443–448. 28. Taskinen HK, Kyyronen P, Sallmen M, et al. Reduced fertility among female wood workers exposed to formaldehyde. Am J Ind Med 1999; 36:206–212. 29. Sallmen M, Lindbohm ML, Kyyronen P, et al. Reduced fertility among women exposed to organic solvents. Am J Ind Med 1995; 27:699–713. 30. Ng TP, Foo SC, Yoong T. Menstrual function in workers exposed to toluene. Br J Ind Med 1992; 49:799–803. 31. Zielhuis GA, Gijsen R, van der Gulden JW. Menstrual disorders among dry-cleaning workers. Scand J Work Environ Health 1989; 15:238. 32. Lemasters G, Hagen A, Samuels S. Reproductive outcomes in women exposed to solvents in 36 reinforced plastics companies. I. Menstrual dysfunction. J Occup Med 1985; 27:490–494. 33. McDonald AD, McDonald JC, Armstrong B, et al. Fetal death and work in pregnancy. Br J Ind Med 1988; 45:148–157. 34. Lindbohm ML, Taskinen H, Sallmen M, Hemminki K. Spontaneous abortions among women exposed to organic solvents. Am J Ind Med 1990; 17:449–463. 35. Kyyronen P, Taskinen H, Lindbohm ML, Hemminki K, Heinonen OP. Spontaneous abortions and congenital malformations among women exposed to tetrachloroethylene in dry cleaning. J Epidemiol Community Health 1989; 43:346–351. 36. Taskinen H, Kyyronen P, Hemminki K, Hoikkala M, Lajunen K, Lindbohm ML. Laboratory work and pregnancy outcome. J Occup Med 1994; 36:311–319.

Effects of Organic Solvents on Reproductive Outcome

101

37. Taskinen H, Anttila A, Lindbohm ML, Sallmen M, Hemminki K. Spontaneous abortions and congenital malformations among the wives of men occupationally exposed to organic solvents. Scand J Work Environ Health 1989; 15:345–352. 38. Heidam LZ. Spontaneous abortions among dental assistants, factory workers, painters, and gardening workers: a follow up study. J Epidemiol Community Health 1984; 38:149–155. 39. Axelsson G, Lutz C, Rylander R. Exposure to solvents and outcome of pregnancy in university laboratory employees. Br J Ind Med 1984; 41:305–312. 40. Lipscomb JA, Fenster L, Wrensch M, Shusterman D, Swan S. Pregnancy outcomes in women potentially exposed to occupational solvents and women working in the electronics industry. J Occup Med 1991; 33:597–604. 41. Taskinen H, Lindbohm ML, Hemminki K. Spontaneous abortions among women working in the pharmaceutical industry. Br J Ind Med 1986; 43:199–205. 42. Holmberg PC, Nurminen M. Congenital defects of the central nervous system and occupational factors during pregnancy. A case-referent study. Am J Ind Med 1980; 1:167–176. 43. Tikkanen J, Heinonen OP. Cardiovascular malformations and organic solvent exposure during pregnancy in Finland. Am J Ind Med 1988; 14:1–8. 44. McDonald JC, Lavoie J, Cote R, McDonald AD. Chemical exposures at work in early pregnancy and congenital defect: a case-referent study. Br J Ind Med 1987; 44:527–533. 45. Holmberg PC, Kurppa K. An epidemiologic approach to the study of possible teratogenic effects of chemical and physical environments. Scand J Work Environ Health 1982; 8:89–91. 46. Laumon B, Martin JL, Collet P, Bertucat I, Verney MP, Robert E. Exposure to organic solvents during pregnancy and oral clefts: a case-control study. Reprod Toxicol 1996; 10:15–19. 47. Garcia AM, Fletcher T. Maternal occupation in the leather industry and selected congenital malformations. Occup Environ Med 1998; 55:284–286. 48. Hemminki K, Mutanen P, Saloniemi I, Luoma K. Congenital malformations and maternal occupation in Finland: multivariate analysis. J Epidemiol Community Health 1981; 35:5–10. 49. Blomqvist U, Ericson A, Kallen B, Westerholm P. Delivery outcome for women working in the pulp and paper industry. Scand J Work Environ Health 1981; 7:114–118. 50. Tikkanen J, Kurppa K, Timonen H, Holmberg PC, Kuosma E, Rantala K. Cardiovascular malformations, work attendance, and occupational exposures during pregnancy in Finland. Am J Ind Med 1988; 14:197–204. 51. Khattak S, Moghtader G, McMartin K, Barrera M, Kennedy D, Koren G. Pregnancy outcome following gestational exposure to organic solvents: a prospective controlled study. JAMA 1999; 281:1106–1109. 52. McMartin KI, Chu M, Kopecky E, Einarson TR, Koren G. Pregnancy outcome following maternal organic solvent exposure: a meta-analysis of epidemiologic studies. Am J Ind Med 1998; 34:288–292. 53. Spyker JM. Assessing the impact of low level chemicals on development: behavioral and latent effects. Fed Proc 1975; 34:1835–1844. 54. Mattson SN, Riley EP, Gramling L, Delis DC, Jones KL. Neuropsychological comparison of alcohol-exposed children with or without physical features of fetal alcohol syndrome. Neuropsychol 1998; 12:146–153. 55. Eskenazi B, Gaylord L, Bracken MB, Brown D. In utero exposure to organic solvents and human neurodevelopment. Dev Med Child Neurol 1988; 30:492–501. 56. Kersemaekers WM, Roeleveld N, Zielhuis GA, Gabreels FJ. Neurodevelopment in offspring of hairdressers. Dev Med Child Neurol 1997; 39:358–362. 57. Till C, Westall CA, Rovet JF, Koren G. Effects of maternal occupational exposure to organic solvents on offspring visual functioning: a prospective controlled study. Teratol 2001; 64:134–141. 58. Till C, Koren G, Rovet JF. Prenatal exposure to organic solvents and child neurobehavioral performance. Neurotoxicol Teratol 2001; 23:235–245. 59. Laslo-Baker D, Barrera M, Knittel-Keren D, et al. Child neurodevelopmental outcome and maternal occupational exposure to solvents. Arch Pediatr Adolesc Med 2004; 158:956–961.

102

Till et al.

60. Gobba F, Galassi C, Imbriani M, Ghittori S, Candela S, Cavalleri A. Acquired dyschromatopsia among styrene-exposed workers. J Occup Med 1991; 33:761–765. 61. Fallas C, Fallas J, Maslard P, Dally S. Subclinical impairment of colour vision among workers exposed to styrene. Br J Ind Med 1992; 49:679–682. 62. Chia SE, Jeyaratnam J, Ong CN, Ng TP, Lee HS. Impairment of color vision among workers exposed to low concentrations of styrene. Am J Ind Med 1994; 26:481–488. 63. Campagna D, Gobba F, Mergler D, et al. Color vision loss among styrene-exposed workers neurotoxicological threshold assessment. Neurotoxicol 1996; 17:367–373. 64. Zavalic M, Mandic Z, Turk R, Bogadi-Sare A, Plavec D. Quantitative assessment of color vision impairment in workers exposed to toluene. Am J Ind Med 1998; 33:297–304. 65. Raitta C, Seppalainen AN, Huuskonen MS. N-hexane maculopathy in industrial workers. Albrecht Von Graefes Arch Clin Exp Ophthalmol 1978; 209:99–110. 66. Mergler D, Belanger S, De Grosbois S, Vachon N. Chromal focus of acquired chromatic discrimination loss and solvent exposure among printshop workers. Toxicol 1988; 49:341–348. 67. Muttray A, Wolff U, Jung D, Konietzko J. Blue-yellow deficiency in workers exposed to low concentrations of organic solvents. Int Arch Occup Environ Health 1997; 70:407–412. 68. Mergler D, Huel G, Bowler R, Frenette B, Cone J. Visual dysfunction among former microelectronics assembly workers. Arch Environ Health 1991; 46:326–334. 69. Frenette B, Mergler D, Bowler R. Contrast-sensitivity loss in a group of former microelectronics workers with normal visual acuity. Optom Vis Sci 1991; 68:556–560. 70. Merigan WH, Wood RW, Zehl D, Eskin TA. Carbon disulfide effects on the visual system. I. Visual thresholds and ophthalmoscopy. Invest Ophthalmol Vis Sci 1988; 29:512–518. 71. Eskin TA, Merigan WH, Wood RW. Carbon disulfide effects on the visual system. II. Retinogeniculate degeneration. Invest Ophthalmol Vis Sci 1988; 29:519–527. 72. Till C, Nulman I, Swartz S, Koren G, Westall CA, Rovet JF. Vision abnormalities in young children exposed prenatally to organic solvents. Neurotoxicology, 2005; 26:599–613. 73. Burbacher TM, Grant KS, Gilbert SG, Rice DC. The effects of methylmercury exposure on visual and auditory functions in nonhuman primates. Toxicologist 1999; 48:362. 74. Rice DC, Gilbert SG. Effects of developmental exposure to methyl mercury on spatial and temporal visual function in monkeys. Toxicol Appl Pharmacol 1990; 102:151–163. 75. Block SS, Moore BD, Scharre JE. Visual anomalies in young children exposed to cocaine. Optom Vis Sc 1997; 74:28–36. 76. Stromland K. Ocular involvement in the fetal alcohol syndrome. Surv Ophthalmol 1987; 31:277–284. 77. Stromland K, Hellstrom A. Fetal alcohol syndrome—an ophthalmological and socioeducational prospective study. Pediatrics 1996; 97:845–850. 78. Hug TE, Fitzgerald KM, Cibis GW. Clinical and electroretinographic findings in fetal alcohol syndrome. J AAPOS 2000; 4:200–204. 79. Katz LM, Fox DA. Prenatal ethanol exposure alters scotopic and photpic components of adult rat electroretinograms. Invest Ophthalmol Vis Sci 1991; 32:2861–2872. 80. Wilkins-Haug L. Teratogen update: Toluene. Teratol 1997; 55:145–151. 81. Toutant C, Lippmann S. Fetal solvents syndrome. Lancet 1979;1. 82. Till C, Koren G, Rovet JF. Agreement between prospective and retrospective reports of maternal exposure to chemicals during pregnancy. J Occup Environ Med 2002; 44:708–713.

SECTION TWO:

MEDICINAL AND RECREATIONAL SUBSTANCE USE

6 The Structural and Functional Teratology of Antiepileptic Medications Jane Adams and Jennifer Anne Lantz Gavin Department of Psychology, University of Massachusetts Boston, Boston, Massachusetts, U.S.A.

Patricia A. Janulewicz Department of Environmental Health, Boston University School of Public Health and Department of Psychology, University of Massachusetts, Boston, Massachusetts, U.S.A.

Understanding the association between exposure to antiepileptic medications during pregnancy and the diverse manifestations of adverse outcome is complicated by the need to first understand basic principles of teratogenic, and particularly neuro-teratogenic, action and basic definitions of adverse outcomes. Conveyance of information specific to the teratogenicity of anticonvulsants is further complicated by the fact that the emergence in the late 1960s of studies raising concern about antiepileptic drug use during pregnancy overlapped with the establishment of the basic scientific principles. As a result, the designs of early studies pre-dated recognition of certain principles of teratogenic action as well as recognition of the need to employ certain design characteristics when neurodevelopmental, especially cognitive, endpoints were examined. In light of these complexities and with the goal of having this chapter serve the needs not only of neurobehavioral teratologists, but of students, health professionals, regulatory scientists, and structurallyfocused teratologists, our approach is to cover certain fundamentals alongside the data specific to the actions of anticonvulsant medications as neurobehavioral teratogens. Thus, we will first examine fundamentals of teratogenic action and the nature of adverse outcomes, and then focus on the risks associated with anticonvulsant exposure during pregnancy. By necessity, opening discussions are brief and make considerable reference to review sources. FUNDAMENTALS OF TERATOLOGY AND NEUROBEHAVIORAL TERATOLOGY Teratogens are agents that are capable of inducing abnormal development when exposures occur during prenatal development. Abnormality may be manifest as spontaneous abortion or stillbirth, or in liveborn infants, as major malformations, growth alterations, or functional developmental deficits (1). For teratogens of the central nervous system, these 103

104

Adams et al.

outcomes have been shown to occur across a continuum from death to functional impairment, and to be manifest in syndromes whereby an individual teratogen produces a signature cluster of major and minor malformations, possible effects on growth, and effects on neurobehavioral and cognitive functioning (2–5). The syndromic expression of teratogenic effects dictates designs aimed at uncovering and measuring multiple endpoints. Thus, early research studies aimed at associating prenatal exposure with a single, fairly rare malformation were often not able to identify teratogens within the sample sizes evaluated in most studies (3). The importance of syndromic manifestation is critical to understanding the difficulty inherent in identifying anticonvulsant teratogenicity in the historical timeframe where studies often focused on single malformation outcomes. First, it is important to provide formal definitions for the categories of endpoints relevant to the syndromic expression of adverse outcomes. Most notably, it is important to distinguish between major and minor malformations, and to examine how minor structural defects, growth impairments, and cognitive impairments have been defined. As used by teratologists (3), “major malformations” are defined as major structural defects that have either cosmetic or functional significance to the child. Examples are heart malformations, cleft lip, or neural tube defects such as anencephaly or spina bifida. “Minor malformations” are defined as minor structural defects that occur infrequently (in less than 4% of the population) and that have no known cosmetic or functional significance to the child. Minor malformations include things such as midface hypoplasia, which has features such as a broad nasal bridge, short nose, or long upper lip. Patterns of minor malformation have been identified for many neurobehavioral teratogens, including antiepileptic drugs. “Patterns” of minor malformations are often defined as a minimum of three specific defects that occur in at least two infants in the exposed group under study. Minor malformations are typically only detected if a formal dysmorphologic exam is done, or when an abridged number of measurements are taken of structures under suspicion as relevant to a particular agent. The incorporation of full dysmorphologic examination into research studies aimed at determining the teratogenic risks associated with drug exposures is not yet routine. Nevertheless, the relevance of patterns of minor malformation as early warning indicators for increased risks for neurobehavioral and cognitive dysfunction is receiving increased attention. Such relationships will be examined later in this chapter when we discuss predictors of outcome. Standardized dysmorphological exams are administered by trained physicians specialized in dysmorphology, generally in association with obstetrics, pediatrics, or genetics. The standardized exams consist of many body measurements including head circumference, length, weight, palpebral fissure length, inner canthal distance, ear length, philtral length, length and width of the nasal bridge, and multiple other measures of all bodily parts. Measurements are then referred to databases of values corresponding with established percentiles for growth at specific ages, such as at term birth. For example, measurements of weight, length, and head circumference can be used to calculate exact percentiles for infants of the same age and sex using the National Center for Health Statistics growth curves (3,6). For most measurements, values that are greater than two standard deviations (2 SD) from the mean are judged to be abnormal and the differential outcomes of study groups are compared on the basis of the proportion of infants having abnormal scores. Thus, an increased prevalence of microcephaly, for example, would mean that following teratogenic exposure, a higher proportion of the exposed infants had occipito-frontal head circumferences that were greater than 2 SD below the established mean. This approach to the characterization of deviant body measurements was frequently applied to the characterization of intellectual dysfunction in early studies of the effects of prenatal exposure to anticonvulsants. As will be described later, such

The Structural and Functional Teratology of Antiepileptic Medications

105

applications considered intellectual outcomes as “normal” or “abnormal” and thereby only judged children with mental retardation (IQ scores greater than 2 SD below the mean) to have been affected by exposure. This early approach to risk ascertainment is contrary to the now established recognition that adverse neurodevelopmental outcomes are typically manifest over a continuum from severe to more subtle dysfunction, and that it is critically important to examine general ability as well as specific abilities across domains of neuropsychological functioning (2). In the field of neurobehavioral teratology, multiple adverse outcomes are often studied alongside and in relationship to effects on nervous system functioning. Since the brain ultimately contributes to or controls the function of all organ systems and since individual systems also have their own adverse responses to exposure, functional developmental deficits may be manifest through dysfunction in any organ system. Nevertheless, deficits mediated by the nervous system primarily and historically have been investigated with respect to behavioral development and cognitive functioning, although more recent attention is developing for social and psychiatric outcome characteristics (7–9), as well as endocrine mediated characteristics such as sexually dimorphic reproductive or cognitive performance abilities (2,10,11). As is the case for teratogenic effects on death, malformation, growth or function, abnormalities in nervous system development are also generally exhibited across a continuum whereby severe effects fall at one end and more subtle, but nevertheless significantly compromising dysfunctions occur at the other (2,5,12,13). Some examples are mental retardation or subtle learning disabilities, motor spasticities or mild coordination deficits, sensory loss such as in blindness or deafness, or more subtle impairments such as changes in visual, auditory, or tactile abilities. Severe effects are generally associated with detectable malformations in the brain, but more subtle dysfunctions are often seen in the absence of currently detectable structural changes (5,13,14). Research in animal models as well as human beings has clearly demonstrated that the specific outcome along the continuum from death to dysfunction or severe to subtle dysfunction is associated with exposure dose, exposure duration, precise time of exposure during development, general health characteristics of the mother, and genetic characteristics of mother and offspring (2). The importance of these variables will be illustrated in the case of antiepileptic medications. It is important to recognize that teratogen-induced adverse outcomes of pregnancy, from death to subtle dysfunction, are preventable causes of abnormal development that can be precluded by better understanding of the many drugs, chemicals, diseases, metabolic disorders, and infections that increase risks. In this chapter, we hope to broaden the understanding of the risks associated with exposure to anti-epileptic drugs during prenatal development. “The risks of major malformations, minor anomalies, and dysmorphic features is twofold to threefold higher in infants of mothers with epilepsy who receive treatment with antiepileptic drugs (AEDs) compared with the risks in infants of mothers without epilepsy” (15). This consensusbased statement from an international group of clinician-investigators followed many years of controversy about the relative roles of the maternal disorder of epilepsy versus the medications themselves as the predominant cause of certain adverse outcomes: a controversy which still continues. In an effort to examine the multiple variables of relevance to the association between exposure and outcome, we will first examine the general role of epilepsy as a disorder and of AED exposure with respect to structural adverse outcomes. We will then evaluate the role of dosage level and polytherapy versus monotherapy exposure upon structural effects of AED use during pregnancy. Next, the effects of AED polytherapy upon behavioral and cognitive outcomes will be presented. Then, we will describe the effects of each of the primary medications that have been

106

Adams et al.

used to treat epilepsy during pregnancy: diphenylhydantoin (PHT), Phenobarbital (PB), carbamazepine (CBZ), and valproic acid (VPA). Individualized predictors of susceptibility to adverse outcome that can be assessed during pregnancy or at birth will then be discussed. Finally, we will identify issues that need to be further explored in future studies.

RESEARCH IN THE EARLY YEARS: THE ROLE OF EPILEPSY AS A DISEASE CONTRIBUTING TO ADVERSE OUTCOMES OF PREGNANCY Epilepsy is a common neurological disorder that affects about 2.3 million people in the United States, one-third of whom are women of childbearing age (6,16). Approximately 95% of women with epilepsy are on AEDs for seizure management, and thus the majority of children born to mothers with seizure disorders are exposed to AEDs in utero (17). Likewise, these children may be exposed to certain untoward effects of epilepsy, particularly if seizures occur during pregnancy. There are several altered physiological characteristics of women with epilepsy that are associated with the disease itself and/or its treatment. Alterations in several reproductive, hormonal, and other physiological parameters are known (17,18). With respect to pregnancy outcome, a primary emphasis has been placed on dissociating the effects of seizures during pregnancy or possible genetic risk factors from the drug-based induction of teratogenicity. The extent of occurrence of seizures among women treated with anticonvulsant medications during pregnancy is unclear and varies across studies in conjunction with the types of seizures that are monitored, the exclusion criteria, as well as other characteristics of the specific samples of epileptic women that are studied. Typically arguments expressing concern about the potential adverse effects of seizures during pregnancy upon infant outcome have been based on possible harm from prolonged generalized convulsions that may pose risk for fetal anoxia or hypoxia (15). Given these concerns, early reports on the adverse outcomes of pregnancy among women with seizure disorders (19–26) met with considerable resistance and difficulty in interpretation. The medical community needed to reconcile competing ideas about the harm from seizures during pregnancy and the harm from other aspects of epilepsy as a maternal disorder with the idea of a dominantly druginduced set of adverse outcomes (15). This dilemma and surrounding controversy has generated multiple animal as well as human studies addressing the role of the maternal disease of epilepsy as a genetic or seizure-related contributor to the reported adverse outcomes. From the late 1960s throughout the 80s, research on the teratogenicity of anticonvulsant drugs was dominated by these themes. This was a period when the typical woman with epilepsy was more often treated by combinations of AEDs than by single medications, and the medications most frequently used were phenytoin, Phenobarbital, carbamazepine, and valproic acid (15). Thus, the human research primarily examined the effects of polytherapy. During that period, animal studies (monotherapy studies of the above named drugs) clearly demonstrated that (1) the administration to healthy animals of individual AEDs during pregnancy increased malformations and behavioral abnormalities in the offspring (27,28), and (2) that rates of malformation were increased even when outcomes in offspring from AED exposed rodents or monkeys were compared to outcomes in epileptic animals that had seizures throughout pregnancy (29–31). The human research of these two decades determined that: (1) The prevalence of structural malformation in infants born to women on anticonvulsant polytherapy was increased relative to the general population, while prevalence in infants born to women with untreated seizure disorders did not differ from

The Structural and Functional Teratology of Antiepileptic Medications

107

the general population (21,32–35), (2) Among pregnant women taking AEDs for seizure control, the type of seizure disorder (hence possible anoxia for some) did not appear to affect malformation rate (36), (3) The prevalence of major and minor malformations was shown to be dose-related (37), and (4) among infants born to epileptic fathers (hence genetic risks), the prevalence of malformation is lower than among infants born to mothers (33,38,39). More recent studies also document that epileptic women who take AEDs during pregnancy have higher malformation rates among infants than are seen in infants born to unmedicated epileptic women (40,41) or to fathers with epilepsy. Further, it has been shown that when AEDs are used to treat non-epileptic conditions during pregnancy, adverse outcomes are elevated compared to controls (40). Nevertheless, debate continues as to whether the occurrence of seizures during pregnancy may exacerbate risks when AED-exposed women with and without seizures during pregnancy are compared. While evidence clearly establishes that anticonvulsant drugs are teratogenic, certain genetic risk factors associated with some forms of epilepsy are known to increase susceptibility to certain drug-induced adverse outcomes, notably cleft palate (4,42), see later discussion on predictors of susceptibility. Animal studies also demonstrate the importance of genotype to the phenotypic expression of adverse outcomes (4).

THE EFFECTS OF POLYTHERAPY VERSUS MONOTHERAPY TREATMENTS Treatment prior to the mid-1990s involved combinations of medications, mainly hydantoins (Dilantine), barbiturates such as Phenobarbital, carbamazepine, and valproic acid. Thus, studies prior to this time largely focused on the effects of polytherapy. The research also focused on pregnancy outcome or birth characteristics such as major malformations or microcephaly, with non-systematic observations of increased mental retardation and developmental delay (27). As is typically the case, the identification of anticonvulsants as teratogenic drugs progressed from early case reports, to retrospective studies (participants are identified or may choose to participate after outcome is known), to studies involving prospective enrollment. A pure prospective enrollment refers to recruitment of infant subjects during gestation via their pregnant mother. Prospective also often refers simply to enrollment at the time of birth based only on knowledge of exposure, but not on any aspect of outcome. As has been the case for the evolution of studies of structural endpoints, studies focused on behavioral outcome progressed from nonsystematic observation of mental retardation or developmental delay, to studies involving systematic testing but focusing on general mental ability (often referred to as IQ), to studies using batteries of neuropsychological tests to ascertain functioning across multiple domains of processing. Much of the early research lacked rigor with respect to cognitive evaluation which in single studies was done at different ages, using different methods, without experimenter blinding, without control for maternal socio-economic or intellectual characteristics, and with typical attention only to the rates of mental retardation (43). Nevertheless, by convergence, the studies began to establish that use of anticonvulsant medications during pregnancy increased risks for central nervous system (CNS), cardiac, craniofacial, skeletal, growth (including microcephaly), and cognitive defects (21,22,40,42–44). A pattern of abnormalities was identified that included the above along with minor anomalies such as hypertelorism, midface hypoplasia (depressed nasal bridge, short nose with anteverted nostrils), low set ears, single transverse palmar creases, and digital and nail hypoplasia. We will first examine the effects of polytherapy on

108

Adams et al.

structural and growth endpoints. We will then present the history of neurobehavioral teratology studies. Effects on Structural and Growth Outcomes The early studies suggested that polytherapy was generally more harmful than monotherapy with respect to effects on malformation rates, rates of microcephaly and growth endpoints (22,37,39), thus leading to a consensus judgment that monotherapy should be employed whenever possible (15). Because most women were treated with drug combinations, studies comparing polytherapy cases with monotherapy cases, by necessity collapsed across the individual medications administered in monotherapy due to insufficient sample sizes for individual drugs. Studies were conflicting regarding which combinations of anticonvulsant drugs appeared most harmful: a controversy that resulted in part due to the small numbers of patients on monotherapy that could be studied as well as possible differences across medications in their potency to induce different specific endpoints. In 1974, Hill and colleagues published a review of 20 prior reports that evaluated a total of 1653 infants from mothers on various combinations of AEDs. Across these studies, a 12.5% incidence of congenital abnormalities was reported including cleft lip or palate, heart defects such as ventricular septal defects or patent ductus arteriosus, skeletal anomalies such as clubfoot, dislocated hips, hypoplastic digits and nails, central nervous system malformations including neural tube defects and hydrocephalus, gastrointestinal malformations, genitourinary anomalies, and facial or ear anomalies, and microcephaly. Based on the convergence of findings from the results of prior studies, several larger studies were undertaken. Two studies are noteworthy for their relatively large sample sizes, attention to important risk factors, and the examination of large enough numbers of monotherapy cases to allow comparison across individual medications. Holmes et al. conducted a large scale study between 1986 and 1993 that was designed to detect a tailored subset of adverse outcomes in line with the reported syndrome of major and minor abnormalities (45). The study sought to address the impact of AED treatment versus no treatment among women with seizure disorders, the effects of polytherapy versus monotherapy, as well as to further examine the role of type of epilepsy upon outcome. While not a pure prospective study in which participants would be identified prior to birth, the study did examine the outcomes of pregnancy based on identification of exposed and control women during the labor and delivery period. Women in the labor and delivery suites of five maternity hospitals in the Boston area were interviewed in part to identify women who suffered from epilepsy during pregnancy and who did or did not take medications during pregnancy for seizure control. For each infant born to a medication treated or untreated woman with epilepsy, a control infant was identified from among the 10 infants born closest in time to that infant. Each infant was examined for the presence or absence of major malformations and was given a dysmorphology exam that focused on the minor malformation features previously associated with AED exposure. Outcomes of interest were major malformations, microcephaly, growth retardation, and hypoplasia of the midface and digits. Midface hypoplasia was defined as the presence of two or more of the following features: short nose, long upper lip, telecanthus, or broad nasal bridge. A feature was considered to be present if the measurement value was more than 1 SD above or below the mean values for the control infants. Digital hypoplasia was defined as marked stiffness of the interphalangeal joints or the presence of 6 or more arches among the 10 dermal ridge patterns. Efforts toward more empirical definitions illustrate the

The Structural and Functional Teratology of Antiepileptic Medications

109

recognition of the subjectivity of some of these judgments. Indeed, digital hypoplasia is now frequently defined on the basis of measurements from hand radiographs (46,47), a technique used by Kelly in 1984 (48). Following the determination of qualification for study after exclusions of women who had multiple gestation pregnancies, the presence of another teratogenic risk factor, or who were non-English speaking, coupled with loss of subjects based on refusals to participate or later failure to show up for scheduled appointments, the following subject pool resulted: 233 of a possible 386 women with monotherapy AED exposures for the treatment of epilepsy, 93 of a possible 123 women exposed to 2 or more AEDs for the treatment of epilepsy, and 98 out of a possible 98 women with epilepsy who were not treated during pregnancy. Among the 233 women on monotherapy, 87 were taking phenytoin, 64 phenobarbital, 58 carbamazepine, 6 valproic acid, and 8 were taking other AEDs. In addition to the women on AEDs for the treatment of epilepsy, there were 35 women who took AEDs for other indications. Across the outcomes that were measured under blind conditions for 93% of the exams, there were no differences between the infants born to control mothers and those born to women with epilepsy who were not treated with AEDs. 28% of the polytherapy exposed infants had 1 or more of the abnormal outcomes of interest, as compared to 20.6% of infants exposed to prenatal monotherapies, and 8.5% of controls. With respect to the specific outcomes of interest (which were often simultaneously manifest in individual children), among the polytherapy-exposed infants, there was an 8.6% incidence of major malformations, 3.3% exhibited microcephaly, 7.6% growth retardation, 12.7% midface hypoplasia, and 7.9% hypoplasia of the fingers. The Holmes et al. study found no difference in the outcome of infants born to women with “familial” versus “acquired” forms of epilepsy, and no differences in major malformation rate in association with the type of seizure disorder or types of seizures that occurred during pregnancy. Among 35 women on AEDs for the treatment of conditions other than epilepsy, abnormal outcomes were increased above control levels with 9.3% of the infants having major malformations and 25.3% having growth retardation or microcephaly. Thus, this study corroborated earlier findings from smaller studies with respect to effects on structural and growth outcomes, and emphasized the need for a better understanding of the effects of monotherapy in order to best counsel and treat pregnant women with epilepsy. Around the same time that data were collected for the Holmes et al. study, Kaneko et al. conducted a prospective study of the outcome of patients who were primarily enrolled for study during the first trimester of pregnancy (49). This study represented a collaboration among medical centers in Japan, Italy, and Canada from 1978–1991. The investigators were interested in studying the influence of seizure frequency and seizure type during pregnancy upon outcome as well as in examining the effects of polytherapy and dosage amounts on outcome. A total of 983 infants were evaluated of which 500 were on monotherapy treatment regimens. In this study, the incidence of major malformations was 9.0% in the AED-exposed infants and 3.1% in infants without gestational drug exposure. The incidence of major malformations increased as a function of the number of drugs the infant was exposed to as well as the total daily dose of the medications. Major malformation rates were not influenced by the type of epilepsy (familial or acquired), the type of seizures that occurred during pregnancy, the occurrence of seizures during the first trimester, or the frequency of seizures. Like the Holmes et al. study, this large-scale effort corroborated the increased risks associated with polytherapy and emphasized the teratogenic role of the medications as more important to outcome than multiple maternal disease variables.

110

Adams et al.

Effects on Developmental and Mental Ability Outcomes Concerns about developmental delay and mental retardation in association with polytherapy exposures throughout gestation were initially raised on the basis of case reports by Hanson and Smith (20) and a retrospective study by Speidel and Meadow (24). Prospective studies followed that also suggested generalized developmental delays. These early studies raised concerns about the risks for mental retardation resulting from exposure to anticonvulsant medications, but did not control for parental socioeconomic or educational characteristics. Indeed, the employment of appropriate controls for these variables did not begin until the 1990s. Table 1 presents certain design characteristics and results obtained in selected prospective studies. To be included in the table, the studies must have (1) examined developmental and behavioral outcomes in AED-exposed children that were prospectively identified, (2) compared the findings with those from nonexposed children born to women from the same “community”, and (3) examined the effects of polytherapy upon 20 or more AED-exposed subjects. The first prospective study, conducted in 1974 by Hill et al. (21), followed 28 prenatally AED-exposed infants from birth to 36 months of age as well as 165 unexposed infants that were delivered at the same hospitals. Physical examinations were performed on the infants on 10 occasions during the first 3 years of life, and Gesell developmental evaluations were performed on six occasions beginning at 9 months of age. Seven of the 28 infants had major malformations, and three had developmental quotients less than 90 at 18 months of age. The authors raised concerns about possible lifelong effects on mental ability, while also indicating that many genetic as well as environmental factors were not addressed. In 1982, Hill et al. (22) reported on a larger group of 59 infants born to epileptic mothers compared to 253 non-epileptic controls. In this cohort, the educational backgrounds of the exposed and control parents were recorded and were similar with an average of 14 years for mothers and 15 years for fathers (in the context of the U.S. K-12 system). This sample therefore represented women from middle to high socio-economic backgrounds. Evaluations were performed at some point between 9 months and 9 years of age using the Gesell Psychometric Scale for 9–36 months olds, the Wechsler Preschool and Primary Scale of Intelligence (WPPSI) for 4 years olds, or the Revised Wechsler Intelligence Scale for Children (WISC-R) for older children. Data were reported as the percentage of exposed versus control children that scored !90 (the low average to below average range), in the average range, in the high average to above average range, and greater than 2 SD above the mean for all of these measures. Scores from these tests showed that when measured at 3 years of age or younger, 25% of the infants born to epileptic women scored less than 90 as compared to only 9% of the controls. The authors pointed out, however, that the incidence of children scoring in this lower range was within population norms although lower than the matched controls. This is an important point in that it dispelled the idea of increased rates of mental retardation among children prenatally exposed to AEDs and instead emphasized milder reductions in mental ability or increased rates of learning disabilities. Indeed, Hill et al. (22) suggested that 48% of the infants born to epileptic mothers had some form of learning disability, although the criteria for defining learning disorders were not presented. The authors suggested that the most prominent type of learning disability was a language-based learning disorder. Further, they reported that 35% of the infants born to epileptic women required “special education” as compared to 8% of the control group. Just as an increased number of children scored in the lower end of the normal distribution, fewer exposed children (18%) scored in the high average to above average ranges than did control infants (33%). At later ages, 19% of the exposed compared

28 AEDexposed

Sample size

165 nonexposed infants

Comparison group Birth– 3 years

Age at testing

Birth– Hanson et al. Data from the 104 AED100 non7 years 1976 (19)a exposed exposed Collaborainfants also all includtive Perinafrom the ing PHT tal Project Collabora(24 PHT prenatal tive Perinatal monotherenrollment Project apy) Birth– 252 nonHill et al. Admission to 59 AED9 years exposed exposed 1982 (22) labor and infants delivery suite

Admission to labor and delivery suite

Time of enrollment

Effects on postnatal growth

Modified Gesell Reduced weight, (! 3 years); length, WPPSI and head (4 years); circ. WISC-R (6–9 years)

Reduced general mental ability

Reduced Gesell Develop- Reduced developmental Evaluweight, mental ations length, score and head circ. Reduced Wechsler Intelli- Reduced head circ. general gence Scale mental for Children ability (age 7 years)

Behavioral methods used

Effects on development or general mental ability

Increased languagebased learning problems and increased need for special education

Not reported

Not assessed

Effects on specific abilities

Prospective Studies on the Behavioral Development of Children Exposed to Anticonvulsant Medications During Gestation

Hill et al. 1974 (21)

Study

Table 1

(Continued)

Neonatal withdrawal syndrome for barbiturateand PHTexposed Suggested relationship between presence of syndrome and occurrence of reduced mental ability Polytherapy worse than monotherapy; PHT monotherapy better than PHT C PB; Identified 5 risk factors for reduced mental ability

Other findings

The Structural and Functional Teratology of Antiepileptic Medications 111

105 nonexposed infants

41 nonexposed infants

104 AEDexposed

43 AED exposed

Prior to Gaily, conception Kantolaor prenatal Sorsa, and Granstrom, 1990 (63)

Leavitt et al. 1992 (96)

Prior to conception or during 1st trimester of pregnancy

28 nonexposed infants

47 (29 MT; 18 PT)

Sample size

Comparison group

Jager-Roman Prenatal et al. 1982 (53)

Study

Time of enrollment

Behavioral methods used

Effects on postnatal growth

No effects on Mildly 8 wk– Audiology; reduced weight or 12 months Bayley Scales Bayley head of Infant MDI circumferDevelopment scores ence

Reduced Reduced 6–24 months Denver Develpsychoweight, opmental motor length, Scale; performand head Munchener ance circ. Funktionelle Entwicklungsdiagnostik Not reported No differExamined at WPPSI and ences in 5.5 years Illinois Test of general Psycholinmental guistic Abilability ities

Age at testing

Effects on development or general mental ability

Increased risks were associated with seizures during pregnancy and with low paternal education Polytherapy lower than monotherapy Increased # with low scores on specific measures

Not evaluated

Polytherapy worse than monotherapy; Improvement in some after age 1

Other findings

Not assessed

Effects on specific abilities

Table 1 Prospective Studies on the Behavioral Development of Children Exposed to Anticonvulsant Medications During Gestation (Continued)

112 Adams et al.

Steinhausen Prenatal et al. 1994 (50)b and Losche et al. 1994 (54)

VanOverloop Labor and delivery et al. 1992 suite (55)

WPPSI (age 4); Not reported Reduced Birth– 98 non20 AEDgeneral WISC-R and 9 years of exposed exposed, mental other age infants with all includability measures of 3 or more ing PHT specific minor (15 were abilities anomalies; PHT-PB (6–9 years) also 20 combimatched on nation) parity, socioeconomic status, and child sex and age Brazelton Neo- Not reported Reduced 34–73 vary- 42–67 varying Birth– general 6 years natal Behaacross ing across mental vioral Assessmeasures; measures ability ment Scale Matched on (NBAS); socioecovideotaped nomic status; maternalmaternal age child interand parity action; and other HOME factors Inventory; Bayley Scales of Infant Development; WPPSI; and other tests of specific abilities as well as psychopathology Reduced motor performance; Reduced verbal and non-verbal performance; Increased speech problems

Reduced nonverbal performance and drawing ability compared to both comparison groups

(Continued)

No effects in infants born to epileptic women not on meds or epileptic fathers; Polytherapy worse than monotherapy.

Combination therapy with PHT and PB appeared to account for these results.

The Structural and Functional Teratology of Antiepileptic Medications 113

99 born to healthy mothers; controlled for parental income, maternal education and age, gestational age of child

66 nonexposed infants

Comparison group

Birth–3 yr

Birth– 5 years

Age at testing

Effects on postnatal growth

Effects on specific abilities

No difference Griffith’s Test of Not reported No differon specific Development ence in subscales general score Reduced Reduced Enjoji’s Analyti- Reduced motor, developlength and cal Developsocial, mental head cirment Test; speech and performcumferneurological language ance at 1.5 ence at exams, strucscores and birth but tured 3 years not at interviews 3 years

Behavioral methods used

PHT (but not polytherapy) reduced locomotor development AED dose and small head size associated with reduced developmental scores at 1.5 years, but effects of socioeconomic background was more important at age 3.

Other findings

Note: In order to be included in this table, studies must have utilized prospective ascertainment prior to known behavioral outcome of the children, must have evaluated a sample size O20 AED-exposed children, and must have evaluated a control/comparison group from a similar population and locale. a Hanson et al. 1976 also reported on 35 cases who were not compared to controls and where cognitive outcome was not formally measured. b Steinhausen et al. 1994 represents an increased sample size in the context of the same sample reported on by Hattig and Steinhausen 1987.

Prenatal or at 71 AEDbirth exposed

Hirano et al. 2004 (98)

67 AEDexposed

Prenatal

Sample size

Wide et al. 2002 (97)

Study

Time of enrollment

Effects on development or general mental ability

Table 1 Prospective Studies on the Behavioral Development of Children Exposed to Anticonvulsant Medications During Gestation (Continued)

114 Adams et al.

The Structural and Functional Teratology of Antiepileptic Medications

115

to 4% of the controls scored less than 90, and 62% exposed versus 69% of controls scored in the higher ranges. Thus, some improvement was seen with age. Hill et al. (22) also reported that mental development scores of children born to women on polytherapy were lower than those on monotherapies, a finding our prior discussion demonstrated as consistent for effects on structural outcomes. They also reported that Phenobarbital and phenytoin in combination acted to reduce scores. Prognostic risk factors for reduced developmental scores included being small-forgestational age at birth, exhibiting failure to thrive in the early months of life, having one or more major malformation, and having nine or more minor anomalies. Hill et al. suggested that these factors should be used to channel infants into programs for identifying learning related problems. This now classic study set the stage for later studies to examine functional outcomes as well as morphologic effects, and to examine effects on general mental ability as well as specific areas of weakness. Despite this early and compelling flag for the importance of monitoring childhood neurobehavioral outcome with respect to general as well as specific mental abilities, only a small subset of past and ongoing research on the teratogenicity of AEDs have included any measures of intellectual functioning. Nevertheless when cognitive abilities are measured, they are now done so alongside measures of performance in specific areas of processing as shown in Table 1. With respect to polytherapy effects on mental ability of the offspring, reduced scores in children born to epileptic mothers on AEDs in comparison to control non-exposed, nonepileptic women have now been reported in multiple prospective studies as shown in Table 1. These effects have been shown to be more pronounced than among “monotherapy” samples in which individual medications were collapsed into a single group and not separately examined. Consistent with effects on structural outcomes, Steinhausen et al. also demonstrated that infants born to non-medicated women with epilepsy or born of epileptic fathers were not at increased risks for reduced mental ability (50). In addition to the prospective studies of cognitive outcome that have been presented, there are also a few studies of the educational performance of AED-exposed children. These studies are based on retrospective ascertainment of subjects, nevertheless, the information about classroom performance is important to consider. As previously noted, Hill et al. reported that 35% of the infants born to AED-exposed epileptic women required “special education” as compared to 8% of the control group. Adab et al. conducted a retrospective survey in the United Kingdom of mothers with epilepsy and the “schooling” characteristics of their children (43). The questionnaire ascertained the need for additional educational supports among these children. Among the 156 non-exposed children born to women with epilepsy, 12.8% required additional supports. Among 188 AED-exposed children, 19.1% required educational supports. The authors also suggested that children exposed to valproate monotherapy were most likely to require extra attention with 43% of the 39 VPA monotherapy-exposed children receiving extra services, while only 2 of 61 carbamazepine-exposed children required additional educational intervention. While parental education characteristics were not taken into account, the difference across medications is striking. In a Scandinavian cohort, Dessens et al. looked retrospectively at children born to 172 AED-exposed mothers versus a group of 168 children matched on parental education and socioeconomic status. They reported that 9.5% of the AEDexposed versus 5.0% of the control children were referred during primary grades to special schools for children with learning problems (51). However, at secondary educational levels, 12% of the exposed and only 1% of the controls received services. These studies suggest the need to understand not only the influence of AED exposure upon cognitive test score performance, but also upon classroom performance, with particular attention to the

116

Adams et al.

relationships between the two. Both types of information will be necessary to identify the full spectrum of adverse outcomes associated with prenatal AED exposure. In light of the recognition that polytherapy as compared to monotherapy with certain drugs increases the rates of adverse structural as well as functional outcomes, the suggestions that certain drugs may be more harmful than others, and the demonstration that AEDs themselves are harmful during pregnancy, research during the last decade has shifted to efforts at identifying the differential risks associated with the use of each medication. The goal for the last decade has been to try to identify the least harmful drug for use in pregnant women with epilepsy. While most of this work focuses on structural and growth endpoints, several studies have examined the effects of monotherapy exposures on cognitive functioning during infancy or childhood.

THE TERATOGENIC EFFECTS OF MONOTHERAPY WITH PHENYTOIN, PHENOBARBITAL, CARBAMAZEPINE, AND VALPROIC ACID Table 2 presents the results of prospective studies that have examined conventional teratogenic endpoints. To be included in Table 2, studies must have utilized prospective designs. In order for data on an individual medication to be included, the study must have examined 19 or more monotherapy-exposed cases. There were occasions where studies may have compared across meds, but where only data on a subset of the drugs is presented in the table due to small sample size. Table 2 reveals that despite the fact that most of these studies have been conducted since 1990; the research is still marked by greater attention to major malformations than to minor anomalies or effects on growth. The absence of reported measures of minor malformations is particularly striking given that these malformations have been identified as important aspects of the “fetal anticonvulsant syndrome.” The research presented in Table 2 also illustrates the infrequency of addressing dose-related effects. One reason for this is that the clinical doses necessary for the control of seizures appear to be above the teratogenic threshold and are also often quite narrow in range. As can be seen in the table, valproic acid appears to represent an exception. Phenytoin (PHT) Despite coining the term “fetal hydantoin syndrome” to describe their results, Hanson and Smith actually reported upon the effects of exposure to combination therapies that included phenytoin (20). Many years of research have been necessary to tease apart the effects of the typical combination therapies from the effects of individual medications used as sole treatment. Table 2 summarizes findings from several studies aimed at examining the differential impact of exposure across monotherapy treatments or at examining effects in single drugs. As can be seen, Holmes et al. report lower major malformation rates than do Kaneko et al. or Vajda et al. with respect to both the PHT-exposed group and the controls (45,49,52). For each study, however, PHT exposure at least doubles the rate of major malformations. Differences across studies may be due to the exact malformations defined as major. As shown, Holmes et al. also reported on microcephaly as a separate category and on growth retardation and minor anomalies (45). Prenatal exposure to PHT markedly increased the occurrence of midface hypoplasia and hypoplasia of the fingers. In the Holmes et al. study, the rate of major malformations in 87 PHT monotherapy patients was 3.4% as compared to a rate of 8.6% in 93 infants exposed to 2 or more anticonvulsant drugs and 1.8% in 508 controls (45). Microcephaly was seen at a rate of 1.1% following

Phenobarbital

Phenytoin

98 unexposed with a seizure history

143 non-exposed with no seizure history 27 non-exposed with seizure disorders 105 non-exposed with no seizure history 508 non-exposed with no seizure history

132

55

64

46

26

508 non-exposed with no seizure history

87

# and nature of controls

4.7% vs. 1.8%

Not reported

10.5% vs. 4.3%

Not reported

No effect on head circumference No effect on head circumference 4.8% vs. 1.6%

No effect

1.1% vs. 1.6%

3.4% vs. 1.8%

9.1% vs. 3.1%

Rate of microcephaly or measurement of reduced head circumference

Rate of major malformations in exposed versus controls

1.6% vs. 1.2%

No effect on birth weight or length Increased birth weight and length Not reported

2.3 vs. 1.2

Growth effects

Not reported

Not present

Yes, with teratogenic effects seen at O200 mg/day Not present

Not examined

Dose-related effects

Not examined Midface hypoplasia: 15.2% vs. 3.8%; Finger hypoplasia: 9.6% vs. 2.3%

Not reported

Not reported

Not reported

Midface hypoplasia: 13.2% vs. 3.8%; Finger hypoplasia: 12.2% vs. 2.3% Not reported

Minor anomalies

(Continued)

Holmes et al. 2001 (45)

Gaily et al. 1990 (102)

Hiilesma et al. 1981 (100) Vajda et al. 2003 (101)

Kaneko et al. 1999 (99)

Holmes et al. 2001 (45)

Source

The Effects of Exposure to Monotherapeutic Treatment with Antiepileptic Drugs During Gestation Upon Structural and Growth Outcomes of Infants

# of exposed AED used in monotherapy monotherapy (MT) cases

Table 2

The Structural and Functional Teratology of Antiepileptic Medications 117

Carbamazepine

98 unexposed with a seizure history 143 non-exposed with no seizure history 27 non-exposed with seizure disorders

158

123

20

508 non-exposed with no seizure history

3.3 vs. 4.3%

Not reported

5.7% vs. 3.1%

5.2% vs. 1.8%

98 unexposed with 5.1% vs. 3.1% seizure disorders 105 unexposed with Not reported no seizure history

# and nature of controls

58

55

80

# of exposed AED used in monotherapy monotherapy (MT) cases

Rate of major malformations in exposed versus controls

Reduced head circumference No effect on head circumference

Not reported

(Reduced head circumference) 3.6% vs. 1.6%

Not reported

Rate of microcephaly or measurement of reduced head circumference

No effect on birth weight or length No effect on birth weight or length

Not reported

5.3% vs. 1.2%

Reduced birth weight

Not reported

Growth effects

Not examined Midface hypoplasia: 5.3% vs. 3.8%; Finger hypoplasia: 0% vs. 2.3% Not reported

Not present

Not present

Not reported

Not reported

Not present

Not examined

Not present

Dose-related effects

Not reported

Not reported

Minor anomalies

Kaneko et al. 1999 (49) Hiilesma et al. 1981 (100) Vajda et al. 2003 (101)

Kaneko et al. 1999 (49) Mastroiacovo et al. 1988 (61) Holmes et al. 2001 (45)

Source

Table 2 The Effects of Exposure to Monotherapeutic Treatment with Antiepileptic Drugs During Gestation Upon Structural and Growth Outcomes of Infants (Continued)

118 Adams et al.

Hospital population 10.7% vs. 1.6% norms

149

Not reported

No effect on head circumference

16.7% vs. 4.3%

27 non-exposed with seizure history

Not reported

Not reported

Not reported

No effect on birth weight or length

Not reported

Not reported

Not reported

Not reported Not reported No effect on head circumference Not reported for Not reported for Not reported MT cases MT cases only only

11.1% vs. 3.1%

97

98 unexposed with a seizure history

Not reported 105 non-exposed with no seizure history 5.3% vs. 2.75% 182 non-exposed women with seizure disorders Matalon et al. 2002 (103)

Kaneko et al. 1999 (49)

Not reported

Malformed infants were exposed to higher maternal levels than non-malformeda Malformed infants were exposed to higher maternal levels than non-malformedb Not reported

Wyszynski et al. 2005 (74)

Vajda et al. 2003 (101)

Gaily et al. 1990 (102)

Not reported

Note: This table only includes prospective studies with sample sizes greater than 19 exposed cases. When a study examined !19 cases for one drug and O 19 cases for another drug, only the data from the drug(s) with the larger sample are included. a Malformed infants were exposed to VPA levels of 77.8 mg/dl (from R1000 mg doses) while non-malformed were exposed to VPA levels of 46.8 mg/dl. b Malformed infants were exposed to mean VPA doses of 2081 mg/day while non-malformed were exposed to 1149 mg/day.

Valproic Acid

Meta-analysis of 16 studies: 797 MT cases 81

19

The Structural and Functional Teratology of Antiepileptic Medications 119

120

Adams et al.

PHT exposure, 3.3% following polydrug therapy, and 1.6% in controls: thus microcephaly was not increased by PHT monotherapy in this cohort. Growth retardation was present in 2.3% of the PHT-exposed versus 7.6% of the polytherapy cases and 1.2% of the controls. Midface hypoplasia occurred at a rate of 13.2% among the PHT-exposed, 12.7% in the polytherapy exposed group, and 3.8% of the controls. Hypoplasia of the fingers was seen in 12.2% of the PHT-exposed, 7.9% of the infants exposed to 2 or more AEDs, and 2.3% of the controls. In this cohort, the incidence of PHT monotherapy-induced effects was lower than the levels associated with monotherapy with Phenobarbital or carbamazepine with respect to major malformations and microcephaly. With respect to growth retardation in this study, PHT appeared to have less impact than carbamazepine. However, midface hypoplasia appeared to be somewhat higher among PHT and PB monotherapy-exposed infants than among carbamazepine infants, and hypoplasia of the fingers appeared higher in PHT monotherapy cases than either of the other drugs examined. Distal phalangeal hypoplasia or hypoplasia of the fingers has been noted as a primary sign of PHT teratogenicity in several studies, including studies using measurements from hand radiographs (46,48). The Kaneko et al. study described above also included 500 infants exposed to monotherapy, of which 132 were PHT-exposed (49). The occurrence of major malformations among the PHT cases was dose-related with most of the mothers having been exposed to 200 mg/day or more of PHT. The rate of major malformations was 9.1% among the PHT-exposed infants, and although values across medications did not differ statistically, this rate was higher than what was associated with Phenobarbital or carbamazepine, and lower than what was recorded for valproic acid or primidone. While it was the early reports of combination therapies including phenytoin as a component that generated concern about cognitive deficits, notably mental retardation (19,21,22), there have been few studies that have assessed the effects of PHT monotherapy upon cognitive outcome of the child. Nevertheless, several studies of phenytoin in combination particularly with Phenobarbital indicate effects on learning ability (51,53,54) and some have suggested that PHT monotherapy led to lesser risks than combination therapies (22,55). Ongoing research by Holmes, Adams, and coworkers is examining the effects of monotherapy with phenytoin, carbamazepine, or Phenobarbital. Our cohort has been drawn from the labor and delivery room surveillance sample that identified exposed cases examined in the Holmes et al. (45) study, as well as from referrals from obstetricians, pediatricians, and neurologists. While the subjects have either been identified in childhood or at the time of birth, a preliminary comparison of the demographic features and educational backgrounds of the parents as well as the children’s intellectual performance when identified at birth versus later in childhood suggest that the two sources of subjects are reasonably comparable. Nevertheless, our sample should be thought of as utilizing a combination of retrospective and prospective ascertainment while also being biased towards favorable outcomes. The parents of the children under study are primarily collegeeducated with a large number of individuals with advanced degrees. Merits of the study are that exposed children have been matched to non-exposed controls according to child sex and age, as well as maternal age and parental educational levels. Maternal evaluations of intelligence are also performed using the Wechsler Adult Intelligence Scale-III. Potential subjects have been excluded based on exposure to other teratogenic risks, occurrence of tonic-clonic seizures leading to loss of consciousness or hospitalization in the mother during pregnancy, maternal or paternal mental retardation, non-English speaking parents, or the presence of other neurological risks in the child (such as head injury or CNS infection). Children in this study have been (and continue to be) evaluated between 6 and

The Structural and Functional Teratology of Antiepileptic Medications

121

16 years of age through a dysmorphological exam, skeletal evaluation through hand radiographs, craniofacial measurements from digital photos and imaging, hearing and visual screening, and a neuropsychological battery. The neuropsychological battery has been designed to measure general mental ability (WISC-III) and specific cognitive domains including language-based processing, visuo-spatial and visuomotor abilities, fine motor speed and coordination, verbal and visual memory, and attention and planning abilities (including performance on auditory and visual continuous performance tasks). Although neuropsychological test performance is measured across the ages of 6 to 16 yr, each AED-exposed child’s performance is compared to that of a control matched on sex, age, parental education, and socioeconomic status. While our target is 40 exposed and 40 control children for each monotherapy-exposed drug group, these targets have not yet been reached. Data for the structural endpoints have not yet been reported for these specific PHT monotherapy-exposed subjects, but cognitive findings were reported when the sample included 17 matched pairs (56). At this sample size, no differences were found in general mental ability, verbal or non-verbal performance on the WISC-III between the PHT-exposed children and their matched controls. Indeed, mean full-scale IQ scores were 111.7 for the PHT-exposed children and 113.5 for the controls. These results suggests that phenytoin when used as monotherapy may be associated with low risks for adverse cognitive outcome, despite its prominent role as a harmful agent in a polytherapy context, and despite early reports that did not control for parental intelligence or socioeconomic status. Phenobarbital (PB) Phenobarbital was widely used during the earlier studies of anticonvulsant polytherapy and continues to be used today. With respect to structural outcomes, monotherapeutic use of PB during pregnancy has been linked to cardiac and craniofacial abnormalities (45), midface hypoplasia (27,42,45,57,58), digital and nail hypoplasia (59), reduced birth weight, and head circumference (60–62). As can be seen in Table 2, both Holmes et al. and Kaneko et al. report major malformation rates around 5% for PB monotherapy-exposed cases against 2–3% rates in controls (45,49). Holmes et al. also reported that the infants born to PB-exposed versus unexposed, non-epileptic women, or polytherapy-exposed women, had rates of microcephaly of 4.8%, 1.6%, and 3.3%, and rates of growth retardation of 1.6%, 1.2%, and 7.6% respectively (45). Midface hypoplasia was noted in 15.2% of the PB group, 3.8% of the unexposed controls, and 12.7% of the polytherapy exposed infants. Hypoplasia of the fingers was seen in 9.6%, 2.3%, and 7.9% respectively across the groups. These findings suggest that PB exposure places an infant at increased risks for most of the adverse structural outcomes with minor abnormalities as prominent hallmarks of the syndrome. Reduced cognitive ability in association with PB exposure has been reported in several studies (63–65). Reinisch et al. emphasized language processing as an area of specific weakness. Deficits in general mental ability, verbal, and non-verbal performance have been reported by Adams and coworkers (66) based on preliminary findings from our ongoing research. As described above, this study involves the examination of the teratogenic effects of PB monotherapy with respect to dysmorphology, additional examination of digital skeletal anomalies by x-ray of the hands and of craniofacial characteristics by imaging and digital photos, as well as a complete neuropsychological exam at a single age between 6 and 16 yr. With respect to cognitive functioning, preliminary data based on 23 PB-exposed and 23 matched controls has shown that on the WISC-III, the leading measure of mental ability in children, PB-exposed children scored

122

Adams et al.

11 points lower in general mental ability than controls matched on age, sex, and parental educational level (66). Despite this reduction, mean performance was well within the average range with a mean among the exposed children of 104.5 as compared to 115.8 in the matched control children. Multivariate analyses performed on processing related measures (language-based tests, visual-perceptual, verbal memory, non-verbal memory, drawing, and motor measures) revealed weaknesses in both language-based processing and visual-perceptual ability. It is important to point out that a neonatal barbiturate withdrawal syndrome is often manifest in infants exposed to PB throughout pregnancy (18). Although described by Hill et al. (21) in 1974, no studies of the effects of PB monotherapy on cognitive outcome are known that have examined the relationship between the occurrence or magnitude of early withdrawal, its treatment, or its association with later cognitive performance (21).

Carbamazepine (CBZ) Carbamazepine has been associated with increased risks for neural tube defects, a pattern of minor malformations involving craniofacial structures, digital and nail hypoplasia, as well as reduced birthweight and head circumference, increased risks for developmental delay, and reduced cognitive performance (34,67,68). As can be seen in Table 2, reported rates of major malformations are generally around 5% but effects on other endpoints have been inconsistent. Holmes et al. did not find significant increases in either midface or digit hypoplasia (45). Adverse effects on cognitive outcome have also been inconsistent with normal intelligence reported by Gaily et al. in a cohort of 86 carbamazepine monotherapy exposed children compared to educationally and socioeconomically matched controls (69). The ongoing collaborative research of Holmes and Adams described above also addresses the effects of monotherapy with CBZ during pregnancy in the context of the same design characteristics. To date, we have evaluated 35 matched pairs of monotherapyexposed and control children (70). The WISC-III performance of these children differed from controls with respect to general mental ability, verbal IQ, and performance IQ. Nevertheless, mean performance of the CBZ-exposed children was within the average range. Analyses of performance in specific domains suggested more consistent effects upon language-based processing. While carbamazepine is established as a teratogen that increases risks for neural tube defects as well as other major malformations, effects on other endpoints are unclear. Thus, a great deal more research is needed to determine whether carbamazepine presents additional risks beyond increased major malformations.

Valproic Acid (VPA) Valproic acid medications (Depakotee, Valproatee, Valreleasee) are widely used for the treatment of convulsive disorders, neuropathic pain, migraine headaches, and bipolar and schizo-affective disorders. First trimester VPA exposure for the treatment of epilepsy has been associated with neural, craniofacial, cardiovascular, skeletal, and urogenital defects (71,72), as well as growth retardation, mental retardation, and reduced intelligence (42,73). This profile is similar to that reported for polytherapy with other anticonvulsant medications and also includes minor anomalies such as hypertelorism, midface hypoplasia, low set ears, and digital and nail hypoplasia [(27,42,73,45); also, Consensus Guidelines for Pregnant Women with Epilepsy (15)].

The Structural and Functional Teratology of Antiepileptic Medications

123

Despite certain similarities across anticonvulsants, however, the developing embryo appears to be more sensitive to VPA than to the older medications that have been most widely examined in polytherapy studies. Research by Wyszynski et al. (74) on registrants in the National Antiepileptic Drug Pregnancy Registry (AED Registry) has focused on the frequency of major birth defects in babies whose mothers were exposed to valproate (VPA) in comparison to other anticonvulsant drug exposures. Participants were prospectively enrolled before the status of their fetuses was known (enrollment usually taking place prior to 16 weeks gestational age), and the study ascertained 149 completed pregnancies exposed to VPA as monotherapy. Wyszynski et al. (74) reported the prevalence of major birth defects in babies following exposure to VPA to be 10.7%, as compared to 2.9% in the control group exposed to other anticonvulsant monotherapies, and 1.6% in the unexposed group drawn from hospital records. Other research has shown that neural tube defects occur in 1–2% of infants exposed to VPA in utero (75), a rate that is 10–20 times the general population rate (76,77). In addition to the increased rate of major malformations, there appear to be a far greater number of VPA-exposed pregnancies that result in children with developmental delays and/or intellectual impairments (78,79). Estimates for these outcomes range from 20–71% (78,80). Adab and colleagues (81) have reported decreased verbal IQs compared to a group of unexposed siblings and to children exposed to other AEDs. These deficits were more often seen in children with dysmorphic features. Increased social, language, and learning disabilities have also been reported (82). Accurate estimation of the number of VPA-exposed children with compromised cognitive and/or social development is difficult, however, because studies typically have involved retrospective (after outcome is known) identification of participants which may have resulted in biased samples, and have not adequately matched exposed and control cases on parental educational and demographic characteristics. The studies may therefore over-represent children with problems, and misinterpret the presence of social deficits under conditions that did not take into account the influence of reduced intelligence. Recent attention has been focused on the possible increased risks for autism spectrum disorders (ASD) among prenatal VPA-exposed children (8,83–86). Rodier highlighted a shared neuropathology among individuals with autism and animals and humans exposed to VPA, summarized case reports of VPA-exposed children who had ASD, and then presented an elegant teratological model that explained and logically accounted for various symptoms of autism. This compelling model and its promising new avenue for investigating a mechanistic etiology of ASD have drawn considerable attention. Nevertheless, the empirical relationship between VPA exposure and autism is quite tenuous and in need of careful, systematic examination. While reports of an association between prenatal VPA exposure and autism have been published, all have been based either on retrospective samples or case reports, and none have controlled for the level of mental retardation in the children (87, 88). This is problematic given that the diagnosis of autism and ASD is far more difficult and far less reliable among moderately to severely mentally retarded children (89), and prevalence rates are estimated to be between 38% to well above 50% among general samples of these low-functioning individuals (90–92). In summary, monotherapeutic treatment with each of the historically prominent AEDs has been associated with teratogenic risks. PB and VPA have been associated with adverse cognitive outcomes across studies. VPA and CBZ have been associated with elevated risks for neural tube defects. The magnitude of reduced mental ability appears less severe for PB but potentially quite severe for VPA. The cognitive effects associated with PHT and CBZ have been demonstrated less consistently following monotherapy

124

Adams et al.

treatment, but when demonstrated, effects have been on the milder end of the continuum. Nevertheless, mild effects on general mental ability may manifest potentially as learning disabilities or specific areas of weakness. Such effects can be quite compromising for the individual child and family. A great deal more research must be done to rank order the safety of these medications during pregnancy, and it is essential that the research examine effects on general and specific mental abilities as well as structural endpoints. At present, however, based on limited information on cognitive outcome, more flags have been raised regarding the increased risks associated with VPA exposure than for the other medications examined in this review.

INDIVIDUALIZED PREDICTORS OF SUSCEPTIBILITY TO ADVERSE OUTCOME THAT CAN BE ASSESSED DURING PREGNANCY OR AT BIRTH While it is of critical importance that the comprehensive teratogenicity of the antiepileptic medications be understood across individual drugs, it is equally important that individual risk factors of the mother and fetus be identified. Research suggests that some women and fetuses may be more susceptible to the teratogenic effects of AEDs on the basis of specific genetic features relevant to drug metabolism. Such knowledge could be quite valuable in the counseling of the individual woman considering pregnancy or the pregnant woman on medication who needs to understand her personal risk status. Likewise, it would be quite valuable if individualized risk factors present at birth that predict risks for long term cognitive outcome could be identified. Such knowledge would facilitate referral of infants into early intervention programs that could facilitate their outcome. The status of knowledge with respect to each of these arenas is discussed below.

Genetic Markers of Increased Susceptibility Research on the molecular basis of teratogenic action has proliferated in recent years. Finnell et al. (4) have presented a review of this research and that is the primary source for the following description of this knowledge. Development from zygote to infant involves the complex orchestration of precisely regulated developmental events. The occurrence of these events involves the exact timing of the production, release, reception, and response to chemical signals that vary according to highly controlled molecular and biochemical “dose” and dose-dependently triggered response cascades. Teratogens interfere with these events and cause alterations in development. Anticonvulsant medications or metabolites of the medications interfere with certain signaling pathways relevant to embryonic and fetal development (4). In animal models, certain genes have been identified that alter drug metabolism and increase the susceptibility to the teratogenic effects of phenytoin, Phenobarbital, and valproic acid. Certain inbred strains of mice have been identified that vary according to their teratogenic response to valproic acid from highly-resistant to highly sensitive, particularly with respect to the occurrence of neural tube defects. It is believed that a genetic component is also important in the induction of neural tube defects in women who take valproic acid during pregnancy. This putative relationship demands further exploration given its potential relevance to the identification, counseling, and treatment of women with seizure disorders and other disorders for which anticonvulsant medications are often administered.

The Structural and Functional Teratology of Antiepileptic Medications

125

Infant Dysmorphological Characteristics as Predictors of Later Neuropsychological Functioning Hanson et al. first suggested that AED-exposed infants with a syndrome of major and minor malformations might be at increased risks for reduced mental ability (19). Hill et al. identified the presence of minor anomalies as among five risk factors that should be used to channel infants into programs for identifying learning related problems (22). Nevertheless, formal studies of this association have been more recent. Mawer et al. reported a predictive association between dysmorphic features and developmental delay among VPA-exposed infants (93). Ornoy and Cohen suggested a similar correlation among a group of CBZexposed infants (68). Holmes et al. examined the relationships between structural outcomes and cognitive ability in 80 children exposed to AEDs during prenatal development (94). They found no predictive association between major malformations (in this cohort, only one child had a major malformation of the CNS) and later mental ability. Not surprisingly, a correlation was found between the presence of microcephaly and reduced mental ability. When children with microcephaly were excluded from an analysis of the association between midface and/or digit hypoplasia and cognitive outcome, these minor malformations were found to be predictive. This correlation between structural defects that can be measured at birth and later general mental ability is potentially of great importance given that midface and digit hypoplasia are the two most common manifestations of the fetal anticonvulsant syndrome. Holmes et al. recommend that the presence of these features warrants referral of the infant into early intervention programs (94).

FUTURE RESEARCH NEEDS It is evident that more research needs to be done to establish the effects of monotherapy with anticonvulsants during pregnancy upon the structural and functional outcomes of the children. This is true not only with respect to the specific AEDs reviewed in this paper but to the new regime of anticonvulsant medications as well. Essential to this research is attention to effects on general mental ability as well as to abilities in specific domains of processing. Moreover, research needs to be undertaken on the influence of exposure upon socio-emotional and psychiatric characteristics of the children. This research is critical in order to fully characterize the relationships between exposure and incidence of reduced mental ability, learning disabilities, social dysfunction, and psychiatric disturbance. It is also important that the academic relevance of mild reductions in ability be examined through research looking at children’s classroom performance as well as performance on standardized achievement tests. Further, given that anticonvulsant medications are known to disrupt thyroid and growth hormone functioning in women and their exposed offspring (18), these effects need to be examined with respect to their contributions to effects on growth and cognition following prenatal exposures. Likewise, a better understanding of the impact of neonatal withdrawal to certain medications upon later childhood functioning is needed. Finally, predictors of individual risks need to be better understood. Research must examine the relationships between individual genetic susceptibilities and cognitive as well as structural outcomes of children prenatally exposed to AED medications. Also, it is paramount that investigators pursue a greater understanding of the infant characteristics at birth that are associated with increased risks for cognitive as well as social or psychiatric dysfunction, and that these data be used to facilitate the outcomes of high risk infants through early referral for targeted intervention. Some will read this proposed research agenda and say “dream on” while others will see it as central to the dream of providing all

126

Adams et al.

families and all children with an equal opportunity for success. The latter group may support undertaking this research at any cost. The former group may be influenced by a few fiscal facts. Finnell et al. highlight research indicating that in the state of California alone, the estimated lifetime costs for children with spina bifida is more than $58 million, for conotruncal heart defects, more than $287 million, and for orofacial clefts, $86 million (4). They conservatively suggested that multiplying those figures by a factor of 10 would provide an estimate for this subset of malformations for the United States as a whole. The Centers for Disease Control estimated that the direct and indirect lifetime economic costs in the United States for persons born in 2000 with mental retardation would be $51.2 billion (95) (valued in 2003 dollars). Across the named endpoints, approximately $55 billion dollars will be spent over a lifetime of care. This alarming figure does not include estimates for providing public school services to children with learning disabilities. These estimates establish the paramount importance of conducting more research on the effects of teratogens on childhood outcome and the many variables that influence susceptibility to adverse outcome following exposure. It appears that an ounce of prevention would be worth a pound of cure.

REFERENCES 1. Wilson J. Environment and Birth Defects. New York: Academic Press, 1973. 2. Adams J, Barone S, LaMantia A, et al. Workshop to identify critical windows of exposure for children’s health: neurobehavioral workshop summary. Environ Health Perspect 2000; 108:535–544. 3. Chambers CD, Braddock SR, Briggs GG, Einarson A, Johnson YR, Miller RK, Polifka JE, Robinson LK, Stepanuk K, Jones KL. Postmarketing surveillance for human teratogenicity: a model approach. Teratology 2001; 64:252–261. 4. Finnell RH, Gelineau van Waes J, Eudy JD, Rosenquist TH. Molecular basis of environmentally induced birth defects. Ann Rev Pharmacol Toxicol 2002; 42:181–208. 5. Rice D, Barone S, Jr. Critical periods of vulnerability for the developing nervous system: evidence from humans and animal models. Environ Health Perspect 2000; 108:511–533. 6. CDC. Epilepsy: One of the Nation’s Most Common Disabling Neurological Conditions [website]. National Center for Chronic Disease Prevention and Health Promotion. Vol. 2005, 2004. 7. Janulewicz PA, Gavin JAL, Meyer E, Adams J. Pattern of psychosocial and attentional characteristics of children exposed to Phenobarbital monotherapy in utero. Neurotoxicol Teratol 2001; 23:294. 8. Rodier PM. The early origins of autism. Sci Am 2000; 282:56–63. 9. Rodier PM, Bryson SE, Welch JP. Minor malformations and physical measurements in autism: data from Nova Scotia. Teratology 1997; 55:319–325. 10. Schantz SL, Widholm JJ. Cognitive effects of endocrine-disrupting chemicals in animals. Environ Health Perspect 2001; 109:1197–1206. 11. Weiss B. Sexually dimorphic nonreproductive behaviors as indicators of endocrine disruption. Environ Health Perspect 2002; 110:387–391. 12. Vorhees CV. Principles of Behavioral Teratology. New York: Plenum Press, 1986. 13. Weiss B. Vulnerability of children and the developing brain to neurotoxic hazards. Environ Health Perspect 2000; 108:375–381. 14. Adams J. Neural and Behavioral Pathology Following Prenatal Exposure to Retinoids. New York: Marcel Dekker, 1993. 15. Delgado-Escueta AV, Janz D. Consensus guidelines: preconception counseling, management, and care of the pregnant woman with epilepsy. Neurology 1992; 42:149–160.

The Structural and Functional Teratology of Antiepileptic Medications

127

16. Fairgrieve SD, Jackson M, Jonas P, et al. Population based, prospective study of the care of women with epilepsy in pregnancy. BMJ 2000; 321:674–675. 17. Yerby MS. Risks of pregnancy in women with epilepsy. Epilepsia 1992; 33:S23–S27. 18. Kaneko S. Antiepileptic drug therapy and reproductive consequences: functional and morphological effects. Reprod Toxicol 1991; 5:179–198. 19. Hanson JW, Myrianthopoulos NC, Harvey MA, Smith DW. Risks to the offspring of women treated with hydantoin anticonvulsants, with emphasis to the fetal hydantoin syndrome. J Pediatr 1976; 89:662–668. 20. Hanson JW, Smith DW. The fetal hydantoin syndrome. J Pediatr 1975; 87:285–290. 21. Hill RM, Verniaud WM, Horning MG, et al. Infants exposed in utero to antiepileptic drugs. A prospective study.. Am J Dis Child 1974; 127:645–653. 22. Hill RM, Verniaud WM, Rettig GM, Tennyson LM, Craig JP. Relationship Between Antiepileptic Drug Exposure of the Infant and Developmental Potential. New York: Raven Press, 1982. 23. Meadow SR. Anticonvulsant drugs and congenital abnormalities. Lancet 1968; 2:1296. 24. Speidel BD, Meadow SR. Maternal epilepsy and abnormalities of the fetus and newborn. Lancet 1972; 2:839–843. 25. Gaily E, Granstrom ML, Hiilesmaa V, Bardy A. Minor anomalies in offspring of epileptic mothers. J Pediatr 1988; 112:520–529. 26. Gaily E, Kantola-Sorsa E, Granstro¨m ML. Intelligence of children of epileptic mothers. J Pediat 1988; 113:677–684. 27. Adams J, Vorhees CV, Middaugh LD. Developmental neurotoxicity of anticonvulsants: human and animal evidence on Phenytoin. Neurotoxicol Teratol 1990; 12:203–214. 28. Middaugh LD. Phenobarbital during pregnancy in mouse and man. Neurotoxicology 1986; 7:287–301. 29. Finnell RH, Abbott LC, Taylor SM. The fetal hydantoin syndrome: answers from a mouse model. Reprod Toxicol 1989; 3:127–133. 30. Finnell RH, Chernoff GF. Mouse fetal hydantoin syndrome: effects of maternal seizures. Epilepsia 1982; 23:423–429. 31. Phillips NK, Lockard JS. A gestational monkey model: effects of phenytoin versus seizures on neonatal outcome. Epilepsia 1985; 26:697–703. 32. Janz D. Antiepileptic drugs and pregnancy: altered utilization patterns and teratogenesis. Epilepsia 1982; 23:S53–S63. 33. Majewski F, Steger M, Richter B, et al. The teratogenicity of hydantoins and barbiturates in humans with consideration on the etiology of malformations and cerebral disturbances in the children of epileptic parents. Biolog Res Pregnancy 1981; 2:37–45. 34. Granstrom ML, Hiilesmaa V. Malformations and minor anomalies in the children of epileptic mothers: preliminary results of the prospective Helsinski study. In: Janz D, Dam M, Richens A, eds. Epilepsy, Pregnancy, and the Child. New York: Raven Press, 1982:303–307. 35. Miyakoshi M, Seino M. In: Sato T, Shinagawa S, eds. Malformations in children born to mothers with epilepsy. In: Antiepileptic Drugs and Pregnancy. Amsterdam: Excerpta Medica, 1984. 36. Seino M, Miyakoshi M. Teratogenic risks of antiepileptic drugs in respect to the type of epilepsy. Folia Psychiatr Neurol Jpn 1979; 33:379–385. 37. Kaneko S, Fukushima Y, Sato T, et al. Teratogenicity of antiepileptic drugs: a prospective study. Jap Psychiatric Neurol 1986; 40:447–450. 38. Bossi L. Fetal effects of anticonvulsants. In: Morselli P, Pippenger C, Penry J, eds. Antiepileptic Drug Therapy in Pediatrics. New York: Raven Press, 1983:37–64. 39. Koch S, Losche G, Jager-Roman E, et al. Major and minor birth malformations and antiepileptic drugs. Neurology 1992; 42:83–88. 40. Holmes L, Adams J, Ryan L. Physical and behavioral consequences of in utero exposure to anticonvulsant monotherapy. Paper presented at the European Teratology Society, Hungary, 2001.

128

Adams et al.

41. Nulman I, Scolnik D, Chitayat D, et al. Findings in children exposed in utero to phenytoin and carbamazepine monotherapy: independent effects of epilepsy and medications. Am J Med Genet 1997; 68:18–24. 42. Dansky LV, Finnell RH. Parental epilepsy, anticonvulsant drugs, and reproductive outcome: epidemiologic and experimental findings spanning three decades; 2: human studies. Reprod Toxicol 1991; 5:301–335. 43. Adab N, Jacoby A, Smith D, Chadwick D. Additional educational needs in children born to mothers with epilepsy. J Neurol Neurosurg Psychiatry 2001; 70:15–21. 44. Beghi E, Annegers JF. Pregnancy registries in epilepsy. Epilepsia 2001; 42:1422–1425. 45. Holmes LB, Harvey EA, Coull BA, et al. The teratogenicity of anticonvulsant drugs. N Engl J Med 2001; 344:1132–1138. 46. Gaily E. Distal phalangeal hypoplasia in children with prenatal phenytoin exposure: results of a controlled anthropometric study. Amer J Med Genet 1990; 35:574–578. 47. Holmes LB, Wyszynski DF, Lieberman E. The AED (antiepileptic drug) pregnancy registry: a 6-years experience. Arch Neurol 2004; 61:673–678. 48. Kelly TE. Teratogenicity of anticonvulsant drugs: iii: radiographic hand analysis of children exposed in utero to diphenylhydantoin. Amer J Med Genet 1984; 19:445–450. 49. Kaneko S, Battino D, Andermann E, et al. Congenital malformations due to antiepileptic drugs. Epilepsy Res 1999; 33:145–158. 50. Steinhausen HC, Losche G, Koch S, Helge H. The psychological development of children of epileptic parents I. Study design and comparative findings. Acta Paediatr 1994; 83:955–960. 51. Dessens AB, Cohen-Kettenis PT, Mellenbergh GJ, et al. Association of prenatal phenobarbital and phenytoin exposure with small head size at birth and with learning problems. Acta Paediatr 2000; 89:533–541. 52. Vajda FOB, O’Brien TJ, Hitchcock A, Graham J. The Australian registry of anti-epileptic drugs in pregnancy: experience after 30 months. J Clin Neurosci 2003; 10:543–549. 53. Jager-Roman ER, Koch S, Gopfert-Geyer I, Jacob S, Helge H. Somatic Parameters, Diseases, and Psychomotor Development in the Offspring of Epileptic Parents. New York: Raven Press, 1982. 54. Losche G, Steinhausen HC, Koch S, Helge H. The psychological development of children of epileptic parents. II. The differential impact of intrauterine exposure to anticonvulsant drugs and further influential factors. Acta Paediatr 1994; 83:961–966. 55. Vanoverloop D, Schnell RR, Harvey EA, Holmes LB. The effects of prenatal exposure to phenytoin and other anticonvulsants on intellectual function at 4 to 8 years of age. Neurotoxicol Teratol 1992; 14:329–335. 56. Adams J, Harvey EA, Holmes LB. Cognitive deficits following gestational monotherapy with phenobarbital and carbamazepine. Program for the Annual Meeting of the Neurobehavioral Teratology Society, Abstract 55, 2000. 57. Holmes LB, Wyszynski D, Mittendorf R. Evidence for an increased risk of birth defects in the offspring of women exposed to valproate during pregnancy: findings from the AED Pregnancy Registry. Am J Obstet Gynecol 2002; 187:S137. 58. Hansen DK, Holson RR. Developmental neurotoxicity of antiepileptic drugs. In: Slikker WS, Chang LW, eds. Developmental Neurotoxicology. San Diego: Academic Press, 1998:643– 660. 59. Holmes LB. The teratogenicity of anticonvulsant drugs: a progress report. J Med Genet 2002; 39:245–247. 60. Majewski F, Steger M. Letter to the editors. Fetal Head Growth Retardation associated with maternal phenobarbitone/primidone and/or phenytoin therapy. Europ J Pediatr 1984; 141:188–189. 61. Mastroiacovo P, Bertollini R, Licata D. Fetal growth in the offspring of epileptic women: results of an Italian multicentric cohort study. Acta Neurol Scand 1988; 78:110–114. 62. van der Pol MC, Hadders-Algra M, Huisjes HJ, Touwen BC. Antiepileptic medication in pregnancy: late effects on the children’s central nervous system development. Am J Obstet Gynecol 1991; 164:121–128.

The Structural and Functional Teratology of Antiepileptic Medications

129

63. Gaily E, Kantola-Sorsa E, Granstrom ML. Specific cognitive dysfunction in children with epileptic mothers. Dev Med Child Neurol 1990; 32:403–414. 64. Holmes LB, Rosenberger PB, Harvey EA, et al. Intelligence and physical features of children of women with epilepsy. Teratology 2000; 61:196–202. 65. Reinisch JM, Sanders SA, Mortensen EL, Rubin DB. In utero exposure to phenobarbital and intelligence deficits in adult men. J Am Med Associ 1995; 274:1518–1525. 66. Adams J, Holmes LB, Janulewicz P. The adverse effect profile of neurobehavioral teratogens: phenobarbital. Neurotoxicol Teratol 2004; 26:507. 67. Jones KL, Lacro RV, Johnson KA, Adams J. Pattern of malformations in the children of women treated with carbamazepine during pregnancy. N Engl J Med 1989; 320:1661–1666. 68. Ornoy A, Cohen E. Outcome of children born to epileptic mothers treated with carbamazepine during pregnancy. Arch Dis Child 1996; 75:517–520. 69. Gaily E, Kantola-Sorsa E, Hiilesmaa V, et al. Normal intelligence in children with prenatal exposure to carbamazepine. Neurology 2004; 62:28–32. 70. Janulewicz P, Adams J, Dhillon R, Holmes LB. Developmental outcome of children prenatally exposed to carbamazepine. Neurotoxicol Teratol. In Press. 71. Ja¨ger-Roman E, Deichl A, Jakob S, et al. Fetal growth, major malformations, and minor anomalies in infants born to women receiving valproic acid. J Pediatr 1986; 108:997–1004. 72. Lindhout D, Schmidt D. In-utero exposure to valproate and neural tube defects [letter]. Lancet 1986; 1:1392–1393. 73. Hansen DK, Holson RR. Developmental Neurotoxicity of Antiepileptic Drugs. New York: Academic Press, 1998. 74. Wyszynski D, Nambisan M, Surve T, Alsdorf RM, Smith CR, Holmes LB. Increased rate of major malformation in offspring exposed to valproate during pregnancy. Neurology 2005; 64:961–965. 75. Robert E, Wells PG, Rosa F. Maternal valproic acid and congenital neural tube defects [letter]. Teratogenicity of Isotretinoin. Lancet 1982; 2:937. 76. Bjerkedal T, Czeizel A, Goujard J, et al. Valproic acid and spina bifida [letter]. Lancet 1982; 2:1096. 77. CDC. Recommendations for the use of folic acid to reduce the number of cases of spina bifida and other neural tube defects. MMWR 1992; 4:1–7. 78. Kozma C. Valproic acid embryopathy: report of two siblings with further expansion of the phenotypic abnormalities and a review of the literature. Am J Med Genet 2001; 98:168–175. 79. Ardinger H, Atkin J, Blackston R, Elsas L. Verification of the fetal valproate syndrome phenotype. Am J Med Genet 1988; 29:171–185. 80. Ardinger HH, Atkin JF, Blackston RD, et al. Verification of the fetal valproate syndrome phenotype. Am J Med Genet 1988; 29:171–185. 81. Adab N, Kini U, Vinten J, et al. The longer term outcome of children born to mothers with epilepsy. J Neurol Neurosurg Psychiat 2004; 75:1575–1583. 82. Adab N, Jacoby A, Smith D, Chadwick D. Additional educational needs in children born to mothers with epilepsy. J Neurol Neurosurg Psychiat 2001; 70:15–21. 83. Rodier PM, Ingram JL, Tisdale B, et al. Embryological origin for autism: developmental anomalies of the cranial nerve motor nuclei. Journal Comp Neurol 1996; 370:247–261. 84. Rodier PM, Ingram JL, Tisdale B, Croog VJ. Linking etiologies in humans and animal models: studies of autism. Reprod Toxicol 1997; 11:417–422. 85. Ingram JL, Peckham SM, Tisdale B, Rodier PM. Prenatal exposure of rats to valproic acid reproduces the cerebellar anomalies associated with autism. Neurotoxicol Teratol 2000; 22:319–324. 86. Williams G, King J, Cunningham M, et al. Fetal valproate syndrome and autism: additional evidence of an association. Dev Med Child Neurol 2001; 43:847. 87. Moore SJ, Turnpenny P, Quinn A, et al. A clinical study of 57 children with fetal anticonvulsant syndromes. J Med Genet 2000; 37:489–497. 88. Bescoby-Chambers N, Forster P, Bates G. Fetal valproate syndrome and autism. Dev Med Child Neurol 2001; 43:847–852.

130

Adams et al.

89. Lord C, Risi S. Frameworks and methods in diagnosing autism spectrum disorders. Mental Retard Develop Disabilities Res Rev 1998; 4:90–96. 90. Gillberg C, Persson E, Grufman M, Themne´r U. Psychiatric disorders in mildly and severely mentally retarded urban children and adolescents: epidemiological aspects. Brit J Psychiat 1986; 149:68–74. 91. Shah A, Holmes N, Wing L. Prevalence of autism and related conditions in adults in a mental handicap hospital. Appl Res Mental Retard 1982; 3:303–317. 92. Lord C, Rutter M, LeCouteur A. Autism Diagnostic Interview–Revised: a revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders. J Autism Develop Disorders 1994; 24:659–668. 93. Mawer G, Clayton-Smith J, Coyle H, Kini U. Outcome of pregnancy in women attending an outpatient epilepsy clinic: adverse features associated with higher doses of sodium valproate. Seizure 2002; 11:512–518. 94. Holmes LB, Coull BA, Dorfman J, Rosenberger PB. The correlation of deficits in IQ with midface and digit hypoplasia in children exposed in utero to anticonvulsant drugs. J Pediatrics 2005; 146:118–122. 95. Honeycutt A, Dunlap L, Chen H, al Homsi G. Economic costs associated with mental retardation cerebral palsy, hearing loss, and vision impairment—United States, 2003. MMWR 2004; 53:57–59. 96. Leavitt AM, Yerby MS, Robinson N, et al. Epilepsy in pregnancy: developmental outcome of offspring at 12 months. Neurology 1992; 42:141–143. 97. Wide K, Henning E, Tomson T, Winbladh B. Psychomotor development in preschool children exposed to antiepileptic drugs in utero. Acta Paediatr 2002; 91:409–414. 98. Hirano T, Fujioka K, Okada M, et al. Physical and psychomotor development in the offspring born to mothers with epilepsy. Epilepsia 2004; 45:53–57. 99. Battino D, Kaneko S, Andermann E, et al. Intrauterine growth in the offspring of epileptic women: a prospective multicenter study. Epilepsy Res 1999; 36:53–60. 100. Hiilesma V, Teramo K, Granstrom ML, Bardy AH. Fetal head growth retardation associated with maternal antiepileptic drugs. Lancet 1981; II:165–167. 101. Vajda F, O’Brien TJ, Hitchcock A, Graham J, Lander C. The Australian registry of antiepileptic drugs in pregnancy: experience after 30 months. J Clin Neurosci 2003; 10:543–549. 102. Gaily EK, Granstrom ML, Hiilesmaa VK, Bardy AH. Head circumference in children of epileptic mothers: contributions of drug exposure and genetic background. Epilepsy Res 1990; 5:217–222. 103. Matalon S, Schechtman S, Goldzweig G, Ornoy A. The teratogenic effect of carbamazepine: a meta-analysis of 1255 exposures. Reprod Toxicol 2002; 16:9–17.

7 Chemotherapy Agents for Treatment of Acute Lymphoblastic Leukemia Christine Mrakotsky and Deborah P. Waber Department of Psychiatry, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A.

INTRODUCTION Acute Lymphoblastic Leukemia (ALL) accounts for 80% of leukemia in children and is the most common type of childhood cancer. Although it is a rare disease, 2400 children are newly diagnosed each year in the U.S., with peak incidence during the preschool period. Significant improvements in treatment have led to markedly improved survival over the past four decades. Just 35 years ago, a diagnosis of ALL was almost always fatal. With treatment advances, long-term event-free survival rates now approach and sometimes exceed 80% (1,2). Central to these dramatic successes was the introduction of central nervous system (CNS) therapy in the early 1970s, which serves to protect children from relapse in the central nervous system. The blood–brain barrier is permeable for leukemia cells but not for (lower dose) systemic chemotherapy agents, thus creating a potential sanctuary for malignant cells. Treatment regimens, therefore, typically include prophylactic intrathecal administration of methotrexate alone, or triple intrathecal therapy (methotrexate, vincristine, and hydrocortisone), with agents directly injected into the spinal fluid. When CNS prophylaxis was first introduced, cranial irradiation therapy (CRT) was the primary modality of treatment, but it soon became apparent that it was associated with adverse cognitive sequelae, which were quite significant in some children. As more has been learned about the disease, however, investigators have been able to eliminate CRT without compromising efficacy for large subgroups of patients. Although the adverse effects of intensive treatment involving high doses of CRT on both physical and cognitive development have been long recognized (3–8), the sequelae of multi-agent chemotherapy protocols without CRT are less well understood. With the improved efficiency of medical treatment and welcome growth in the number of long-term survivors, quality of life, health status, and cognitive outcome become more central considerations (9). As the majority of patients with ALL are now treated without CRT, evaluation of the potential for long-term cognitive effects in children treated on these “chemotherapy only” protocols is of increasing importance. 131

132

Mrakotsky and Waber

This chapter will address current issues of leukemia therapy, focusing on chemotherapy-only protocols, and associated neurotoxicity. An overview of disease factors, past and current treatment protocols, and their efficacy and neurotoxicity will be provided. The potential impact of systemic and intrathecal (administered directly in the spinal column) chemotherapy on the CNS and associated development of cognitive functions will be reviewed.

PATHOPHYSIOLOGY AND PREVALENCE OF ALL ALL evolves from uncontrolled growth and proliferation of immature leukocytes into the bone marrow and lymphoids (malignant lymphoblasts). These malignant cells originate most often from B-cells early in development (precursors), and less frequently from T-cells. Common symptoms at onset can include bone and joint pain, intermittent fevers, fatigue, pallor, frequent infections as well as swollen lymph nodes, liver, and spleen, due to these cells’ infiltration of healthy bone marrow and lymphoids. Although less common at diagnosis, CNS involvement can occur, with lymphoblasts invading the cerebrospinal fluid; headaches, nausea and vomiting are associated clinical presentations (10). Although ALL is a rare disease, with an occurrence of one in every 29,000 children in the United States (11), incidence rates vary largely with age. The age of onset peaks between 2 and 4 years, with current rates of 6 in 100,000 per year, decreasing to a rate of 1.6 in 100,000 for 10 to 14 year-olds (12). Because of the early age of onset, aggressive treatments must be undertaken at a time when the CNS is developing and may be especially vulnerable to insult. Gender distributions are relatively even, with a slightly higher incidence for boys than girls (ratio 1.3:1). Differences related to ethnicity are greater. Caucasian children are on average twice as likely to be affected as African American children, with a 3-fold higher incidence in children younger than 5 years of age (11). Incidence rates for Hispanic children are equal to or slightly higher than that of Caucasians (12).

EVOLUTION OF TREATMENT PROTOCOLS As indicated above, the marked increase in survival rates over the past four decades for children newly diagnosed with ALL is one of the great successes of cancer therapy. Until the late 1960s, this disease was almost always fatal, with very few children surviving more than 5 years. Death frequently followed relapse in the CNS. With the introduction of CNS therapies, survival rates increased to 40% in the 1970s, and with further improvement of treatment to 65–70% in the 1980s. Long-term survival rates can now exceed 80%. This success has been achieved in part by tailoring treatment protocols specifically to riskgroups based on prognostic factors at diagnosis, including leukocyte counts, age, and disease phenotype (13). The majority of children in the United States and Europe are treated for ALL on multi-center clinical trials that are broadly comparable regarding their key elements. The initial phase of treatment focuses on remission induction, whereby blood counts are restored and the disease is brought into complete remission. This typically takes about one month, with an induction regimen comprised of vincristine, corticosteroids, and asparaginase. Once remission has been achieved, an extended period of therapy is necessary to prevent relapse. Subsequent consolidation or intensification therapy is a period of intensified treatment that begins with CNS-directed intervention, including multi-agent intrathecal chemotherapy and, for more high-risk patients, adjunct CRT or

Treatment of Acute Lymphoblastic Leukemia

133

additional doxorubicin. Drug regimens most often include high-dose methotrexate, 6-mercaptopurine, high-dose asparaginase, steroids, and an epipodophyllotoxin with cytarabine (14). Consolidation therapy takes several months to a year and is followed by an additional 18–24 months of maintenance therapy, which commonly includes daily 6-mercaptopurine, weekly methotrexate, and periodic pulses of steroids and vincristine (10,14). These agents, although effective in treating the leukemia, can have significant side-effects, both acute and long-term. The most common late adverse events associated with chemotherapy include endocrine abnormalities such as growth delay, obesity, or reproductive problems (associated with corticosteroids and alkylating agents) (15–18), bone necrosis (corticosteroids) (19,20), cardiovascular defects (anthracylines) (21,22), peripheral neuropathy (vincristine) (23), CNS toxicity (methotrexate and corticosteroids) (24–28) and related learning problems (29–32). Although the long-term neurocognitive outcomes associated with chemotherapyonly protocols are generally milder than those that were seen when CRT was routinely used, eliminating CRT does not necessarily eliminate the risk for late effects of treatment. Understanding the potential burden or neurocognitive morbidity among children treated on these protocols thus remains important. Because of the young age of onset and thus the potential for longevity of late effects on the developing nervous system, delayed consequences of therapy can have a greater impact on a child’s life and overall well-being than acute morbidities of anti-leukemic therapies. Whereas short-term morbidities can be expected with virtually all therapeutic agents, the risk for long-term sequelae—especially involving the CNS—appears to be greater for some agents (e.g., methotrexate, steroids, or a combination) than others. In evaluating the costs of treatment-associated toxicity, it is of course important to always be mindful of the significant benefits of therapy in terms of long-term survival and improved health outcomes for a child diagnosed with an aggressive malignancy. Unlike many other neurotoxic exposures, such as environmental pollutants, exposure to antileukemic agents is critical for the child’s survival, with the benefits obviously outweighing the costs. Evaluation of outcomes is therefore better understood in terms of comparing equally efficacious therapies for differential toxicities, or in terms of appreciating the risk for sequelae so that their impact can be well managed and hopefully minimized.

CHEMOTHERAPY AGENTS: EFFICACY AND TOXICITY Chemotherapy agents are highly effective cytotoxic and antineoplastic drugs. They can be generally classified as: (1) antimetabolites (e.g., methotrexate, 6-mercaptopurine); (2) alkaloids (e.g., vincristine); (3) alkylating agents (e.g., cyclophosphamide); (4) anthracyclines (e.g., doxorubicin); (5) epidophyllotoxins (e.g., etoposide); and (6) immune modulators: enzymes (e.g., L-asparaginase) and corticosteroid hormones modulating metabolism and immune functions (e.g., prednisone). Agents within the same group, however, are not necessarily associated with similar toxicities. The risk for toxicity is naturally greater with higher doses, and can also depend on the mode of administration—oral, intravenous, or intrathecal (directly into the spinal fluid)—with the latter causing the highest risk for delayed neurotoxicity. The two chemotherapy agents that are the most likely sources of CNS late effects are methotrexate and, possibly, corticosteroids. Methotrexate (MTX) is a folic acid antagonist that inhibits formation of a lipid component of myelin (3) and blocks DNA synthesis stopping rapidly growing cells. It is administered both intrathecally and intravenously, and constitutes a main component of

134

Mrakotsky and Waber

chemotherapy throughout all treatment phases. High-dose MTX may improve long-term event-free survival as compared to low dose (33); however, MTX can also cause severe neurophysiological toxicities, including leukoencephalopathy with or without seizures and cortical atrophy, particularly when administered intravenously or intrathecally in high doses and repeatedly (14,24–25). Methotrexate also appears to be associated with subclinical levels of demyelination during therapy, whose functional consequences are not clear (34). Corticosteroids or glucocorticoids (GCs) are steroid hormones regulating homeostasis, including peripheral metabolism, immune functions, cardiovascular integrity, and neuronal growth. Clinically, GCs have proven to be very effective anti-leukemic agents, particularly when given early in treatment, and are thus a mainstay of therapy. Their anti-leukemic properties are assumed to be based on their preferential target of lymphoid cells, leading to inhibition of DNA and protein synthesis, and induction of programmed cell death (35,36). Historically, the corticosteroid component of treatment has typically been prednisone. More recently, dexamethasone has been substituted for prednisone as a more potent agent due to its increased bioavailability, superior anti-leukemic activity at equivalent doses, and higher CNS penetration (37,38). Randomized controlled trials have demonstrated lower CNS relapse rates and higher 6-year event-free survival rates for dexamethasone than for prednisone (39,40). Along with its enhanced therapeutic efficacy, dexamethasone may be more toxic than other GCs. It can cause more weight gain, increased bone morbidity, higher risk for proximal myopathy (likely associated with cushingoid appearance) and hyperglycemia, and possibly more cognitive, particularly memory, problems than prednisone (17,20,31,40,41). In addition, dexamethasone toxicity can be enhanced by other treatment agents, such as concurrent daunomycin, and can exacerbate toxicity of other agents such as vincristine-induced neuropathies (17,40), complicating the delineation of steroid-related from other adverse events. More accurate assessment of the risks associated with dexamethasone therapy awaits evaluation of outcomes of randomized controlled trials. Other agents can also be associated with neurotoxicity, but these effects are less frequent. Vincristine (VCR) is a cytotoxic alkaloid that binds to the protein tubulin and interferes with microtubule assembly in cells, hence blocking mitosis and DNA synthesis. Due to its anti-leukemic properties, it is another therapeutic backbone of leukemia therapy; it is administered in combination with multiple agents during remission induction, and in frequent pulses coupled with corticosteroids during post-remission therapy. Vincristine can only be given intravenously, and is lethal if given intrathecally causing progressive encephalopathy leading to brain death. High-dose vincristine therapy is typically associated with peripheral neuropathies occurring early in therapy, with loss of deep-tendon reflex, weakness in muscles, and gait abnormalities (ataxia). These symptoms typically subside when doses are reduced or the agent is discontinued. Short-term difficulties in fine motor control (e.g., buttoning, handwriting) have also been observed (42,43), but these may not always be transient. Asparaginase is an enzyme that breaks down L-asparagine, an amino acid needed for production of protein. Since rapidly growing malignant cells require vast amounts of asparagine, asparaginase deprives these cells of essential nutrients causing cell death without interfering with growth of normal cells. Acute CNS toxicity, presenting as lethargy and somnolence, has been associated with a fall in asparagines in the cerebrospinal fluid. In rare cases, coagulation defects can lead to intracranial hemorrhage and thrombosis, which usually occurs about three weeks after start of asparaginase during induction (44,45). Reversible neuropsychological deficits have also been reported (46,47).

Treatment of Acute Lymphoblastic Leukemia

135

Other agents that are commonly used in therapy, namely, 6-mercaptopurine (6-MP), cytarabine arabinoside (Ara-C), and doxorubicin, are not known to carry a particular risk for acute or late CNS toxicity. In the therapeutic context, however, exposure to each of these agents does not occur in isolation; anti-leukemic drugs are typically given in multiple combinations. Thus, drug– drug interactions are likely to be at least as important in the consideration of late CNS sequelae as single agent effects. In an animal model investigating various doses and combinations of CRT, methotrexate, and steroids, combined exposures caused more behavioral perturbations than single-agent exposures, with steroids playing a major role in limiting or enhancing negative treatment effects of other agents (48). With the refinement of multi-agent chemotherapy and frequent elimination of CRT, many drug combinations remain to be evaluated for interactions, with similarly modulating effects. Moreover, interpretation of findings needs to consider that single agents (e.g., methotrexate, CRT) are typically given in a therapeutic context, and so their potential adverse effects may be understood only within that context.

NEUROPSYCHOLOGICAL SEQUELAE OF TREATMENT FOR ALL WITH CHEMOTHERAPY ONLY Neuropsychological sequelae of leukemia treatment have been most extensively documented in relation to CRT administered in combination with intravenous or intrathecal drugs. Adverse effects on physical and cognitive development are most clearly seen with a dose of 24 Gy (8,49,50), which is rarely if ever used clinically anymore, but is more equivocal with the more typical lower dose of 18 Gy or even 12 Gy (7,51). The possible independent contribution of chemotherapy to the development of CNS late effects in the context of protocols that do not include CRT is less well established. There is emerging evidence of an increased occurrence of neuropsychological deficits and cognitive complaints after systemic and intrathecal chemotherapy (without CRT), although the general consensus is that these effects are milder than those seen when CRT is employed, particularly at the higher doses that were prevalent in older protocols. In a relatively recent review, Moleski (2000) (52) cited a variety of sequelae, including mild declines in IQ, attention, non-verbal memory, visual-motor function, and mathematics. The research literature suggests, however, a relatively diffuse pattern of chemotherapyinduced cognitive changes. Some studies report relatively subtle declines following chemotherapy only (53–58), whereas others report no effects (59–62). An overview of recent outcome research of chemotherapy-only protocols is provided in Table 1. Drug-Related Effects Methotrexate-Related Effects Adverse cognitive outcomes have most frequently been associated with high-dose MTX or triple intrathecal therapy (including MTX). One large randomized controlled trial comparing CRT (18 Gy) combined with intravenous MTX to intrathecal and intravenous MTX without radiation showed no group difference, but poorer performance by both groups than normative populations, suggesting that chemotherapy alone—particularly MTX—can cause cognitive problems (7,29). Similarly, lower levels of intellectual and academic functioning were found in both CRT/chemotherapy (including MTX)-combined and chemotherapy-only groups when compared to asthmatic and healthy controls (32).

136

Table 1 Study authors Copeland et al. (1996) (53)

Mrakotsky and Waber

Neuropsychological Late Effects of Chemotherapy-Only Treatment CNS therapy

Methods

IT - MTX, Ara-C, NZ99 non-CNS cancer CORT various patients prospective: systemic agents baseline, 1,2, 3, & R on different 5 years post diagnoprotocols sis treatment groups: ITC vs. No ITC

Findings 3 & 5 years post change: † no significant group differences at years 3& 5; interaction: † perceptual-motor skills Z ITC, \ NITC † academic achievement Z in both groups † age and SES effects in both groups

NZ63 non-CNS cancer patients ALL (29), AML (9) various cancers (25) prospective: baseline, 1, 2, 3, 4 years post diagnosis treatment groups: ITC vs. No ITC

1, 2, 3, 4 years post change: † No sign. group differences at year 1&2 † at year 3: ITC group sign. Z than No ITC group in reading, spelling, and arithmetic † at year 4: ITC Z No ITC in reading † ITC lack to improve on IQ compared to No ITC

Espy et al. (2001) (57)

IT - MTX, Ara-C, NZ30 ALL patients prospective: baseline, CORT C IV2,3, & 4 years post MTX (POG diagnosis treatment protocols) IT groups: ITCIV vs. MTX only (ITO) (CCG ITO protocols)

4 years post change rates: † modest declines in arithmetic, verbal fluency, visual-motor integration † IT C IV therapy Z visual-motor 4 years. post, faster decline than ITO † arithmetic, verbal fluency decline unrelated to treatment type

Kingma et al. (2002) (61)

IT MTX, 6-MP, NZ20 ALL patients prospective: baseline, HD IV MTX (ALL-7) IT2 & 7 years. postdiagnosis healthy, MTX, PRDL, sibling, historic HD IV MTX; DEX, VCR controls (ALL-6 (post-remission) vs. ALL-7) (ALL-6)

7 years post-diagnosis: † VIQ Z, set shifting Z than healthy controls † auditory memory \, sustained attention \ than earlier protocol with omission of DEX † referral to SPED \ than sibling controls

Rodgers et al. (2003) (62)

HD IV MTX or IT NZ17 ALL patients MTX retrospectivea: mean Z 5 years post treatment sibling controls

† no significant group differences in attention - patients vs. siblings † trend: - sustained attention Z (omissions \, hit rate Z) - focused attention Z - working memory Z

Brown IT-MTX et al. (ALL, AML) (ALL5, 6, and (1996, ANLL 1 proto1999) cols) No IT (54, 56) CNS therapy (other cancer types)

a Retrospective, no baseline available, assessments conducted prospectively.

Abbreviations: CNS, central nervous system; IT, intrathecal; ITC, intrathecal chemotherapy; IV, intravenous; HD, high dose; SD, standard dose; MTX, methotrexate; Ara-C, cytarabine; CORT, hydrocortisone; VCR, vincristine; 6-MP, 6-mercaptopurine; PRD, prednisone; PRDL, prednisolone; DEX, dexamethasone; SES, socioeconomic status; SPED, special education; VIQ, verbal intelligence quotient.

Treatment of Acute Lymphoblastic Leukemia

137

IT combined with IV MTX therapy was associated with decreased perceptual-motor skills over time relative to therapy without IT MTX (53); however, delineation of the effects of other agents (e.g., vincristine, steroids) on visual-motor skills is often difficult. Despite its greater therapeutic efficacy, IV MTX is more frequently associated with neurotoxicity than low-dose oral MTX (25). Associated cognitive problems have been found to be dosedependant and augmented by other agents. Waber and colleagues (63) reported a poorer cognitive outcome in girls treated with high-dose (vs. low dose) IV MTX, but only when MTX was followed by CRT. Similarly, Iuvone and colleagues (64) reported cognitive sequelae to be associated with cumulative MTX doses when MTX was given in the context of CRT. Perhaps more significantly, recent results from the Pediatric Oncology Group (POG 9005) indicate that 30% of children treated on several treatment arms receiving either intravenous or oral MTX in combination with other agents and intrathecal CNS prophylaxis without CRT demonstrated IQ scores lower than 85, with 5% lower than 70 (D. Armstrong, personal communication, 2004). The high prevalence of general intellectual impairment is concerning, reminiscent of outcomes that were seen on older protocols that employed higher doses of CRT. The specific features of this protocol (e.g., dose levels, timing of administration of various components, and potential adverse drug–drug interactions) should be carefully reviewed in comparison to other protocols, where neuropsychological sequelae have generally been milder. Indeed, most studies have reported mild declines one years after chemotherapy with performance that is, nevertheless, still within the normal range (53,57,58). Visual/ perceptual-motor skills seem to be particularly vulnerable to effects of CNS prophylaxis that includes IV and IT MTX. Espy and colleagues (57) describe an additive effect of MTX on the CNS. That is, children who received CRT and MTX demonstrated the most dramatic declines in visual-motor skills, followed by children treated with IT and systemic MTX, with no significant changes in children treated with IT MTX only. Kingma and colleagues (58,61) found gradual improvement in function over time among children treated on consecutive chemotherapy-only protocols as compared to earlier protocols. When compared to children treated on a previous protocol that included CRT, those treated on a chemotherapy-only protocol including intrathecal MTX had a generally more favorable outcome several years post-treatment. Their memory and fine motor skills, however, were poorer than those of healthy controls (58). Results based on a more recent chemotherapyonly protocol, which included IT and high-dose IV MTX, documented no major cognitive deficits compared to healthy and sibling controls, as well as better memory abilities than the previous chemotherapy protocol, attributed potentially to the lower cumulative doses of dexamethasone, vincristine, and IT MTX in the later protocol (61). Corticosteroid-Related Effects Corticosteroids are known for their potential impact on hippocampal, prefrontal, and hypothalamic structures and associated deficits in memory, learning, and mood regulation in adult healthy and ill populations. Although they are known to acutely affect behavioral regulation during treatment in pediatric medical populations (65–67), less is known about long-term behavioral morbidities. A double-blind, placebo-controlled trial of early postnatal therapy for lung disease in premature children showed poorer motor, visualmotor and cognitive ability in the dexamethasone-exposed group at school-age (68), highlighting the potential of dexamethasone for cognitive late effects. Waber and colleagues (31,63) have also speculated that steroids, especially dexamethasone, may be a factor in cognitive late effects of leukemia therapy, specifically

138

Mrakotsky and Waber

affecting verbal declarative memory. Children treated on Dana-Faber Cancer Institute (DFCI) ALL Consortium Protocol 87-01, who received chemotherapy with or without CRT (18 Gy), demonstrated a profile of relative deficits in rote verbal memory and symbolic processing (especially spelling) that were independent of CRT, potentially related to the steroid component prednisone (63). Subsequent comparison of two DFCI protocols, one using prednisone (87-01) and one dexamethasone (91-01) as the steroid component of treatment, revealed more neuropsychological morbidities associated with the latter, again affecting verbal memory and symbolic function (31). Similarly, as mentioned above, Kingma and colleagues (58) found mild deficits in auditory memory in children treated with dexamethasone, whereas children treated with prednisone on a later protocol did not show these deficits. Evaluation of late effects among participants in randomized controlled trials evaluating the relative efficacy of prednisone and dexamethasone for ALL therapy should help to resolve this question. Vincristine-Related Effects Vincristine-associated peripheral neuropathies and related problems with fine motor skills are believed to account, at least partially, for subtle late effects in motor function among children treated on chemotherapy-only regimens. Fine motor difficulties among leukemia survivors were related to high cumulative doses of VCR in post-remission therapy (58). Similarly, problems with hand-writing, persisting after cessation of therapy, have been observed in a small sample of children whose therapy included VCR (43). Compared to healthy controls, children treated with weekly VCR demonstrated slower graphomotor output, longer pause durations, and increased drawing pressure, thought to be a compensation for peripheral neuropathies common with VCR exposure. Asparaginase-Related Effects As described earlier, asparaginase can, in rare cases, cause coagulation defects with cerebrovascular complications, such as thrombosis and associated cognitive deficits. Reports are sparse, however, and available only from case reviews (46,47). Although Ott and colleagues (46) found that asparaginase-related cerebrovascular damage and associated neurological problems were reversible, systematic neuropsychological evaluation might have revealed more persistent deficits. Most commonly, patients experience aphasia and other serious language impairments along with cognitive deficits during the first weeks after the injury. This typically occurs within a few weeks of start of asparaginase therapy. Despite typical improvement of symptoms several months later, some children may have irreversible deficits (47). Child-Related Risk Factors Chemotherapy regimens may not affect all children equivalently; certain subpopulations of children may be at higher risk for poorer cognitive outcomes. Younger children can be more vulnerable to intensive treatments during critical periods of brain development. Young age at diagnosis (!6 years) has been a robust risk factor for late effects of combined regimens of CRT and chemotherapy (51,69). Copeland and colleagues (53) also found younger age at diagnosis to be correlated with poorer nonverbal and perceptual-motor skills in children treated on chemotherapy-only protocols. Children treated as infants (!6 months at diagnosis) with very high-dose IV MTX and IT Ara-C and MTX exhibit relatively poor cognitive outcomes 5 years later (70).

Treatment of Acute Lymphoblastic Leukemia

139

Another potential risk factor for long-term morbidities is gender. Girls appear to be more vulnerable than boys to cognitive late effects when they are treated with high doses of MTX (both IV and IT) and CRT (8,30,63,71). In contrast, Balsom and colleagues (72) found that pre-irradiation MTX ameliorates the negative cognitive late effects of highdose CRT only in girls. Such sex differences have generally not emerged, however, for protocols that do not include CRT or where CRT doses are relatively lower (%18 Gy). Neuropathological Changes Despite the emerging evidence that chemotherapy with or without CRT can cause subtle cognitive late effects, the underlying pathophysiological mechanisms are poorly understood. The neuroimaging research to date, although limited, has shown demyelination of white matter, particularly in frontal areas, calcifications, and atrophy or reduction of cerebral gray matter (3,47,58) that may result directly from chemotherapy. Methotrexate is thought to be a primary contributor to white matter abnormalities (34,73). Surtees, Clelland, and Hahn (34) documented subclinical demyelination in ALL patients during treatment with MTX with or without CRT. Young children (under 6 years) are particularly vulnerable to these abnormalities because insults to the brain early in life, such as exposure to a neurotoxin, can have detrimental effects on myelinization, synaptogenesis, and dendritic branching (74). For example, Pa¨a¨kko¨ and colleagues (73) found young children to be more vulnerable to white matter changes during and after repeated MTX injections. Brain changes may, however, not only be treatment-related. An intriguing report is given by Kahkonen and colleagues (75), who found that cerebral glucose utilization in the brain among long-term survivors was associated with leukocyte counts at diagnosis, suggesting that cognitive late effects may reflect not only the effects of therapy, but also the direct impact of the disease itself on the developing brain. Accounts of the direct relationship between treatment-related cognitive and neuropathological changes are still scarce. Intellectual decline following CRT (24 Gy) combined with MTX in earlier protocols was associated with necrotizing leukencephalopathy and mineralizing microangiopathy (76). Results from more contemporary therapy protocols are inconsistent, however; some studies report no significant or systematic correlation between MRI results and test scores or educational level (58,73), whereas others have documented such relationships. Children treated with chemotherapy alone (MTX in combination with other agents and intrathecal CNS prophylaxis without CRT) on POG protocol 9005 are reported to demonstrate high rates (40%) of CT abnormalities, including calcification, white matter abnormalities, or both, 5 years after diagnosis. Children with CT abnormalities performed more poorly on a range of cognitive and achievement tasks. The presence of calcifications was associated with impairment of Performance IQ but not Verbal IQ, and white matter changes were associated with poorer Verbal and Performance IQ (D. Armstrong, personal communication, 2004). More research in this area is needed to establish the relationship between treatment-related brain changes and cognitive outcome more clearly.

METHODOLOGICAL CONSIDERATIONS, LIMITATIONS, OUTSTANDING ISSUES AND DATA NEEDS Systematic research on outcomes of leukemia therapy within and outside the United States is typically conducted within the context of multicenter collaborations. These provide for standardized protocols across institutions as well as sufficient numbers of patients to

140

Mrakotsky and Waber

evaluate clinical trials. These trials, however, are generally designed with efficacy as the study endpoint, and are rarely framed to answer questions about neurotoxicity. Posing questions and drawing valid inferences about cognitive late effects is therefore difficult, depending on designs that are often less than optimal. Although many of these constraints are unavoidable, given pragmatic considerations, the research nevertheless needs to be evaluated in light of these limitations. Limited Sample Size One of the key limitations of most studies is the sample size. Inadequate sample size can occur for a number of reasons, including the low prevalence of the disease in the general population, the multi-site nature of most clinical trials, and the need to exclude many children from studies because of factors such as medical complications. Added to this is the possibility that outcomes may differ for subgroups of patients (e.g., younger vs. older), further reducing the potential sample size and the statistical power to detect meaningful differences, especially interactions. The risk for a Type II error, that is, failing to reject a false null hypothesis, is greater with smaller samples. Within this context, studies that have reported no effects or only trend effects of chemotherapy on cognitive performance could underestimate a true effect. Longitudinal research, potentially very important in terms of fully describing the course and development of cognitive late effects, is quite limited because of decreased sample sizes and the pragmatic difficulties of recruiting and retaining patients for multiple visits and over long periods of time. Heterogenous Treatment Protocols Another challenge is the variation in treatment protocols across and within institutions, with varying types and doses of chemotherapy agents. One chemotherapy-only protocol may differ in dose, combinations and sequence of administration of agents from another. The effect of these protocol differences on outcome is uncertain and can, along with other factors cited, be a source of apparent inconsistency in results. Some authors provide little detail about the therapy protocols on which children were treated, further limiting the utility of their findings. In order to truly interpret behavioral findings, treatment regimens should be adequately described, so that variations in findings can be correlated with different drugs, doses, and combinations. Because of problems with sample size, moreover, some studies may combine patients on different protocols, but at some cost in terms of interpretability. Lack of Appropriate Comparison Groups Another potential limitation of treatment outcome research can be the lack of adequate controls. Precisely what constitutes an appropriate control group is uncertain in this context. Studies based on randomized controlled trials can legitimately compare performance of children on different arms of the protocol, but drawing inferences about comparison to “normal” is more treacherous. Some studies elect not to use a control group, choosing to compare patient performance to age-normed expectations. Others may include historic patient controls, healthy controls, or sibling controls that theoretically take into consideration environmental/genetic factors. Each choice, however, represents some compromise. Results from studies that use historic controls may be confounded by cohort effects, for example, and sibling controls may not adequately account for age and sex factors. Siblings of ALL patients typically have higher average mean IQs than normative

Treatment of Acute Lymphoblastic Leukemia

141

expectation would predict (77), suggesting that their performance may better reflect pre-morbid potential of the index sample. An average performance among such patients might actually reflect a loss of potential function. Comparison with patient controls is another strategy that has occasionally been employed to increase validity. However, these children typically receive different drug therapies that may themselves have consequences, or they may be developmentally different from the leukemia patients because of different modal ages at diagnosis (53). Ultimately, investigators need to design their studies to be as informative and interpretable as possible, while acknowledging the limitations incurred by the particular choice of strategy.

Retrospective vs. Prospective Design Most late effects studies employ a quasi-retrospective or cross-sectional design at time of post-treatment evaluation, with no information on pre-treatment functioning. Although pre- and post-testing is theoretically ideal, there are some pragmatic barriers to proceeding in this way. Children are typically quite ill when diagnosed, and families are emotionally upset, with uncertain impact on test performance. Moreover, because the modal age of diagnosis for ALL is during the preschool period, the measures that must be used are relatively imprecise and the extent to which they might constitute a valid baseline is uncertain. This is especially true for the youngest children, two to four years old, who are likely to experience the greatest adverse impact of the disease and its therapy.

Non-Disease Related Moderator Variables Female gender and young age at diagnosis and treatment are widely accepted as potential risk factors for poor cognitive outcome. There are, however, other variables that potentially influence cognitive performance and should be included in study designs. Parent socioeconomic status and education level are highly correlated with child IQ and neuropsychological performance in general, and frequently account for substantial variability in performance of ALL survivors (53,57,78). These influences should be systematically considered in study designs.

Measurement Problems Measuring cognitive functions over time and across different age ranges is a challenge to most neuropsychological research assessing development and change longitudinally. Test versions can change over time (e.g., WISC-III to WISC-IV) and forms may differ for individuals of different ages (e.g., WISC to WAIS), compromising the comparability of data points when comparing results from historic protocols, following children longitudinally, or when following samples that are heterogeneous for age. Butler and Copeland (79) point out that the use of global measures, such as intelligence tests, may account in part for the inconsistency of findings between studies. Reliance on such global measures can risk missing the more subtle deficits that may be associated with the less intensive contemporary protocols. Although these deficits may be relatively more subtle, in context of the growing numbers of long-term survivors they may have a broader impact on later adaptive, academic, and vocational functioning.

142

Mrakotsky and Waber

Difference vs. Deficit Another consideration is the clinical significance of group differences in performance (“difference” vs. “deficit”). Although statistically significant differences may be meaningful in scientific terms, a finding of difference in specific neuropsychological functions between groups does not necessarily imply a clinically meaningful deficit. It can be informative to report not only whether groups differ, but also the proportions of children falling into a ‘clinically significant’ range, as well as exploring potential risk and protective factors. Developmental Considerations Over the past several decades, the risk for neuropsychological impairment among children who survive leukemia has diminished steadily as advances in medical treatment have permitted some relaxation of the intensity of therapies. Nonetheless, there appear to be subtle long-term deficits, which may not emerge in basic cognitive skills, but rather in their efficiency and application. These may not be apparent in young children, but may become more apparent as the child develops and seeks to respond to increasing cognitive demands. Since brain development does not occur in isolation but in constant interaction with the environment, a child’s neurobehavioral response (e.g., cognitive performance) is always context-dependent. It is the failure to adapt to this context (e.g., the learning environment) in a developmentally appropriate way that constitutes a deficit or learning disorder (80). Thus, although a child with relatively minor compromise of executive control processes or visuomotor skills may be able to perform comfortably in the lower grades, he/she may not be able to sustain that level of function in higher grades, as demands for attention, organization, and written output increase. This developmental phenomenon can give the appearance of decline, if one attends only to standardized scores, but in fact represents adaptive problems as age-level demands increasingly stress areas of more subtle weakness. Adult Survivors’ Quality of Life (QOL) Despite the growing numbers of adult survivors of pediatric leukemia and higher expectations for long-term quality of life (QOL), data on adult QOL, including vocational and adaptive functions, is sparse. Current reports on adult survivors are based on data from patients treated 10, 15, or even 20 years ago. Adult cancer survivors, particularly those who received radiation, are more likely to report adverse general health, activity limitations, and functional impairment than sibling controls or non-irradiated patients (81,82). Long-term outcome research, however, is a “moving target.” Adults eligible for evaluation now, who were treated as children, are likely to have been exposed to more intensive therapies (often involving CRT) than children who are newly diagnosed. One cannot predict from current adults to future adults. In this regard, the Pui et al. study (81) is encouraging; non-irradiated leukemia survivors did not show any major impairments in adaptive living skills relative to the general population.

CONCLUSION In summary, advances in leukemia therapy over the past decades have led to excellent efficacy in treating a disease that was once nearly uniformly fatal. Toxicity and outcome have notably improved with the reduction or elimination of CRT. With these

Treatment of Acute Lymphoblastic Leukemia

143

improvements and growing numbers of survivors, long-term quality of life and the transition into young adulthood will become increasingly important considerations. Although cognitive deficits appear more subtle on contemporary chemotherapy protocols than they were on older protocols using high doses (24 Gy) of CRT, they need to be monitored carefully over time. This is particularly of concern as reported outcomes of difficulties with executive control, attention, and memory may become more troublesome with increasing academic and social demands, potentially affecting a child’s ability to become a self-organized independent learner who is well-equipped to meet expectable educational and career goals later in life. Reduction or elimination of CRT has not eliminated cognitive late effects, and changes in protocols can increase risks in ways that might not have been predicted. These protocols therefore need to be consistently monitored with respect to cognitive late effects. Moreover, it remains important to monitor individual children so that behavioral interventions can be undertaken before problems escalate.

ACKNOWLEDGMENTS We gratefully acknowledge the support of NCI P01 68484 and the continuing guidance of Stephen E. Sallan, M.D. and Lewis B. Silverman M.D., of the Dana Farber/Children’s Hospital ALL Consortium.

REFERENCES 1. Jemal A, Clegg LX, Ward E, et al. Annual report to the nation on the status of cancer, 1975– 2001, with a special feature regarding survival. Cancer 2004; 101:3–27. 2. Silverman LB, Declerck L, Gelber RD, et al. Results of the Dana-Farber Cancer Institute Consortium protocols for children with newly diagnosed acute lymphoblastic leukemia (1981–1995). Leukemia 2000; 14:2247–2256. 3. Packer RJ, Meadows AT, Rorke LB, Goldwein JL, D’Angio G. Long-term sequelae of cancer treatment on the central nervous system in childhood. Med Pediatr Oncol 1987; 15:241–253. 4. Pui C-H. Acute lymphoblastic leukemia. Pediatr Clin North Am 1997; 44:831–846. 5. Brouwers P, Riccardi R, Poplack D, Fedio P. Attentional deficits in long-term survivors of childhood acute lymphoblastic leukemia (ALL). J Clin Neuropsychol 1984; 6:325–336. 6. Copeland DR, Dowell RE, Jr., Fletcher JM, et al. Neuropsychological effects of childhood cancer treatment. J Child Neurol 1988; 3:53–62. 7. Mulhern RK, Fairclough D, Ochs J. A prospective comparison of neuropsychologic performance of children surviving leukemia who received 18 Gy, 24 Gy, or no cranial irradiation. J Clin Oncol 1991; 9:1348–1356. 8. Waber DP, Urion DK, Tarbell NJ, Niemeyer C, Gelber R, Sallan SE. Late effects of central nervous system treatment of acute lymphoblastic leukemia in childhood are sex-dependent. Dev Med Child Neurol 1990; 32:238–248. 9. Kadan-Lottick N, Neglia JP. Pediatric acute lymphoblastic leukemia: Redefining outcomes. Comment. J Ped Hematol Oncol 2002; 24:88. 10. Steuber CP, Poplack DG. Acute lymphoblastic leukemia. In: Rudolph CD, Rudolph AM, Hostetter MK, Lister G, Siegel N, eds. Rudolph’s Pediatrics. 21st ed. New York: McGraw-Hill Medical Publishing Division, 2002 electronic manuscript. 11. Smith MA, Gloeckler-Ries LA, Gurney JG et al., eds. Cancer Incidence and Survival among Children and Adolescents: United States SEER Program 1975–1995. Bethesda, MD: National Cancer Institute, 1999:17–34. SEER Program. NIH Pub. No. 99-4649.

144

Mrakotsky and Waber

12. Surveillance, Epidemiology, and End Results (SEER) Program SEER*Stat Database: Incidence-SEER 9 Regs Public-Use, Nov 2002 Sub (1973–2000), National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2003. 13. Smith M, Arthur D, Camitta B, et al. Uniform approach to risk classification and treatment assignment for children with acute lymphoblastic leukemia. J Clin Oncol 1996; 14:18–24. 14. Pui CH, Evans WE. Acute lymphoblastic leukemia. N Engl J Med 1998; 339:605–615. 15. Wolthers OD. Growth problems in children with asthma. Hormone Res 2002; 57:83–87. 16. Van Dongen-Melman JE, Hokken-Koelega AC, Hahlen K, De Groot A, Tromp CG, Egeler RM. Obesity after successful treatment of acute lymphoblastic leukemia in childhood. Pediatr Res 1995; 38:86–90. 17. Belgaumi AF, Al-Bakrah M, Al-Mahr M, et al. Dexamethasone-associated toxicity during induction chemotherapy for childhood acute lymphoblastic leukemia is augmented by concurrent use of daunomycin. Cancer 2003; 97:2898–2903. 18. Glantz JC. Reproductive toxicology of alkylating agents. Obstetr Gyn Survey 1994; 49:709–715. 19. Mattano LA, Jr., Sather HN, Trigg ME, Nachman JB. Osteonecrosis as a complication of treating acute lymphoblastic leukemia in children: a report from the Children’s Cancer Group. J Clin Oncol 2000; 18:3262–3272. 20. Strauss AJ, Su JT, Dalton VM, Gleber RD, Sallan SE, Silverman LB. Bony morbidity in children treated for acute lymphoblastic leukemia. J Clin Oncol 2001; 19:3066–3072. 21. Lipshultz SE, Colan SD, Gelber RD, Perez-Atayde AR, Sallan SE, Sanders SP. Late cardiac effects of doxorubicin therapy for acute lymphoblastic leukemia in childhood. N Engl J Med 1991; 324:808–815. 22. Lipshultz SE, Lipsitz SR, Mone SM, et al. Female sex and drug dose as risk factors for late cardiotoxic effects of doxorubicin therapy for childhood cancer. N Engl J Med 1995; 332:1738–1743. 23. Postma TJ, Benard BA, Huijgens PC, Ossenkoppele GJ, Heimans JJ. Long-term effects of vincristine on the peripheral nervous system. J Neuro-Oncol 1993; 15:23–27. 24. Ochs JJ, Bowman WP, Pui CH, Abromowitch M, Mason C, Simone JV. Seizures in childhood lymphoblastic leukaemia patients. Lancet 1984; 2:1422–1424. 25. Mahoney DH, Shuster JJ, Nitschke R, et al. Acute neurotoxicity in children with B-precursor acute lymphoid leukemia: An association with intermediate-dose intravenous methotrexate and intrathecal triple therapy—a Pediatric Oncology Group study. J Clin Oncol 1998; 16:1712–1722. 26. Wolkowitz OM. Prospective controlled studies of the behavioral and biological effects of exogenous corticosteroids. Psychoneuroendocrinology 1994; 19:233–255. 27. Murphy BP, Inder TE, Huppi PS, et al. Impaired cerebral cortical gray matter growth after treatment with dexamethasone for neonatal chronic lung disease. Pediatrics 2001; 107:217–221. 28. Macdonald DR. Neurologic complications of chemotherapy. Neurol Clin 1991; 9:955–967. 29. Ochs J, Mulhern R, Fairclough D, et al. Comparison of neuropsychologic functioning and clinical indicators of neurotoxicity in long-term survivors of childhood leukemia given cranial radiation or parenteral methotrexate: a prospective study. J Clin Oncol 1991; 9:145–151. 30. Waber DP, Tarbell NJ, Kahn CM, Gelber RD, Sallan SE. The relationship of sex and treatment modality to neuropsychologic outcome in childhood acute lymphoblastic leukemia. J Clin Oncol 1992; 10:810–817. 31. Waber DP, Carpentieri SC, Klar N, et al. Cognitive sequelae in children treated for acute lymphoblastic leukemia with dexamethasone or prednisone. J Pediatr Hematol Oncol 2000; 22:206–213. 32. Raymond-Speden E, Tripp G, Lawrence B, Holdaway D. Intellectual, neuropsychological, and academic functioning in long-term survivors of leukemia. J Pediatr Psychol 2000; 25:59–68. 33. Niemeyer CM, Gelber RD, Tarbell NJ, et al. Low-dose versus high-dose methotrexate during remission induction in childhood acute lymphoblastic leukemia (Protocol 81-01 update). Blood 1991; 78:2514–2519.

Treatment of Acute Lymphoblastic Leukemia

145

34. Surtees R, Clelland J, Hann I. Demyelination and single-carbon transfer pathway metabolites during the treatment of acute lymphoblastic leukemia: CSF studies. J Clin Oncol 1998; 16:1505–1511. 35. Baxter GD, Collins RJ, Harmon BV, et al. Cell death by apoptosis in acute leukaemia. J Pathol 1989; 158:123–129. 36. Cohen JJ. Lymphocyte death induced by glucocorticoids. In: Schleimer RP, Claman HN, Oronsky AL, eds. Anti-inflammatory Steroid Action: Basic and Clinical Aspects. San Diego: Academic Press, 1989:110. 37. Balis FM, Lester CM, Chrousos GP, Heideman RL, Poplack DG. Differences in cerebrospinal fluid penetration of corticosteroids: possible relationship to the prevention of meningeal leukemia. J Clin Oncol 1987; 5:202–207. 38. Kaspers GJ, Veerman AJ, Popp-Snijders C, et al. Comparison of the antileukemic activity in vitro of dexamethasone and prednisolone in childhood acute lymphoblastic leukemia. Med Pediatr Oncol 1996; 27:114–121. 39. Jones B, Freeman AI, Shuster JJ, et al. Lower incidence of meningeal leukemia when prednisone is replaced by dexamethasone in the treatment of acute lymphoblastic leukemia. Med Pediatr Oncol 1991; 19:269–275. 40. Bostrom BC, Sensel MR, Sather HN. Dexamethasone versus prednisone and daily oral versus weekly intravenous mercaptopurine for patients with standard-risk acute lymphoblastic leukemia: a report from the Children’s Cancer Group. Blood 2003; 101:3809–3817. 41. Dropcho EJ, Soong SJ. Steroid-induced weakness in patients with primary brain tumors. Neurology 1991; 4:1235–1239. 42. Copeland DR, Dowell RE, Jr., Fletcher JM, et al. Neuropsychological test performance of pediatric cancer patients at diagnosis and one year later. J Pediatr Psychol 1988; 13:183–196. 43. Reinders-Messelink HA, Schoemaker MM, Snijders TA, Goeken LN, Bokkerink JP, Kamps WA. Analysis of handwriting of children during treatment for acute lymphoblastic leukemia. Med Pediatr Oncol 2001; 37:393–399. 44. Feinberg WM, Swenson MR. Cerebrovascular complications of L-asparaginase therapy. Neurology 1988; 38:127–133. 45. Priest JR, Ramsay NK, Steinherz PG, et al. A syndrome of thrombosis and hemorrhage complicating L-asparaginase therapy for childhood acute lymphoblastic leukemia. J Pediatr 1982; 100:984–989. 46. Ott N, Ramsay NKC, Priest JR, Lipton M, Pui CH. Sequelae of thrombotic or hemorrhagic complications following L-asparaginase therapy for childhood lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1988; 10:191–195. 47. Kingma A, Tamminga RYJ, Kamps WA, Le Coultre R, Saan RJ. Cerebrovascular complications of L-asparaginase herapy in children with leukemia: Aphasia and other neuropsychological deficits. Pediatr Hematol Oncol 1993; 10:303–339. 48. Mullenix PJ, Kernan WJ, Schunior A, et al. Interactions of steroid, methotrexate, and radiation determine neurotoxicity in an animal model to study therapy for childhood leukemia. Pediatr Res 1994; 35:171–178. 49. Robison LL, Nesbit ME, Sather HN, Meadows AT, Ortega JA, Hammond GD. Factors associated with IQ scores in long-term survivors of childhood acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1984; 6:115–121. 50. Brouwers P, Riccardi R, Fedio P, Poplack DG. Long-term neuropsychological sequelae of childhood leukemia: correlation with CT brain scan abnormalities. J Pediatr 1985; 6:723–728. 51. Waber DP, Shapiro BL, Carpentieri SC, et al. Excellent therapeutic efficacy and minimal late neurotoxicity in children treated with 18 grays of cranial radiation therapy for high-risk acute lymphoblastic leukemia: a 7 years follow-up study of the Dana-Farber Cancer Institute Consortium Protocol 87-01. Cancer 2001; 92:15–22. 52. Moleski M. Neuropsychological, neuroanatomical, and neurophysiological consequences of CNS chemotherapy for acute lymphoblastic leukemia. Arch Clin Neuropsychol 2000; 15:60–630.

146

Mrakotsky and Waber

53. Copeland DR, Moore BD, Francis DJ, Jaffe N, Culbert SJ. Neuropsychologic effects of chemotherapy on children with cancer: a longitudinal study. J Clin Oncol 1996; 14:2826–2835. 54. Brown RT, Sawyer MB, Antoniou G, et al. A 3 years follow-up of the intellectual and academic functioning of children receiving central nervous system prophylactic chemotherapy for leukemia. J Dev Behav Pediatr 1996; 17:392–398. 55. Brown RT, Madan-Swain A, Walco GA, et al. Cognitive and academic late effects among children previously treated for acute lymphocytic leukemia receiving chemotherapy as CNS prophylaxis. J Pediatr Psychol 1998; 23:333–340. 56. Brown RT, Sawyer MG, Antoniou G, Toogood I, Rice M. Longitudinal follow-up of the intellectual and academic functioning of children receiving central nervous systemprophylactic chemotherapy for leukemia: a four-year final report. J Dev Behav Pediatr 1999; 20:373–377. 57. Espy KA, Moore IM, Kaufmann PM, Kramer JH, Matthay K, Hutter JJ. C.N.S. Chemotherapeutic prophylaxis and neuropsychologic change in children with acute lymphoblastic leukemia: a prospective study. J Pediatr Psychol 2001; 26:1–9. 58. Kingma A, van Dommelen RI, Mooyaart EL, Wilmink JT, Deelman BG, Kamps WA. Slight cognitive impairment and magnetic resonance imaging abnormalities but normal school levels in children treated for acute lymphoblastic leukemia with chemotherapy only. J Pediatr 2001; 139:413–420. 59. Anderson V, Smibert E, Ekert H, Godber T. Intellectual, educational, and behavioural sequelae after cranial irradiation and chemotherapy. Arch Dis Child 1994; 70:476–483. 60. Butler RW, Hill JM, Steinherz PG, Meyers PA, Finlay JL. Neuropsychologic effects of cranial irradiation, intrathecal methotrexate, and systemic methotrexate in childhood cancer. J Clin Oncol 1994; 12:2621–2629. 61. Kingma A, Van Dommelen R, Mooyaart EL, Wilmink JT, Deelman BG, Kamps WA. No major cognitive impairment in young children with acute lymphoblastic leukemia using chemotherapy only: a prospective longitudinal study. J Pediatr Hematol Oncol 2002; 24:106–114. 62. Rodgers J, Marckus R, Kearns P, Windebank K. Attentional ability among survivors of leukaemia treated without cranial irradiation. Arch Dis Child 2003; 88:147–150. 63. Waber DP, Tarbell NJ, Fairclough D, et al. Cognitive sequelae of treatment in childhood acute lymphoblastic leukemia: Cranial radiation requires an accomplice. J Clin Oncol 1995; 13:2490–2496. 64. Iuvone L, Mariotti P, Colosimo C, Guzzetta F, Ruggiero A, Riccardi R. Long-term cognitive outcome, brain computed tomography scan, and magnetic resonance imaging in children cured for acute lymphoblastic leukemia. Cancer 2002; 95:2562–2570. 65. Drigan R, Spirito A, Gelber RD. Behavioral effects of corticosteroids in children with acute lymphoblastic leukemia. Med Pediatr Oncol 1992; 20:13–21. 66. Harris JC, Carel CA, Rosenberg LA, Joshi P, Leventhal BG. Intermittent high dose corticosteroid treatment in childhood cancer: behavioral and emotional consequences. J Am Acad Child Psychiatry 1986; 25:120–124. 67. Soliday E, Grey S, Lande MB. Behavioral effects of corticosteroids in steroid-sensitive nephrotic syndrome. Pediatrics 1999; 104:956–957. 68. Yeh TF, Lin YJ, Lin HC, et al. Outcomes at school age after postnatal dexamethasone therapy for lung disease of prematurity. NEJM 2004; 350:1304–1313. 69. Cousens P, Waters B, Said J, Stevens M. Cognitive effects of cranial irradiation in leukemia: A survey and meta-analysis. Child Psychol Psychiatr 1988; 29:839–852. 70. Kaleita TA, Reaman GH, MacLean WE, Sather HN, Whitt JK. Neurodevelopmental outcome of infants with acute lymphoblastic leukemia: a Children’s Cancer Group report. Cancer 1999; 85:1859–1865. 71. Precourt S, Robaey P, Lamothe I, Lassonde M, Sauerwein HC, Moghrabi A. Verbal cognitive functioning and learning in girls treated for acute lymphoblastic leukemia by chemotherapy with or without cranial irradiation. Dev Neuropsychol 2002; 21:173–195.

Treatment of Acute Lymphoblastic Leukemia

147

72. Balsom WR, Bleyer WA, Robison LL, et al. Intellectual function in long-term survivors of childhood acute lymphoblastic leukemia: protective effect of pre-irradiation methotrexate? A Childrens Cancer Study Group study Med Pediatr Oncol 1991; 19:486–492. 73. Pa¨a¨kko¨ E, Harila-Saari A, Vanionpaa L, Himanen S, Pyhtinen J, Lanning M. White matter changes on MRI during treatment in children with acute lymphoblastic leukemia: correlation with neuropsychological findings. Med Pediatr Oncol 2000; 35:456–461. 74. Aram DM, Eisele JA. Plasticity and recovery of higher cognitive functions following early brain injury. In: Rapin I, Segalowitz SJ, eds. In: Handbook of Neuropsychology, Vol. 6. New York: Elsevier Science, 1992:73–92. 75. Kahkonen M, Metsahonkala L, Minn H, et al. Cerebral glucose metabolism in survivors of childhood acute lymphoblastic leukemia. Cancer 2000; 88:693–700. 76. Bleyer WA, Griffin TW. White matter necrosis, mineralizing microangiopathy, and intellectual abilities in survivors of childhood leukemia: Associations with central nervous system irradiation and methotrexate therapy. In: Gilbert HA, Kagan AR, eds. Radiation Damage to the Nervous System. New York: Raven Press, 1980:155. 77. Giralt J, Ortega JJ, Olive T, Verges R, Forio J, Salvador L. Long-term neuropsychologic sequelae of childhood leukemia: comparison of two CNS prophylaxis regimens. Int J Radiat Oncol Biol Phys 1992; 24:49–53. 78. Trautman PD, Erickson C, Shaffer D, et al. Prediction of intellectual deficits in children with acute lymphoblastic leukemia. J Dev Behav Pediatr 1988; 9:122–128. 79. Butler RW, Copeland DR. Neuropsychological effects of central nervous system prophylactic treatment in childhood leukemia: methodological considerations. J Pediatr Psychol 1993; 18:319–338. 80. Holmes-Bernstein J, Waber DP. Developmental neuropsychological assessment: The systemic approach. In: Boulton AA, Baker GB, Hiscock M, eds. In: Neuromethods, Vol. 17. Clifton, NJ: Humana Press, 1990. 81. Pui CH, Cheng C, Leung W, et al. Extended follow-up of long-term survivors of childhood acute lymphoblastic leukemia. N Engl J Med 2003; 349:640–649. 82. Hudson MM, Mertens AC, Yasui Y. Childhood Cancer Survivor Study Investigators. Health status of adult long-term survivors of childhood cancer: a report from the Childhood Cancer Survivor Study. JAMA 2003; 290:1583–1592.

8 Prenatal Tobacco and Postnatal Environmental Tobacco Smoke Exposure and Children’s Cognitive and Behavioral Functioning Michael Weitzman, Todd A. Florin, and Megan Kavanaugh American Academy of Pediatrics, Center for Child Health Research and Department of Pediatrics, The University of Rochester School of Medicine and Dentistry, Rochester and Department of Pediatrics, New York University School of Medicine, New York, New York, U.S.A.

A significant and still growing body of literature indicates that maternal smoking during pregnancy is associated with neurotoxic effects on children. Both animal model studies and human epidemiologic studies demonstrate similar effects in terms of increased activity, decreased attention, and diminished intellectual abilities. Epidemiologic studies also suggest that prenatal tobacco exposure is associated with higher rates of behavior problems and school failure.

INTRODUCTION Tobacco smoke exposure of fetuses and children remains common in the United States, despite a 33% reduction in smoking during pregnancy over the past decade (1). It has been convincingly established that parental smoking contributes to many child health problems, such as low birth weight, asthma, respiratory infections, otitis media, and Sudden Infant Death Syndrome (SIDS) (2,3). Both animal model and human epidemiologic studies also strongly suggest that prenatal and early passive exposure to tobacco smoke leads to negative behavioral and neurocognitive effects in children, and there are plausible biologic mechanisms through which this may occur. To date, these negative neurodevelopmental and behavioral associations are far less well recognized by the pediatric, child development and public health communities than are the respiratory and SIDS associations. An expanding body of literature indicates that maternal smoking during pregnancy and early childhood is associated with neurotoxic effects on children. This literature suggests that such exposure results in increased rates of children’s behavior problems and psychiatric disorders, and may also lead to subtle intellectual decrements and 149

150

Weitzman et al.

neurocognitive impairments. This exposure is common, rarely confined to the prenatal period, and is often associated with other factors adversely affecting behavioral and cognitive outcomes in children. Lower maternal educational achievement and socioeconomic status, increased rates of maternal depression and anxiety disorder, and alcohol and psychoactive drug use are all more common among women who smoke during pregnancy or during the child bearing and rearing years. Each of these factors, and possibly others as of yet uncovered, may compound the negative effects of children’s tobacco exposure on child development. Other childhood environmental exposures, such as to lead, cause neurodevelopmental problems that are subtle, but serious. This has resulted in substantial clinical, public health, and environmental policy development and implementation, which in turn have been associated with profound reductions in children’s exposure. Children’s prenatal and early passive exposure to tobacco smoke is extremely prevalent and may result in neurocognitive and behavioral problems of similar type and magnitude to those of lead. This paper reviews the literature on the effects of tobacco exposure, both pre- and postnatal, on children’s behavior and cognition.

CHARACTERISTICS OF WOMEN ASSOCIATED WITH MATERNAL SMOKING The prevalence of smoking during pregnancy is estimated to be between 11.4% and 30% of all pregnant women in the United States, with the estimates varying depending on the source of the data and maternal characteristics of the study population, and has been declining in the recent past (4–9). In 2001, approximately 23% of women of childbearing age smoked (7). The most recent national data based on birth records indicates that 11.4% of pregnant women smoke, a 42% decline from 1989 (6). In addition, the proportion of mothers who smoked 11 cigarettes (half a pack) or more per day has declined from 41% in 1989 to 26% in 2002 (6). Smoking during pregnancy varies by ethnicity: American Indians, including Aleuts and Eskimos (19.7%), Hawaiian Asians (13.7%), White (12.3%), Black (8.7%), Hispanic (3.0%), and Asian (2.5%) (6). Prenatal smoking is greatest during late teen years (18.2%) (6). The percentage of pregnant women who smoke varies greatly by education, with more years of education associated with lower rates of smoking: 9–11 years of education (24.1%), 12 years (15.8%), 13–15 years (8.7%), and 16 or more years (1.7%) (6). In addition, more than one-third of all U.S. children are regularly exposed to environmental tobacco smoke as assessed by serum cotinine levels (10,11). Pirkle et al. report that in a survey of 10,642 subjects 4 years and older, 87.9% had detectable levels of serum cotinine (10). Many epidemiologic studies rely on parent reported smoking behavior rather than assessment of biomarkers to characterize fetal and child exposure. This is likely to result in underreporting of exposure, introducing an important bias for studies: some children who are exposed are misclassified or their exposure is underestimated, thereby likely underestimating the effect of tobacco exposure (12). Infants who are exposed to maternal smoking during pregnancy are at increased risk for other toxic exposures. Smoking mothers are more likely to drink and use illicit drugs (13–16). Women smokers differ from nonsmoking women in a number of important psychosocial characteristics (17). Higher rates of unwanted pregnancies (18), difficulties coping with stress, and lower self-esteem (19) are found more commonly among smoking

Tobacco Exposure and Children

151

women compared to nonsmoking women. Women who smoked during pregnancy have been found to be more likely to report marital difficulties and more likely to physically discipline their infants (20). Smoking mothers are less likely to breastfeed their infants (12,21,22). Women who smoke also are more anxious, change jobs more frequently, and divorce more often than women who do not smoke (23). Associations of cigarette smoking and mental illness are not confined to women who smoke during pregnancy. Cigarette smoking is associated with psychiatric disorders among adolescents and adults in the general population (24–27). Persons with mental illness are about twice as likely to smoke as other persons (28). Data suggest that dopamine-related genes are associated with smoking (29,30), and abnormalities in the dopaminergic reward pathways have been implicated in substance abuse and addictive behaviors (29). Dopamine D2 receptor gene variants have been found to be associated with alcoholism, drug dependency, obesity, smoking, pathological gambling, attention-deficit hyperactivity disorder (ADHD), Tourette syndrome, as well as other compulsive behaviors (31). Linkage studies indicate that there are several possible smoking-associated genes, which include cytochrome P450 subfamily polypeptide 6 (CYP2A6), dopamine D1, D2, and D4 receptors, dopamine transporter, and serotonin transporter genes (32–36). Genetic aspects of behavioral problems that are autosomally linked to smoking are likely to be transmitted equally from smoking mothers and smoking fathers. Studies to date, however, have found that maternal smoking is more strongly associated with adverse developmental outcomes than is paternal smoking (37,38). If parental smoking were simply a marker for genetically driven behavioral problems in children, then studies should find that maternal and paternal smoking contribute equally to adverse outcomes in children, which has not been the case. A number of seminal epidemiologic studies of child development have demonstrated that the number of risk factors present in a child’s life increases the likelihood of adverse outcomes for that child (39–41). Both nature and nurture are responsible for developmental outcomes. Socioeconomic and familial factors sometimes overshadow the role of biology in producing emotional difficulties and intellectual retardation (42). The “transactional model” (43) of child development supposes that a child’s initial biological makeup, including genotype, is not fully expressed at birth, but only develops during an interactive process with the environment. Particularly relevant to the topic at hand are recent findings that demonstrate that the relationship between smoking during pregnancy and low birth weight babies is modified by polymorphisms in two maternal metabolic genes, CYP1A1 and GSTT1 (44), and that children with two copies of a dopamine transporter (DAT) polymorphism are the ones likely to develop ADHD in the presence of parental maternal smoking (45).

POTENTIAL PATHWAY FOR ADVERSE EFFECTS Low Birth Weight In 1957, Simpson (46) reported on the adverse effect of maternal smoking on birth weight, and multiple studies have confirmed this finding (47–50). These studies show a direct dose–response (47,51). Magee and colleagues found the relative risk of low birth weight among smokers was 1.58. In this population, the rate of low birth weight increased from 6.38% in nonsmokers to 11.72% among heavy smokers (52). The effect of prenatal exposure on birth weight is more attributable to intrauterine growth retardation than to pre-term delivery. Kramer et al. (53) estimated the effect of prenatal maternal smoking as a

152

Weitzman et al.

5% reduction in relative weight per pack of cigarettes smoked per day, and Meyer and Comstock (54) reported that the effect of maternal cigarette smoking on infant birth weight was an average reduction of 150 g to over 300 g. Both maternal smoking and paternal smoking are associated with lower birth weight, but maternal smoking is associated with more of an effect on birth weight than paternal smoking (37,55). Ong et al. reported that infants of maternal smokers were symmetrically small at birth when compared to infants of nonsmokers, but showed complete catch-up growth by 12 month of age. The mechanisms by which catch-up growth occurs are unknown, but the authors of this study suggest that appetite programming and early childhood nutrition may play significant roles (56). Cigarette smoking is the single most important factor affecting birth weight in developed countries (57). Only one study we are aware of has found no effect of passive exposure to cigarette smoking among nonsmoking mothers (58). Reduction of smoking during pregnancy has been shown to improve infant birth weight (59). Prenatal maternal smoking affects the fetus in a number of ways that may result in chronic hypoxia and thus low birth weight. Placental vascular resistance increases when women smoke during pregnancy (60,61). Along with the known direct vasoconstrictive effect of nicotine, nitric oxide and prostacyclin deficiency may affect the uteroplacental blood flow and contribute to the impaired fetal nutrition of babies born to women who smoke (62). Maternal smoking during pregnancy also is associated with alterations of protein metabolism and enzyme activity in fetal cord blood (63). These may be secondary to irreversible changes in cellular function of the trophoblast and may contribute to fetal growth restriction. Cigarette smoking during pregnancy also transiently lowers maternal uterine blood flow and reduces the flow of oxygen from the uterus to the placenta (64). Increased levels of carboxyhemoglobin are found in both maternal and fetal blood when the mother smokes during pregnancy, and this can lead to fetal hypoxia because carboxyhemoglobin replaces oxyhemoglobin that normally releases oxygen to the fetal tissues (65). The fetus suffers chronic hypoxic stress as a consequence of maternal smoking, as evidenced by elevated neonatal hematocrit levels (66). Poor intrauterine growth has a lasting effect on subsequent development of children (67). Low birth weight infants are at increased risk of emotional and behavioral problems (68–70). Gray and colleagues recently reported that in a sample of low birth weight infants followed over 8 years, the odds of having clinically apparent behavioral problems was 57% higher in low birth weight children whose mother smoked during pregnancy compared to those with non-smoking mothers (71). The sequelae of low birth weight also include lowered cognitive abilities and hyperactivity (72). Breslau et al. (73) found an increase in neurologic soft signs among low birth weight children. These soft signs were in turn associated with increased risk for subnormal IQ and learning disorders among children with normal IQ. Low birth weight also is associated with an increased risk of reading and math disabilities (74). It remains unclear whether the modest decrements in birth weight associated with maternal smoking have neurobehavioral consequences among those who are not born prematurely or of substantially low birth weight, but data exist showing that even among all children born with normal birth weight, on average those with greater birth weights have higher IQs than those with lower birth weights (75). In Utero Brain Growth Maternal smoking increases the likelihood a child will be born with a small head circumference (76). Children born to smoking mothers experience catch-up growth in weight and partial catch-up growth in length, but the differences in head circumference

Tobacco Exposure and Children

153

persist to at least 5 years of age (77). No difference in head circumference measurements was found when women who are pregnant stop smoking prior to 32 wk gestation, as compared to children of nonsmoking mothers (78).

NEUROCOGNITIVE AND BEHAVIORAL OUTCOMES ASSOCIATED WITH MATERNAL SMOKING At least four comprehensive reviews of the animal model and epidemiologic studies of maternal prenatal tobacco smoke’s effects on brain development, behavior, and neurocognitive functioning of children have been published in the past decade (12,79–81). Cigarette smoke is comprised of more than 2000 chemical compounds (82), and only a relative few of these have been studied for their biologic activity (83,84). Both animal model and human epidemiologic studies have focused primarily on nicotine. Animal Models Animal studies provide experimental models where a toxic exposure can be isolated to the prenatal period and isolated to tobacco exposure. The majority of the animal research has focused on the toxic effects of nicotine, all have focused on exposures in the prenatal period, and the findings have been quite consistent despite the wide variability of study designs employed (79). Animal studies have confirmed that nicotine at doses not causing growth retardation is a neuro-teratogen and alters rodent brain development and behavior (85). Nicotine exposure to the prenatal brain may prematurely stimulate the shift from neuronal proliferation to differentiation, a shift normally occurring later in development. Thus, in the developing brain nicotine may act as a cholinergic signal mimicking the trophic effects of acetylcholine. This finding has been further elucidated in Rhesus macaque monkeys exposed to environmental tobacco smoke in gestation and the early neonatal period, demonstrating that nicotinic acetylcholine receptors are upregulated to sufficient degree to elicit damage to the developing brain not only in rodents, but also in primates (86). As the relative density and distribution of nicotine receptors change during the course of prenatal development, nicotine may elicit different effects at different developmental stages of the human brain (85). Prenatal nicotine exposure significantly increases adrenergic receptor binding in the cerebral cortex of adult animals (87–90). In mouse experiments where the dose and timing of nicotine were varied, Nasrat et al. (91) demonstrated that doses of nicotine equivalent to 20 cigarettes per day in humans resulted in shortening of the gestational period, particularly when that exposure occurred during the second and third trimester. Similar to the findings of human studies, prenatal exposure to nicotine in animal studies consistently is associated with lower birth weight in offspring (15,92–94,103). Animal studies also demonstrate that in multiple species (rats, mice, and guinea pigs), there is increased postnatal motor activity associated with in utero nicotine exposure (95,96,103). Such studies also have found attention and memory deficits in maze task performance (97–99) and mild deficits in learning (96,97,100,103). In two studies in rats, animals exposed to nicotine during gestation showed increased levels of anxiety, poor adaptation, and increased impulsivity in maze tasks with high doses of nicotine (103,104). In many cases, these studies have found alterations in attention, memory, and learning that are consistent with ADHD, but not all studies have found such findings (101–103).

154

Weitzman et al.

Nicotine suppresses DNA synthesis in newborn rat brains, especially in the cerebellum (105). Studies in rats have also demonstrated that chronic nicotine exposure alters levels of N-methyl-D-aspartate receptor (NMDAR) subunits in the auditory cortex, thus disrupting auditory cortex development by over-stimulating these glutamatergic synapses. This finding could be the foundation of the auditory-cognitive processing deficits in children exposed to prenatal nicotine (106). Nicotine also reduces dopaminergic activity in the offspring of nicotine-exposed pregnant females in the ventral tegmental area, nucleus accumbens, and striatum. It also reduces the uptake of serotonin (107) and is typically associated with rat hyperactive behavior (29,30). Thus, nicotine in experimental animals has been shown to alter in utero growth, and offspring’s cognitive and motor performance, DNA synthesis, and neurotransmitter function associated with mood. Epidemiologic Studies Observational studies involving humans using both cross-sectional and longitudinal data also suggest negative developmental consequences of children’s prenatal and early passive exposure to tobacco smoke. These studies have used samples that are ethnically, culturally, and socially diverse. These studies support the view that children’s behavior and cognition are adversely affected by prenatal and early childhood tobacco exposure. Many employ multivariate statistical analyses to control for numerous potential confounders of the association between prenatal tobacco exposure and children’s behavior and cognition. Adverse Behavior Outcomes in Children of Smoking Mothers Studies of children whose mothers smoked during pregnancy have consistently demonstrated that such children have higher rates of behavior problems than those not exposed. Olds (12), for example, notes in his paper of 1997 that 10 of 11 human studies reviewed found increased rates of child behavior problems and ADHD-like behaviors even after controlling for many potential confounders (108–121). These studies vary from the newborn period up through adolescence. Newborns and Preschoolers Fried et al. (122) reported increases in hypertonicity and heightened tremors and startles among neonates who were prenatally exposed to tobacco compared to neonates born to nonsmokers. Longo (123) found evidence for neonatal hyperactivity among offspring of smoking mothers. Law et al. found dose–response relationships between higher maternal salivary cotinine and newborn hyperexcitability, hypertonicity, and increased stress/ abstinence signs (124). Reijneveld et al. in a study of 3179 children aged 1–6 month from Holland, found maternal smoking to be the strongest and most consistent predictor of excessive crying (125). They concluded that if maternal smoking indeed causes excessive crying, it may explain between 2.1% and 17.2% of excessive crying during this age period depending upon the definition of excessive crying employed (125). Kelmanson et al. conducted a study involving 250 randomly selected 2–4 month old infants whose mothers completed the Early Infancy Temperament Questionnaires (126). Infants born to smoking mothers had more frequent fussy periods, more protesting behavior at face washing and bath, indifferent attitudes to the mother when held by a new person, extreme reactions during diapering and bowel movements, less attention to the parent during parent/infant play activity, and more sensitivity to wet diapers. They also were

Tobacco Exposure and Children

155

characterized by more intensive reactions compared to the babies of nonsmokers (126). Wakschlag and Hans examined a group of urban African American infants and found that prenatal smoking was associated with low sociability and negative emotionality scores in boys, but not in girls. These boys were only at risk for these symptoms if they had unresponsive maternal parenting during infancy, demonstrating the synergistic role of environment and prenatal nicotine exposure in the development of conduct problems in infants (127). In a study by Brook et al. (128), maternal smoking during pregnancy was associated with negativity in 2-year-old children. Rauh et al. reported that 2-year-old children exposed to environmental tobacco smoke without active maternal smoking during pregnancy were twice as likely to be cognitively delayed when compared with non-exposed children. Children with both environmental tobacco smoke exposure in pregnancy and a higher level of material hardship exhibited the greatest cognitive deficit. This study demonstrates that passive smoking alone in the home has cognitive effects on the developing fetus (129). Using the National Maternal and Infant Health Survey, Faden and Graubard reported that maternal cigarette smoking during pregnancy was associated with less well developed language, higher activity level, greater difficulty of management, fearfulness, decreased ability to get along with peers, and increased tantrums in a population of three-years old children (130). Williams et al. (131) reported on a prospective longitudinal study of 5342 5-year-old children in which maternal smoking during pregnancy was associated with externalizing behavior problems. These studies, like the studies noted below of older children, suggest that prenatal tobacco exposure may increase the risk for ADHD, Oppositional Defiant Disorder, and Conduct Disorder. School Age Children and Adolescents Weitzman et al. (121), in a longitudinal study involving 2256 U.S. children ages 4–11 years, found that children who were exposed postnatally and both prenatally and postnatally were more likely to have behavior problems, even after controlling for numerous potential confounders. In this study, there was evidence of a dose–effect response and the tobacco exposure effect was not limited to any particular area of children’s behavior, such as antisocial behaviors, anxiety, depression, hyperactivity, or easy distractibility. Fergusson et al. (132) also found maternal smoking during pregnancy to be associated with increased rates of behavior problems in a longitudinal study of a birth cohort of 1265 children up to age 18 years in New Zealand. This study used both teacher and mother reports, thereby eliminating the potential problem that smoking mothers may be less tolerant of children’s behaviors and more likely to report them as abnormal. A clear dose–response relationship between cigarettes smoked during pregnancy and disruptive behaviors, conduct disorder, and attention deficit was found after adjusting for confounding variables. Rantakallio et al. (116) found an association between prenatal cigarette smoking and later delinquency in the Finnish birth cohort study, and Wakschlag et al. (119), in a more recent prospective study in the United States, found that boys ages 7–12 were more likely to be referred for psychiatric care for conduct disorder if their mothers smoked during pregnancy. Wakschlag and Hans later demonstrated that maternal smoking was significantly associated with higher conduct disorder symptoms scores in urban African American 10-year-old boys, but not in girls (127). A study by Weissman et al. also demonstrated the increased risk of conduct disorder in boys, noting that males born to mothers who smoked during pregnancy had a four-fold increased risk of prepubertal-onset of conduct disorder. This study also demonstrated a five-fold increased risk of adolescentonset drug abuse/dependence in females prenatally exposed to nicotine (133). In a recent

156

Weitzman et al.

Dutch cohort, Batstra et al. found that after adjusting for confounding variables, children of mothers who smoked during pregnancy showed more signs of attention deficit and higher levels of externalizing behavior compared to non-exposed children, with no evidence against a dose–response relationship (134). In contrast, Silberg et al. found no increased rates of conduct disorder among adolescents with mothers who smoked (135). Maughan et al. used data from the 1970 British Birth Cohort Study to examine the links between maternal smoking in pregnancy and antisocial behavior in offspring (136). This study found a strong dose–response relationship between the extent of pregnancy smoking and childhood-onset conduct disorders, but no links with adolescent-onset antisocial behaviors. Controls for mothers’ subsequent smoking history modified this picture, suggesting that the prime risks for early-onset conduct problems may be associated with persistent maternal smoking or correlates of persistent smoking rather than with pregnancy smoking per se (136). In young adult subjects, Brennen et al. found a dose–response relationship between the amount of maternal prenatal smoking and criminal arrest and psychiatric hospitalization for substance abuse in both male and female offspring (137). Wakschlag et al. reviewed the literature on in utero exposure to maternal smoking as a risk factor for conduct disorder and delinquency and concluded that the association is independent of confounders, present across diverse contexts, and consistent with basic science, but that methodologic limitations preclude causal conclusions (138). They say that the research to date provides consistent support for, but not proof of, an etiologic role for prenatal smoking in the development of antisocial behavior (138). Maughn et al. also explored the role of confounding variables, specifically parental antisocial behavior, maternal depression, family disadvantage, and genetic influences, in the development of conduct problems in 5- and 7-year-olds. They found that prenatal smoking showed a strong dose–response relationship with child conduct problems, however when the confounding genetic and environmental variables were taken into account, the effects of prenatal smoking were reduced by at least 75% (139). This study highlights the multifactorial nature of childhood behavior problems, but does note that prenatal exposure to nicotine remains one of the many factors contributing to conduct difficulties in children. Cognitive Impairments Prenatal exposure to maternal smoking has been shown to adversely affect children’s performance on intelligence and achievement tests, as well as performance in school, although the findings in this area are not as consistent as those for increased rates of behavior problems. Butler and Goldstein (140) demonstrated that children whose mothers smoked 10 or more cigarettes per day were on average between 3 and 5 month delayed in reading, mathematics, and general ability when compared to offspring of nonsmokers. In a population of 10-year-old children, Cornelius et al. demonstrated that prenatal tobacco exposure was associated with a decrease in ability to learn a list of words, reproduce geometric designs from memory, use feedback to change problem-solving strategies, and complete manual dexterity tasks (141). A number of studies (109,110,115,142–145) demonstrated similar effects, while some found effects to virtually disappear after controlling for confounders (146–148). In families in which mothers smoked during some but not all pregnancies, exposed children performed worse on intelligence tests than their unexposed siblings (147). Similarly, children of women who quit smoking during pregnancy have been found to score higher on tests of cognitive ability than children whose mothers smoke throughout pregnancy (111). The Ottawa Prenatal Prospective Study provided longitudinal data regarding auditory processing, reading, and language development. Fried and Watkinson (149)

Tobacco Exposure and Children

157

found that infants born to maternal smokers have decreased rates of auditory habituation and increased sound thresholds. The children in this study at 12 and 24 month showed decreased responsiveness on auditory-related items on the Bayley Scales of Infant Development associated with prenatal cigarette exposure. By ages 3 and 4 years, language development as assessed by standardized tests was found to be adversely affected by maternal cigarette smoking (110). These findings were dose-related, and persisted in follow-up studies through age 12 years (142,150–152). In 13- to 16-year-old adolescents in this cohort, the strongest relationship between maternal smoking in pregnancy and cognitive variables were seen in overall general intelligence and auditory memory. Using the Wechsler Intelligence Scale for Children, the IQ was linearly and negatively associated with in utero cigarette exposure (153). Using the same cohort to study the effects of prenatal cigarette exposure on various facets of attention, the authors demonstrated that in 152 13- to 16-year-old adolescents, prenatal cigarette exposure was associated with a decreased ability to encode and retain information and an increased impulsivity element (154). A study by Olds et al. (155) estimated the effect of prenatal smoking on cognitive function after controlling for many potential confounders: smoking 10 or more cigarettes per day during pregnancy was independently associated with an average 4.35 point decrease in Stanford–Binet IQ scores. The same investigators also demonstrated that intervention with quality, long-term prenatal nurse home visitation can offset the impairment in intellectual functioning associated with substantial maternal smoking during pregnancy (156). While Olds et al. were unable to estimate how much of the prevention in intellectual impairment was due to smoking cessation itself, the data suggest that the increase in IQ was in part due to a reduction in maternal smoking during pregnancy. Denson et al. (157), in a case control study, showed hyperactivity to be associated in a dose–response manner with maternal smoking. Milberger et al. (114,158) also employed a case control study and found that prenatal tobacco exposure contributes to children’s ADHD. Linnet et al. reviewed literature assessing the relationship between prenatal exposure to nicotine and the risk of developing behavior problems related to ADHD in childhood (159). They concluded that the literature, 24 studies published between 1973 and 2002, demonstrated that nicotine exposure during pregnancy is a risk factor for ADHDrelated disorders among children (159). Mick et al. investigated the association between ADHD and prenatal exposure to maternal cigarette smoking using a retrospective, hospital-based, case-control study of 280 ADHD cases and 242 non-ADHD cases (160). ADHD cases were 2.1 times more likely to have been exposed to cigarettes in utero than were the non-ADHD controls (160). In a more recent twin study, Thapar et al. found a dose–response relationship between the number of cigarettes smoked during pregnancy and ADHD symptom scores, regardless of whether the child was rated by parents or teachers. Maternal smoking during pregnancy showed a significant association with ADHD symptoms that is additional to the influence of genetic factors contributing to the development of ADHD (161). Wasserman et al. examined the association between children’s behavior problems using the Child Behavior Checklist and maternal smoking in Yugoslavia, a country where there is no smoking–social class gradient (162). Mothers were enrolled during pregnancy in a prospective study of lead exposure and the children were assessed using the Child Behavior Checklist (CBCL) at ages 4, 41⁄2 and 5 years. Blood lead level measured twice a year was associated slightly with the delinquent scale. Smoking was associated with worse scores on almost all of the CBCL subscales for boys and with somatic complaints for girls,

158

Weitzman et al.

controlling for lead, HOME scores, maternal education, age, gender, lead exposure and birth weight. This study adds to the literature because it controls for lead and because it includes non-economically or socially disadvantaged families (162). School Performance Rantakallio (144) reported data from a 1966 birth cohort of 1819 Finnish children demonstrating that parental smoking is associated with lower mean scores on ‘theoretical subjects based on school reports.’ Byrd and Weitzman (163) demonstrated that children of smoking parents are more likely to repeat kindergarten or first grade, using the U.S. National Health Interview Survey. This study, however, was cross-sectional, and did not distinguish prenatal from childhood passive exposure to tobacco. A recent study of 1186 children aged 5.5–11 years demonstrated that children of smoking mothers performed worse on spelling and arithmetic tasks, and spelling problems were more pronounced when the mother continued to smoke after the child’s birth (134).

RESEARCH IMPLICATIONS Listed below are a series of 10 questions that emanate from the findings to date in this profoundly important field: 1. What additional information can be learned from animal model studies, what are the limitations of extrapolations from such data to effects on children, and what additional animal model studies are indicated? What hypotheses suggested by human studies can better be tested in animal studies? 2. What additional information can be derived from further human epidemiologic investigation, what samples and study designs are indicated, and how would we develop such studies? What hypotheses suggested by animal studies can be tested in epidemiologic studies? 3. Are the adverse neurocognitive and behavioral effects associated with tobacco exposure due to prenatal or early childhood passive exposure, or to both prenatal and postnatal exposure? This is a very difficult question to answer using human observational epidemiologic data, as a very small minority of women smoke only during pregnancy. Animal model studies to date have largely focused on prenatal smoke exposure’s effects on subsequent neurocognitive functioning of offspring. Studies in animals who are not exposed in utero, but are exposed after birth, would help answer this very important question. Moreover, there are different and more chemicals that youngsters are exposed to from environmental tobacco smoke than from direct smoking. Therefore, it is possible that some of these differential exposures may have differing effects. 4. Is there a unique neurobehavioral signature associated with prenatal and early passive exposure to tobacco smoke, or is it variable and does it vary by stage of development at which the fetus or child is exposed? 5. Do tobacco effects vary by whether the exposure is acute or chronic, and is there a critical period of exposure? Similarly, is the effect modified by other environmental exposures, such as to lead and mercury? 6. Do neurobehavioral problems change over the course of childhood, or are they static, and whether changing or static are their effects likely to be more deleterious at different stages of children’s lives?

Tobacco Exposure and Children

159

7. There is evidence of a dose–effect response between children’s tobacco exposure and various domains of behavior and neurocognitive functioning. Do the slopes of these relationships change over the range of exposure? 8. Are there children who are especially vulnerable to tobacco exposure, such as those growing up in poverty, or with mothers who are poorly educated or depressed, or children with intellectual impairments such as already low IQ, ADHD, or specific learning disabilities? Recent studies, as noted earlier, suggest that there are genetic as well as social characteristics that make certain children more vulnerable than others, but a more robust literature clearly is needed. 9. Are the adverse neurocognitive and behavioral effects reversible if mothers reduce or stop smoking, and if yes, are there times when, or by when it is especially important to reduce or curtail exposure? For example, does reduction or curtailment of tobacco exposure lead to decreased rates of behavior problems or decrements in IQ? Similarly, are there environmental control mechanisms that effectively could reduce children’s exposure to tobacco smoke? 10. Are there parenting strategies or social support interventions that can overcome the biologic effects of tobacco exposure? What are the effects of early intervention on children who have been exposed to maternal smoking?

SUMMARY A causal argument rests on the accumulation of evidence along five major domains: biologic plausibility, consistency, temporality, dose–response gradient, and strength of association. The literature on maternal smoking and subsequent child neurocognitive functioning provides evidence in each of these domains, yet the knowledge base remains incomplete. Biologic plausibility is difficult to demonstrate in human studies because of limitations in human epidemiologic research, but plausibility has been demonstrated in strong animal-model studies. However, the epidemiologic studies do provide a broad base of consistency across populations and across various study designs and endpoint measures. The temporal sequence of exposure preceding outcome, while somewhat cloudy in some human studies, is clearly evident in the animal models. Dose–response relationships remain unclear in many respects. However, again, the animal data provide a basis for concern. In any causal argument, a strong association is a key criterion largely because such associations are unlikely to be explained away by other factors. This is a very important consideration in the association between prenatal tobacco exposure and adverse behavioral and neurocognitive effects on children. Some remain skeptical (12,79,80,164) of research demonstrating adverse effects of smoking during pregnancy. As noted elsewhere, differences exist between smoking and nonsmoking mothers that might explain adverse outcomes among the offspring of smoking mothers, i.e., heavy and moderate smokers receive less prenatal care, recognize their pregnancies later, and report more symptoms of depression than do nonsmokers and light smokers (40). These differences suggest that smokers are more likely to be depressed and less likely to practice health-promoting behaviors for themselves or their children. However, a significant number of observational studies have controlled for numerous potential confounders, including depression and substance abuse, and the association between smoking and adverse child neurocognitive outcomes remained. Intervention

160

Weitzman et al.

studies and animal models also potently contradict the contention that maternal smoking simply is a proxy for other factors responsible for adverse neurocognitive and behavioral outcomes. The magnitude of the adverse behavioral and neurocognitive effects of tobacco exposure for individual children remains unclear, and some available measurements seem modest, i.e., a decrement of 4–5 IQ points and an odds ratio of approximately 1.5 for adverse developmental or behavioral outcomes. Yet, considerable public resources have been directed to other problems having a similar magnitude of effect, and given the vast numbers of children effected the net effect on a societal level would be expected to be quite large. It also is essential that we recognize that an insult of this type and magnitude, when coupled with other risks that tend to cluster among a significant percentage of exposed children, may have substantial effects on functioning and quality of life across the life span. While many questions remain, both animal model and human epidemiologic data clearly suggest a causal relationship between prenatal tobacco exposure and adverse behavioral and neurocognitive effects on children.

REFERENCES 1. Mathews TJ. Smoking During Pregnancy in the 1990s. National Vital Statistics Reports, editor. vol. 49 no. 7, 1-16. 8-28-2001. Hyattsville, MD, National Center for Health Statistics. National Vital Statistics Reports. 2. Etzel RA, Balk SJ, Bearer CF, Miller MD, Shea KM, Simon PR, et al. Environmental tobacco smoke: A hazard to children. Pediatrics 1997; 99:639–642. 3. Jacobs EA, Joffe A, Knight JR, Kulig J, Rogers PD. Tobacco’s toll: Implications for the pediatrician. Pediatrics 2001; 107:794–798. 4. Pregnancy risks determined from birth certificate data—United States, 1989. MMWR— Morbidity & Mortality Weekly Report 1992; 41:556–563. 5. Chandra A. Health Aspects of Pregnancy and Childbirth: United States, 1982–88. 23 No. 18. 1995. Hyattsville, Maryland, Center for Disease Control and Prevention. Vital Health Statistics. 6. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, Munson ML. Births: Final data for 2002. National Vital Statistics Reports; vol 52 no 10. Hyattsville, Maryland: National Center for Health Statistics, 2003. 7. Fried VM, Prager K, MacKay AP, Xia H. Chartbook on the Trends in the Health of Americans. Hyattsville, MD: National Center for Health Statistics, 2003. 8. Department of Health and Human Services. Healthy People 2010. McLean, VA: International Medical Publishing, 2001. 9. Ventura SJ, Martin JA, Mathews TJ, Clarke SC. Advance Report of Final Natality Statistics, 1994 Monthly Vital Statistics Report. 42, 12-13-60-63. 1996. Hyattsville, MD, National Center for Health Statistics. 10. Pirkle JL, Flegal KM, Bernett JT, Brody DJ, Etzel RA, Maurer KR. Exposure of the US population to environmental tobacco smoke: The Third National Health and Nutrition Examination Survey 1988–1991. JAMA 1996; 275:1233–1240. 11. Schuster MA, Franke T, Pham CB. Smoking patterns of household members and visitors in homes with children in the United States. Arch Pediatr Adolesc Med 2002; 156:1094–1100. 12. Olds D. Tobacco exposure and impaired development: A review of the evidence. Ment Retard Dev Disab Res Rev 1997; 3:257–269. 13. Hoffman JH, Welte JW, Barnes GM. Co-occurrence of alcohol and cigarette use among adolescents. Addictive Behav 2001; 26:63–78.

Tobacco Exposure and Children

161

14. Jacobson JL, Jacobson SW, Sokol RJ, Martier SS, Ager JW, Shankaran S. Effects of alcoholuse, smoking, and illicit drug-use on fetal growth in black infants. J Pediatr 1994; 124:757–764. 15. Miller JM, Boudreaux MC, Regan FA. A case-control study of cocaine use in pregnancy. Amer J Obstet Gynecol 1995; 172:180–185. 16. National Institute on Drug Abuse (NIDA). Drug Abuse and Drug Research.: US Department of Health and Human Services, 1993. 17. Koren G. The association between maternal cigarette smoking and psychiatric diseases or criminal outcome in the offspring: a precautionary note about the assumption of causation. Reprod Toxicol 1999; 13:345–346. 18. Hellerstedt WL, Pirie PL, Lando HA, Curry SJ, McBride CM, Grothaus LC, et al. Differences in preconceptional and prenatal behaviors in women with intended and unintended pregnancies. Amer J Public Health 1998; 88:663–666. 19. Siqueira L, Diab M, Bodian C, Rolnitzky L. Adolescents becoming smokers: The roles of stress and coping methods. J Adolesc Health 2000; 27:399–408. 20. Morales AW, Marks MN, Kumar R. Smoking in pregnancy: a study of psychosocial and reproductive risk factors. J Psychosom Obstet Gynecol 1997; 18:247–254. 21. Lanting CI, Fidler V, Huisman M, Touwen BCL, Boersma ER. Neurological differences between 9-year-old children fed breast-milk or formula-milk as babies. Lancet 1994; 344:1319–1322. 22. Niemela A, Jarvenpaa AL. Is breastfeeding beneficial and maternal smoking harmful to the cognitive development of children? Acta Paediatr 1996; 85:1202–1206. 23. Yerushalmy J. The relationship of parents’ cigarette smoking to outcome of pregnancy— implications as to the problem of inferring causation from observed associations. Amer J Epidemiol 1971; 93:443–456. 24. Brook JS, Cohen P, Brook DW. Longitudinal study of co-occurring psychiatric disorders and substance use. J Amer Acad Child Adolesc Psychiatry 1998; 37:322–330. 25. Hill SY, Lowers L, Locke-Wellman J, Shen SA. Maternal smoking and drinking during pregnancy and the risk for child and adolescent psychiatric disorders. J Studies Alcohol 2000; 61:661–668. 26. Kandel DB, Johnson JG, Bird HR, Canino G, Goodman SH, Lahey BB, et al. Psychiatric disorders associated with substance use among children and adolescents: Findings from the methods for the epidemiology of child and adolescent mental disorders (MECA) study. J Abnorm Child Psychol 1997; 25:121–132. 27. Yerushalmy J. The relationship of parents’ cigarette smoking to outcome of pregnancy— implications as to the problem of inferring causation from observed associations. Am J Epidemiol 1971; 93:443–456. 28. Lasser K, Boyd JW, Woolhandler S, Himmelstein DU, McCormick D, Bor DH. Smoking and mental illness—A population-based prevalence study. JAMA 2000; 284:2606–2610. 29. Muneoka K, Ogawa T, Kamei K, Mimura Y, Kato H, Takigawa M. Nicotine exposure during pregnancy is a factor which influences serotonin transporter density in the rat brain. Europ J Pharmacol 2001; 411:279–282. 30. Richardson SA, Tizabi Y. Hyperactivity in the offspring of nicotine-treated rats—role of the mesolimbic and nigrostriatal dopaminergic pathways. Pharmacol Biochem Behav 1994; 47:331–337. 31. Blum K, Wood RC, Braverman ER, Chen TJH, Sheridan PJ. The D2 Dopamine-receptor gene as a predictor of compulsive disease—Bayes Theorem. Functional Neurol 1995; 10:37–44. 32. Arinami T, Ishiguro H, Onaivi ES. Polymorphisms in genes involved in neurotransmission in relation to smoking. Europ J Pharmacol 2000; 27:215–226. 33. Blum K, Sheridan PJ, Wood RC, Braverman ER, Chen TJH, Comings DE. Dopamine D2 receptor gene variants—association and linkage studies in impulsive-addictive-compulsive behavior. Pharmacogenetics 1995; 5:121–141.

162

Weitzman et al.

34. Comings DE, Blum K. Reward deficiency syndrome: genetic aspects of behavioral disorders. Cognition, emotion and autonomic responses: the integrative role of the prefrontal cortex and limbic structures. Prog Brain Res 2000; 126:325–341. 35. Comings DE, Gade R, Wu S, Chiu C, Dietz G, Muhleman D, et al. Studies of the potential role of the dopamine D-1 receptor gene in addictive behaviors. Molec Psychiatry 1997; 2:44–56. 36. Comings DE, Ferry L, Bradshaw Robinson S, Burchette R, Chiu C, Muhleman D. The dopamine D-2 receptor (DRD2) gene: A genetic risk factor in smoking. Pharmacogenetics 1996; 6:73–79. 37. Eskenazi B, Castorina R. Association of prenatal maternal or postnatal child environmental tobacco smoke exposure and neurodevelopmental and behavioral problems in children. Environ Health Perspect 1999; 107:991–1000. 38. Trasti N, Vik T, Jacobsen G, Bakketeig LS. Smoking in pregnancy and children’s mental and motor development at age 1 and 5 years. Early Hum Develop 1999; 55:137–147. 39. Escalona SK. Babies at double hazard—Early development of infants at biologic and social risk. Pediatrics 1982; 70:670–676. 40. Sameroff AJ, Seifer R. Familial risk and child competence. Child Dev 1983; 54:1254–1268. 41. Sameroff AJ, Chandler MJ. Reproductive risk and the continuum of caretaking casuality. In: Horowitz FD, Hetherington M, Scarr-Salapetak S, Siegel R, eds. Review of Child Developmental Research. Chicago: University Press, 1975:187–244. 42. Ramey CT, Bryant DM, Sparling JJ, Wasik BH. A biosocial systems perspective on environmental interventions for low-birth-weight infants. Clin Obstet Gynecol 1984; 27:672–692. 43. Sameroff AJ. The social context of development. In: Eisenberg N, ed. Contemporary Topics in Developmental Psychology. Wiley, 1987:273–291. 44. Wang X, Zuckerman B, Pearson C, Kaufman G, Chen C, Wang G, et al. Maternal cigarette smoking, metabolic gene polymorphism, and infant birth weight. JAMA 1999; 287:195–202. 45. Kahn RS, Khoury J, Nichols WC, Lanphear BP. Role of dopamine transporter genotype and maternal prenatal smoking in childhood hyperactive-impulsive, inattentive, and oppositional behaviors. J Pediatr 2003; 143:104–110. 46. Simpson WJ. A preliminary report on cigarette smoking and the incidence of prematurity. Am J Obstet Gynecol 1957; 73:808–815. 47. Kleinman JC, Madans JH. The effects of maternal smoking, physical stature, and educationalattainment on the incidence of low birth-weight. Am J Epidemiol 1985; 121:843–855. 48. Kramer MS. Determinants of low birth-weight—Methodological assessment and metaanalysis. Bull WHO 1987; 65:663–737. 49. Macarthur C, Knox EG. Smoking in pregnancy—effects of stopping at different stages. Brit J Obstet Gynaecol 1988; 95:551–555. 50. MacMahon B, Alpert M, Salber EJ. Infant weight and parental smoking habits. Am J Epidemiol 1965; 82:247–261. 51. Kline J, Stein Z, Hutzler M. Cigarettes, alcohol and marijuana—varying associations with birth-weight. Int J Epidemiol 1987; 16:44–51. 52. Magee BD, Hattis D, Kivel NM. Role of smoking in low birth weight. J Reprod Med 2004; 49:23–27. 53. Kramer MS, Olivier M, Mclean FH, Dougherty GE, Willis DM, Usher RH. Determinants of fetal growth and body proportionality. Pediatrics 1990; 86:18–26. 54. Meyer MB, Comstock GW. Maternal cigarette smoking and perinatal mortality. Am J Epidemiol 1972; 96:1–10. 55. Matsubara F, Kida M, Tamakoshi A, Wakai K, Kawamura T, Ohno Y. Maternal active and passive smoking and fetal growth: A prospective study in Nagoya, Japan. J Epidemiol 2000; 10:335–343. 56. Ong KKL, Preece MA, Emmett PM, Ahmed ML, Dunger DB. Size at birth and early childhood growth in relation to maternal smoking, parity and infant breast-feeding: longitudinal birth cohort study and analysis. Pediatr Res 2002; 52:863–867.

Tobacco Exposure and Children

163

57. Kramer MS. Intrauterine growth and gestational duration determinants. Pediatrics 1987; 80:502–511. 58. Haug K, Irgens LM, Skjaerven R, Markestad T, Baste V, Schreuder P. Maternal smoking and birthweight: effect modification of period, maternal age and paternal smoking. Acta Obstet Gynecol Scand 2000; 79:485–489. 59. Sexton M, Hebel JR. A clinical-trial of change in maternal smoking and its effect on birthweight. JAMA 1984; 251:911–915. 60. Howard RB, Hosokawa T, Maguire MH. Hypoxia-induced fetoplacental vasoconstriction in perfused human placental cotyledons. Am J Obstet Gynecol 1987; 157:1261–1266. 61. Lehtovirta P, Forss M. Acute effect of smoking on intervillous blood-flow of placenta. Brit J Obstet Gynaecol 1978; 85:729–731. 62. Ulm MR, Plockinger B, Pirich C, Gryglewski RJ, Sinzinger HF. Umbilical arteries of babies born to cigarette smokers generate less prostacyclin and contain less arginine and citrulline compared with those of babies born to control subjects. Am J Obstet Gynecol 1995; 172:1485–1487. 63. Jauniaux E, Biernaux V, Gerlo E, Gulbis B. Chronic maternal smoking and cord blood amino acid and enzyme levels at term. Obstet Gynecol 2001; 97:57–61. 64. Morrow RJ, Ritchie JWK, Bull SB. Maternal cigarette-smoking—the effects on umbilical and uterine blood-flow velocity. Am J Obstet Gynecol 1988; 159:1069–1071. 65. Soothill PW, Morafa W, Ayida GA, Rodeck CH. Maternal smoking and fetal carboxyhaemoglobin and blood gas levels. Brit J Obstet Gynaecol 1996; 103:78–82. 66. Bush PG, Mayhew TM, Abramovich DR, Aggett PJ, Burke MD, Page KR. Maternal cigarette smoking and oxygen diffusion across the placenta. Placenta 2000; 21:824–833. 67. Dunn HG, Mcburney AK, Ingram S, Hunter CM. Maternal cigarette-smoking during pregnancy and childs subsequent development.1. physical growth to age of 6 1/2-years. Canadian J Public Health-Revue Canadienne de Sante Publique 1976; 67:499–505. 68. Barros FC, Huttly SRA, Victora CG, Kirkwood BR, Vaughan JP. Comparison of the causes and consequences of prematurity and intrauterine growth-retardation—a longitudinal-study in southern Brazil. Pediatrics 1992; 90:238–244. 69. McCarton C. Behavioral outcomes in low birth weight infants. Pediatrics 1998; 102:1293–1297. 70. Pharoah POD, Stevenson CJ, Cooke RWI, Stevenson RC. Prevalence of behavior disorders in low-birth-weight infants. Arch Dis Child 1994; 70:271–274. 71. Gray RF, Indurkhya A, McCormick MC. Prevalence, stability, and predictors of clinically significant behavior problems in low birth weight children at 3, 5, and 8 years of age. Pediatrics 2004; 114:736–743. 72. Breslau N, Chilcoat HD. Psychiatric sequelae of low birth weight at 11 years of age. Biolog Psychiatry 2000; 47:1005–1011. 73. Breslau N, Chilcoat HD, Johnson EO, Andreski P, Lucia VC. Neurologic soft signs and low birthweight: their association and neuropsychiatric implications. Biolog Psychiatry 2000; 47:71–79. 74. Johnson EO, Breslau N. Increased risk of learning disabilities in low birth weight boys at age 11 years. Biolog Psychiatry 2000; 47:490–500. 75. Matte TD, Bresnahan M, Begg MD, Susser E. Influence of variation in birth weight within normal range and within sibships on IQ at age 7 years: cohort study. BMJ 2001; 323:310–314 [erratum appears in BMJ 2001 Sep 22;323(7314):684]. 76. Kallen K. Maternal smoking during pregnancy and infant head circumference at birth. Early Hum Develop 2000; 58:197–204. 77. Vik T, Jacobsen G, Vatten L, Bakketeig LS. Pre- and post-natal growth in children of women who smoked in pregnancy. Early Human Develop 1996; 45:245–255. 78. Lindley AA, Becker S, Gray RH, Herman AA. Effect of continuing or stopping smoking during pregnancy on infant birth weight, crown-heel length, head circumference, ponderal index, and brain:body weight ratio. Am J Epidemiol 2000; 152:219–225.

164

Weitzman et al.

79. Ernst M, Moolchan ET, Robinson ML. Behavioral and neural consequences of prenatal exposure to nicotine. J Amer Acad Child Adolesc Psychiatry 2001; 40:630–641. 80. National Cancer Institute. Health Effects of Exposure to Environmental Tobacco Smoke: The Report of the California Environmental Protection Agency. NIH Pub. 99-4645. Bethesda, MD, US Department of Health and Human Services, National Institutes of Health, National Cancer Institute. Smoking and Tobacco Control Monograph No. 10, 1999. 81. Weitzman M, Byrd RS, Aligne CA, Moss M. The effects of tobacco exposure on children’s behavioral and cognitive functioning: Implications for clinical and public health policy and future research. Neurotoxicol Teratol 2002; 24:397–406. 82. Johnson JG, Cohen P, Pine DS, Klein DF, Kasen S, Brook JS. Association between cigarette smoking and anxiety disorders during adolescence and early adulthood. JAMA 2000; 284:2348–2351. 83. Dube MF, Green CR. Methods of collection of smoke for analytical purposes. Recent Adv Tobacco Sci 1982; 8:42–102. 84. Naeye RL. Cognitive and behavioral abnormalities in children whose mothers smoked cigarettes during pregnancy. J Dev Behav Pediatr 1992; 13:425–428. 85. Hellstrom-Lindahl E, Nordberg A. Smoking during pregnancy: A way to transfer the addiction to the next generation? Respiration 2002; 69:289–293. 86. Slotkin TA, Pinkerton KE, Auman JT, Qiao D, Seidler FJ. Perinatal exposure to environmental tobacco smoke upregulates nicotinic cholinergic receptors in monkey brain. Develop Brain Res 2002; 133:175–179. 87. Navarro HA, Mills E, Seidler FJ, Baker FE, Lappi SE, Tayyeb MI, et al. Prenatal nicotine exposure impairs beta-adrenergic function—persistent chronotropic subsensitivity despite recovery from deficits in receptor-binding. Brain Res Bull 1990; 25:233–237. 88. Navarro HA, Slotkin TA, Tayyeb MI, Lappi SE, Seidler FJ. Effects of fetal nicotine exposure on development of adrenergic-receptor binding in rat-brain regions—Selective changes in alpha-1-receptors. Res Comm Substances Abuse 1990; 11:95–103. 89. Peters DAV. Prenatal nicotine exposure increases adrenergic-receptor binding in the rat cerebral-cortex. Res Comm Chem Pathol Pharmacol 1984; 46:307–317. 90. Tizabi Y, Perry DC. Prenatal nicotine exposure is associated with an increase in [I125]epibatidine binding in discrete cortical regions in rats. Pharmacol Biochem Behav 2000; 67:319–323. 91. Nasrat HA, Alhachim GM, Mahmood FA. Perinatal effects of nicotine. Biology Neonate 1986; 49:8–14. 92. Bassi JA, Rosso P, Moessinger AC, Blanc WA, James LS. Fetal growth retardation due to maternal tobacco smoke exposure in the rat. Pediatr Res 1984; 18:127–130. 93. Leichter J. Decreased birth-weight and attainment of postnatal catch-up growth in offspring of rats exposed to cigarette-smoke during gestation. Growth Develop Aging 1995; 59:63–66. 94. Leichter J. Effect of maternal exposure to smoke from low and high-yield cigarettes on fetal growth in the rat. Biochem Arch 1993; 9:1–6. 95. Ajarem JS, Ahmad M. Prenatal nicotine exposure modifies behavior of mice through early development. Pharmacol Biochem Behav 1998; 59:313–318. 96. Johns JM, Louis TM, Becker RF, Means LW. Behavioral-effects of prenatal exposure to nicotine in guinea-pigs. Neurobehav Toxicol Teratol 1982; 4:365–369. 97. Levin ED, Briggs SJ, Christopher NC, Rose JE. Prenatal nicotine exposure and cognitive performance in rats. Neurotoxicol Teratol 1993; 15:251–260. 98. Peters MA, Ngan LLE. The effects of totigestational exposure to nicotine on prenatal and postnatal-development in the rat. Archives Internationales de Pharmacodynamie et de Therapie 1982; 257:155–167. 99. Yanai J, Pick CG, Rogelfuchs Y, Zahalka EA. Alterations in hippocampal cholinergic receptors and hippocampal behaviors after early exposure to nicotine. Brain Res Bull 1992; 29:363–368. 100. Genedani S, Bernardi M, Bertolini A. Sex-linked differences in avoidance-learning in the offspring of rats treated with nicotine during pregnancy. Psychopharmacol 1983; 80:93–95.

Tobacco Exposure and Children

165

101. Bertolini A, Bernardi M, Genedani S. Effects of prenatal exposure to cigarette-smoke and nicotine on pregnancy, offspring development and avoidance-behavior in rats. Neurobehav Toxicol Teratol 1982; 4:545–548. 102. Paulson RB, Shanfeld J, Vorhees CV, Sweazy A, Gagni S, Smith AR, et al. Behavioral-effects of prenatally administered smokeless tobacco on rat offspring. Neurotoxicol Teratol 1993; 15:183–192. 103. Vaglenova J, Birru S, Pandiella NM, Breese CR. An assessment of the long-term developmental and behavioral teratogenicity of prenatal nicotine exposure. Behav Brain Res 2004; 150:159–170. 104. Sobrian SK, Marr L, Ressman K. Prenatal cocaine and/or nicotine exposure produces depression and anxiety in aging rats. Prog Neuro-Psychopharmacol Biolog Psychiatry 2003; 27:501–518. 105. Slotkin TA, Greer N, Faust J, Cho H, Seidler FJ. Effects of maternal nicotine injections on brain-development in the rat—ornithine decarboxylase activity, nucleic-acids and proteins in discrete brain-regions. Brain Res Bull 1986; 17:41–50. 106. Hsieh CY, Leslie FM, Metherate R. Nicotine exposure during a postnatal critical period alters NR2A and NR2B mRNA expression in rat auditory forebrain. Dev Brain Res 2002; 133:19–25. 107. Thomas JD, Garrison ME, Slawecki CJ, Ehlers CL, Riley EP. Nicotine exposure during the neonatal brain growth spurt produces hyperactivity in preweanling rats. Neurotoxicol Teratol 2000; 22:695–701. 108. Day NL, Richardson GA, Goldschmidt L, Cornelius MD. Effects of prenatal tobacco exposure on preschoolers’ behavior. J Dev Behav Pediatr 2000; 21:180–188. 109. Dunn HG, Mcburney AK, Ingram S, Hunter CM. Maternal cigarette-smoking during pregnancy and childs subsequent development—neurological and intellectual maturation to age of 6 1/2 years. Canadian Journal of Public Health-Revue Canadienne de Sante Publique 1977; 68:43–50. 110. Fergusson DM, Horwood LJ, Lynskey MT. Maternal smoking before and after pregnancy— effects on behavioral outcomes in middle childhood. Pediatrics 1993; 92:815–822. 111. Fried PA, Watkinson B. 36-month and 48-month neurobehavioral follow-up of children prenatally exposed to marijuana, cigarettes, and alcohol. J Dev Behav Pediatr 1990; 11:49–58. 112. Hardy JB, Mellits ED. Does maternal smoking during pregnancy have a long-term effect on the child? Lancet 1972; 2:1332–1336. 113. Kristjansson EA, Fried PA, Watkinson B. Maternal smoking during pregnancy affects childrens’ vigilance performance. Drug Alcohol Depend 1989; 24:11–19. 114. Milberger S, Biederman J, Faraone SV, Chen L, Jones J. Maternal smoking during pregnancy a risk factor for attention deficit hyperactivity disorder in children? Am J Psychiatry 1996; 153:1138–1142. 115. Naeye RL, Peters EC. Mental-development of children whose mothers smoked during pregnancy. Obstet Gynecol 1984; 64:601–607. 116. Rantakallio P, Laara E, Isohanni M, Moilanen I. Maternal smoking during pregnancy and delinquency of the offspring—an association without causation. Int J Epidemiol 1992; 21:1106–1113. 117. Streissguth AP, Barr HM, Sampson PD, Parrishjohnson JC, Kirchner GL, Martin DC. Attention, distraction and reaction-time at age 7 years and prenatal alcohol exposure. Neurobehav Toxicol Teratol 1986; 8:717–725. 118. Streissguth AP, Martin DC, Barr HM, Sandman BM, Kirchner GL, Darby BL. Intrauterine acohol and nicotine exposure—attention and reaction-time in 4-year-old children. Dev Psychol 1984; 20:533–541. 119. Wakschlag LS, Lahey BB, Loeber R, Green SM, Gordon RA, Leventhal BL. Maternal smoking during pregnancy and the risk of conduct disorder in boys. Arch Gen Psychiatry 1997; 54:670–676.

166

Weitzman et al.

120. Walker A, Rosenberg M, Balaban-Gil K. Neurodevelopmental and neurobehavioral sequelae of selected substances of abuse and psychiatric medications in utero. Child Adolesc Psychiatric Clin North Amer 1999; 8:845–855. 121. Weitzman M, Gortmaker S, Sobol A. Maternal smoking and behavior problems of children. Pediatrics 1992; 90:342–349. 122. Fried PA, Watkinson B, Dillon RF, Dulberg CS. Neonatal neurological status in a low-risk population after prenatal exposure to cigarettes, marijuana, and alcohol. J Dev Behav Pediatr 1987; 8:318–326. 123. Longo LD. Biological effects of carbon-monoxide on pregnant woman, fetus, and newborninfant. Am J Obstet Gynecol 1977; 129:69–103. 124. Law KL, Stroud LR, LaGasse LL, Niaura R, Liu J, Lester BM. Smoking during pregnancy and newborn neurobehavior. Pediatrics 2003; 111:1318–1323. 125. Reijneveld SA, Brugman E, Hirasing RA. Excessive infant crying: definitions determine risk groups. Arch Dis Child 2002; 87:43–44. 126. Kelmanson IA, Erman LV, Litvina SV. Maternal smoking during pregnancy and behavioural characteristics in 2–4-month-old infants. Klin Padiatr 2002; 214:359–364. 127. Wakschlag LS, Hans SL. Maternal smoking during pregnancy and conduct problems in highrisk youth: A developmental framework. Dev Psychopath 2002; 14:351–369. 128. Brook JS, Brook DW, Whiteman M. The influence of maternal smoking during pregnancy on the toddler’s negativity. Arch Pediatr Adolesc Med 2000; 154:381–385. 129. Rauh VA, Whyatt RM, Garfinkel R, et al. Developmental effects of exposure to environmental tobacco smoke and material hardship among inner-city children. Neurotoxicol Teratol 2004; 26:373–385. 130. Faden VB, Graubard BI. Maternal substance use during pregnancy and developmental outcome at age three. J Substance Abuse 2000; 12:329–340. 131. Williams GM, O’Callaghan M, Najman JM, Bor W, Andersen MJ, Richards D, et al. Maternal cigarette smoking and child psychiatric morbidity: A longitudinal study. Pediatrics 1998; 102:e11. 132. Fergusson DM, Woodward LJ, Horwood LJ. Maternal smoking during pregnancy and psychiatric adjustment in late adolescence. Arch Gen Psychiatry 1998; 55:721–727. 133. Weissman MM, Warner V, Wickramaratne PJ, Kandel DB. Maternal smoking during pregnancy and psychopathology in offspring followed to adulthood. J Amer Acad Child Adolesc Psychiatry 1999; 38:892–899. 134. Batstra L, Hadders-Algra M, Neeleman J. Effect of antenatal exposure to maternal smoking on behavioural problems and academic achievement in childhood: prospective evidence from a Dutch birth cohort. Early Hum Develop 2003; 75:21–33. 135. Silberg JL, Parr T, Neale MC, Rutter M, Angold A, Eaves LJ. Maternal smoking during pregnancy and risk to boys’ conduct disturbance: An examination of the causal hypothesis. Biolog Psychiatry 2003; 53:130–135. 136. Maughan B, Taylor C, Taylor A, Butler N, Bynner J. Pregnancy smoking and childhood conduct problems: a causal association? J Child Psychol Psychiatr 2001; 42:1021–1028. 137. Brennan PA, Grekin ER, Mortensen EL, Mednick SA. Relationship of maternal smoking during pregnancy with criminal arrest and hospitalization for substance abuse in male and female adult offspring. Am J Psychiatry 2002; 159:48–54. 138. Wakschlag LS, Pickett KE, Cook E, Benowitz NL, Leventhal BL. Maternal smoking during pregnancy and severe antisocial behavior in offspring: A review. Am J Public Health 2002; 92:966–974. 139. Maughn B, Taylor A, Caspi A, Moffitt TE. Prenatal smoking and early childhood conduct problems: Testing genetic and environmental explanations of the association. Arch Gen Psychiatry 2004; 61:836–843. 140. Butler NR, Goldstein H. Smoking in pregnancy and subsequent child development. Brit Med J 1973; 4:573–575. 141. Cornelius MD, Ryan CM, Day NL, Goldschmidt L, Willford JA. Prenatal tobacco effects on neuropsychological outcomes among preadolescents. J Dev Behav Pediatr 2001; 22:217–225.

Tobacco Exposure and Children

167

142. Fogelman KR, Manor O. Smoking in pregnancy and development into early adulthood. Brit Med J 1988; 297:1233–1236. 143. Fried PA, Oconnell CM, Watkinson B. 60-month and 72-month follow-up of children prenatally exposed to marijuana, cigarettes, and alcohol—cognitive and language assessment. J Dev Behav Pediatr 1992; 13:383–391. 144. Rantakallio P. A follow-up-study up to the age of 14 of children whose mothers smoked during pregnancy. Acta Paediatr Scandinav 1983; 72:747–753. 145. Rantakallio P, Koiranen M. Neurological handicaps among children whose mothers smoked during pregnancy. Prev Med 1987; 16:597–606. 146. Baghurst PA, Tong SL, Woodward A, McMichael AJ. Effects of maternal smoking upon neuropsychological development in early childhood: importance of taking account of social and environmental factors. Paediatr Perinat Epidemiol 1992; 6:403–415. 147. Fergusson DM, Lloyd M. Smoking during pregnancy and its effects on child cognitive ability from the ages of 8 to 12 years. Paediatr Perinat Epidemiol 1991; 5:189–200. 148. Mcgee R, Stanton WR. Smoking in pregnancy and child-development to age 9 years. J Paediatr Child Health 1994; 30:263–268. 149. Fried PA, Watkinson B. 12-month and 24-month neurobehavioural follow-up of children prenatally exposed to marihuana, cigarettes and alcohol. Neurotoxicol Teratol 1988; 10:305–313. 150. Fried PA, Watkinson B. Visuoperceptual functioning differs in 9- to 12-years olds prenatally exposed to cigarettes and marihuana. Neurotoxicol Teratol 2000; 22:11–20. 151. Fried PA, Watkinson B, Gray R. Differential effects on cognitive functioning in 9- to 12-years olds prenatally exposed to cigarettes and marihuana. Neurotoxicol Teratol 1998; 20:293–306. 152. Fried PA, Watkinson B, Siegel LS. Reading and language in 9- to 12-years olds prenatally exposed to cigarettes and marijuana. Neurotoxicol Teratol 1997; 19:171–183. 153. Fried PA, Watkinson B, Gray R. Differential effects on cognitive functioning in 13- to 16year-olds prenatally exposed to cigarettes and marihuana. Neurotoxicol Teratol 2003; 25:427–436. 154. Fried PA, Watkinson B. Differential effects on facets of attention in adolescents prenatally exposed to cigarettes and marihuana. Neurotoxicol Teratol 2001; 23:421–430. 155. Olds DL, Henderson CR, Tatelbaum R. Intellectual impairment in children of women who smoke cigarettes during pregnancy. Pediatrics 1994; 93:221–227. 156. Olds DL, Henderson CR, Tatelbaum R. Prevention of intellectual impairment in children of women who smoke cigarettes during pregnancy. Pediatrics 1994; 93:228–233. 157. Denson R, Nanson JL, Mcwatters MA. Hyperkinesis and maternal smoking. Can Psychiatric Assoc J 1975; 20:183–187. 158. Milberger S, Biederman J, Faraone SV, Jones J. Further evidence of an association between maternal smoking during pregnancy and attention deficit hyperactivity disorder: Findings from a high-risk sample of siblings. J Clin Child Psychol 1998; 27:352–358. 159. Linnet KM, Dalsgaard S, Obel C, Wisborg K, Henriksen TB, Rodriguez A, et al. Maternal lifestyle factors in pregnancy risk of attention deficit hyperactivity disorder and associated behaviors: Review of the current evidence. Am J Psychiatry 2003; 160:1028–1040. 160. Mick E, Biederman J, Faraone SV, Sayer J, Kleinman S. Case-control study of attentiondeficit hyperactivity disorder and maternal smoking, alcohol use, and drug use during pregnancy. J Amer Acad Child Adol Psychiatry 2002; 41:378–385. 161. Thapar A, Fowler T, Rice F, et al. Maternal smoking during pregnancy and attention deficit hyperactivity disorder symptoms in offspring. Am J Psychiatry 2003; 160:1985–1989. 162. Wasserman GA, Liu XH, Pine DS, Graziano JH. Contribution of maternal smoking during pregnancy and lead exposure to early child behavior problems. Neurotoxicol Teratol 2001; 23:13–21. 163. Byrd RS, Weitzman ML. Predictors of early grade retention among children in the United States. Pediatrics 1994; 93:481–487. 164. Ramsay MC, Reynolds CR. Does smoking by pregnant women influence IQ, birth weight, and developmental disabilities in their infants? A methodological review and multivariate analysis Neuropsychol Rev 2000; 10:1–40.

9 Lessons Learned: Fetal Alcohol Spectrum Disorders Christie L. McGee, Susanna L. Fryer, and Sarah N. Mattson Center for Behavioral Teratology, Department of Psychology, San Diego State University, San Diego, California, U.S.A.

INTRODUCTION Fetal Alcohol Syndrome (FAS) is a devastating developmental disorder resulting from heavy prenatal alcohol exposure. Worldwide incidence rates of FAS average about 1 case per 1000 live births (1), which makes FAS the leading preventable cause of mental retardation (2). The diagnosis of FAS requires a constellation of three symptom classes: (1) growth deficiency, (2) cranio-facial abnormalities, and (3) central nervous system (CNS) dysfunction (3–5). Research from animal models of prenatal ethanol exposure and from observational human exposure studies has shown evidence of a dose–response relationship to alcohol’s teratogenic capacity (6). FAS falls at the most severe end of the outcome spectrum, and is associated with high levels of maternal alcohol consumption, such as those associated with alcohol abuse or dependence. However, in other cases, prenatal alcohol exposure can produce neurobehavioral dysfunction in the absence of the gross physical abnormalities required for clinical recognition of FAS (7). The reason for this disparity of outcome is unknown but may be related to dose or timing of exposure or other maternal or fetal considerations. However, it is apparent that more individuals are adversely affected by prenatal alcohol exposure than those meeting the diagnostic criteria for FAS. Recognizing the wide-ranging effects of prenatal alcohol exposure, the National Task Force on Fetal Alcohol Syndrome and Fetal Alcohol Effect has adopted the umbrella term Fetal Alcohol Spectrum Disorders (FASD) to describe the outcome range resulting from prenatal alcohol exposure (8). These effects vary from full-blown FAS to subtle neurobehavioral, growth, or physical deficits. Using this classification, FAS would be a case of dysmorphic FASD (i.e., all three of the diagnostic criteria described above are met, including the characteristic pattern of facial anomalies). Individuals with a documented history of prenatal alcohol exposure and some neurobehavioral deficits thought to be related to that exposure, who lack the distinct pattern of structural anomalies in FAS, would be considered to have nondysmorphic FASD. Historically, several terms have been used to classify individuals with histories of alcohol exposure who do not meet full criteria for FAS. These terms include Fetal Alcohol Effects (FAE), Alcohol-Related Birth Defects (ARBD), 169

170

McGee et al.

Alcohol-Related Neurodevelopmental Disorder (ARND), Prenatal Exposure to Alcohol (PEA), and static encephalopathy. For reasons of comparison and coherence of this chapter, we will adopt the use of the classification system proposed by the National Task Force discussed above. Thus, individuals affected by prenatal alcohol exposure will be identified collectively as FASD. As noted previously, nondysmorphic and dysmorphic individuals have been found to perform similarly on neuropsychological measures; thus, the distinction between nondysmorphic and dysmorphic FASD will only be noted when differences have emerged. This methodologically focused review should provide the reader with a good understanding of what is currently known about the teratogenic effects of alcohol on neuropsychological functioning and identify important areas of functioning that may be vulnerable to other teratogens. For each neuropsychological domain of interest, discussion of results from prospective studies will be followed with discussion of results from retrospective studies, so that the reader may better understand the ways in which methodological differences in study design might influence data interpretation. It is our hope that the study of other teratogens may benefit from this knowledge so that future studies can hone research designs and analyses, to optimally answer questions of interest to the field of behavioral teratology.

METHODOLOGY Before describing neuropsychological outcomes of prenatally exposed individuals, examination of possible methodological concerns of such studies should be highlighted. Researchers must carefully consider the ascertainment method used when designing studies, as both prospective and retrospective designs have advantages and disadvantages and implications for conclusions and generalizability. In this discussion, methodological concerns of prospectively designed studies will be presented first, followed by those of retrospectively designed studies. It is the behavioral teratogenicity of alcohol, that is, the drug’s ability to cause birth defects of a cognitive or behavioral nature, which has been the focus of population-based research of prenatal alcohol exposure at mostly moderate or “social-drinking” levels. The goal of such studies is systematic investigation of a fundamental hypothesis in behavioral teratology, which is that in cases where high levels of exposure to a substance lead to severe developmental impairment, lower exposure levels will lead to more subtle deficits (9). Usually such studies have adopted a prospective, longitudinal study design. In prospective studies, the study participants are identified before the outcome of interest occurs. Typically, prospective subject recruitment entails screening pregnant women to identify those that are likely to give birth to offspring exposed to the teratogen of interest, or identifying exposed infants at the time of birth (or very shortly thereafter). Because of the prospective nature of these studies, the number of cases with heavy exposure (or FAS) are typically very small and thus sample sizes must be very large to detect alcohol effects. In addition, the clinical significance of small effect sizes must be considered. Prospective studies have several advantages, and can be especially useful in the case of prenatal alcohol exposure, given potential reliability problems with retrospective maternal recall of drinking (10). Moreover, because alcohol leaves no long-lasting metabolite behind in the body, postnatal physiological verification of maternal alcohol dosage is not possible. Additionally, prospective identification of study participants offers greater capability to control for confounding factors, because environmental and demographic information can be collected more accurately at or near the time of

Fetal Alcohol Spectrum Disorders

171

exposure. Most prospective studies of prenatal alcohol exposure are also longitudinal. In a longitudinal design, study participants are repeatedly assessed on the response measures of choice. In contrast to a cross-sectional design, where outcomes are measured at one time point, in a longitudinal study, subjects are studied throughout the course of development, and conclusions can be drawn regarding changes in outcome over time. For these reasons, a longitudinal study design with prospective subject recruitment is advantageous over a cross-sectional study design with retrospective subject recruitment. However, prospective, longitudinal studies are time-consuming and expensive to conduct and so are often not feasible to undertake. Prospective longitudinal designs allow for more accurate exposure measurement. In the Seattle Study on Alcohol and Pregnancy, a large-scale population based study of prenatal alcohol exposure, Streissguth and colleagues highlighted the importance of measuring more than a single dimension of prenatal exposure (11). The Seattle study used a quantity-frequency-variability structured interview to assess maternal alcohol intake, thereby obtaining several measurements about the dose, timing, and nature of the exposure to alcohol. Outcome measure selection is yet another important consideration in study design. Whenever possible, it is advisable to consult with clinicians who have observed or treated both the physical and psychological symptoms of individuals with known exposures to the teratogen under study. Of course, this ideal may not be possible in the initial phases of investigation. In such cases, clinical expertise should be incorporated at a time when the teratogenic effects of the compound become more widely recognized (11). There are also special considerations when undertaking a prospective, longitudinal study of behavioral teratogenesis. For instance, causal inferences between teratogenic exposure and neurobehavioral outcome must be considered in light of potentially confounding influences such as demographic information, prenatal exposure to other drugs, postnatal environment, and child’s age (12). Additionally, the failure to detect effects that actually exist, often referred to as “false negatives” or Type II error, is especially problematic in neurotoxicity research, given the potential for studies with negative results to be interpreted in public policy arenas as demonstrating little or no public health threat of the exposure in question. When designing a behavioral neurotoxicity study, a priori power analyses can be used to determine the sample size necessary to detect an effect. Jacobson and Jacobson also list common sources of Type II error in behavioral teratology, including failure to sample adequately high-level exposures, over-controlling for confounding influences, and statistical overcorrection of multiple comparisons (12). Thus, one must strike a balance between controlling confounds and Type II error reduction. In order to detect adverse outcomes, it is frequently necessary to oversample individuals who have experienced high levels of exposure; failure to do so can lead to inaccurate conclusions regarding the teratogenic capacity of an agent (13). Vorhees underscored the utility and efficiency, for government regulatory purposes, of focusing research on the high-level end of teratogenic exposure (9). Retrospective studies on the effects of prenatal alcohol exposure typically focus on individuals with heavy exposure levels, thus filling an important niche in FASD research. In contrast to prospectively-focused studies, in retrospective studies subjects are included based on past events (i.e., prenatal alcohol exposure). For example, retrospective recruitment for a study of alcohol’s teratogenic effects would identify subjects after the alcohol exposure occurred, that is, anytime after the birth of the child. Because retrospective studies do not typically screen a large, population-representative cohort, they often identify potential research subjects on the basis of known alcohol exposure or neurodevelopmental problems. Such samples are sometimes referred to as “clinically referred,” because the subjects may come to the attention of the researchers because of

172

McGee et al.

some clinically significant manifestation of prenatal alcohol exposure. In addition, these studies are more subject to selection bias than studies with a prospective design. However, some retrospective studies involve children identified at or near birth based on a history of heavy prenatal alcohol exposure or features of FAS; selection bias is less of a problem in these cases. Unlike most prospective studies, retrospective studies of FASD typically evaluate individuals exposed prenatally to heavy amounts of alcohol, although the precise dose is often unknown. The design of retrospective studies in this field generally involves the comparison of groups of heavily exposed individuals and nonexposed controls rather than treating alcohol as a continuous variable. More recent studies have found that individuals exposed to large amounts of alcohol prenatally who do not meet the criteria for a diagnosis of FAS present with a similar neuropsychological profile as individuals with FAS (7). Thus, in retrospective studies, groups of children and adolescents with heavy prenatal alcohol exposure with and without a diagnosis of FAS are often combined into an alcohol-exposed (or FASD) group and compared with nonexposed controls. Retrospective studies may be characterized by various types of bias and confounding influences. Reporting bias is a common problem in retrospective studies in general; patient reports of past behavior are often inaccurate. As antenatal reporting of alcohol consumption is more accurate than retrospective recall (10), the reliability of maternal alcohol consumption measures poses a major difficulty to retrospective studies in this field. Because children are identified after birth and medical records and/or contact with the biological mother are often unavailable, precise measurement of the alcohol exposure is difficult to attain. Even when the biological mother is available, it is unlikely she will be able to accurately recall her drinking pattern during her pregnancy, especially since alcohol consumption typically varies throughout pregnancy. Thus, reliable evaluation of dose–response variables is typically not possible in retrospective studies. When children with a diagnosis of FAS are evaluated, this problem of reporting bias is largely circumvented since characteristic facial features are indicative of this exposure. However, when children with nondysmorphic FASD are included in a research sample, there is more of a reliance on prenatal alcohol history. In addition to reporting bias, retrospective studies may face ascertainment bias. Subjects are often recruited from specialty clinics or are referred by doctors or other community agencies rather than by random selection and random group assignment. In addition to these biases, factors such as cultural or socioeconomic effects may confound results, although this is not restricted to retrospective studies. Other potential confounding variables such as maternal IQ, paternal effects, medication and other drug usage during pregnancy, and other comorbid factors may not be addressed in retrospective studies. Well-controlled studies with well-matched comparison groups can help reduce a number of these confounding factors. Although retrospective studies have several disadvantages in comparison to prospective studies, they offer several advantages to researchers. The largest advantage is that sample size and length of time needed to conduct the study, and therefore the overall cost, are reduced in comparison to prospective studies. Additional benefits include the ability to include distinct control or comparison groups and increased sampling of more severe cases. Thus, larger effect sizes and therefore clinically significant group differences are possible. In addition, problems such as subject attrition and instrument revisions rarely affect retrospective studies, since the majority of studies are cross-sectional in design. Another important methodological consideration in behavioral teratology concerns dosage of the compound of interest. Distillation of dose–response relationships in behavioral teratology is complex due to the interaction of many influences. Such factors include critical periods in fetal development, which are specific times during gestation of elevated susceptibility to teratogenic effect. By definition, critical periods imply that the

Fetal Alcohol Spectrum Disorders

173

dose–response relationship is bound to be of a complicated, non-linear nature. Moreover, the timing of critical periods can vary by brain region, meaning that exposure to alcohol at any given time point may not uniformly affect the developing fetal brain. Generally speaking, there is evidence from both animal models and human exposure research that the effects of prenatal alcohol exposure fall on a dose–response continuum, with the most severe outcomes associated with heavy exposure, and moderate or low levels of exposure associated with more subtle detriments (6,14,15). The Maternal Health Practice and Child Development study is a longitudinal project that has provided evidence for a dose– response relationship with regard to the physiological effects of prenatal alcohol exposure, including decreases in height, weight, and head circumference along with palpebral fissure and skinfold thickness abnormalities. These findings persist into adolescence and are apparent at exposure levels less than one drink per day (16). In addition, studies have suggested that the pattern of alcohol exposure influences developmental outcome. For instance, in one study a binge drinking style (defined as five or more alcoholic drinks per occasion) during pregnancy, especially during early pregnancy, was reported as the best indicator of child outcome (17). Yet the same authors point out that many manifestations of fetal alcohol effects seem unassociated with timing or course of alcohol dose (15).

REVIEW OF NEUROPSYCHOLOGICAL LITERATURE Cognitive Processing The majority of investigations on the effects of maternal drinking have linked prenatal alcohol exposure to compromised neuropsychological outcomes, even at low levels of exposure. The general consensus from the prospective literature is that low to moderate levels of exposure lead to deficits in a wide array of neuropsychological domains including attention, memory, learning, visuo-spatial cognition, motor ability, and executive function; these deficits do not dissipate with time, and they exhibit dose-dependent qualities (14,18). A few studies have failed to show effects associating neurobehavioral detriment with prenatal alcohol exposure (19,20). The authors of these studies make a salient point that prenatal alcohol exposure does not necessarily lead to a poor cognitive outcome. However, though studies failing to show an effect of maternal drinking on infant outcome exist, they contradict the general consensus of the research literature, which provides evidence for alcohol’s capacity as a behavioral teratogen, even at low exposure levels. A point of debate in the neuropsychological literature is the degree to which the putative neurocognitive deficits associated with prenatal alcohol exposure represent general or specific disabilities. Results from a cohort being followed longitudinally in Finland did not support the notion of a specific profile, contrary to initial hypotheses of the investigators (21). Assessment with the WISC-III (Finnish version) and NEPSY-A subtests demonstrated alcohol exposure effects in attention and executive functioning, but the deficits were no greater than those of several other neurocognitive domains, including language, manual motor function, visuo-spatial skills, verbal memory, and learning. The authors concluded that the neuropsychological profile of prenatal alcohol exposure is one of general, rather than selective, impairment (21). This particular study may be capturing a different range of the alcohol exposure spectrum than other studies, though cross-cultural and socioeconomic status considerations are also possible explanations for these results. A meta-analysis of the effects of prenatal alcohol exposure on infant cognition examined impairments on the Bayley Scale’s Mental Development Index (MDI) (22). Results from infants 12–13 months of age revealed that prenatal alcohol exposure was

174

McGee et al.

associated with decreases in MDI scores at the three different dosage levels examined. Assessment of the same children at 6–8 months and 18–26 months, however, did not reveal any association between exposure and MDI scores, possibly reflecting differences in the sensitivity and content of the MDI leading to variation in the measurement’s capability to detect prenatal alcohol exposure effects (22). Assessment of 482 second graders from the Seattle cohort indicated that exposure to one or more absolute ounces of alcohol per day was associated with a decrease in IQ of 6.7 points, which is almost half of one standard deviation (23). Additionally, these analyses indicated that environmental factors (paternal IQ and number of young children in the household) interacted with alcohol exposure to influence general cognitive outcome. More specifically, decreased paternal IQ and/or increased numbers of young children in the household were associated with poorer outcomes in alcohol-exposed offspring as measured by IQ tests (23). In the school-aged analysis, Digit Span and Arithmetic subtests from the WISC-R Verbal Scale were most closely associated with alcohol exposure (23). There was no evidence for differential alcohol effects between the verbal and performance IQ scores. Interestingly, when this cohort had been assessed at four years of age, performance IQ showed greater deficits than verbal IQ (18). This discrepancy between assessment time points might mean that the effects of alcohol on general cognitive capacity are dynamic in response to developmental time point, or, alternatively, it could simply reflect psychometric differences between test versions. At 14 years of age, the Arithmetic subtest from the Verbal Scale of the WISC-R continued to have the strongest association with prenatal alcohol exposure, suggesting a potent longitudinal effect of exposure on arithmetic ability (18), although the relative contribution of attention and short-term memory to this subtest cannot be discounted. The most frequently used tests in the evaluation of general intellectual functioning of children and adolescents with histories of heavy alcohol exposure have been the Wechsler Intelligence Scales. The mean Full Scale IQ (FSIQ) for children with FAS has been estimated to be between 65 and 72 (24), and results have demonstrated that children with dysmorphic features tend to have somewhat lower IQ scores than those without dysmorphic features (25). Verbal IQ and Performance IQ were also evaluated, and both alcohol-exposed groups had significantly lower mean IQ scores on both indices than controls, but did not differ significantly from each other. Language/Verbal Skills Language impairments have been found in a variety of studies of FASD. For instance, in the Seattle cohort, 462 fourteen-year olds showed impaired performance on a phonological processing task (26), suggesting that language may be affected even at low exposure levels. Yet, no evidence of an association between alcohol exposure and deficits in expressive language, receptive language, or speech was found in a populationbased cohort study of low socioeconomic status urban children exposed to moderate levels of alcohol prenatally (27). The children were assessed at one, two, and three years of age. Possible reasons for the discrepancy between these two studies include age range of child assessed, socioeconomic status, and nature and specificity of language deficit. In contrast, both receptive and expressive language skill deficits have been reported in children with heavy prenatal alcohol exposure (7,28–30), as have speech disorders, including mild problems such as production errors on one or two consonant phonemes, or more severe problems such as marked difficulty with production that resulted in low speech intelligibility (28). In addition, due to the inherent craniofacial abnormalities in children

Fetal Alcohol Spectrum Disorders

175

with FAS, it is not rare for children with this disorder to have impairments in speech mechanisms and production arising from physical anomalies of the maxilla or palate (31). One study, using the Test of Language Development, examined grammatical components of receptive and expressive language. Results suggested that, compared with matched controls, older children with FAS presented with primarily syntactic deficits, whereas younger children displayed more global language deficits (32). Because children with FAS are generally smaller for their age, deficits in language may not be recognized because people underestimate the alcohol-exposed child’s actual age and deficits in language may be consistent with their overall lower cognitive functioning. Becker, Warr-Leeper, and Leeper (33) evaluated grammatical and semantic abilities to determine if deficits were consistent with mental age. In this study, children with FAS displayed delayed grammatical skills that were consistent with their general cognitive impairments, but articulation and speech production were impaired beyond what would be expected based on cognitive deficits. Learning and Memory A substantial body of literature has begun to accumulate on the learning and memory skills of children affected by prenatal alcohol exposure. Memory is often conceptualized as consisting of two main divisions: explicit memory and implicit memory. Explicit memory involves the conscious storage and retrieval of information, whereas implicit memory relies on unconscious recall of previously performed tasks. Both implicit and explicit memory have been examined in alcohol-exposed children. When compared to nonexposed controls matched by age, sex, and ethnicity, children prenatally exposed to alcohol displayed relatively intact implicit memory on a priming task, but impaired explicit memory skills (34). Explicit memory was assessed through both a free recall task and a recognition task. In comparison with controls, alcohol-exposed children demonstrated impaired performance on the free recall task but intact recognition memory skills, which suggested retrieval deficits in addition to the encoding deficits demonstrated in other studies (35,36). Although other studies have found recognition deficits, the type of recognition task used in this study aimed to eliminate the influence of positive response bias by utilizing a two-alternative forced choice paradigm rather than yes–no. Verbal Learning and Memory The majority of explicit memory studies in children with FASD have focused on impairments in verbal learning and memory. In a series of studies, Mattson and colleagues have evaluated the verbal learning (acquisition) and memory (retention) skills of children exposed to large amounts of alcohol prenatally using the California Verbal Learning Test for Children (CVLT-C). In the first of this series of studies, the authors (35) found that in comparison with controls, children with FAS recalled fewer words on five immediate recall trials, displayed poorer performance on both the long delay free recall and cued recall, and were less able to discriminate between target items from foils on a recognition task. Thus, children with FAS learned fewer words and had more difficulty recalling these words after a delay. Importantly, there was no significant difference on retention measures, indicating that deficits were at the level of encoding rather than retrieval. Some, but not all, deficits persisted when compared to a mental age control group, suggesting that while many of the verbal learning tasks seen in FAS may be due to a global decline in cognitive ability, some features may be specific to prenatal alcohol exposure. Subsequent studies replicated the pattern of verbal learning deficits and spared retention in children with

176

McGee et al.

nondysmorphic FASD (7) and indicated that the spared retention may be due to the use of an implicit learning strategy (37). In another study, Mattson and Roebuck (36) compared both verbal and nonverbal learning and memory in children with histories of heavy prenatal alcohol exposure. In this study, a direct comparison was made between verbal and nonverbal learning and memory skills using tests of similar structure: the CVLT-C, which measures verbal skills and the Biber Figure Learning Test (BFLT), which tests nonverbal skills. Results suggested that, in addition to initial encoding difficulties for both verbal and nonverbal information, children exposed prenatally to alcohol had storage or retrieval deficits for nonverbal information. These results were recently confirmed in a prospective study with moderate prenatal exposure to alcohol (38). The study by Mattson and Roebuck (36) also demonstrated reduced learning on the learning subtests of the Wide Range Assessment of Memory and Learning (WRAML). Overall, the pattern of results suggested that alcoholexposed subjects benefit from repeated exposure to material, but to a lesser degree than controls. These findings were recently replicated (39). Spatial memory has been evaluated in several studies. In general, results suggest intact object recall but impaired spatial location recall (40,41). However, one study indicated that once deficits in perceptual skills and verbal memory impairment were accounted for, no significant differences were present on spatial memory (42). In a recent study, the Morris Water Maze task was adapted for the computer and used to evaluate spatial learning and memory in children with prenatal alcohol exposure (43). This task, which has been used to evaluate animal models of FASD, is sensitive to hippocampal damage (44,45). Alcohol-exposed children demonstrated impaired place learning relative to controls but were equally proficient during the cue-navigation phase. These results are consistent with animal studies, which suggest the possible involvement of the hippocampus in place learning. Visuospatial Ability In addition to deficits in spatial learning and memory, children exposed to heavy amounts of alcohol prenatally have been shown to have deficits in visuospatial construction skills. Visuospatial skills are often assessed with a figure copy task called the Beery-Buktenica Developmental Test of Visual-Motor Integration (VMI). Impairments on this measure have been consistently shown for children with FASD (7,29,30). However, when visual perceptual abilities were examined without the motor drawing component in one study, no differences were found, suggesting that impairment on the VMI in alcohol-exposed children may be due to motor difficulties rather than perceptual problems (30). Visuospatial skills have further been examined using the Global–Local Task (46). On this task, alcohol-exposed children demonstrated difficulties recalling local features of hierarchical stimuli, but showed little to no impairment in recalling the global features when compared with controls. These deficits were not due to the smaller stimulus size of the local features or to the memory component of the task as they were apparent only when the targets were hierarchical figures and persisted on a copy condition. Attention Attention is an important area of focus in FASD research, given the characteristic attention deficits exhibited by individuals with histories of prenatal alcohol exposure. Attention deficits and hyperactivity are associated with prenatal alcohol exposure, regardless of cognitive impairment; in fact, hyperactive disorders are among the most frequent

Fetal Alcohol Spectrum Disorders

177

comorbid diagnoses with FAS (47). One longitudinal study of the effects of lower levels of prenatal alcohol exposure on attention assessed children with vigilance tasks at ages four, seven, and fourteen (48). Such tasks are thought to tap sustained attention skills, requiring study participants to respond to the infrequent and sporadic presentation of a target stimulus. These tasks are not cognitively challenging, but are designed to be repetitive and therefore demanding on attention systems. This study found that prenatal alcohol exposure was associated with increased variability in the time it took for an individual to respond to items on the vigilance task (49). Moreover, variability of reaction time was correlated with behavior ratings of attention deficits made by the participant’s parent, teacher, and test examiner, suggesting that low to moderate levels of prenatal alcohol exposure can lead to clinically significant attention deficits. However, a separate study of low alcohol exposure failed to demonstrate a relationship between prenatal alcohol exposure and attention variables (50). A recent study evaluated both auditory and visual sustained attention in prospectively identified alcohol-exposed individuals in comparison with nonexposed controls (51). The dysmorphic alcohol-exposed group performed significantly worse than controls and the nondysmorphic group when stimuli were presented visually, whereas no group differences were evident in the auditory modality. A similar study comparing visual and auditory attention indicated that children with FASD have pervasive deficits in visual focused attention but that their ability to attend to auditory information was only affected when required to attend over longer periods of time (52). Given the relatively high rate of co-occurrence of prenatal alcohol exposure and attention deficits, research has focused on comparing the attention deficits associated with prenatal alcohol exposure to those seen in Attention Deficit Hyperactivity Disorder (ADHD). A sub-sample drawn from a longitudinal study of prenatal alcohol exposure compared alcohol-exposed African-American children from low socioeconomic status backgrounds to children diagnosed with ADHD (53). The study methodology relied on a stratified assessment strategy where the alcohol-exposed and ADHD group performance were compared on: (1) a test battery including behavioral, psychiatric, and externalizing measures of attention, and (2) test measures based on Mirsky’s four-factor neurologically based etiological model of attention. The Mirsky model poses that four primary factors comprise attention: focus, encoding, sustaining, and shifting (54). Group comparisons from both test batteries suggested different attention deficit profiles for alcohol exposure and ADHD; the alcohol-exposed group had trouble with tests tapping the encoding and shifting aspects of attention, while the ADHD group had greater difficulty with the focusing and sustaining of attention (53). In addition, a discriminant function analysis indicated that the neuropsychological measures from the traditional battery could differentiate the effects of alcohol exposure from ADHD with a reliability rate of 71%. Results of the discriminant analysis suggested that FAS was best identified by tests tapping visuospatial skills, information encoding, and problem solving skills, while ADHD was best identified by tasks that require focusing and sustaining attention and by the Achenbach Child Behavior Checklist (CBCL). Interestingly, the ADHD group was correctly classified at a much higher rate (85% accuracy) than the FAS group (44% accuracy). Only 12% of the FAS group, however, was misclassified as ADHD (53). The study authors concluded that it is possible that the discrepant neuropsychological profile between alcohol-exposed and ADHD groups can be explained by underlying attention deficits caused by compromises in different brain regions (53). Ultimately, this suggests that attention deficits associated with prenatal alcohol exposure may respond best to different treatment approaches than those designed for children with ADHD. Specifically, psychopharmacological and behavioral interventions aimed at mitigating

178

McGee et al.

attention problems in FASD might do better focusing on encoding, and the complex relationship between memory, learning, and attention, rather than the ADHD-driven clinical foci on sustaining and focusing attention (53). Additionally, an earlier study compared alcohol-exposed children to children with ADHD and controls on a series of computer-based tests of attention, and found different patterns of performance for all three groups (55). Both the alcohol-exposed and ADHD group exhibited significantly higher impulsive errors than controls. Children with ADHD were able to respond as quickly as controls but made more errors. Alcohol-exposed children were slower but produced equal errors to the ADHD group. In addition, the alcohol-exposed children had slower response time than the ADHD and control groups on a vigilance task. Furthermore, the Connors Abbreviated Parent Teacher Questionnaire, a commonly used measure to assess hyperactive behavior and evaluate treatment response, was administered to the sample. Alcohol-exposed children were rated as having significantly higher levels of hyperactivity by their parents and teachers than control children, and scores were not significantly different on this measure from children with ADHD (55). A recent study used logistic regression to classify retrospectively identified heavyexposed children versus controls based on performance on standard, widely used measures of attention (56). The final model included the Freedom from Distractibility index from the WISC-III and the Attention Problems scale from the CBCL. Overall classification accuracy was 91.7%, with 93.3% of the alcohol-exposed children and 90% of the control children classified accurately. These data indicate that children prenatally exposed to heavy amounts of alcohol can be distinguished from nonexposed controls with a high degree of accuracy based on two commonly used measures of attention. Reaction Time Reaction time (RT) is an indicator of cognitive processing efficiency. In general, both prospective and retrospective studies have demonstrated increased or more variable RT, which suggests that individuals prenatally exposed to alcohol suffer from inefficient cognitive processing systems. Prenatal alcohol exposure was found to have a negative impact on information processing in a sample of 403 African-American infants, as evidenced by longer fixation times during testing (57). Longer fixation times on cognitive tasks are thought to reflect slowed and/or inefficient information processing abilities. Prenatal alcohol exposure, however, was not associated with performance impairments in visual recognition memory or information transfer across sensory modalities (57). The same authors have shown an association between maternal alcohol consumption and both increased infant RT and decreased fast responses (57). These RT findings were dose-dependent and independent of confounding factors that could potentially influence RT such as maternal depression and infant intellectual stimulation. Cognitive processing has been examined in older study participants with prenatal alcohol exposure histories. In a follow-up of the Seattle cohort during young adolescence, cognitive processing was examined using three separate computer-administered tasks that each tapped one of the following domains: (1) procedural and declarative learning (Nissen sequence learning task), (2) spatial-visual reasoning, and (3) reading speed, memory, and comprehension (Rapid Single Visual Presentation task). On the spatial-visual reasoning task, there was a speed-accuracy tradeoff in favor of speed and this tradeoff was related to alcohol exposure (58). Several retrospective studies have demonstrated that RT measures are dependant on motor performance and have attempted to separate the cognitive and motor components involved in these tasks. In a recent study, RT was fractionated into

Fetal Alcohol Spectrum Disorders

179

two components: premotor RT and motor RT (59). The premotor component involves the time necessary to perceive the stimulus and formulate motor action and is a measure of central processing speed. Motor RT is conceptualized as a measure of peripheral processing speed and involves the impulse propagation along motor pathways and the motor unit recruitment needed to respond. Furthermore, both simple RT (SRT) and choice RT (CRT) tasks were included. During the SRT task, subjects were instructed only to use their dominant hand and only one light stimulus was used, whereas both hands and two light stimuli were used in the CRT task. Similar to an earlier study (29), total RTs did not differ between groups on the SRT task when using the dominant hand. However, when RT was fractionated, group differences emerged. Alcohol-exposed children had slower motor RT on both the SRT and CRT tasks but exhibited significantly slower premotor RT than controls only during the CRT task. In addition, alcoholexposed children exhibited greater variability in their responding compared with control subjects, but this variability was not greater than predicted, given their lengthened RT. Thus, alcohol-exposed children demonstrated deficits on peripheral processing speed when cognitive demands were low but demonstrated deficits in both central and peripheral processing during more complex tasks. Cognitive processing speed was further characterized in relation to the effects of repeated practice. In a serial RT task, participants were asked to push the button that corresponded with a light stimulus that was illuminated (60). Previous studies have used this paradigm and found that alcohol-exposed children have a longer mean response time than control children, but that their learning profiles do not differ (58,61). Based on findings of relatively intact implicit memory of children exposed prenatally to alcohol (34) and studies with populations with damage to the striatum or cerebellum (62–64), the authors proposed that by administering a larger number of trials, deficits may emerge during the later stages of the acquisition period when the skills are believed to become automatic. Results indicated that groups did not differ significantly on accuracy. In addition, results supported earlier findings that the alcohol-exposed children did not demonstrate incremental learning after the first set of learning trials, whereas controls continued to improve their RT with practice. Motor Skills Early studies identified motor difficulties in children prenatally exposed to heavy amounts of alcohol, including tremors, weak grasp, and poor hand/eye coordination (4). The majority of studies investigating motor abilities have supported these early descriptions (24). Prospective studies have yielded mixed results in both fine and gross motor skills, which may be due to the sensitivity of the measures used and degree of exposure. A cohort of moderately alcohol-exposed children was assessed at ten years of age using the Grooved Peg Board Test, a task of fine motor skills (65). Alcohol exposure was not a significant predictor of hand-eye coordination and psychomotor speed. In addition, gross motor ability at three years of age was not affected in this same population (66). In many of these cases, exposure decreased or stopped after the first trimester of pregnancy; however, exposure during any pregnancy trimester was not associated with decreased gross motor performance in this sample. In contrast, the Seattle study has documented both gross and fine motor deficits at various points across the developmental span in its cohort (67). Additionally, there is some evidence from the study of a Finnish cohort of 70 children that motor learning is impaired more than motor precision following prenatal alcohol exposure (21).

180

McGee et al.

Motor skills have also been examined as part of larger neuropsychological evaluations in retrospective studies although as in the prospective studies, reports have been mixed. Deficits have been reported in some measures of motor functioning, including finger tapping, grip strength, and fine motor speed and precision (7,29) but not others (30). Recently, cognitive motor development of children with FAS from a community in the Western Cape Province in South Africa was examined using the Griffiths Mental Development Scales (68). Children with FAS showed significant impairments on the Eye and Hand Coordination Scale, which contains items related to fine-motor coordination, but not on the locomotor scale, which measures gross motor skills. Gross-motor functioning was also examined in two studies using more sensitive measures. Postural balance was assessed in a group of alcohol-exposed children in comparison with matched nonexposed controls under conditions of systematic variation of visual and somatosensory information (69). Alcohol-exposed children were more reliant on somatosensory input and when this input was not reliable, these children displayed significant anterior-posterior body sway and were unable to compensate by using available visual or vestibular information. These deficits may be related to cerebellar anomalies reported in children with heavy prenatal alcohol exposure. In a follow-up study, corrective postural reactions in response to rapid toe-up movements of the support surface were examined in alcohol-exposed children (70). Three latency measures were measured by an electromyograph and analyzed by computer. Both the short- and medium-latency responses are considered peripheral responses whereas the long-latency response is considered central in nature. While analyses revealed no group differences on short- and medium-latency responses, in comparison to controls, alcohol-exposed children exhibited increased long-latency responses. These results suggest that balance deficits seen in these children are, at least in part, central in nature. In general, retrospective studies have found more significant motor impairments than prospective studies. This may be due to the sensitivity of the tasks used or to exposure levels which are typically higher in retrospective studies. Executive Functioning Much research has recently focused on executive functioning, a complex construct that has been broadly defined as “those capabilities that enable a person to engage successfully in independent, purposive, self-serving behavior” (71). A variety of cognitive domains are subsumed under this general definition including inhibition, set shifting and set maintenance, planning, working memory, and the ability to integrate information across time and space (72). Such skills are associated with the frontal lobe and individuals exposed to alcohol prenatally often present with clinically significant behaviors similar to those exhibited by patients with frontal lobe lesions. Specifically, it has been noted that individuals prenatally exposed to alcohol have deficits in planning, problem-solving, and judgment skills. They also have trouble with abstraction and understanding the potential consequences of their behavior. The first retrospective study to evaluate several domains of executive functioning was conducted by Kodituwakku and colleagues (73). In this study, three domains were assessed under the rubric of self-regulatory behavior: planning, regulation of behavior, and utilization of feedback. The alcohol-exposed group performed significantly worse on a measure of planning ability than nonexposed controls. In the regulation of behavior domain, alcohol-exposed children performed equally well as controls on a delayed response task, category fluency, and the Subject-Ordered Task, while demonstrating significant impairments on letter fluency. Last, within the utilization of feedback domain,

Fetal Alcohol Spectrum Disorders

181

alcohol-exposed children displayed impaired performance on both number of categories and perseverative errors on the Wisconsin Card Sorting Test (WCST), but did not differ significantly from controls on a Go-No-Go task. The authors propose that the particular nature of these deficits may be due to the selective disruption of some of the multiple working memory circuits during development that depend on the timing of teratogenic insult. Executive functioning in children with heavy prenatal exposure to alcohol was comprehensively evaluated in two studies using the Delis-Kaplan Executive Functioning System (D-KEFS, 74). In the first study (75), four domains of executive functioning were evaluated: cognitive flexibility, response inhibition, planning, and concept formation and reasoning. Group differences emerged across all four domains even when accounting for deficits in more basic skills. In a second study, both verbal and nonverbal fluency were examined using two additional measures from the D-KEFS. When compared to nonexposed controls, alcohol-exposed children displayed deficits in both fluency domains, demonstrating impaired verbal and nonverbal fluency (76). These deficits may also be related to deficits in spatial memory, perseverative tendencies, and attention problems. Recently, Kodituwakku and colleagues have differentiated cognition-related and emotion-related executive skills (77). Cognition-related skills “involve holding, updating, and manipulation of non-emotional, complex information in working memory,” whereas emotion-related abilities “involve establishment and modification of behavioral sets based on emotionally significant information” (77). On a visual discrimination reversal task that taps emotion-related executive functioning, the alcohol-exposed group made fewer reversals and demonstrated more variability in extinction of the reward-response association than the control group. In addition, alcohol-exposed subjects made more commission and omission errors. Emotion-related executive functioning was found to be relatively independent of nonverbal intellectual ability and conceptual set shifting. Alcohol-exposed children displayed poorer performance on the WCST and displayed a greater number of perseverative errors. Thus, emotion-related and cognition-related executive functioning were demonstrated to be independent of each other, though impaired to a comparable magnitude. Parents of alcohol-exposed children in this sample also reported significantly increased executive functioning and behavioral problems in their children and these parent reports were related to test performance. One recent study examined executive functioning in a clinically diagnosed group of 30 men with FAS or FAE (78). Tasks requiring shifting and maintenance of attention, visuo-spatial skills, or working memory in the face of distraction were identified by the study authors as tests that directly tap executive function independently of general cognition (78). This FAS/E group displayed deficits on the executive function tasks that were worse than their IQ scores would have predicted, suggesting that executive function deficits can occur independently of IQ deficits and may be especially vulnerable to alcohol’s teratogenic effects. Thus, heavy prenatal alcohol exposure is consistently related to deficits in executive functioning. Low levels of exposure have also been associated with executive functioning deficits at some ages but not at others (65).

LONG-TERM CONSEQUENCES In general, as individuals with prenatal alcohol exposure age, behavioral and adaptive problems are increasingly manifested and neuropsychological deficits, including problems with attention, motor function, verbal skills, learning, and memory, continue to be

182

McGee et al.

problematic (26). Typically, older individuals with FASD do more poorly on achievement tests than their IQ scores would predict, and arithmetic scores are a particular weakness. Adaptive functioning is seemingly even more affected than cognition, including the areas of communication, daily living, and social skills. Many individuals with FASD are unable to live or work independently. Such disability takes a tremendous toll on the individual, family members, caregivers, and ultimately society (26). In addition, many individuals prenatally exposed to alcohol have additional mental health problems, are suspended or expelled from school, get in trouble with the law, show inappropriate sexual behavior, and have alcohol and drug problems. Streissguth and colleagues have termed such difficulties “secondary disabilities” as they arise after birth and could presumably be ameliorated through better understanding and appropriate interventions (79). One reason for evaluating secondary disabilities in addition to overall intellectual functioning is that individuals with FASD who have IQ scores above 70 (i.e., not defined as mentally retarded) are often unable to qualify for special services. Kerns and colleagues (80) described sixteen adults with FAS and IQ scores above 70 who performed more poorly on tests of complex attention, verbal learning, and executive function than their IQ scores would have predicted. These results indicate that non-retarded adults with FAS may suffer from compromising neuropsychological deficits, especially on tasks that require cognitive sophistication or mental processing efficiency. Anecdotal evidence suggests such difficulties are likely to impair daily functioning, even if the individual is not mentally retarded. Thus, IQ performance alone is not a sufficient indicator of special service need in the alcohol-exposed population and identification and further understanding of secondary difficulties may improve the services offered to these individuals. Some of the most notable difficulties in childhood involve academic performance. Long-term consequences of prenatal alcohol exposure were evaluated in an adolescent and adult group of 61 clinically referred individuals with FASD of predominantly Native American ethnicity (81). Only six percent of the sample was placed in a regular school classroom without need for extra help. Average achievement scores for reading, spelling, and arithmetic were at the fourth, third, and second grade levels, respectively, and there was no evidence for academic outcome improvement with age (81). A population-based longitudinal study examined the role of moderate prenatal alcohol exposure on scholastic achievement in 432 second-grade children (23). Decreased achievement test scores on the Wide Range Achievement Test (WRAT) were associated with maternal binge drinking prior to pregnancy recognition, compared to other drinking styles. In general, achievement test scores indicated that prenatal alcohol exposure was associated with compromised reading and arithmetic skills, but spelling skills were relatively preserved. In addition to achievement test scores, the association of prenatal alcohol exposure with learning problems was indicated by teacher and parent behavior ratings of academic difficulties, and remedial classroom placements (23). School performance problems, requiring additional support or resources, were also noted in a Finnish cohort of 70 prenatally exposed individuals being followed longitudinally, and the authors cite consideration of prenatal alcohol exposure as a possible factor in the etiology of learning disabilities (14). Individuals exposed to alcohol prenatally have been repeatedly shown to have academic difficulties, especially in the domain of mathematical skills. An evaluation of number processing abilities in adolescent and adult individuals with prenatal alcohol exposure revealed difficulties especially in calculation and estimation tests, with intact number reading and writing ability. However, the pattern of impairment was variable among alcohol-exposed subjects, with 10 of 29 participants exhibiting no clear impairment on the numerical task battery (82).

Fetal Alcohol Spectrum Disorders

183

Parent-Rated Behavior A number of studies have compared the behavioral functioning of children exposed prenatally to alcohol with nonexposed control children by having parents complete behavior checklists. The most frequently used behavior checklist in research with alcoholexposed children has been the CBCL and several studies have reported elevations in both summary scales and individual problem scales (30,55,83,84). One study reported higher overall CBCL scores and hyperactivity scores in comparison to nonexposed controls but not in comparison to children with attention deficits (55). Importantly, in a comparison of children with FAS and nondysmorphic FASD, no significant differences were found (84). Other parent-report checklists have been used in research with children prenatally exposed to alcohol. The Personality Inventory for Children was administered to parents of children prenatally exposed to alcohol and controls (85). The alcohol-exposed children were rated significantly higher (more difficulty) on all scales, with the exception of the Hyperactivity scale. Elevations were found to be particularly high (O2 SD) on the Intellectual Screening, Development, Psychosis, and Delinquency scales when compared to control children. The profiles of the children with FAS and nondysmorphic FASD were similar, with significant differences only present on one validity scale and the cognitive triad scales (Achievement, Intellectual Screening, and Development). This is consistent with the fact that, in general, children with FAS had lower FSIQ scores than the children who did not meet criteria for a diagnosis. The Vineland Adaptive Behavior Scales (VABS) is a parent-completed interview that assesses the adaptive functioning across three domains: communication, daily living skills, and socialization. Long-term adaptive functioning was evaluated in a group of Native American adolescents and adults with FASD (81). Although the mean chronological age was 17, the mean adaptive age was approximately seven years. Among the subscales of the VABS, subjects had most difficulty with socialization skills, performing at an average age of six years. The authors point out that poor social skills such as failure to consider consequences of one’s actions and inappropriate responsiveness to social cues may be a problem for alcohol-exposed individuals, even those who score within normal ranges on IQ tests. The most common maladaptive behaviors reported were poor concentration/attention, stubbornness, and social withdrawal. The study authors concluded that FAS presents the greatest treatment challenges in adulthood, given the widespread behavioral problems that tend to occur in this population independent of marked IQ deficits. Several retrospective studies have evaluated adaptive functioning using the VABS. Alcohol-exposed children displayed significant deficits in adaptive functioning across all three domains when compared with the normative sample (86) but were not significantly different from nonexposed controls matched by age, sex, inpatient status, and FSIQ. Both groups were functioning in the borderline range. Further analysis of the alcohol-exposed group revealed a greater age-related decline in standard scores in the socialization domain compared with the nonexposed group. A previous study documented a similar age-related decline in socialization scores for alcohol-exposed children on the VABS (87). In this study, the three sub-domains of the socialization scale were examined and children with FAS were the most impaired on the interpersonal skills sub-domain. A significant correlation was found between age and age-equivalent discrepancy socialization scores only for children in the FAS group; discrepancy scores were greatest in older children. This result suggests that social skills of children with FAS are deficient below what could be expected due to low IQ scores. Moreover, the authors point out that the salience of adaptive skill deficits may increase as children with prenatal

184

McGee et al.

alcohol exposure mature, thus suggesting that adaptive skill deficits are arrested rather than delayed in this population. In addition to behavioral and adaptive difficulties, other secondary disabilities are commonly faced by individuals with prenatal alcohol exposure. For example, an examination of attachment behavior in 4- and 5-year-old children prenatally exposed to alcohol revealed that prenatal alcohol exposure was highly related to attachment insecurity and a predictor of negative affect in the child. In addition, 80% of the alcohol-exposed children but only 36% of control were rated as insecure. However, when the mothers of the alcohol-exposed children provided high levels of support, these children evidenced better coping skills and more secure attachment relations compared to children of mothers who provided a less supportive environment (88). The same research group examined prenatal alcohol exposure and infant negative affect as precursors of depressive features in 6-year old children (89). Significant associations were found between prenatal alcohol exposure, infant negative affect at 1 year, and depressive symptoms at 6 years. Specifically higher levels of alcohol exposure were related to high levels of negative affect at 1 year and more depressive symptoms at six years of age. These effects were almost entirely driven by the girls in the sample. In addition to a direct relationship between alcohol and depression, an indirect relationship, via higher levels of emotionality in infancy, was proposed. Other studies have evaluated the psychological functioning of individuals exposed to alcohol prenatally. Psychopathology was evaluated in a group of children in Berlin, Germany via a structured psychiatric interview (90). Compared to controls, children with FAS had greater instances of outpatient therapy, eating and sleeping problems, head and body rocking, stereotyped habits, reduced vocabulary and clarity of speech, speech impairment, hearing impairments, and a visual condition known as strabismus. Children with FAS were also clumsier, hyperactive, and inattentive, had more difficulty with peer relationships, and tended to be more dependent and had problems with phobias. In a follow-up study (83) psychiatric symptoms were compared at multiple time points to evaluate the course of symptoms seen in alcohol-exposed individuals over time. Namely, hyperkinetic disorders persisted over three time points, and emotional disorders, sleep disorders, and abnormal habits and stereotypes were noted especially during school age. Other age-specific disorders such as enuresis, encopresis, and eating disorders remitted over time whereas conduct disorders remained relatively constant due to a mixed pattern of remissions and new manifestations. Adults with histories of prenatal alcohol exposure seem to be at high risk for developing psychopathology. A study of 25 adults with FASD but average IQ revealed that 23 out of the 25 subjects met criteria for an Axis I mental health disorder on the basis of the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) (91). The most common disorders were alcohol/drug dependence, depression, and psychotic disorders. Eighteen of the 25 subjects had previously sought mental health treatment, a third of which cases had necessitated inpatient care in a psychiatric hospital. Suicidality was present in six out of the twenty-five participants (92). The results were tempered by the small number of subjects studied, the need to exclude below-average IQ participants because of SCID administration criteria, the fact that many study participants had previously been identified with having “adaptational difficulties,” and the lack of a control group (91). Still the percentage of subjects in this study with mental health symptoms is staggering and suggests that clinically significant psychopathogy should be considered among the potential constellation of long-term effects of prenatal alcohol exposure. Prenatal alcohol exposure was associated with alcohol and drug problems in young adulthood by two separate studies. In a sample of 433 young adults, alcohol exposure was

Fetal Alcohol Spectrum Disorders

185

related to higher levels of drug and alcohol use in young adulthood (93). This relationship was independent of various potentially confounding factors such as sex, family history of alcohol problems, other prenatal exposures, and parental use of drugs postnatally. The study authors concluded that prenatal alcohol exposure is a risk factor for adulthood alcohol problems and should be considered as a potential factor in the etiology of alcoholism (93). In the second study (94), adult adoptees were interviewed for substance abuse disorders including nicotine, alcohol, and other drugs. Retrospective analysis of prison, hospital, and adoption agency records were used to determine prenatal alcohol exposure. Individuals prenatally exposed to alcohol reported higher dependence symptom counts than nonexposed controls even after controlling for sex, biological parent diagnosis, birth weight, gestational age, and other environmental variables. In addition to substance abuse and psychiatric disorders, individuals with FASD may be at increased risk of delinquency (92). A recent study found alcohol-exposed children to have higher rates of delinquent behaviors and probable diagnoses of Conduct Disorder than nonexposed controls (95). This study also demonstrated that alcoholexposed children perform at a lower level of moral maturity than controls, especially in their relationships with others. While overall moral maturity did not significantly predict delinquency, the moral judgment values of relations with others, stealing, and obeying the law significantly added to the model predicting delinquency beyond the effects of age and group membership. In contrast, no association between prenatal alcohol exposure and increased delinquency was found in a low socioeconomic status, predominantly AfricanAmerican adolescent sample (96). Rather, several factors were noted to relate to increased delinquent acts across subject classification: increased life stress, drug use, and decreased parental supervision. These results underscore the importance of considering environmental factors in addition to prenatal alcohol exposure.

CONCLUSIONS The purpose of this chapter was to present a review of what is known about the neuropsychological and behavioral functioning of individuals prenatally exposed to alcohol for several reasons: (1) to provide an methodologically-focused review of the literature, (2) to identify areas where further research is still needed, and (3) to identify and discuss shortcomings in research so that the study of other teratogens may benefit from this knowledge and capitalize on effective research designs and analyses. Two general methods of ascertainment, prospective and retrospective, were addressed along with the advantages and disadvantages of both. The presentation of neuropsychological and behavioral functioning was organized into three main sections: (1) general intellectual functioning, (2) specific neuropsychological abilities, and (3) long-term consequences of prenatal alcohol exposure. The specific neuropsychological domains discussed included learning and memory, visuospatial processing, attention, reaction time, fine and gross motor abilities, language skills, and executive functioning. A variety of long-term consequences were presented including academic and behavioral difficulties and additional psychological problems commonly encountered. This review should provide the reader with a good understanding of what is currently known about the teratogenic effects of alcohol on neuropsychological functioning and identify important areas of functioning that may be vulnerable to other teratogens. Inconsistencies in terminology and methods have brought confusion to the field. This is especially apparent in differences in terminology and criteria used in forming alcohol-exposed groups in research studies. As indicated in the introduction, individuals

186

McGee et al.

with histories of alcohol exposure have been classified with a variety of terms including FAS, FAE, ARND, ARBD, and PEA. The latter four classifications, which refer to nondysmorphic alcohol-exposed individuals, are not equivalent and do not all have welldefined or established criteria. In addition, different research groups will use the same term (e.g., FAE) but apply different selection criteria to define the subject group. This makes comparison of study results difficult and sometimes inaccurate. In addition, criteria for a diagnosis of FAS have varied across studies. A further problem that has made comparison across studies difficult has been the measurement and reporting of prenatal alcohol exposure levels. In addition to mitigating the inaccuracies associated with self-reports of alcohol use, there is a need for a way to categorize drinking behavior that simplifies this complex behavior but yet still accurately captures the variability, dose, and timing over the course of pregnancy. As binging behavior may be more harmful to the developing fetus than a non-binge pattern of alcohol exposure (17), failure to capture differential patterns of alcohol intake may produce measurement error with real world significance. As discussed above, it is preferable to use multiple measures of alcohol exposure, such as dose, timing, and variability, to optimally tap this complicated construct. Of course, retrospective studies may suffer from inaccurate or missing information on alcohol exposure, and for these studies classifications of “heavily exposed” based on record review or maternal report must suffice. While these studies lack the detail to examine the important variables of dose and timing, they benefit from including the more severe end of the spectrum of effect. In summary, the effects of heavy prenatal alcohol exposure are well documented. Although differences in terminology and methodology plague the field, it remains that alcohol is a known human teratogen capable of producing devastating consequences on brain development and subsequent behavioral, social, cognitive, and neuropsychological development. Over 30 years of research has resulted in a multitude of studies on these effects. However, much remains to be learned about the specificity of alcohol’s teratogenic effects and the potential for amelioration of these effects. ACKNOWLEDGMENTS The authors acknowledge the support of Edward P. Riley and the Center for Behavioral Teratology. Supported by the National Institute on Alcohol Abuse and Alcoholism grants AA10820 and AA13525.

REFERENCES 1. Abel EL. An update on incidence of FAS: FAS is not an equal opportunity birth defect. Neurotoxicol Teratol 1995; 17:437–443. 2. Pulsifer MB. The neuropsychology of mental retardation. J Int Neuropsychol Soc 1996; 2:159–176. 3. Jones KL, Smith DW. Recognition of the fetal alcohol syndrome in early infancy. Lancet 1973; 2:999–1001. 4. Jones KL, Smith DW, Ulleland CN, Streissguth AP. Pattern of malformation in offspring of chronic alcoholic mothers. Lancet 1973; 1:1267–1271. 5. Stratton K, Howe C, Battaglia F. Fetal Alcohol Syndrome: Diagnosis, Epidemiology, Prevention, and Treatment. Washington, DC: National Academy Press, 1996. 6. Meyer LS, Riley EP. Behavioral teratology of alcohol. In: Riley EP, Vorhees CV, eds. Handbook of Behavioral Teratology. New York: Plenum Press, 1986:101–140.

Fetal Alcohol Spectrum Disorders

187

7. Mattson SN, Riley EP, Gramling LJ, Delis DC, Jones KL. Neuropsychological comparison of alcohol-exposed children with or without physical features of fetal alcohol syndrome. Neuropsychology 1998; 12:146–153. 8. Bertrand J, Floyd RL, Weber MK, O’Connor M, Riley EP, Johnson KA, et al. National Task Force on FAS/FAE: Guidelines for Referral and Diagnosis. Atlanta, GA: Centers for Disease Control and Prevention, 2004. 9. Vorhees CV. Principles of behavioral teratology. In: Riley EP, Vorhees CV, eds. Handbook of Behavioral Teratology. New York: Plenum Press, 1986:23–48. 10. Jacobson SW, Chiodo LM, Sokol RJ, Jacobson JL. Validity of maternal report of prenatal alcohol, cocaine, and smoking in relation to neurobehavioral outcome. Pediatrics 2002; 109:815–825. 11. Streissguth AP, Sampson PD, Barr HM, Bookstein FL, Olson HC. The effects of prenatal exposure to alcohol and tobacco: contributions from the Seattle Longitudinal Prospective Study and implications for public policy. In: Needleman HL, Bellinger D, eds. Prenatal Exposure to Toxicants: Developmental Consequences. Baltimore, MD: The Johns Hopkins University Press, 1994:148–183. 12. Jacobson JL, Jacobson SW. Prospective, longitudinal assessment of developmental neurotoxicity. Environ Health Perspect 1996; 104:275–283. 13. Jacobson JL, Jacobson SW. Methodological considerations in behavioral toxicology in infants and children. Dev Psychol 1996; 32:390–403. 14. Autti-Ra¨mo¨ I. Twelve-year follow-up of children exposed to alcohol in utero. Dev Med Child Neurol 2000; 42:406–411. 15. Streissguth AP, Sampson PD, Barr HM. Neurobehavioral dose-response effects of prenatal alcohol exposure from infancy to adulthood. Ann NY Acad Sci 1989; 562:145–158. 16. Day NL, Leech SL, Richardson GA, Cornelius MD, Robles N, Larkby C. Prenatal alcohol exposure predicts continued deficits in offspring size at 14 years of age. Alcohol Clin Exp Res 2002; 26:1584–1591. 17. Sampson PD, Streissguth AP, Barr HM, Bookstein FL. Neurobehavioral effects of prenatal alcohol: Part II. Partial least squares analysis. Neurotoxicol Teratol 1989; 11:477–491. 18. Streissguth AP, Barr HM, Sampson PD, Bookstein FL. Prenatal alcohol and offspring development: The first fourteen years. Drug Alcohol Depend 1994; 36:89–99. 19. Ernhart CB, Greene T, Sokol RJ, Martier SS, Boyd TA, Ager JW. Neonatal diagnosis of fetal alcohol syndrome: Not necessarily a hopeless prognosis. Alcohol Clin Exp Res 1995; 19:1550–1557. 20. Greene T, Ernhart CB, Ager JW, Sokol RJ, Martier SS, Boyd T. Prenatal alcohol exposure and cognitive development in the preschool years. Neurotoxicol Teratol 1991; 13:57–68. 21. Korkman M, Kettunen S, Autti-Ramo I. Neurocognitive impairment in early adolescence following prenatal alcohol exposure of varying duration. Child Neuropsychol 2003; 9:117–128. 22. Testa M, Quigley BM, Eiden RD. The effects of prenatal alcohol exposure on infant mental development: a meta-analytical review. Alcohol & Alcoholism 2003; 38:295–304. 23. Streissguth AP, Barr HM, Sampson PD. Moderate prenatal alcohol exposure: Effects on child IQ and learning problems at age 7 1/2 years. Alcohol Clin Exp Res 1990; 14:662–669. 24. Mattson SN, Riley EP. A review of the neurobehavioral deficits in children with fetal alcohol syndrome or prenatal exposure to alcohol. Alcohol Clin Exp Res 1998; 22:279–294. 25. Mattson SN, Riley EP, Gramling LJ, Delis DC, Jones KL. Heavy prenatal alcohol exposure with or without physical features of fetal alcohol syndrome leads to IQ deficits. J Pediatr 1997; 131:718–721. 26. Streissguth AP, Barr HM, Olson HC, Sampson PD, Bookstein FL, Burgess DM. Drinking during pregnancy decreases word attack and arithmetic scores on standardized tests: adolescent data from a population-based prospective study. Alcohol Clin Exp Res 1994; 18:248–254.

188

McGee et al.

27. Greene T, Ernhart CB, Martier SS, Sokol RJ, Ager JW. Prenatal alcohol exposure and language development. Alcohol Clin Exp Res 1990; 14:937–945. 28. Church MW, Eldis F, Blakley BW, Bawle EV. Hearing, language, speech, vestibular, and dentofacial disorders in fetal alcohol syndrome. Alcohol Clin Exp Res 1997; 21:227–237. 29. Conry J. Neuropsychological deficits in fetal alcohol syndrome and fetal alcohol effects. Alcohol Clin Exp Res 1990; 14:650–655. 30. Janzen LA, Nanson JL, Block GW. Neuropsychological evaluation of preschoolers with fetal alcohol syndrome. Neurotoxicol Teratol 1995; 17:273–279. 31. Abel EL. Fetal Alcohol Syndrome. New Jersey: Medical Economics Company, Inc., 1990. 32. Carney LJ, Chermak GD. Performance of American Indian children with fetal alcohol syndrome on the test of language development. J Commun Disord 1991; 24:123–134. 33. Becker M, Warr-Leeper GA, Leeper HA. Fetal alcohol syndrome: A description of oral motor, articulatory, short-term memory, grammatical, and semantic abilities. J Commun Disord 1990; 23:97–124. 34. Mattson SN, Riley EP. Implicit and explicit memory functioning in children with heavy prenatal alcohol exposure. J Int Neuropsychol Soc 1999; 5:462–471. 35. Mattson SN, Riley EP, Delis DC, Stern C, Jones KL. Verbal learning and memory in children with fetal alcohol syndrome. Alcohol Clin Exp Res 1996; 20:810–816. 36. Mattson SN, Roebuck TM. Acquisition and retention of verbal and nonverbal information in children with heavy prenatal alcohol exposure. Alcohol Clin Exp Res 2002; 26:875–882. 37. Roebuck-Spencer TM, Mattson SN. Implicit strategy affects learning in children with heavy prenatal alcohol exposure. Alcohol Clin Exp Res 2004; 28:1424–1431. 38. Willford JA, Richardson GA, Leech SL, Day NL. Verbal and visuospatial learning and memory function in children with moderate prenatal alcohol exposure. Alcohol Clin Exp Res 2004; 28:497–507. 39. Kaemingk KL, Mulvaney S, Tanner Halverson P. Learning following prenatal alcohol exposure: Performance on verbal and visual multitrial tasks. Arch Clin Neuropsychol 2003; 18:33–47. 40. Uecker A, Nadel L. Spatial locations gone awry: Object and spatial memory deficits in children with fetal alcohol syndrome. Neuropsychologia 1996; 34:209–223. 41. Uecker A, Nadel L. Spatial but not object memory impairments in children with fetal alcohol syndrome. Am J Ment Retard 1998; 103:12–18. 42. Kaemingk KL, Halverson PT. Spatial memory following prenatal alcohol exposure: More than a material specific memory deficit. Child Neuropsychol 2000; 6:115–128. 43. Hamilton DA, Kodituwakku P, Sutherland RJ, Savage DD. Children with fetal alcohol syndrome are impaired at place learning but not cued-navigation in a virtual Morris water task. Behav Brain Res 2003; 143:85–94. 44. Johnson TB, Goodlett CR. Selective and enduring deficits in spatial learning after limited neonatal binge alcohol exposure in male rats. Alcohol Clin Exp Res 2002; 26:83–93. 45. Sutherland RJ, Weisend MP, Mumby D, et al. Retrograde amnesia after hippocampal damage: recent vs. remote memories in two tasks. Hippocampus 2001; 11:27–42. 46. Mattson SN, Gramling L, Delis DC, Jones KL, Riley EP. Global-local processing in children prenatally exposed to alcohol. Child Neuropsychol 1996; 2:165–175. 47. Steinhausen H-C, Willms J, Spohr H-L. Long-term psychopathological and cognitive outcome of children with fetal alcohol syndrome. J Am Acad Child Adolesc Psychiatry 1993; 32:990–994. 48. Streissguth AP, Bookstein FL, Sampson PD, Barr HM. Neurobehavioral effects of prenatal alcohol: Part III. PLS analyses of neuropsychologic tests. Neurotoxicol Teratol 1989; 11:493–507. 49. Streissguth AP, Bookstein FL, Sampson PD, Barr HM. Attention: Prenatal alcohol and continuities of vigilance and attentional problems from 4 through 14 years. Dev Psychopathol 1995; 7:419–446.

Fetal Alcohol Spectrum Disorders

189

50. Fried PA, Watkinson B, Gray R. A follow-up study of attentional behavior in 6-year-old children exposed prenatally to marihuana, cigarettes, and alcohol. Neurotoxicol Teratol 1992; 14:299–311. 51. Coles CD, Platzman KA, Lynch ME, Freides D. Auditory and visual sustained attention in adolescents prenatally exposed to alcohol. Alcohol Clin Exp Res 2002; 26:263–271. 52. Mattson SN, Calarco KE, Lang AR. Focused and shifting attention in children with heavy prenatal alcohol exposure. submitted 2004. 53. Coles CD, Platzman KA, Raskind-Hood CL, Brown RT, Falek A, Smith IE. A comparison of children affected by prenatal alcohol exposure and attention deficit, hyperactivity disorder. Alcohol Clin Exp Res 1997; 21:150–161. 54. Mirsky AF, Anthony BJ, Duncan CC, Ahearn MB, Kellam SG. Analysis of the elements of attention: A neuropsychological approach. Neuropsychol Rev 1991; 2:109–145. 55. Nanson JL, Hiscock M. Attention deficits in children exposed to alcohol prenatally. Alcohol Clin Exp Res 1990; 14:656–661. 56. Lee KT, Mattson SN, Riley EP. Classifying children with heavy prenatal alcohol exposure using measures of attention. J Int Neuropsychol Soc 2004; 10:271–277. 57. Jacobson SW, Jacobson JL, Sokol RJ, Martier SS, Ager JW. Prenatal alcohol exposure and infant information processing ability. Child Dev 1993; 64:1706–1721. 58. Sampson PD, Kerr B, Olson HC, et al. The effects of prenatal alcohol exposure on adolescent cognitive processing: a speed-accuracy tradeoff. Intelligence 1997; 24:329–353. 59. Simmons RW, Wass T, Thomas JD, Riley EP. Fractionated simple and choice reaction time in children with prenatal exposure to alcohol. Alcohol Clin Exp Res 2002; 26:1412–1419. 60. LaForce R, Hayward S, Cox LV. Impaired skill learning in children with heavy prenatal alcohol exposure. J Int Neuropsychol Soc 2001; 7:112–114. 61. Carmichael Olson H, Feldman JJ, Streissguth AP, Sampson PD, Bookstein FL. Neuropsychological deficits in adolescents with fetal alcohol syndrome: Clinical findings. Alcohol Clin Exp Res 1998; 22:1998–2012. 62. Doyon J, Gaudreau D, Laforce R, et al. Role of the striatum, cerebellum, and frontal lobes in the learning of a visuomotor sequence. Brain Cogn 1997; 34:218–245. 63. Doyon J, Laforce R, Bouchard G, et al. Role of the striatum, cerebellum and frontal lobes in the automatization of a repeated visuomotor sequence of movements. Neuropsychologia 1998; 36:625–641. 64. Karni A. The acquisition of perceptual and motor skills: a memory system in the adult human cortex. Cog Brain Res 1996; 5:39–48. 65. Richardson GA, Ryan C, Willford J, Day NL, Goldschmidt L. Prenatal alcohol and marijuana exposure: Effects on neuropsychological outcomes at 10 years. Neurotoxicol Teratol 2002; 24:309–320. 66. Chandler LS, Richardson GA, Gallagher JD, Day NL. Prenatal exposure to alcohol and marijuana: effects on motor development of preschool children. Alcohol Clin Exp Res 1996; 20:455–461. 67. Streissguth AP, Barr HM, Sampson PD. Alcohol use during pregnancy and child development: A longitudinal, prospective study of human behavioral teratology. In: Greenbaum CW, Aurbach JG, eds. Longitudinal Studies of Children at Psychological Risk: Cross-National Perspectives. Westport, CT: Greenwood Publishing Group, Inc., 1992:174–200. 68. Adnams CM, Kodituwakku PW, Hay A, Molteno CD, Viljoen D, May PA. Patterns of cognitive-motor development in children with fetal alcohol syndrome from a community in South Africa. Alcohol Clin Exp Res 2001; 25:557–562. 69. Roebuck TM, Simmons RW, Mattson SN, Riley EP. Prenatal exposure to alcohol affects the ability to maintain postural balance. Alcohol Clin Exp Res 1998; 22:252–258. 70. Roebuck TM, Simmons RW, Richardson C, Mattson SN, Riley EP. Neuromuscular responses to disturbance of balance in children with prenatal exposure to alcohol. Alcohol Clin Exp Res 1998; 22:1992–1997.

190

McGee et al.

71. Lezak MD. Neuropsychological Assessment. 3rd ed. New York: Oxford University Press, 1995. 72. Pennington BF, Ozonoff S. Executive functions and developmental psychopathology. J Child Psychol Psychiatry 1996; 37:51–87. 73. Kodituwakku PW, Handmaker NS, Cutler SK, Weathersby EK, Handmaker SD. Specific impairments in self-regulation in children exposed to alcohol prenatally. Alcohol Clin Exp Res 1995; 19:1558–1564. 74. Delis DC, Kaplan E, Kramer JH. In: Manual for the Delis-Kaplan Executive Function System. San Antonio, TX: Psychological Corporation, 2001. 75. Mattson SN, Goodman AM, Caine C, Delis DC, Riley EP. Executive functioning in children with heavy prenatal alcohol exposure. Alcohol Clin Exp Res 1999; 23:1808–1815. 76. Schonfeld AM, Mattson SN, Lang AR, Delis DC, Riley EP. Verbal and nonverbal fluency in children with heavy prenatal alcohol exposure. J Stud Alcohol 2001; 62:239–246. 77. Kodituwakku PW, May PA, Clericuzio CL, Weers D. Emotion-related learning in individuals prenatally exposed to alcohol: An investigation of the relation between set shifting, extinction of responses, and behavior. Neuropsychologia 2001; 39:699–708. 78. Connor PD, Sampson PD, Bookstein FL, Barr HM, Streissguth AP. Direct and indirect effects of prenatal alcohol damage on executive function. Dev Neuropsychol 2000; 18:331–354. 79. Streissguth AP, Kanter J. In: Streissguth AP, Kanter J, eds. The challenge of Fetal Alcohol Syndrome: Overcoming Secondary Disabilities. Seattle, WA: University of Washington Press, 1997. 80. Kerns KA, Don A, Mateer CA, Streissguth AP. Cognitive deficits in nonretarded adults with fetal alcohol syndrome. J Learn Disabil 1997; 30:685–693. 81. Streissguth AP, Aase JM, Clarren SK, Randels SP, LaDue RA, Smith DF. Fetal alcohol syndrome in adolescents and adults. J Am Med Assoc 1991; 265:1961–1967. 82. Kopera-Frye K, Dehaene S, Streissguth AP. Impairments of number processing induced by prenatal alcohol exposure. Neuropsychologia 1996; 34:1187–1196. 83. Steinhausen H-C, Spohr H-L. Long-term outcome of children with fetal alcohol syndrome: Psychopathology, behavior and intelligence. Alcohol Clin Exp Res 1998; 22:334–338. 84. Mattson SN, Riley EP. Parent ratings of behavior in children with heavy prenatal alcohol exposure and IQ-matched controls. Alcohol Clin Exp Res 2000; 24:226–231. 85. Roebuck TM, Mattson SN, Riley EP. Behavioral and psychosocial profiles of alcohol-exposed children. Alcohol Clin Exp Res 1999; 23:1070–1076. 86. Whaley SE, O’Connor MJ, Gunderson B. Comparison of the adaptive functioning of children prenatally exposed to alcohol to a nonexposed clinical sample. Alcohol Clin Exp Res 2001; 25:1018–1024. 87. Thomas SE, Kelly SJ, Mattson SN, Riley EP. Comparison of social abilities of children with fetal alcohol syndrome to those of children with similar IQ scores and normal controls. Alcohol Clin Exp Res 1998; 22:528–533. 88. O’Connor MJ, Kogan N, Findlay R. Prenatal alcohol exposure and attachment behavior in children. Alcohol Clin Exp Res 2002; 26:1592–1602. 89. O’Connor MJ. Prenatal alcohol exposure and infant negative affect as precursors of depressive features in children. IMHJ 2001; 22:291–299. 90. Steinhausen H-C, Nestler V, Spohr H-L. Development and psychopathology of children with the fetal alcohol syndrome. J Dev Behav Pediatr 1982; 3:49–54. 91. Famy C, Streissguth AP, Unis AS. Mental illness in adults with fetal alcohol syndrome or fetal alcohol effects. Am J Psychiatry 1998; 155:552–554. 92. Streissguth AP, O’Malley K. Neuropsychiatric implications and long-term consequences of fetal alcohol spectrum disorders. Semin Clin Neuropsychiatry 2000; 5:177–190. 93. Baer JS, Sampson PD, Barr HM, Connor PD, Streissguth AP. 21-year longitudinal analysis of the effects of prenatal alcohol exposure on young adult drinking. Arch Gen Psychiatry 2003; 60:377–385.

Fetal Alcohol Spectrum Disorders

191

94. Yates WR, Cadoret RJ, Troughton EP, Stewart M, Giunta TS. Effect of fetal alcohol exposure on adult symptoms of nicotine, alcohol, and drug dependence. Alcohol Clin Exp Res 1998; 22:914–920. 95. Schonfeld AM, Mattson SN, Riley EP. Moral maturity and delinquency after prenatal alcohol exposure. J Stud Alcohol. In Press. 96. Lynch ME, Coles CD, Corley T, Falek A. Examining delinquency in adolescents differentially prenatally exposed to alcohol: The role of proximal and distal risk factors. J Stud Alcohol 2003; 64:678–686.

10 Neurodevelopmental Sequelae of Prenatal Cocaine Exposure Linda C. Mayes Yale Child Study Center, New Haven, Connecticut, U.S.A.

Marjukka Pajulo Department of Child Psychiatry, University of Tampere, Tampere, Finland and Yale Child Study Center, New Haven, Connecticut, U.S.A.

In-utero exposure to cocaine and its derivatives pose major perinatal health problems for a large number of infants and their mothers. Since the late 1980s when cocaine and crack use last peaked in the United States, many investigators have focused on the physical, neurodevelopmental, and neuropsychological effects of prenatal cocaine exposure on infants and young children. Although inconclusive on many crucial issues, published studies reveal the beginnings of a profile of possible cocaine-related effects on neuropsychological functions subserving arousal and attention regulation. That profile is informed by preclinical studies in which important factors such as duration and type of exposure as well as environmental conditions may be more adequately controlled. In this chapter, we review the state of knowledge regarding the neurobiological effects of prenatal cocaine exposure on the developing nervous system as elaborated through use of preclinical and clinical models. Studies in this area represent an important interface between basic studies of normal neural ontogeny and those of neuroteratology. Because cocaine is a potent central nervous system stimulant, understanding its effects on developing brain not only explicates the possible neurodevelopmental impairments resulting from intrauterine exposure to a potential neurotoxin but also sheds light on mechanisms of normal neural ontogeny. Cocaine is also a remarkably addictive drug and hence, the adults caring for children prenatally exposed are often themselves hindered in their parenting by their addiction and all the associated environmental perturbations that accompany a drug-using lifestyle. Hence, we also present a model by which the biological vulnerabilities potentially conveyed by prenatal exposure to cocaine interact with the environmental handicaps accompanying life in a substance dependent world. First we turn to the central tenets of a teratological perspective.

MODELS OF NEUROBEHAVIORAL TERATOLOGY The field of teratology in general and neurobehavioral teratology specifically is a very recent invention reflecting combinations of several different fields including 193

194

Mayes and Pajulo

pharmacology, pathology, embryology, genetics, anatomy, and cytology. Teratology is the study of birth defects, their etiology, pathophysiology, and epidemiology (1). Teratology is the study of perinatal developmental injury or abnormal development and of the factors including birth accidents and genetic mutations that increase the risk of developmental injury (2). In this broader interpretation, the field covers a vast range encompassing the impact of exposures to exogenous agents or events during specific developmental phases on physical and functional outcomes. The outcomes may range from death, physical malformation, growth abnormalities, and disruption in function in all organ systems including the central nervous system. Among the agents or events of interest are both prescribed and illicit drugs, industrial chemicals, environmental pollutants, irradiation, viral infections, traumas such as preterm birth, and psychological conditions such as increased perinatal stress or maternal depression/deprivation. An even more recent research specialty is neurobehavioral teratology, which investigates the developmental impact of exposure to similar exogenous agents or events during different critical periods on the developing brain, and hence, on the child’s psychological development (3). The impact of individual exposures is assessed at varying distances in time from the original fetal exposure and varying levels of analyses (e.g., behavioral, neuropsychological, neurochemical, neuropathological). Behavioral teratology traces its origins not only to studies of CNS birth defects but also to preclinical animal models on the effects of early experiential differences in, for example, handling on behavior (3). As a field, behavioral teratology adds developmental psychology and developmental neuroscience to the multidisciplinary mixture of fields studying birth defects and developmental injuries. Some of the earliest questions in this field were about the effects of maternal malnutrition and of irradiation on fetal and infant neurodevelopmental outcome (4), and the most contemporary questions are about the impact of maternal stress and anxiety on long-term infant and child neurodevelopment (5,6). Questions regarding the impact of exposures to exogenous agents or events during specific developmental phases are implicit in nearly every consideration of a developmental or psychiatric disorder or the manifestations of that disorder across development. Interest in developmental injury from a neurodevelopmental/neurobehavioral point of view has provided a significant incentive to understanding the normal or expectable features of early perceptual, social-emotional, and cognitive processes and how to measure more accurately and specifically these developmental domains. One of the major thrusts of studies of physically defined perinatal risks such as prematurity has been to compare normal and abnormal psychological development, predict later outcome, and design interventions to alter the predicted developmental trajectory (7). These kinds of comparative studies are also common in studies of perinatal exposures to licit and illicit drugs or to environmental toxins: there is less of an emphasis on understanding mechanisms of developmental injury than on describing differential outcome. Questions regarding developmental injury related to prenatal exposures may be approached from two broad perspectives. One perspective examines a question by its presumed causes, that is, by grouping subjects into those with and without the presumed cause of injury such as exposure to a specific drug that produces mutagenesis or other somatic injury or malformation (e.g., alcohol). Another perspective focuses on questions of development injury not through the putative cause but rather through the target organ, system, or function of injury. Hence, subjects of study are grouped by their functional impairment and the programmatic focus is to understand the various mechanisms and routes to that particular functional impairment. These two orientations may lead to different research questions, findings, and interpretations inasmuch as one focuses on

NDS of Prenatal Cocaine Exposure

195

outcome while the other emphasizes mechanism. The most productive approach to any investigation of developmental injury brings a combination of these two perspectives and a plurality of methods to the research questions. Studies of prenatal cocaine exposure to date are accomplished largely through the first perspective using longitudinal, prospective designs of exposed and non-exposed cohorts. However, as the potential possible mechanisms of effect of prenatal cocaine exposure are better elucidated, it is clear that the second perspective also has some merit and applicability to the problem especially as we will illustrate below with examples from studies of dopaminergic system function. In either perspective, the fundamental logic of questions from behavioral teratology is does exposure to X during a specific phase of development cause A, B, C and/or D immediately or later in development. (Or in the language of mechanism, does disruption in process X during a specific phase of development lead to disruptions in functions A, B, C, and/or D later in development.) Importantly, one agent, Y, may produce several different outcomes depending on dose or amount of exposure and there may be different doseresponse curves for the different outcomes. The shape of the dose-response curves may also be different depending on the outcome of interest. That is, one outcome may reflect a linear relationship with exposure—more impairment with more exposure, while another outcome may reflect a threshold effect. Only exposures above a certain amount or time are teratogenic. Perhaps most importantly for studies of neurocognitive development is the observation that functional behavioral or neuropsychological changes usually occur at lower doses than abnormal growth or major disruptions in organogenesis (8). This latter point is key to understanding why assurances regarding no association with birth defects or physical malformations are not sufficient to assuage concerns regarding effects on neurobehavioral development, outcomes that are usually of equal if not greater interest to child psychiatrists and developmentalists. There are several classic examples of direct causative links between prenatal exposures and functional/physical outcomes. These include prenatal rubella exposure and its association with deafness and mental retardation or the classic example of prenatal thalidomide exposure and severe malformations of limb development. In these models, there is a clear association between a specific exposure to a discrete toxin and a clearly defined and easily identified endpoint or outcome. However, many, if not most, of the more contemporary questions capturing the interest of clinical scholars are those that involve far more complex models of exposure, timing, and outcome assessment. These are not always clearly direct causality models and hence are not easily approached with standard research designs. The exposures are neither specific nor discrete, the outcomes are not uniformly present even with documented exposure, and the severity or extent of the deformation or developmental abnormality is variable. In these more complex models, interactions between the exposure agent or event and the environment are central. That is, does the environment in one way or another moderate the fetus’s or child’s risk of exposure as well as vulnerability to the potentially toxic effects of exposure and at the same time, determine other risk factors that may also mediate the severity of any exposure related outcome. For example, even the question of malnutrition and its effect on fetal outcome is not a straightforward question of exposure and effect. Malnutrition more often occurs among very poor or displaced populations who are usually considerably stressed and isolated from adequate medical and prenatal care. These conditions may further compound the impact of malnutrition on fetal development in a way that would not occur theoretically if malnutrition occurred in isolation or in the absence of social displacement and chronic stress. Similarly, a more contemporary question regarding the effects of maternal anti-depressants on infant neurodevelopmental integrity is also made more complex by the possible relations between maternal depression, heightened perinatal stress, and altered maternal-infant care. Exposure

196

Mayes and Pajulo

to drugs of abuse is one of the clearest examples of an interactive exposure model. While cocaine may be associated with direct toxic effects on aspects of neural development, how these effects are expressed functionally across development depends on many other intervening and mediating conditions. Within these broad tenets of a teratological perspective, there are at least ten basic principles that guide studies seeking to causally link a putative exposure to a given set of outcomes (9–11). These principles, also outlined in Table 1, include: (1) Delineating the possible mechanisms of teratogenic effect; (2) Defining the specific teratogenic agent; (3) Specifying the timing of the exposure; (4) Defining the nature of the exposure; (5) Delineating the range of susceptibility and response relationships; (6) Selecting those groups at greater or lesser risk for exposure; (7) Considering the environmental context and conditions most related to the exposure; (8) Defining the outcomes most likely related to the mechanism of action of the exposure agent or event; (9) Considering when exposure related outcomes are most likely to be apparent; and (10) Taking into account those Table 1

Principles of Behavioral Teratology

Delineating the possible mechanisms of teratogenic effect: Agents that are behaviorally teratogenic should act on the developing CNS by specific mechanisms Defining the specific teratogenic agent: Not all agents that produce malformations are necessarily behavioral teratogens. Only those agents that produce either teratogenic or psychoactive CNS effects are capable of producing behavioral teratogenic effects Specifying the timing of the exposure: Based on the principle of critical periods, the type and magnitude of the behavioral teratogenic effect will depend on the stage of CNS development when exposure occurs Defining the nature of the exposure: The type and magnitude of the behavioral teratogenic effect depends on the type of agent, frequency and amount of use, and route of administration Delineating dose response relationships and the range of susceptibility: The type and severity of the behavioral effects depends on the dose of the agent reaching the developing central nervous system. Behavioral teratogenic effects are usually demonstrable at levels of exposure below that causing other malformations if the exposure agent is capable of causing behavioral changes Selecting those groups at greater or lesser risk for exposure and susceptibility to effects: How exposed groups are identified influences the likelihood of finding greater or lesser behavioral teratogenic effects. Individual genetic differences in the exposed individual or organism also influence the type and magnitude of behavioral teratogenic effect Considering the environmental context and conditions most related to the exposure: The magnitude and type of a behavioral teratogenic effect (and the likelihood of finding such an effect) depends on environmental factors Defining the outcomes most likely related to the mechanism of action of the exposure agent or event: Behavioral teratogenic effects are expressed as impaired cognitive, perceptual, or social emotional function or delayed maturation of capacities in these domains and the chosen outcomes for study should be linked to the proposed mechanism of CNS teratogenesis rather than selecting “broad band” measures of CNS function Considering when exposure related outcomes are most likely to be apparent: Not all behavioral teratogenic effects are apparent in the perinatal period. Some are evident later in development when environmental demands on specific functional domains are higher or when periods of developmentally time CNS reorganization are occurring Taking into account those conditions that ameliorate or exacerbate any exposure related functional outcomes: Some behavioral teratogenic effects may be exacerbated or ameliorated by other exposures or environmental conditions such as how the organism is handled or parented or other unexpected events such as illnesses that occur after the exposure period Source: Mayes and Ward, 2003; Vorhees, 1989; Vorhees, 1986b.

NDS of Prenatal Cocaine Exposure

197

conditions that ameliorate or exacerbate any exposure related functional outcomes. Importantly, these principles cut across animal and human models. Indeed, principles of behavioral teratology developed in animal models should also apply to studies of humans (8) although the methodological challenges may be more complex in the human model. The problem of prenatal cocaine exposure in particular illustrates the principles of defining the independent variable (Table 1, principle 4), delineating dose-response relations (principle 5), defining cohorts based on risk of exposure (principle 6), and specifying the environmental context and conditions most related to the exposure (principle 7). Linking putative outcome to proposed mechanisms of effect is perhaps the most important of these principles. While it may seem obvious that it is important to consider the possible mechanisms of teratogenic effects of any given agent, it is not always the case particularly in studies of the human model that mechanisms of effect are either specified or hypothesized beyond the general expectation or assumption that, for example, psychoactive drugs administered during active CNS neurogenesis should be potentially teratogenic. This assumption not only ignores consideration of specificity of effect on particular CNS regions and functions but also does not permit a more hypothesis driven consideration of the possible domains of outcome to study. Often these possible mechanisms of action are defined not through investigations of the specific agent but rather through studies of other agents with similar mechanisms of action in the brain. For example, the potential effects of cocaine on developing monoaminergic systems in the fetal brain have been delineated through in vitro and in vivo studies of monoaminergic regulation of neurogenesis, neuronal migration, and synaptogenesis as well as through direct study of disruptions in fetal brain structure-function relations with cocaine exposure. Furthermore, the environmental context related to the exposure may have profound impact on how and when the outcomes of exposure are expressed postnatally. Studies of prenatal cocaine exposure in human models that are informed by preclinical investigations are paradigmatic of contemporary neurobehavioral teratology perspectives which take into account both the biological mechanisms related to the exposure and the environmental conditions in which that exposure occurs and the child develops.

PRENATAL COCAINE EXPOSURE Magnitude of the Problem Systematic information about illicit drug use, and specifically cocaine use, across the country has been available since 1971 through the National Household Survey on Drug Abuse (NHSDA), now the National Survey on Drug Use and Health (NSDUH). This survey is the primary source of information on the use of illicit drugs, alcohol, and tobacco by the civilian, noninstitutionalized population of the United States aged 12 years old or older and interviews approximately 67,500 persons each year (12). In 2002 in the NHSDA, an estimated 19.5 million Americans aged 12 or older were current illicit drug users, meaning they had used an illicit drug during the month prior to the survey interview. This estimate represents 8.3% of the population aged 12 years old or older. Of these, an estimated 2.0 million persons (0.9% of total U.S. population) were current cocaine users (reporting use in the previous month), 567,000 of whom used crack during the same time period (0.2% of total U.S. population). This is a slight increase from the 1997 estimate of 1.5 million Americans who were current cocaine users, an estimate that had not changed significantly since 1992, but remains less than the 1985 peak of 5.7 million cocaine users (3% of the population in 1985). Compared to any other age group, adults 18 to 25 years old had a higher rate of illicit drug use (20.2% of the total population) and a higher rate of

198

Mayes and Pajulo

current cocaine use (2.0% of the total population and 15.4% of the total number reporting illicit drug use). Overall, among those in the 18 to 25 years old group reporting drug use, men had a higher rate of current cocaine use than do women (18.1 vs. 12.7%). More dramatically, among the 18 to 25 years old age group, 6.7% of the total population reported use in the last year and 15.4% report some use of cocaine over their lifetime. The average age of first cocaine use in 2001 was 20.8 years, a peak childbearing age. In 1992, the first national study of drug-use during pregnancy found that more than 5% of the 4 million women who gave birth in the United States in 1992 used illegal drugs while they were pregnant (13). Marijuana and cocaine were the most frequently used illicit drugs—2.9%, or 119,000 women, used marijuana and another 1.1%, or 45,000 women, used cocaine at some time during their pregnancy. The survey found a high incidence of cigarette and alcohol use among pregnant women. At some point during their pregnancy, 20.4% of pregnant women smoked cigarettes and 18.8% drank alcohol. There was also a strong link between cigarette smoking and alcohol use and the use of illicit drugs in this population. Among those women who used both cigarettes and alcohol, 20.4% also used marijuana and 9.5% took cocaine. Ethnic differences in patterns of drug use were also apparent in the survey, that is, about 4.5% of African–American women used cocaine compared with 0.4% of white women and 0.7% of Hispanic women who did so. In 2002, 6.6% of pregnant women aged 18 to 25 years (compared to 17% of nonpregnant women in the same age group) used illicit drugs in the past month (12). Across the child bearing years (ages 15 to 44 years), 3.3% of pregnant women reported illicit drug use in the previous month and 0.9% reported cocaine use. Of those women reporting any illicit drug use, 9.1% reported crack/cocaine use in the last month. The peak period for crack– cocaine use during pregnancy was between 1985 and 1990. In 1989, the Select Committee on Children, Youth, and Families conducted a survey of hospitals in which fifteen of the eighteen hospitals surveyed reported a three- to four-fold increase in drug-exposed births since 1985 (14). It was estimated in these peak years, that approximately 375,000 children each year were born exposed to cocaine (15). Since the peak period for crack–cocaine use in 1985, the rates of cocaine use among pregnant women in the general population has declined though it has remained relatively steady in the last few years. Nonetheless, in the years of peak use between 1985 and 1990, a number of the cohorts of prenatally cocaine-exposed children currently reaching early adolescence were recruited. However, despite declining prevalence rates in the use of illicit drugs during pregnancy, of all the illicit street drugs, crack-cocaine continues to be widely available and in some areas, powdered cocaine may actually be less expensive than in previous years, a factor that surely contributes to the continued use among many cocainedependent individuals (16). In some regions of the country such as Atlanta, Washington, DC, and New Orleans, cocaine-crack related admissions to hospitals are as high as 68 to 40%. Among women arrestees in New York, for example, 58% test positive for cocaine. Indeed, of the 21,276 cocaine-related arrests in New York City from January to October 2001, 83% involved crack. Furthermore, among women entering substance abuse treatment units, the rates of cocaine and crack addiction remain significantly high (16). Data from the Treatment Episode Data Set (TEDS) indicate that in 2000, more than a decade after the introduction of crack cocaine, smoked cocaine was the primary substance of abuse for 14% of all adult women admitted to substance abuse treatment (17). Indeed, in 2000, cocaine (including both smoked and other routes of administration) was the third most common illicit drug responsible for admissions to substance abuse treatment facilities. In 2000, the average adult woman entering treatment for primary use of smoked cocaine was 35 years old. The average length of smoked cocaine use was

NDS of Prenatal Cocaine Exposure

199

12 years prior to admission. Adult women entering treatment for smoked cocaine abuse were disproportionately black (58%, compared with 25% of all women entering treatment). About one-third (32%) of adult women enter entering treatment for smoked cocaine abuse were white, and 5% were hispanic. The proportion of women 35 years or older entering treatment for primary use of smoked cocaine more than doubled over time, from 20% in 1992 to 52% in 2000 (17). These data indicate that despite the reduction in overall crack–cocaine use among pregnant women by 2002, there remains a significant problem among substance dependent women of child-bearing and child-rearing age such that a considerable number of children who may have been also prenatally exposed continue to live in substance using environments.

Studies of Prenatal Cocaine Exposure Mechanisms of Effect In developing brain, there are a number of candidate mechanisms that account for how prenatal cocaine exposure may interfere with neural ontogeny. These are effects on monoaminergic system development, changes in neural growth factors, alteration of ion channel and monoamine transport development, effects on other neurotransmitter systems such as the cholinergic system and on neuropeptides including substance P, dynorphin, GABA and glutamate sites, or alteration of immediate early gene expression (18–20). Cocaine also has potent vasoconstrictive effects that may have a more generalized and less system-specific hypoxia-related deleterious effect on neural growth. The effect on central and peripheral noradrenergic systems accounts for the cardiovascular effects of cocaine including tachycardia, hypertension, and increased risks for strokes and myocardial infarctions and arrhythmias. Thus, in-utero cocaine exposure may compromise fetal brain development through an effect on placental vascular function with decreased placental blood flow and consequent fetal hypoxemia and possibly ischemic injury. Similarly, cocaine-related noradrenergic effects on developing fetal vasculature may also compromise blood-flow to developing brain (21,22). This more general level of effect through fetal hypoxemia or relative ischemia may not be as specifically directed to one region of the brain over another and the severity of effect, if any, will depend particularly on the timing of the exposure and the level of relative hypoxemia. Another group of mechanisms of in-utero cocaine effect on neurodevelopmental outcome is also more general but unique to the human model. Substance abusing adults rarely use cocaine in isolation. Most often, cocaine abuse is combined with other potential neuroteratogens including alcohol, marijuana, and tobacco. While there are no studies modeling various combinations of polydrug exposure in animals, it is reasonable to hypothesize that interactions among different drugs may both produce potentially neurotoxic metabolites as well as directly additive effects on neural development. Related to human substance abuse are other more general health and environmental factors that have effects on fetal health and development. These include poor maternal nutrition (in part reflecting the appetite suppression effect of cocaine) and more often than not, maternal poverty and environmental chaos that may contribute to postnatal neglect and poor parenting, factors that are themselves developmentally compromising. These levels of effect unique to humans cannot be adequately modeled in animals but do confound efforts to attribute any neurobehavioral impairments found in prenatally cocaine-exposed infants and young children to cocaine alone. The methodologic dilemmas in human studies of prenatal cocaine-exposure are reviewed in detail elsewhere outcomes (20,23–32).

200

Mayes and Pajulo

Here we focus primarily on the effects of prenatal cocaine exposure on monoaminergic system function and ontogeny [including the catechoamines dopamine (DA) and norepinephrine (NE) and the indoleamine, serotonin] though several of the aforementioned candidate mechanisms may also be influential in the neurobehavioral profiles of altered arousal and attention regulation. Although inconsistent, inconclusive, even contradictory on many crucial issues, and marked by a number of methodological problems, collected findings from published studies to date suggest a profile of possible cocaine-related effects on arousal and attention regulation and reactivity to stressful conditions. Also implicated are aspects of neurocognitive performance reflecting prefrontal cortical functioning including capacity for sustained focus, inhibiting nonsalient responding, and working memory (27,33–44). This profile is further elaborated by findings from several preclinical models in which important factors such as duration, route, and amount of exposure as well as environmental conditions may be more adequately controlled (45–48). In particular, recent preclinical studies of prenatal and preweaning cocaine-exposure in the rat point to direct and enduring effects on monaminergic mediated circuitry that is important in arousal regulation of attentional and prefrontal cortical executive systems (20,49–55). Importantly, the non-specific but frequent mixture of acute and chronic stress characterizing the postnatal caregiving environments for many prenatally-exposed children may also have enduring effects on attention and arousal system ontogeny perhaps through some of the same monoaminerelated mechanisms. Cocaine and the Monamine System In the mature brain, cocaine/crack act at the level of the synapse to block reuptake of monoaminergic neurotransmitters (DA, NE, and serotonin) at the presynaptic junction. Presynaptically, cocaine is more efficacious in blocking the reuptake of serotonin (5-HT) as compared to the catechoamines DA and NE (56). The mechanism by which cocaine blocks monoamine reuptake involves blocking the action of the presynaptic transporter protein that takes neurotransmitter back into the presynaptic neuron. In the short term, blocking reuptake effectively leaves a greater concentration of monoamines in the synaptic cleft and thus the potential for increased synaptic activity and increased neurotransmitter turnover. In addition, stimulants such as cocaine also accelerate presynaptic release of catecholamines (DA and NE) (57). The DA system is particularly central to the effects of cocaine on both adult and developing brain. DA neurons have long been associated with the initiation of behavior, reward, and motivational processes (58). Two major dopaminergic circuits are involved in the action of stimulants. The nigrostriatal DA system originates in the substantia nigra and projects to the caudate nucleus and putamen (dorsal striatum). The degeneration of this system is implicated in Parkinson’s disease and is also important for the focused repetitive behaviors associated with high doses of stimulants (58,59). These pathways are also associated with the planning, initiation, and coordination of voluntary movement and complex behavioral repertoires (60). The mesolimbic DA system projects from the ventral tegmental area of the midbrain to the limbic forebrain (including the nucleus accumbens, olfactory tubercle, amygdala, and frontal cortex). This second dopaminergic system has been most implicated in activation and locomotor behavior (61) and stimulant induced locomotion (58). The regions involved in the mesolimbic system, especially the amygdala and the prefrontal cortex projections from the amygdala, are also important in the regulation of stress response and emotional reactivity. The mesolimbic dopaminergic circuits are probably critical for the stimulant-induced euphoria and for the reinforcing

NDS of Prenatal Cocaine Exposure

201

(addictive) properties of drugs (62–64). The nucleus accumbens (ventral striatum) within the mesolimbic system seems particularly important for the reinforcing properties of stimulants because animals will self-administer large amounts of cocaine to this particular region even to the point of seizures and death (65). DA acts at specific DA receptor sites and there are at least five DA receptor types that cluster into D1-like (D1 and D5) and D2-like (D2, D3, and D4) subfamilies (66). D1 and D2 receptors are the most extensively studied. Dopaminergic neurons from the dorsal and ventral striatum have efferent neurons to the substantia nigra (the striatonigral system) and to the thalamus by way of the globus pallidus (the striatopallidal system). Each of these systems differs in DA receptor subtypes and in neuropeptide receptors. The striatonigral system has predominantly D1 receptors and the neuropeptides dynorphin and substance P while the striatopallidal system has D2 receptors and the neuropeptide enkaphalin. These two DA systems may function differently in response to cocaine exposure in developing brain (see below). The serotonin system originating from the raphe nuclei of the brainstem project widely to the cortex, and to memory and emotion regulation areas such as hippocampus, thalamus, and amygdala. Just as with DA, the effects of cocaine on serotonin systems appear to be through the inhibition of reuptake processes and increased release (56). There are at least four classes of serotonin receptors in the brain (5-HT1–4) with subtypes of 5HT1. Most studies of the action of cocaine have investigated effects on 5-HT1a and 5-HT2, the former present in high density throughout the hippocampus, amygdala, hypothalamus, and cortex while the latter is more concentrated in the cortex. The NE system is crucial to the regulation of behavioral state, sensory processing, aspects of memory, and attention regulation (67). The noradrenergic system has projections throughout the limbic and corticothalamic system with wide distribution of a and b-adrenergic receptors. Earlier work had suggested that the NE system was important for the reinforcing effects of cocaine and related drugs but subsequent studies have suggested DA as the more critical catecholamine for that role (57,64). The noradrenergic pathways may be most critical for the non-specific activating effects of cocaine and the peripheral effects on overall vascular tone and cardiac function (68).

Preclinical Models of Prenatal Cocaine Exposure Cortical Morphology Growing evidence is available for cocaine-related effects on the earliest phases of cell proliferation and neuronal migration—and thus, on cortical morphology. At the stage on neuronal formation and proliferation in cultured preparations of glioblastoma and neocortical cells, cocaine exposure delays uridine and thymidine incorporation (69). Similarly, in cultured preparations, differentiation of neuroblastoma cells by nerve growth factor (NGF) and cell proliferation stimulated by insulinlike growth factor (IGF-I) are both delayed by cocaine. Prenatal exposure in the intact pregnancy with rats early in gestation interferes with radial gliogenesis and thus disrupts neuronal migration and resulting cortical architecture (70–73). In rhesus monkeys, intermittent prenatal cocaine exposure results in cerebral cortices with highly abnormal structural characteristics including disrupted cortical laminar architecture with an increased number of cells in the underlying white matter consistent with markedly impaired neuronal migration (74–77). Many cortical cells do not reach their proper destination and there is a decrease in the density of cortical glial (connective) elements. Mice embryos exposed to cocaine very early in gestation also show a decrease in glial formation, glial density, and a disorganization of

202

Mayes and Pajulo

axon/dendrite bundles (73). Findings such as these may reflect disrupted monoaminergic system regulated processes that control the genesis of radial glial cells necessary for proper neuronal migration to cortical layers and regulate the formation of connections among neurons (74,75,78,79). In addition to the effects of cocaine on monoaminergic regulated glial cell formation and maturation, cocaine may have a direct toxic effect of early stages of neuronal proliferation and also early connectivity as suggested by cell culture preparations of both glia and neurons. With cocaine exposure to growing neurons in culture, there is a decrease in the number of neurons and a degeneration of the connective processes through which communication with other neurons occurs. There is no apparent reduction in glial cell numbers but there is also a change in structure suggesting again a cocaine-related effect on the morphology of cells involved in neural connectivity. Closer examination of cocaineexposed neurons show morphologic changes in nuclei indicative of apoptosis or neuronal death after approximately 2 days of exposure, suggesting that cocaine exposure initiates apoptotic processes within the neuron by yet undefined mechanisms. One possibility may be through monoamines since in neuronal cell cultures, exposure to catecholamines can lead to neuronal death through action as a glutamate agonist (80). Prenatal cocaine exposure is also associated with abnormally elongated dendrites in corticolimbic pyramidal neurons of cocaine-exposed animals. Dendrites of pyramidal neurons in layers III and V of anterior cingulate cortex are 30 to 50% longer in exposed compared to non-exposed, saline-treated animals (81–83). Confocal analysis shows that these dendrites course abnormally through the cortex. Rather than the expected straight distribution, the dendrites course in and out of the plane of a section. Since the exposed animals show normal lamination and thickness of cortical layers in the anterior cingulated. The trajectory pattern of the dendrites suggest less-controlled growth with the extended dendritic projections undulating to fit within the limits of the cortical layers. These morphologic changes persist into adulthood. Similar observations are made in the prefrontal cortex, another cortical region targeted by dense dopaminergic afferents but these changes in dendritic growth are not found in regions that receive sparse dopaminergic input such as the primary sensory cortex (81–83). These effects on dendritic growth appear rapidly. In vitro studies show that after only two weeks of exposure to cocaine, fetal neurons from the anterior cingulate plated in culture without further cocaine exposure show marked increased in dendritic growth, a finding suggesting that the cocaine-related effect on growth is not simply a direct exposure effect but rather through an alteration in growth-regulating mechanisms. Direct Effects on Monoamine Function Findings from preclinical rodent models point to specific effects on DA D1/D2 mediated circuitry critical to arousal regulation of the prefrontal cortex (84) and to upregulation of NE (and serotonergic) systems at a time in early development that could permanently distort the excitatory/inhibitory balance between arousal systems necessary for optimal cortical function. Dopamine. Adult rats exposed prenatally to cocaine show a reduction in basal DA release and decreased activity in dopaminergic cells in the substantia nigra and ventral tegmentum (52). Reductions in glucose metabolic activity are particularly marked in prenatally exposed male rats in the hypothalamus, nigrostriatal pathway, and structures within the limbic system (85). Prenatally exposed rabbits show reduced striatal DA levels at birth (54) and rhesus at 60 days prenatally show reduced tyrosine hydroxylase mRNA (86) in the substantia nigra and ventral tegmental area.

NDS of Prenatal Cocaine Exposure

203

A reduction in DA synthesis may lead to increased receptor sensitivity and proliferation which is the case in both the prenatally exposed rhesus and rat (87–89). Other studies examining receptor sensitivity and density following prenatal exposure have examined postnatal brains following early or late exposure and have reported conflicting results with either a decrease, no change, or an increase in the expression of D1 and D2 receptors and ligand binding to those receptors (53,88,90–92). These conflicting results may reflect compensatory changes in DA neural transmission following withdrawal from cocaine or cessation of exposure, a suggestion that has been confirmed in adult animals (93). When cocaine is administered postnatally to rat pups in either the first or second week postnatally (equivalent to later gestation in the human and during a period of rapid synaptogenesis), there is widespread increase in glucose metabolic activity across several different regions of the brain including cortex (89) but with particular increases in highly dopaminergic innervated regions. These effects are most dramatic in the female. For the male, there are no substantial areas of increased metabolism and the nucleus accumbens continues to show decreased activity as was true prenatally. When ligands specific for DA D1 and D2 receptors are injected into postnatally treated animals, areas showing marked increase in glucose uptake are those with greater dopaminergic activity (94). Earlier postnatal exposure (days 1 to 10) affects the mesolimbic DA system and other structures within the limbic system differentially more than later postnatal exposure (days 11 to 20) which results in more general activation (95). Such general activation consistent with the relation between monoamines and synaptic formation suggests that exposure to a monoaminergic agonist such as cocaine during these periods of cortical maturation results in an overall increase in monoaminergic (and specifically dopaminergic) activity. More specific studies of receptor functioning have shown that prenatal cocaine exposure is associated with a functional decoupling of the D1 receptor from G protein, the second messenger within the cell that begins a cascade of intracelluar events. Prenatal cocaine exposure impairs coupling in the frontal and cingulate areas of the dopaminergic D1 receptor system to one of the second messengers, GaS protein, a functional decrease that persists into adulthood while coupling between D2 and Gi protein remains normal (50,53,92). This decoupling of D1 to GaS protein may be secondary to increased phosphorylation of the receptor that in turn renders the receptor desensitized to G-protein (96). Also, mRNA for the DA transporter is reduced in these same regions following both prenatal and postnatal exposure (49,91). Taken as a group these findings suggest a selective or at least a more specific attenuating effect of cocaine on the development of the D1 system. In the normal individual, D1 and D2 receptor systems function interactively and in some ways reciprocally to regulate thalamic and cortical inputs. In particular, a reciprocal relation between D1 and D2 activity coupled also with a balance between noradrenergic a2 and a1 activity regulates and protects prefrontal cortical activity during increasing arousal or stressful states (97–99). Increased D2 (and a1) (100) activity may in effect take certain prefrontal cortical activities off-line and in effect interfere with aspects of attention and other higher order reasoning and discriminative abilities (functions that typically also deteriorate in hyperaroused states—see above). An increase in D2 activity might be expected to parallel attenuation in D1 systems; and indeed, prenatal cocaine exposure does appear associated with an increase in D2 binding affinity, receptor density, and mRNA for the D2 receptor in the striatum (91,101,102). Postnatal exposure (days 1–9) also results in an increased density of D2 in striatum (51). Additionally, challenges to the D2 system using a D2 antagonist, haloperidol, result in an exaggerated response in postnatally exposed male rats (103).

204

Mayes and Pajulo

Serotonin In relation to serotonin, in the developing brain cocaine appears to have greater affinity for the serotonin transporter in contrast to the adult brain in which cocaine binds predominantly to DA transporters (104). Effects such as these have led several to suggest that in the developing brain, the relationship between cocaine and serotonin systems may be more important to understanding mechanisms for teratogenic effects while in mature, and perhaps in postnatal brain, DA systems may be more critical (90). While less work has been done to date on serotonin systems and cocaine exposure, the findings are nonetheless consistent with hypotheses based on serotonin as an early trophic factor and cocainerelated effects on serotonin metabolism. During development, serotonin neurons use a negative feedback mechanism for regulating growth. Both tissue culture and whole animal studies show that early in development, increased extracellular serotonin causes a decrease in serotonin neuronal outgrowth. It has been demonstrated that early on cocaine inhibits the development of reuptake mechanisms (105). Thus, based on cocaine’s effect on synaptic reuptake, more serotonin should be present extracellularly in the synaptic cleft and thus reduce the growth of serotonin terminals (106). These predictions have been supported. Studies of prenatally exposed rats have revealed a decrease in the serotonin terminal density and an accompanying reduction in the normal growth of serotonin fibers in the hippocampus. The effect is transient, however, and by a month postnatally no differences between exposed and non-exposed animals are apparent (107). On the other hand, by a month postnatally, late-onset increased growth of serotonin fibers is observed in the striatum (108). It may be that if growth is inhibited or delayed in one region, it is increased elsewhere (106). But what about the return to normal by a month postnatally of serotonin fiber density in cortex and hippocampus? Rather than suggesting transient to no effect of cocaine, even early, brief disruption of one aspect of serotonin system ontogeny may disrupt other areas of the serotonin system development. Delayed serotonin terminal development may lead to decreased serotonin levels in later phases even after exposure has stopped. Decreased serotonin levels in the later prenatal and early postnatal period may have marked consequences for neuronal connectivity. For example, depletion of serotonin by selective synthesis inhibitors results in delayed neuronal differentiation and decreased synaptic connectivity in both cortex and hippocampus in the chick model (109). In rats, depletion of serotonin on the third postnatal day by a selective neurotoxin results in decreased density of granule cells dendritic spines or decreased connectivity which actually continues to worsen as the animals mature (110). Depletion at postnatal day 10 to 20 causes marked loss in dendrites up with reductions still apparent up to a year of age (111). Moreover, delaying serotonin terminal development has been associated with behavioral deficits in the adult animal (112). In other words, by altering serotonin terminal development, prenatal exposure alters events weeks downstream from the exposure. In the first weeks postnatally in the rat, serotonin is required for astroglial maturation, and in the second to third postnatal week, serotonin plays a critical role in neuronal maturation and synaptogenesis (113). Postnatal synaptogenesis and neuronal connectivity depend on the maturation of a number of cell types including astroglial cells. Astroglial cells that play a role early in gestation as the radial glial cells for neuronal migration are also crucial for later connectivity and the production of neural growth factors. Serotonin has a regulatory role in the maturation of the astroglial cell (114,115); and the rapid increase of serotonin 1A receptors (5-HT1A) early in gestation in mammalian brain is largely through serotonin receptors on astroglial cells (116). One aspect of that

NDS of Prenatal Cocaine Exposure

205

maturation is a morphologic change from the long processes needed for guiding neuronal migration to shorter but more processes with greater connectivity among cells. A second aspect of serotonin-mediated astroglial maturation is the release of growth factor S100b (117). This particular growth factor causes neuronal outgrowth and neuronal plasticity in response to experience (such as activity, handling, enriched stimulation environments) (115,118). Thus, S-100b may be one factor involved in potentiation of learning and experience-dependent synaptogenesis. Depletion or decrease in S-100b release from astroglial cells should result in decreased dendritic development and delayed neural outgrowth. Thus, prenatal cocaine exposure may interfere with astroglial cell maturation and in the release of S-100b. Prenatally exposed rat pups examined on postnatal day 6 show a delay in the morphologic change in astroglial cells from their radial glial function to more mature cells (119). There are also fewer astroglial cells in the cortex and hippocampal regions with overall cortical thinning. (Similar reductions in cortical astroglial cells have also been reported for the rhesus model at 2 months postnatal age (74).) Immunochemical staining for S100b revealed significant reductions (71,119). It is possible then that even if serotonin levels returned to normal in the subsequent postnatal period, the reduction of S100b during that critical period may have more enduring effects. Conversely, treatment of prenatally exposed rat pups with 5-HT1a agonists on postnatal days 1 through 5 showed increases in S100b in both cortex and hippocampus and a significant increase in brain growth (71). Prenatal cocaine exposure associated delay in serotonin terminal growth, delay in astroglial maturation and release of S-100b may lead to loss of synapses in the adult animal and thus to deficits in learning and memory (106). In contrast to prenatal exposure, postnatal or preweaning exposure appears to increase serotonin levels. Postnatal exposure to serotonin agonists in the rat from birth to postnatal day 20 increases brain activity during active synaptogenesis leading to accelerated brain maturation with attendant behavioral changes including increased activity [and behavioral changes in the mature animal including increased anxiety and activity, (106,120)]. In addition, preweaning administration cocaine decreases 5-HT1a receptor density in the raphe nuclei (121,122) where the action of the 5-HT1a is inhibitory. Thus, decrease in receptor density is in effect an increase in afferent or ascending serotonergic activity (123). No models to date have combined prenatal and preweaning exposure. Serotonin neurons innervating the forebrain region also send projections to the hypothalamus and mediate changes in plasma ACTH, corticosterone, renin, and prolactin (124). Disruptions in serotonin system ontogeny in both receptor distribution and density of serotonin projections may also alter hypothalamic response to neuroendocrine challenge tests. Adult rats treated prenatally between day 13 and day 20 of gestation and studied at postnatal day 30 show increased ACTH and renin response to a serotonin receptor (5-HT1A) agonist (125,126). In contrast, ACTH and renin response is attenuated to a serotonin releaser acting at the presynapse. These changes in response to neuroendocrine challenge are not paralleled by changes in 5-HT1A receptor density in hypothalamus or midbrain or by any alteration in serotonin uptake sites at the presynapse or in levels of serotonin. However, by postnatal day 70, 5-HT1A receptor density is increased in cortex and midbrain but not hypothalamus (127), findings that speak to ongoing alterations in monoamine related systems after exposure has ceased and the alteration of effects downstream after early changes in system function. Furthermore, findings such as these also suggest that certain alterations in monoaminergic system function related to prenatal cocaine exposure may be only apparent when systems are challenged.

206

Mayes and Pajulo

Norepinephrine Far fewer studies have been done on the relation between prenatal cocaine exposure and the developing noradrenergic system and the majority have been in the postnatal but preweaning period. Preweaning cocaine administration appears to increase activity within the overall noradrenergic system (128). Also demonstrated has been increased fiber density in NE rich regions, increased receptor binding (b1 in cortex at 30 days postnatal but not later; a1 and a2 in cerebellum and forebrain), increased presynaptic synthesis of NE (90,128,129). Rat pups exposed to cocaine on postnatal days 4 through 9 and then exposed to a noradrenergic agonist (clonidine) showed enhanced response to the locomotor stimulating effects of the agonist (130). Noradrenergic neurons arising from the locus coeruleus innervate the entire forebrain with many projections to the posterior and prefrontal cortex. The locus coeruleus appears critical to vigilance or the response to a salient target (131) possibly through projections to the prefrontal cortex. Of special significance is the afferent input from the prefrontal cortex to the locus coeruleus that suggests a feedback, modulating link between cortical activity and ongoing noradrenergic activity (132). Excessive stimulation of adrenergic a1 receptors in effect take the prefrontal cortex offline (98). This in turn disrupts the regulatory loop between the locus coeruleus and the prefrontal cortex and may disrupt the capacity for vigilance and the ability to distinguish salient from non-salient information (133), a behavioral finding in cocaine-exposed animal models (134,135) and suggested in human studies (37,136). These findings underscore the importance for more preclinical work to examine cocaine-related effects on noradrenergic system ontogeny. Behavioral Correlates in Preclinical Models The neurobehavioral areas most frequently studied are: (1) locomotor behavior in open field exploration and responses to monoaminergic agonists and antagonists; (2) attention and conditioned learning; and (3) stress reactivity. In each of these areas, a pattern of neurobehavioral correlates has emerged. Several laboratories studying cocaine and monoaminergic function in the rat, mouse, or rabbit are also examining related behavioral alterations in those domains of behavior most likely related to alterations of mesolimbic or nigrostriatal systems. These include changes in locomotor behavior and in attention. The findings regarding locomotion are mixed depending on the assessment paradigm, the exposure period, and the time of assessment. Prenatal exposure appears to increase spontaneous locomotor activity in open-field exploration during the first 3 weeks of life but not by 60 days of age (137). Others have reported the same lack of relation between prenatal exposure and locomotor activity after the first weeks of life (138–140). Female rats treated on postnatal days 1 through 10 and studied between 60 and 65 days of age show decreased levels of activity (94). Prenatally exposed animals also inhibit approaching novel conditions and open-field exploration (141); and wall-climbing behavior, a dopaminergic regulated behavior, is attenuated with foot-shock in prenatally exposed animals (142). Exposure on postnatal days 11–20 reduced the locomotor response to amphetamine challenge in rats (143). Similarly, amphetamine challenge in prenatally exposed rabbits failed to elicit a stereotypic head bobbing, usually associated with activation of the D1 receptor, and this effect persisted through adulthood (81). The aforementioned hypothesis regarding the uncoupling of the D1 receptor rendering it dysfunctional has support in related behavioral studies. For one, D1 receptor activation regulates the increased locomotor response to cocaine (144). Reduced D1 activity would be expected to reduce behaviors usually stimulated by DA agonists,

NDS of Prenatal Cocaine Exposure

207

especially selective D1 agonists. D1 deficient mice do not show the usual increase in locomotor behavior on cocaine challenge (145,146), an observation perhaps consistent with the reduced response of pre- and postnatal cocaine-exposed rabbit and rat respectively to amphetamine challenge (143). D1 deficient mice also show reduced rearing behavior, a complex motor activity that is an essential part of an exploratory repertoire (145) and the reduction of which is similar to behavioral patterns seen in prenatally exposed animals (see below). While other motor behaviors are variously altered in D1 deficient mice (including overall activity level, grooming, and sniffing) (146), rearing may be a more sensitive behavioral marker of D1 receptor dysfunction (145). Also, studies of locomotor response to novelty suggest that high responding animals show greater D1 binding in the nucleus accumbens and less D2 binding in both accumbens and striatum (147), again a possible parallel with the reduced open field exploration behaviors of cocaine-exposed animals (see above). Second, in rodents, cocaine or stimulant administration in adulthood elicits a number of characteristic responses including head bobbing, increased sniffing, and activity, behaviors mediated in large part through the D1 system. Selective D1 antagonists abolish this characteristic behavioral response to cocaine as does prenatal cocaine exposure presumably at least in part through the decoupling effect on the D1-GaS system. Similarly, behavioral responses to agonists selective for the D1 receptor are reduced following both prenatal and postnatal exposure (148). There may also be evidence suggestive of a role for the D1 uncoupling in the learning deficits. Prenatally and/or preweaning exposed animals inhibit approaching novel conditions and open-field exploration (141). Deficits in learning are suggested by impairments in classical conditioning (142,149–151), deficiencies in active and passive avoidance tasks (149), poor short-term memory (152), increased response perseveration (153), diminished proximal cue learning (154), impaired learning on serial reversal tasks (155), impaired habituation (156), impaired reversal learning (157), increased susceptibility to distractors in a visual attention task (158), and less efficient error detection (159). Prenatally exposed animals appear unable to attend preferentially to less salient but relevant stimuli in the context of more salient but distracting and non-relevant background stimuli (135,160) and have difficulty changing previously learned conditions (18,161). This collection of findings cuts across prenatal and preweaning exposure. Prenatally exposed rabbits show pronounced anatomical changes in the anterior cingulate cortex (a richly dopaminergic region). Specifically, neurons in this region show marked increase in dendrite outgrowth (81), a deficit possibly connected to the decoupling of the D1 receptor inasmuch as D1 receptor activation in normal animals decreases dendritic outgrowth (162). Indeed, D1 receptor-G protein coupling has been shown reduced in the cingulate cortex in prenatally exposed animals (50,53). Also, the anterior cingulate cortex plays a central role in processes of learning and memory and is especially crucial to situations that demand attention preferentially to less salient but relevant stimuli when more salient, but not necessarily relevant, stimuli occur in the same context. Lesions in the anterior cingulate cortex impair attentional processes and in intact animals, attentional tasks lead to activation of this region (163). Uncoupling the D1 receptor from its second messenger systems dampens D1 mediated responses within striatal neurons and within the cingulate cortex regulated functions and impairs learning (134). Prenatally cocaine exposed rabbits with changes in the anterior cingulate region show reduced learning ability when the positive conditioning stimulus less salient than the negative stimulus (e.g., a soft tone versus a loud tone) (81,135). In other words, these animals cannot attend preferentially to less salient but relevant stimuli in the context of more salient background (and irrelevant) stimuli. These deficits persisted into adulthood

208

Mayes and Pajulo

(135). Other studies have also suggested an association between prenatal cocaine exposure and impaired attentional processes that impede the animal’s ability to change previously learned conditions or to screen out distracting but non-relevant information (18,161). Findings of this nature parallel reports from studies of prenatally cocaine-exposed preschool and school-aged children suggesting deficits in selective attention and in information processing (see below). Altered stress responsivity is suggested by several behavioral observations and is the domain most relevant to the earlier suggestion that early compensatory mechanisms which “normalize” basal conditions may nonetheless leave the adult animal with a vulnerability to novel, challenging, or stressful contexts (164). While neural “stress circuitry” is complex and involves many interactive systems including the corticosteroid-based HPA axis, dopaminergic and noradrenergic pathways in the amygdala and other limbic and corticothalamic systems are central (165). Furthermore, stress selectively stimulates many of the same dopaminergic regions as does cocaine. Prenatally exposed rats, while initially showing less motor response to open field exploration, do engage in exploration after a delay and then do so with markedly increased activity suggesting overarousal (90,166,167). Similarly, adult rats exposed to cocaine prenatally demonstrate less expected stress-induced immobility during a forced swim, intermittent foot-shock, or when placed in a novel condition following footshock (142,153,164,168–170). This alteration in stress responsivity as manifest by immobility changes with age. Prepubertal or preweaning animals exposed prenatally to cocaine show greater immobility in stressful conditions such as forced swim or footshock (170–172). Other behaviors suggest increased arousal and altered stress responsivity. Prenatally exposed animals are more sensitive to environmental demands and show more aggression than non-treated animals in conditions of water deprivation and competition for water (164,173). Increased aggression is not apparent however if the animals are not stressed (164). Far more work is required to link these behavioral observations of altered stress responsivity with the alterations in neurochemical function related to prenatal cocaine exposure. Finally, there is some suggestion that prenatally cocaine-exposed animals may be more sensitive to the effects of early experience (stressful and non-stressful). With chronic stress (such as repeated footshock over time), preadolescent cocaine exposed animals “reverse” their acute stressor pattern of reduced immobility and show more immobility as is expected in non-exposed animals (171). Non-exposed but repeatedly stressed animals show less significant changes in the pattern of their stress response. Findings such as these about the effects of “early experience” point to potentially productive lines of work using more traditional early experiences such as increased handling that have been studied in relation to adrenocortical function and stress (48,174).

HUMAN MODEL OF PRENATAL EXPOSURE Outcomes of prenatally cocaine-exposed children seen through adolescence range from assessments of physical growth and motoric maturation to standardized assessments of cognition and language, neurophysiological indices of arousal and sensory response, emotional regulation, maladaptive behavior, and a range of executive control functions. However, in both extensive descriptive reviews and meta-analyses, appropriate concerns have been raised about singularly attributing any one disruption in developmental capacities to the prenatal effects of cocaine on emerging neural systems (27). Prenatal exposure to cocaine is often paired with a simultaneous exposure to other drugs and to severe postnatal environmental deprivation (28,29,177–180). Only a few studies have

NDS of Prenatal Cocaine Exposure

209

addressed apparent cocaine-related effects on children’s development and behavior that are moderated or mediated (181) via pathways of maternal health or quality of caregiving (182,183) or have taken into consideration parental attentional or learning difficulties that may serve as markers of possible genetic contributions to infant functioning (184). Nonetheless, bearing in mind these cautions, a number of findings in human studies have appeared across laboratories and are consistent with those in preclinical models.

Neurochemical Findings in Infants and Children In human infants, neuro-chemical studies following prenatal exposure are scant but two areas of work point to effects on those neurochemical systems most related to arousal regulation and altered response to stress. Three studies have examined metabolites, precursors, or levels of NE or DA in the serum or cerebrospinal fluid of newborns exposed to cocaine prenatally and have found a significant increase in plasma NE (185) and catecholamine precursors in CSF (186) and a decrease in the metabolites of DA (187), findings suggesting an alteration in monoaminergic system function at least neonatally. Further, for the cocaine-exposed infants, NE levels were inversely related to features of the infant’s neurobehavioral profile (186). Second, cocaine exposed infants exhibit an attenuated cortisol response to noninvasive (neurobehavioral exam) and invasive (heelstick) manipulations despite no differences in baseline cortisol levels (188) suggesting that glucocorticoid mediated arousal regulatory systems are altered by in utero cocaine exposure or by the chronically stressful conditions associated with cocaine exposure (189).

Neurobehavioral and Developmental Findings There are at least two pathways by which disruptions in monoaminergic system ontogeny might effect neurocognitive functioning in prenatally exposed children (190). One relates to arousal modulated attention and related cognitive functioning (27,37). In this model, neurocognitive impairments in a range of prefrontal cortical functions are expected especially as children are more emotionally aroused. The second model is based on a direct effect of cocaine on cortical morphology and specialization (74–77). In this model, neurocognitive deficits might be expected in even non-stressed conditions in a range of functions including impaired inhibition and slowed reactions times (191). Findings from several groups are consistent with both of these models. A number of investigators have reported difficulties among prenatally cocaine-exposed children in arousal regulatory capacities ranging from increased excitability with poor state regulation, rapid changes in arousal with stimulation (measured by changes in heart rate), to increased arousal from sleep, and greater physiological lability (136,192–199), with some findings persisting at least through one year of follow-up (200,201) with some findings studied longitudinally and persisting at least through one year of follow-up (200). These differences in behavioral and physiological arousal are expanded upon by a few studies examining the relation between arousal modulated attention and impaired information processing in infants and toddlers including diminished responsiveness to novel stimuli and recognition memory, poor impulse control and task persistence, diminished sustained attention, and greater emotional lability and/or behavioral disorganization. (37,197,202–211) as well as other aspects of the stress response system, including the studies cited above of baseline and peak cortisol response to stressors in which cocaine-exposed infants show a depressed cortisol response to challenge (188) or lower baseline levels (202).

210

Mayes and Pajulo

Thus, evidence suggests that cocaine-exposed children may be more easily aroused and hence vulnerable to difficulties in information processing and learning in novel conditions consistent with one model of the developmental impact of prenatal cocaine exposure. At the same time, a number of labs are also reporting neurocognitive deficits under even optimal testing conditions using narrow band assessments to test specific functions such as task switching and inhibition. Collected findings include slower reaction times to visual stimulus presentation (38) slowed reaction times in continuous performance tasks (39), greater perseveration, diminished response inhibition (40,41,212,213), increased errors of comission or omission on A-not-B or continuous performance tasks (37,39–43), diminished capacity for sustained attention (44,214), and deficits in spatial learning (215). The range of attentional/executive control function effects that seem related to prenatal cocaine exposure may be consistent with preclinical work suggesting a cocainerelated effect on the development of the anterior cingulated cortex (216) and, as discussed in earlier sections, on dopaminergic D1/D2 mediated circuitry critical to regulation of prefrontal cortex (84) and to upregulation of noradrenergic (and serotonergic) systems at a time in early development that could permanently distort the excitatory/inhibitory balance between regulatory systems necessary for optimal cortical function. Consistent with findings discussed earlier regarding the decoupling of the D1 system, in preclinical models, behavioral responses to agonists selective for the D1 receptor are reduced following both prenatal and postnatal exposure (148). Similarly, differential reduction in cingulate activity in cocaine-exposed animals is associated with an impaired ability to discriminate among salient and/or relevant stimuli (134,135), a finding consistent with emerging data cited above in human studies. Also, because as cited above, cocaineassociated disruptions in neuronal migration with attendant disruptions in cortical lamination (74–77,191) might interfere with regional specialization and performance on certain cortically related tasks might be slowed (191). This suggestion is supported by findings from prenatallly cocaine-exposed monkeys followed through six years of age who were tested on their ability to learn a new set of responses following a change in the rules of reinforcement for a simple visual discrimination task (157,217). The task required inhibition of a previously learned and rewarding response. The prenatally cocaine-exposed animals performed much more poorly, that is, they took more sessions to attain or never attained pre-reversal type responding when compared to the drug-naive control animals. These data are also consistent with findings from human studies using event related potential methods in which prenatally cocaine-exposed 7 to 9 years old children used more diffuse regions of the cortex to complete a simple response inhibition task (190). Though the children were able to respond correctly in an untimed testing situation, their responses were slower and they required more time to complete the task as might be considered consonant with more diffuse activation of the cortex. Thus, although the findings in overall cognitive development in preschool and school aged-children are inconsistent, (43,218,219) the evidence increasingly documents somewhat greater cognitive impairments in CE children after controlling for other drug exposures. (197,220,221) While the effects on IQ may be subtle, (222) moderate to severe delays and impairments in both expressive and receptive language development are more consistently reported (175,176,223–230). Some of these include controls for environmental stimulation. Studies of developmental language differences among cocaineexposed children are also beginning to focus on more specific outcome measures such as differences in phonological patterns (229) that are less confounded by environmental deprivation.

NDS of Prenatal Cocaine Exposure

211

In terms of behavior and school performance, a few studies document a suggested increase in behavioral and academic problems among cocaine-exposed children, though these findings are confounded by reporting bias of often non-blinded observers (230). Importantly, several investigators have reported apparent cocaine-related effects on children’s development and behavior that are moderated or mediated (181) via pathways of maternal health or quality of caregiving (182,183). Conversely, others have examined infants exposed prenatally who were adopted into stable environments and have documented apparent direct effects of the prenatal exposure on neurocognitive outcomes (221,231). Also, several studies have differentiated between heavy and moderate to light exposure and found dose-response outcomes Several studies have also documented doseresponse effects on attention control, executive function, and general cognitive deficits (197,201,202,206,232) as well as an interactive effect with the level of environmental risk (212). Thus, despite a number of methodological concerns and still scant data for schoolaged children, the profile emerging for those children prenatally exposed to cocaine suggests a range of neurocognitive, neuropsychological dysfunction that may be a combination of effects, partially mediated through adverse environment, partially through the perinatal impact of the exposure, and partially through a direct effect on neural structure-function relationships.

THE INTERACTION OF PRENATAL COCAINE EXPOSURE AND A DRUG-DEPENDENT ENVIRONMENT Adults participating in substance abuse treatment programs, seeking individual treatment for their addiction, or involved in chronic substance abuse without ever seeking treatment come from a wide range of environmental conditions, each of which contribute to the adult’s ability to parent children. Surely the most studied and most reported on are the confluence of conditions relating to extreme poverty, homelessness, prostitution, and violence (184). Multiple studies from substance abuse treatment programs document the high incidence of unemployment and less than a high school education among participating substance-abusing women (233). Among this population, the rate of unemployment has been shown as high as 96% (234). Many report few to no friendships or contacts with supportive persons who are not also substance abusers, and substanceabusing adults often describe long-standing social detachment (235). The level of violence in substance-abusing families particularly between women and their spouses or male friends is markedly high and exposes children to a considerable amount of witnessed violence (236). Notably, there are few data about how often children in substance-abusing families are being reared by a single mother, although the quoted percentages usually exceed 70% (237), or how often and in what ways fathers are involved (238). The reluctance of many substance-abusing adults to reveal details about their households contributes in part to this lack of knowledge, but it also reflects in part the broader lack of adequate data about the family structure in substance-abusing households—how many adults usually care for a child, how many households may a child move among, how often are substance-abusing mothers and their children virtually homeless. Those coordinating ongoing treatment programs for substance-abusing adults and their children report that participating women have great difficulty understanding their infants’ and children’s communications as expressions of needs and not as demanding and inappropriate (239). Often deprived and neglected themselves, many substance-abusing mothers have unrealistic expectations of what infants or children can do for their mothers

212

Mayes and Pajulo

(240,241). The infant may be seen as a gift or as an extension of the mother’s own needs. Those women who deny that their drug use has any effect on their infant or child have been reported to be at much greater risk for impaired parenting manifest, for example, by neglect (242). On psychological testing, substance-abusing women often score high on externalizing traits, a finding reflecting a commonly reported belief that their lives and fates are controlled by forces and persons outside of themselves and their control (243,244). Feelings of worthlessness, poor self-esteem, anxiety, and depression are commonly reported (240,242,245) although these may also be exacerbated by chronic substance abuse. If there is heterogeneity among studies of the neurobehavioral outcome among prenatally cocaine exposed children, there is even greater variability in studies of the caregiving interactions between cocaine-using mothers and their children. To date, fewer than 20 studies (Table 2) have examined the interactions between cocaine-using mothers and their infants and young children. The findings are varied in part because of variations in sample size (from 5 to 364 mother–child pairs across the studies), use of a comparison or control group, inclusion of the average amount of maternal cocaine use in the analyses and taking into account postnatal changes in amount of use, place of the assessment and age of the child, and the interactive behaviors assessed. For example, amount and form of cocaine exposure and use of other drugs was carefully estimated and reported in only six studies (246–251). Across the studies, the youngest children at the time of assessment of interaction were newborns (1–2 days of age), and the oldest children were three years of age. Interaction quality was assessed during a feeding situation in six studies, in structured play or task situation in seven studies, and in free play situation in eight of the studies. Four studies have used a stress-inducing situation such as short separation-reunion or interruption of the interaction as an additional information basis. Duration of the interactive situation assessed varied from three minutes to approximately one hour. Most focused on mother and child separately and only also included dyadic behavior. With these caveats in mind, findings across the available studies point to a general disengagement, lack of pleasure in the interaction or attention to the infant, and poor attention to the infant’s cues. Cocaine-using mothers, whose children were six months of age or less, were found to spend more time passively looking at their infant but were more disengaged from the infant in terms of responding to the infant’s cues (252). They were lacking in social initiative and resourcefulness (253), showed less flexibility, engagement (254), and higher non-contingency in interaction (247), had shorter feeding episodes (254) and were found less attentive to interaction. Non-attentiveness and tendency to interrupt the interaction also increased towards the six month point (248). With young infants, mothers who relapsed back to cocaine use after birth of the child had more negative interaction behavior as a whole than those who remained drug free following their prenatal use (255). Infants less than six months of age of cocaine-using mothers showed longer periods drowsy, asleep, or distressed as newborns (252), showed less enjoyment during play, continued to show negative expressions and slow recovery after short interruption of interaction (201), higher stress to novelty (247), and less readiness to interact with the mother at six months compared to three months of age (248). In their dyadic interactions, there was a notable lack of enthusiasm and mutual enjoyment (253,256), fewer engagement in dyadic interactions (248), higher dyadic conflict (247), and less mutual arousal within the dyad (256). Cocaine-using mothers of children six months to three years of age continued to show less enjoyment and pleasure in interaction (256), less emotional engagement (250), were more intrusive and hostile, showed lower self confidence, and tended to give commands or instructions not appropriate for the child’s developmental age (249). Maternal interaction

NDS of Prenatal Cocaine Exposure Table 2

213

Studies of Interaction Between Cocaine-Using Mothers and Their Children

Study (Neuspiel, Hamel, Hochberg et al. 1991) (Burns, Chetnik, Burns et al. 1991) (Gottwald and Thurman, 1994) (Burns, Chetnik, Burns et al. 1997) (Hagan and Myers, 1997) (Mays, Feldman, Granger et al. 1997) (Blackwell, Kirkhart, Schmitt et al. 1998) (Bendersky and Lewis, 1998) (Swanson, Beckwith and Howard, 2000) (Eiden, 2001) (Ukeje, Bendersky and Lewis, 2001) (Espinosa, Beckwith, Howard et al. 2001) (Johnson, Morrow, Accornero et al. 2002) (Elden, Lewis, Croff et al. 2002) (Molltor, Mayes and Ward, 2003) (Beeghly, Frank, Rose-Jacobs et al. 2003) (LaGasse, Messenger, Lester et al. 2003)

Sample Control size group

Child’s age

Method

Positive findings

16

Yes

7–16 weeks

Feeding (NCAFS)

No

5

No

4 months

Play (P-CERA)

20

Yes

12–48 hr

Play, tasks

10

Yes

8–11 months

Play (P-CERA)

13

Yes

2–2.5 years

43

Yes

3 & 6 months

Play (T-CPS, P/CTS) Play

Yes (Mother, Dyad) Yes (Mother, Child) Yes (Mother, Dyad) No

21

Yes

4 & 6 months

41

Yes

4 months

51

No

18–23 months Play, tasks

Yes (Mother)

19

Ye

2 months

Feeding(MIFS)

49

Yes

12 months

35

No

6 & 7 months

157

Yes

3 years

Play, separationreunion Feeding, home obs. Play, tasks (TTCS)

Yes (Mother, Child, Dyad) Yes (Mother, Child) Yes (Mother)

19

Yes

2 & 7 months

26

Yes

18 months

90

Yes

5,11, & 12 months

364

Yes

7 months

Feeding (NCAFS) Play, still face

Feeding, play, tasks Play, separationreunion Play, tasks (NCAT), separationreunion Feeding

Yes (Mother, Child, Dyad) Yes (Mother) Yes (Child)

Yes (Mother, Child, Dyad) Yes (Child) Yes (Mother, Child) Yes (Child)

Yes (Mother)

was found to be most impaired with mothers continuing cocaine use during three-year postnatal follow-up than with mothers who were drug free (249). The children of these mothers more often ignored their mothers’ departures (251), cried less during separationreunion and showed more avoidance in reunion (246), showed either lower (250) or higher negative affect in response to stress (257), less emotional engagement in follow-up play after short interruption (250), and diminished ability to persist in task (249).

214

Mayes and Pajulo

A few studies have explored issues of attachment between cocaine-using mothers and their children. Sensitive maternal behavior in interaction has been considered the crucial component leading to secure attachment in the child. Insecure attachment is a potent risk factor for child’s later socio-affective and behavioral maladaptation (258,259). Disorganized attachment pattern is considered most worrisome, because it appears to be associated with higher stress, aggression, externalizing problem behavior and psychiatric symptomatology in later childhood (260–262). Findings relating maternal cocaine-use status and child attachment organization have varied. Three studies report insecure and especially disorganized attachment pattern as clearly more prevalent among cocaineexposed children compared to normative samples, measured at 15 to 18 month of child’s age (263–265). No difference was found in the distribution of child attachment status between biological, kinship and foster care mothers (263,265). In contrast, Beeghly and colleagues explored child attachment among 89 children with lighter or heavier cocaine exposure and without severe psychopathology among the substance-using mothers, compared to 61 non-exposed control children (246) and found the percentage of infants with insecure or disorganized attachment to be consistent with normative samples. Cocaineexposed children were considered no more likely than medically and demographically similar control children to show insecure or disorganized attachment patterns. Thus, while not entirely consistent, what seems clear from these few studies is that the combination of maternal depression, early abuse and neglect, unstable early and current attachments, and continued substance use come together to markedly impair an adult’s ability to care for their infant. In turn, the infant and young children may have a biologically conveyed vulnerability to becoming easily overaroused, behaviorally disorganized or withdrawn, more impulsive, and less attentive—a behavioral profile that would be challenging for the most competent of caregivers and surely stressful for a mother who is herself more fragile and easily disorganized. A fruitful perspective for considering the substance-using environment and its effect on a child is to consider three levels of risk as illustrated in Figure 1. A macroscopic level involves the factors most often discussed and measured from both the maternal and the infant perspective. These includes issues for the mother of the health impact of polydrug use, co-morbid psychiatric and physical health issues, genetic contributions to her vulnerability to substance use, poverty, homelessness, and poor social support. For the infant, at the macroscopic level are similar issues of medical status, genetic contributions, and potential neurobehavorial vulnerabilities from the prenatal substance exposure. These macroscopic variables come together at a more proximal or microscopic level to the behaviors described above in which maternal passivity, disengagement, less contingent responding, less flexible and adaptive parenting style come together with the infant’s easy irritability, less engaged, less attentive behavior to create a tense, non-mutual, often in conflict dyadic interaction. How these kinds of interactive behaviors at a microscopic level translate into enduring social-affective problems for a child growing up in a substance using environment requires thinking yet a third level of how mother and infant experience the interactions. As shown in the figure, substance using mothers by virtue of the factors at the other two levels often experience themselves as poor mothers not able to care for a baby and their infant as not interested in them. When the infant turns away in frustration or fatigue and often in an uncomforted, distressed state, the mother also feels as if she has been rejected and that the baby does not care about her. Their mutual disappointment/frustration is cyclical. Importantly, only a handful of investigators are now beginning to study interactions between substance-using mothers and their infants at this more experiential level. One set of data suggest that such a level is a potentially productive avenue to explore. This experiential level may be operationalized from the maternal point of view as

NDS of Prenatal Cocaine Exposure

215

MACROSCOPIC LEVEL: COMBINATIONS OF PAST AND CURRENT RISK FACTORS Maternal Risks

Infant Risks Cocaine-related effects on arousal and Cocaine-related effects on mood and attention regulatory capacities Mother: Poor resources attention Polydrug effects for Polydrug effects Genetic loading parenting Poor physical health Genetic loading Infant: Increased need for Irritability, lability, poor state regulation Psychiatric comorbidity care Poor physical health Homelessness Unemployment Poor education Poor social support MICROSCOPIC LEVEL: DISRUPTIONS OF BEHAVIORS DURING PARENT-CHILD INTERACTION Mother More passive, disengaged, noncontingent, interrupting Less resourceful, attentive, flexible

Mother and Infant: More dyadic conflict Less mutual pleasure, synchrony, and mutual affect regulation

Infant Poor sleep-wake regulation, more irritable and sensitive to change or novelty Less positive emotion expressed in the interaction and less able to recover from distress (e.g., poorer selfregulatory abilities)

EXPERIENTIAL LEVEL: UNDERSTANDING THE FEELINGS AND NEEDS OF INFANT AS SEPARATE FROM THOSE OF PARENT Mother Less able to think of child’s experience Experiences child’s distress or failure to respond as “rejecting”

Figure 1

Infant Mutual Frustration Mutual Disappointment Negative expectation for future interactive moments

Gets easily tired Inconsistent or unclear cues Needs help in state regulation Easily distressed and overwhelmed Withdraws from interaction when distressed

The substance-using environment and its impact on the parent-child relationship.

a capacity to understand oneself and others in terms of mental states (feelings, beliefs, intentions and desires) and to reason about one’s own and another’s behavior in relation to these (266). This ability, termed reflective functioning, is considered the capacity that makes it possible for a mother to behave sensitively with her child in interaction. In the study by Levy and colleagues (267), 58 cocaine-using mothers were assessed regarding their reflective functioning and compared to a non-cocaine group of mothers. The age of their children was 2–4 years. Cocaine mothers were found to have a significantly lower reflective capacity than the comparison mothers. Reflective capacities correlated positively with child’s social skills, negatively with child’s attention problems, tendency to withdraw, parental distress and dysfunctional parent–child interaction. In both the controls and the substance-using groups, as mothers increased in their abilities to reflect on their children’s needs and feelings, their children were more pro-social, responsive, able to modulate their emotional state, and the dyadic relationship was more congruent and less frustrated and stressful (267). Thus, while substance use may significantly impair a

216

Mayes and Pajulo

mother’s ability to reflect on her child’s needs apart from her own, for those substanceusing women who are able to do so as a result of intervention or on their own, this capacity may be a powerful buffer against the negative weight of many of the conditions at the macroscopic level of experience for substance-using mothers and their children. Continued systematic study of the following areas are crucial for advancing understanding about how substance abuse as a condition and how particular substances of abuse influence parenting: (1) How substance abuse affects the day-to-day activities of parents with their children that are known to mediate early social, affective, and cognitive development (e.g., parental use of language, encouraging exploration, attention directing, joint attention, social referencing); (2) The factors (e.g., social support, history of being abused as a child, chronic exposure to and witnessing of violence, children with special physical demands, sense of competence as a parent) that increase or decrease the risks for a substance-abusing parent and child for impairments in these day-to-day activities and in the more global measures of parenting disruption such as abuse and neglect; (3) The contribution of pre-morbid factors to substance abuse and addiction and to the individual’s ability to parent adequately (e.g., major depression, anxiety disorders, severe childhood trauma, substance abuse in parent’s parents); and (4) the protective factors that allow substanceabusing parents to care for their children adequately despite the demands or compromises of their addiction (e.g., involvement in substance abuse treatment program, supportive partner, sense of self-efficacy, and motivation to change) and especially those factors that allow a parent despite considerable adversity to experience her own parenting role as positive and to be able to think about her child’s needs and feelings apart from her own. SUMMARY While far more work is needed at the preclinical level, findings are accumulating to suggest a targeted and enduring effect of prenatal cocaine exposure on monoaminergic system ontogeny and function. Early gestation exposure to cocaine may affect the most basic processes of neuronal proliferation and migration while later exposure influences neuronal maturation and synaptogenesis. It is probable that there are differential effects on the three monoaminergic systems and that impairment in one system may be compensated for through up or down regulation of another. This interactive and compensatory mechanism may account in part for variations in findings depending on the age of the animal at the time of study. Combining prenatal and preweaning exposure models will be important to model the more common human condition of exposure throughout pregnancy. In addition, promising work suggests that with maturation some of the baseline behavioral effects of prenatal exposure may be less apparent but deficits continue to be manifest under stressful or novel conditions. The preclinical models are extremely useful for suggesting hypothesis and potentially fruitful avenues of study in the human. In summary, three lines of evidence suggest a plausible link between prenatal cocaine exposure and specific effects on the mechanisms subserving arousal and attention regulation in infants and preschool aged children. These are: 1. The association of prenatal cocaine exposure with alterations in monoaminergic system ontogeny with suggestive findings pointing to a functional decoupling of DA D1/D2 interaction, an increase in D2 and NE activity, and a dysregulation of serotonin-mediated neuronal connectivity and neural growth factors. 2. Neurobehavioral effects of prenatal cocaine exposure in animals including inhibited exploration and altered responses to stress suggesting overarousal in the face of novel/stressful situations and disrupted attention and learning.

NDS of Prenatal Cocaine Exposure

217

3. Neurobehavioral findings in infants and preschool aged children suggestive of disrupted arousal regulation in the face of stress or novelty, increased distractibility, impulsivity, and consequent impaired attention to novel, structured tasks. As outlined above, while crack–cocaine use has declined since the peak years in the late 1980s, there continues to be a steady prevalence of regular use among women of child bearing age. Thus, infants continue to be exposed, and those children born earlier continue to live in substance-using environments. As prenatally cocaine-exposed children reach school age, the cumulative effects of the postnatal substance-abusing environment may account for a large portion of the variance in the child’s developmental and psychological outcome. How early deficits in arousal and attention regulation may be expressed later in children’s cognitive development, learning, school performance, social functioning, capacity to metabolize stressful conditions, and psychiatric/psychological dysfunctions of attention (e.g., ADHD), anxiety, and/or conduct disorders is a long-term question that can only be addressed as the longitudinal cohorts accumulated across the country become adolescents and young adults. There is also the issue of compensatory processes after exposure has ended. While it is true that early neural events related to cocaine exposure do alter other ontogenetic events downstream long after exposure has ended, it is also true that for young children there is an enormously rapid, experience-dependent phase of neural development in the first three to four years. Synaptic formation and synaptic remodeling are occurring at extraordinarily rapid rates during these early years and the possibility for remodeling and compensation is surely present. Indeed, with sufficiently nurturing environments, it may well be that some neurocognitive effects either attenuate or appear as children near puberty (48,142,153,168–170). Regretfully, however, many if not the majority of infants exposed to cocaine prenatally grow up in less than adequate environments that may not be able to provide the sufficient experiential substrate to facilitate neural compensatory processes. Rather, it may be that the very vulnerability to dysfunctional arousal systems is further compromised by the inadequate to chaotic environments that these children grow up in and we have proposed one such model of continued biologic-environment vulnerability. Beginning early and comprehensive interventions at all levels of care for the prenatally exposed infant is critical to interrupting the interactive cycle illustrated in our proposed model. Regretfully, the tragic reality is there are usually markedly insufficient psychosocial services available for substance-using families and their children and the design of available services does not always facilitate access by a severely addicted, often homeless and depressed substance-using parent with a difficult to care for child. Mental health interventions informed by these models as well as by those of behavioral teratology and developmental psychopathology will be critical to breaking the continued cycle of substance use, poverty, and environmental chaos that so often tracks the lives of so many children from substance using homes (55,232). ACKNOWLEDGMENTS The work has been supported by NIDA grants ROI-DA-06025 (LCM), KO2-DA00222 (LCM), KO5-DA20091 (LCM), and RO1-DA017863 (LCM). The work has also been supported by The Finnish Medical Foundation, National Institute of Drug Abuse (NIDA) Invest Research Fellowship, International Psychoanalytical Association (IPA), and Academy of Finland. This work also draws extensively on other reviews completed by the first author including references 9, 20, and 27.

218

Mayes and Pajulo

REFERENCES 1. Wilson JG. Environment and Birth Defects. New York: Academic Press, 1973. 2. Wilson JG. Current status of teratology-general principles and mechanisms derived from animal studies. In: Wilson JG, Fraser FC, eds. In: Handbook of Teratology, Vol. 1. New York: Plenum Press, 1977:47–74. 3. Vorhees CV. Origins of behavioral teratology. In: Riley EP, Vorhees CV, eds. Handbook of Behavioral Teratology. New York: Plenum Press, 1986:3–22. 4. Butcher RE. A historical perspective on behavioral teratology. Neurobehav Toxicol Teratol 1985; 7:537–540. 5. O’Connor TG, Heron J, Glover V. Antenatal anxiety predicts child behavioral/emotional problems independently of postnatal depression. J Amer Acad Child Adoles Psychiatry 2002; 41:1470–1477. 6. O’Connor TG, Heron J, Golding J, Beveridge M, Glover V. Maternal antenatal anxiety and children’s behavioural/emotional problems at 4 years: report from the Avon Longitudinal Study of Parents and Children. Brit J Psychiatry 2002; 180:502–508. 7. Kopp CB. Risk factors in development. In: Mussen P, Haith M, eds. In: Handbook of Child Psychology: Infancy and Biological Bases, Vol. 2. New York: Wiley, 1983. 8. Vorhees CV, Mollnow E. Behavioral teratogenesis: long-term influences on behavior from early exposure to environmental agents. In: Osofsky J, ed. Handbook of Infancy, 1987:913–971. 9. Mayes LC, Ward A. Principles of neurobehavioral teratology. In: Cicchetti D, Walker E, eds. Neurodevelopmental Mechanisms in Psychopathology. New York: Cambridge University Press, 2003:3–33. 10. Vorhees CV. Principles of behavioral teratology. In: Riley EP, Vorhees CV, eds. Handbook of Behavioral Teratology. New York: Plenum Press, 1986:23–48. 11. Vorhees C. Concepts in teratology and developmental toxicology derived from animal research. Ann NY Acad Sci 1989; 562:31–41. 12. SAMHSA. Results from the 2002 National Survey on Drug Use and Health: National findings (DHHS Publication No. SMA 03-3836, NHSDA Series H-22). Rockville, MD, Substance Abuse and Mental Health Services Administration, 2003. 13. NIDA. National Pregnancy and Health Survey: Drug Use Among Women Delivering Live Births, 1992. Rockville, MD, U.S. Department of Health and Human Services, 1993. 14. Miller G. Addicted infants and their mothers. Zero to Three 1989; 9:20–23. 15. Viadero D. Study of drug-exposed infants finds problems in learning as late as age 3. Educ Week 1990; 10:15. 16. NIDA. Epidemiological Trends in Drug Abuse Advance Report. Rockville, MD, U.S. Department of Health and Human Services, 2002. 17. SAMHSA. The DASIS Report. Women in Treatment for Smoked Cocaine. Rockville, MD, U.S. Department of Health and Human Services, 2001. 18. Kosofsky BE, Wilkins AS. A mouse model of transplacental cocaine exposure: clinical implications for exposed infants and children. Ann NY Acad Sci 1998; 846:248–261. 19. Nestler EJ. Molecular basis of addiction states. Neuroscientist 1995; 1:212–220. 20. Mayes LC. Developing brain and in-utero cocaine exposure: effects on neural ontogeny. Dev Psychopath 1999; 11:685–714. 21. Woods JR, Plessinger MA. Maternal-fetal cardiovascular system: a target of cocaine. NIDA Res Monogr 1991; 108:7–27. 22. Woods JR, Plessinger MA, Clark KE. Effect of cocaine on uterine blood flow and fetal oxygenation. JAMA 1987; 257:957–961. 23. Neuspiel DR. The problem of confounding in research on prenatal cocaine effects on behavior and development. In: Lewis M, Bendersky M, eds. Mothers, Babies, and Cocaine: The Role of Toxins in Development. Hillsdale, NJ: Lawrence Erlbaum Associates, 1995:95–110. 24. Lester BM, Freier K, LaGasse L. Prenatal cocaine exposure and child outcome: what do we really know?. In: Lewis M, Bendersky M, eds. Mothers, Babies, and Cocaine: The Role of Toxins in Development. Hillsdale, NJ: Lawrence Erlbaum Associates, 1995:19–40.

NDS of Prenatal Cocaine Exposure

219

25. Brooks-Gunn J, McCarton C, Hawley T. Effects of in utero drug exposure on children’s development. Review and recommendations. Arch Pediatr Adolesc Med 1994; 148:33–39. 26. Frank DA, Augustyn M, Zuckerman B. Neonatal neurobehavioral and neuroanatomic correlates of prenatal cocaine exposure. Problems of dose and confounding. Ann NY Acad Sci 1998; 846:40–50. 27. Mayes LC, Fahy T. Prenatal drug exposure and cognitive development. In: Sternberg RJ, Grigorenko EL, eds. Environmental Effects on Cognitive Abilities. Mahwah, NJ: Lawrence Erlbaum Associates, 2001:189–220. 28. Frank DA, Augustyn M, Knight WG, Pell T, Zuckerman B. Growth, development, and behavior in early childhood following prenatal cocaine exposure: a systematic review. JAMA 2001; 285:1613–1625. 29. Tronick EZ, Beeghley M. Prenatal cocaine exposure, child development, and the compromising effects of cumulative risk. Clin Perinatol 1999; 26:151–171. 30. Chiriboga CA. Neurological correlates of fetal cocaine. Ann NY Acad Sci 1998; 846:109–125. 31. Olson HC, Toth SK. Samples in research on prenatal cocaine exposure: vexing problems and practical solutions. J Drug Issues 1999; 29:237–253. 32. Singer L. Advances and redirections in understanding effects of fetal drug exposure. J Drug Issues 1999; 29:253–262. 33. Mayes LC, Bornstein MH. Developmental dilemmas for cocaine-abusing parents and their children. In: Lewis M, Bendersky M, eds. Mothers, Babies, and Cocaine: The Role of Toxins in Development. Hillsdale, NJ: Lawrence Erlbaum Associates, 1995:251–272. 34. Frank DA, Bresnahan K, Zuckerman BS. Maternal cocaine use: impact on child health and development. Adv Pediatr 1993; 40:65–99. 35. Richardson GA, Day NL, McGauhey PJ. The impact of prenatal marijuana and cocaine use on the infant and child. Clin Obstet Gynecol 1993; 36:302–318. 36. Singer L, Farkas K, Kliegman R. Childhood medical and behavioral consequences of maternal cocaine use. J Pediatr Psychol 1992; 17:389–406. 37. Mayes LC, Grillon C, Granger R, Schottenfeld R. Regulation of arousal and attention in preschool children exposed to cocaine prenatally. Ann NY Acad Sci 1998; 846:126–143. 38. Heffelfinger A, Craft S, Shyken J. Visual attention in children with prenatal cocaine exposure. J Inter Neuropsychol Soc 1997; 3:237–245. 39. Eghbalieh B, Crinella FM, Hunt LE, Swanson JM. Prenatal stimulant exposure affects alerting and executive control in six- and seven-year-olds. J Attention Disorders 2000; 4:5–13. 40. Noland JS, Singer LT, Mehta SK, Super DM. Prenatal cocaine/polydrug exposure and infant performance on an executive functioning task. Dev Neuropsych 2003; 24:499–517. 41. Noland JS, Singer LT, Arendt RE, Minnes S, Short EJ, Bearer CF. Executive functioning in preschool-age children prenatally exposed to alcohol, cocaine, and marijuana. Alcohol Clin Exp Res 2003; 27:647–656. 42. Espy KA, Kaufmann PM, Glisky ML. Neuropsychologic function in toddler exposed to cocaine in utero: a preliminary study. Dev Neuropsych 1999; 15:447–460. 43. Richardson G, Conroy M, Day N. Prenatal cocaine exposure: effects on the development of school-age children. Neurotoxicol Teratol 1996; 18:627–634. 44. Bandstra ES, Morrow CE, Anthony JC, Accornero VH, Fried PA. Longitudinal investigation of task persistence and sustained attention in children with prenatal cocaine exposure. Neurotoxicol Teratol 2001; 23:545–559. 45. Dow-Edwards DL. Cocaine effects on fetal development: a comparison of clinical and animal research findings. Neurotoxicol Teratol 1991; 13:347–352. 46. Dow-Edwards D. The puzzle of cocaine’s effects following maternal use during pregnancy: still unsolved [comment]. Neurotoxicol Teratol 1993; 15:295–296 discussion 311–312. 47. Spear LP. Alterations in cognitive function following prenatal cocaine exposure: studies in an animal model. In: Lewis M, Bendersky M, eds. Mothers, Babies, and Cocaine: The Role of Toxins in Development. Hillsdale, NH: Lawrence Erlbaum Associates, 1995:207–228.

220

Mayes and Pajulo

48. Spear LP, Campbell J, Snyder K, Silveri M, Katovic N. Animal behavior models. Increased sensitivity to stressors and other environmental experiences after prenatal cocaine exposure. Ann NY Acad Sci 1998; 846:76–88. 49. Dow-Edwards DK, Yin R, Hurd YL. Alterations in dopamine transporter mRNA in adult rats following developmental exposure to cocaine. Soc Neurosci 1997. Abstr 432.4. 50. Friedman E, Yadin E, Wang H-Y. Effect of prenatal cocaine on dopamine-receptor-G protein coupling in mesocortical regions of the rabbit brain. Neuroscience 1996; 70:739–747. 51. Howard SG, Fisher R, Landry CF. Alterations in the spontaneous release of dopamine and the density of the DA D2 receptor mRNA after chronic postnatal exposure to cocaine. Brain Res Bull 1997; 43:101–106. 52. Minabe Y, Ashby CR, Heyser C, Spear LP, Wang RY. The effects of prenatal cocaine exposure on spontaneously active midbrain dopamine neurons in adult male offspring: an electrophysiological study. Brain Res 1992; 586:152–156. 53. Wang H-Y, Runyan S, Yadin E, Friedman E. Prenatal exposure to cocaine selectively reduces D1 dopamine receptor-mediated activation of striatal G proteins. J Pharm Exp Therap 1995; 273:492–498. 54. Weese-Mayer DE, Silvestri JM, Lin D, Buhrfiend CM, Lo ES, Carvey PM. Effect of cocaine in early gestation on striatal dopamine and neurotrophic activity. Pediatr Res 1993; 34:389–392. 55. Mayes LC. A developmental perspective on the regulation of arousal states. Sem Perinatol 2000; 24:267–279. 56. Lakoski JM, Cunningham KA. The interaction of cocaine with central serotonergic neuronal systems: cellular electrophysiologic approaches. NIDA Res Monog 1988; 88:78–91. 57. Kosten T. Neurobiology of abused drugs. J Nerv Ment Dis 1990; 178:217–227. 58. Koob G. Hedonic valence, dopamine, and motivation. Mol Psychiatry 1996; 1:186–189. 59. Creese I, Iverson SD. A role of forebrain dopamine systems in amphetamine-induced and stereotyped behavior in the rat. Psychopharmacology 1974; 39:345–347. 60. Graybeil AM. Neurotransmitters and neuromodulators in the basal ganglia. Trends Neurosci 1990; 13:244–254. 61. LeMoal M, Simon H. Mesocorticolimbic dopaminergic network: functional and regulatory roles. Physiol Rev 1991; 71:155–234. 62. Goeders NE, Smith JE. Cortical dopaminergic involvement in cocaine reinforcement. Science 1983; 221:773–775. 63. White FJ. Synaptic regulation of mesocorticolimbic dopamine neurons. Ann Rev Neurosci 1996; 19:405–436. 64. Wise R. Neural mechanisms of the reinforcing action of cocaine. NIDA Res Monog 1984; 50:15–33. 65. Corbett D, Wise RA. Intracranial self-stimulation in relation to the ascending dopaminergic systems of the midbrain: a moveable electrode mapping study. Brain Res 1980; 185:1–15. 66. Civelli O, Bunzow JR, Grandy DK. Molecular diversity of the dopamine receptors. Ann Rev Pharmacol Toxicol 1993; 33:281–307. 67. Grace AA, Gerfen CR, Aston-Jones G. Catecholamines in the central nervous system. Adv Pharmacol 1998; 42:655–670. 68. Gawin FH, Ellinwood EH. Cocaine and other stimulants: actions, abuse, and treatment. N Engl J Med 1988; 318:1173–1182. 69. Garg UC, Turndorf H, Bansinath M. Effect of cocaine on macromolecular syntheses and cell proliferation in cultured glial cells. Neurosci 1993; 57:467–472. 70. Yablonsky-Alter E, Gleser I, Carter C, Juvan M. Effects of prenatal cocaine treatment on postnatal development of neocortex in white mice: immunocytochemistry of calbindin- and paralbumin-positive populations of gabaergic neurons. Soc Neurosci Abstr 1992; 18:367. 71. Akbari HM, Whitaker-Azmitia PM, Azmitia EC. Prenatal cocaine decreases the trophic factor S-100 beta and induced microcephaly: reversal by postnatal 5-HT1A receptor agonist. Neurosci Lett 1994; 170:141–144 [published erratum appears in Neurosci Lett., 1994;175:176].

NDS of Prenatal Cocaine Exposure

221

72. Gressens P, Gofflot F, Van Maele-Fabry G, et al. Early neurogenesis and teratogenesis in whole mouse embryo cultures. Histochemical, immunocytological, and ultrastructural study of the premigratory neuronal-glial units in mouse embryo influenced by cocaine and retinoic acid. J Neuropathol Exp Neurol 1992; 51:206–219. 73. Gressens P, Kosofsky BE, Evrard P. Cocaine-induced disturbances of corticogenesis in the developing murine brain. Neurosci Lett 1992; 140:113–116. 74. Lidow M. Prenatal cocaine exposure adversely affects development of the primate cerebral cortex. Synapse 1995; 21:332–341. 75. Lidow MS, Rakic P. Neurotransmitter receptors in the proliferative zones of the developing primate occipital lobe. J Comp Neurol 1995; 360:393–402. 76. Lidow MS, Song Z-M. Effect of cocaine on cell proliferation in the cerebral wall of monkey fetuses. Cerebral Cortex 2001; 11:545–551. 77. He N, Bai J, Champoux M, Suomi SJ, Lidow MS. Neurobehavioral deficits in neonatal rhesus monkeys exposed to cocaine in utero. Neurotoxicol Teratol 2004; 26:13–21. 78. Lidow M, Rakic P. Unique deployment of adrenergic receptor subtypes during development of the primate cerebral cortex including transient subplate and proliferative zones. J Neurosci 1994; 14:4064–4078. 79. Seidler FJ, Temple SW, McCook EC, Slotkin TA. Cocaine inhibits central noradrenergic and dopaminergic activity during the critical developmental period in which catecholamines influence cell development. Brain Res Dev Brain Res 1995; 85:48–53. 80. Nassogne MC, Evrard P, Courtoy P. Selective direct toxicity of cocaine on fetal mouse neurons. Ann NY Acad Sci 1998; 846:51–68. 81. Levitt P, Harvey JA, Friedman E, Simansky K, Murphy EH. New evidence for neurotransmitter influences on brain development. Trends Neurosci 1997; 20:269–274. 82. Jones LB, Stanwood GD, Reinoso BS, et al. In utero cocaine-induced dysfunction of dopamine D1 receptor signaling and abnormal differentiation of cerebral cortical neurons. J Neurosci 2000;4606–4614. 83. Stanwood GD, Washington RA, Levitt P. Identification of a sensitive period of prenatal cocaine exposure that alters the development of the anterior cingulate cortex. Cereb Cortex 2001; 11:430–440. 84. Seamans JK, Floresco SB, Phillips AG. D1 receptor modulation of hippocampal-prefrontal cortical circuits integrating spatial memory with executive functions in the rat. J Neurosci 1998; 18:1613–1621. 85. Dow-Edwards DL, Freed Malen LA, Hughes HE. Long-term alterations in brain function following cocaine administration during the preweanling period. Dev Brain Res 1993; 72:309–313. 86. Ronnekleiv OK, Naylor BR. Chronic cocaine exposure in fetal rhesus monkey: consequences for early development of dopamine neurons. J Neurosci 1995; 15:7330–7343. 87. Ronnekleiv OK, Fang Y, Choi WS, Chai L. Changes in the midbrain-rostral forebrain dopamine circuitry in the cocaine-exposed primate fetal brain. Ann NY Acad Sci 1998; 846:165–181. 88. Scalzo FM, Ali SF, Frambes NA, Spear LP. Weanling rats exposed prenatally to cocaine exhibit an increase in striatal D2 dopamine binding associated with an increase in ligand affinity. Pharmacol Biochem Behav 1990; 37:371–373. 89. Dow Edwards DL, Freed Malen LA, Hughes HE. Long-term alterations in brain function following cocaine administration during the preweanling period. Dev Brain Res 1993; 72:309–313. 90. Henderson MG, McConnaughey MM, McMillen BA. Long-term consequences of prenatal exposure to cocaine or related drugs: effects on rat brain monoaminergic receptors. Brain Res Bull 1991; 26:941–945. 91. Leslie CA, Robertson MW, Jung AB, Liebermann J, Bennett JP. Effects of prenatal cocaine exposure upon postnatal development of neostriatal dopaminergic function. Synapse 1994; 17:210–215.

222

Mayes and Pajulo

92. Wang HY, Yeung JM, Friedman E. Prenatal cocaine exposure selectively reduces mesocortical dopamine release. J Pharmacol Exp Ther 1995; 273:1211–1215. 93. Bonci A, Williams JT. A common mechanism mediates long-term changes in synaptic transmission after chronic cocaine and morphine. Neuron 1996; 16:631–639. 94. Dow Edwards DL. Long-term neurochemical and neurobehavioral consequences of cocaine use during pregnancy. Ann NY Acad Sci 1989; 562:280–289. 95. Dow-Edwards DL. Developmental effects of cocaine. NIDA Res Monogr 1988; 88:290–303. 96. Zhen X, Torres C, Wang H-Y, Friedman E. Prenatal exposure to cocaine disrupts D1A dopamine receptor via selective inhibition of protein phosphatase 1 pathway in rabbit frontal cortex. J Neurosci 2001; 21:9160–9167. 97. Arnsten AFT. The biology of being frazzled. Science 1998; 280:1711–1712. 98. Arnsten AFT. Catecholamine regulation of the prefrontal cortex. J Psychopharmacol 1997; 11:151–162. 99. Arnsten AFT. Catecholamine modulation of prefrontal cognitive function. Trends Cog Sci 1998; 2:436–447. 100. Arnsten AFT, Jentsch JD. The alpha-1 adrenergic agonist, cirazoline, impairs spatial working memory performance in aged monkeys. Pharm Biochem Behav 1997; 58:55–59. 101. Moody CA, Frambes NA, Spear LP. Psychopharmacological responsiveness to the dopamine agonist quinpirole in normal weanlings and in weanling offspring exposed gestationally to cocaine. Psychopharmacology 1992; 109:256–262. 102. Stadlin A, Choi HL, Tsim KW, Tsang D. Prenatal cocaine exposure revealed minimal postnatal changes in rat striatal dopamine D2 receptor sites and mRNA levels in the offspring. Mol Neurobiol 1995; 11:67–76. 103. Dow-Edwards DL, Shi X, Yin R, Frick G. DA receptors in adults rats exposed to preweaning cocaine or GBR: Effects of chronic haloperidol. CPDD Abstr 1998;81. 104. Shearman LP, Collins LM, Meyer JS. Characterization and localization of I25-I-RTI-55labeled cocaine binding sites in fetal and adult rat brain. JPET 1996; 277:1770–1783. 105. DeGeorge G, Hochberg S, Murphy R, Azmitia E. Serotonin neurons may be a major target for the action of cocaine: evidence from in vitro and tissue culture studies. Soc Neurosci Abstr 1989; 15:418. 106. Whitaker-Azmitia PM. Role of the neurotrophic properties of serotonin in the delay of brain maturation induced by cocaine. Ann NY Acad Sci 1998; 846:158–164. 107. Akbari HM, Kramer HK, Whitaker-Azmitia PM, Spear LP, Azmitia EC. Prenatal cocaine exposure disrupts the development of the serotonergic system. Brain Res 1992; 572:57–63. 108. Snyder-Keller AM, Keller RW. Prenatal cocaine increases striatal serotonin innervation without altering the patch/matrix organization of intrinsic cell types. Dev Brain Res 1993; 74:261–267. 109. Cheng L, Hamaguchi K, Ogawa M, Hamada S, Okado N. PCPA reduces both monoaminergic afferents and nonmonoaminergic synapses in the cerebral cortex. Neurosci Res 1993; 19:111–116. 110. Yan W, Wilson CC, Haring JH. Effects of neonatal serotonin depletion on the development of rate dentate granule cells. Dev Brain Res 1997; 98:177–184. 111. Mazer C, Muneyyirci J, Taheny K, Raio N, Borella A, Whitaker-Azmitia PM. Serotonin depletion during synaptogenesis leads to decreased synaptic density and learning deficits in the adult rat. A model of neurodevelopmental disorders with cognitive deficits. Brain Res 1997; 760:68–74. 112. Shemer AV, Azmitia EC, Whitaker-Azmitia PM. Dose-related effects of prenatal 5-methoxytryptamine (5-MT) on development of serotonin terminal density and behavior. Brain Res 1991; 59:59–63. 113. Chubakov AR, Tsyganova VG, Sarkisova EF. The stimulating influence of the raphe nuclei on the morphofunctional development of the hippocampus during their combined cultivation. Neurosci Behav Physiol 1993; 23:271–276. 114. Lauder JM, Liu J. Glial heterogeneity and developing neurotransmitter systems. Perspect Dev Neurobiol 1994; 2:239–250.

NDS of Prenatal Cocaine Exposure

223

115. Whitaker-Azmitia PM, Azmitia EC. Astroglial 5-HT1a receptors and S-100 beta in development and plasticity. Perspect Dev Neurobiol 1994; 2:233–238. 116. Hellendall RP, Schambra U, Liu J, Breese GR, Millhorn DE, Lauder JM. Detection of serotonin receptor transcripts in the developing nervous system. J Chem Neuroanat 1992; 5:299–310. 117. Whitaker-Azmitia PM, Murphy R, Azmitia EC. S-100 protein is released from astroglial cells by stimulation of 5-HT1a receptors and regulates development of serotonin neurons. Brain Res 1990; 528:155–160. 118. Muller CM, Akhavan AC, Bette M. Possible role of S-100 in glia-neuronal signalling involved in activity dependent plasticity in the developing mammalian cortex. J Chem Neuroanat 1993; 6:215–220. 119. Clarke C, Clarke K, Muneyyirci J, Azmitia E, Whitaker-Azmitia PM. Prenatal cocaine delays astroglial maturation: immunodensitometry shows increased markers of immaturity (vimentin and GAP-43) and decreased proliferation and production of growth factor S-100. Dev Brain Res 1996; 91:268–273. 120. Borella A, Bindra M, Whitaker-Azmitia PM. Role of the 5-HTIA receptor in development of the neonatal rat brain: preliminary behavioral studies. Neuropharmacology 1997; 36:445–450. 121. Dow-Edwards DL. Preweaning cocaine administration alters the adult response to quipazine: comparison with fluoxetine. Neurotoxicol Teratol 1998; 20:133–142. 122. Shi X, Goeders N, Irby B, Dow-Edwards D. Gender-specific effects of inhibition of the serotonin or dopamine transporters on 8-OH-DPAT binding in developing rat brain. Soc Neurosci Abstr 1998;538. 123. Pennington NJ, Kelly JS, Fox AP. A study of the mechanism of Ca2C current inhibition produced by serotonin in rat dorsal raphe neurons. J Neurosci 1991; 11:3594–3609. 124. Van de Kar LD, Richardson-Morton KD, Rittenhouse PA. Stress: neuroendocrine and pharmacological mechanisms. In: Cantin JG, ed. In: Stress Revisited. 1. Neuroendocrinology of Stress: Methods and Achievements in Experimental Pathology, Vol. 14. Basel: Karger, 1991. 125. Battaglia G, Cabrera TM. Potentiation of 5-HT1A receptor-mediated neuroendocrine responses in male but not female rat progeny after prenatal cocaine: evidence for gender differences. J Pharmacol Exp Ther 1994; 271:1453–1461. 126. Battaglia G, Cabrera-Vera TM, Van De Kar LD, Garcia F, Vicentic A, Pinto W. Prenatal cocaine exposure produces long-term impairments in brain serotonin function in rat offspring. Ann NY Acad Sci 1998; 846:355–357. 127. Cabrera TM, Yracheta JM, Li Q, Levy AD, Van de Kar LD, Battaglia G. Prenatal cocaine produces deficits in serotonin mediated neuroendocrine responses in adult rat progeny: evidence for long-term functional alterations in brain serotonin pathways. Synapse 1993; 15:158–168. 128. Seidler FJ, Slotkin TA. Fetal cocaine exposure causes persistent noradrenergic hyperactivity in rat brain regions: effects on neurotransmitter turnover and receptors. J Pharmacol Exp Ther 1992; 263:413–421. 129. Akbari H, Azmitia E. Increased tyrosine hydroxylase immunoreactivity in the rat cortex following prenatal cocaine exposure. Dev Brain Res 1992; 66:277–281. 130. Barron S, Staton CM, Willford JA, Hansen-Trench LS. The effects of clonidine on rat pups neonatally exposed to cocaine. Neurotoxicol Teratol 1998; 20:401–410. 131. Delagrange P, Canu MH, Rougeul A, Buser P, Bouyer JJ. Effects of locus coeruleus lesions on vigilance and attentive behavior in cat. Behav Brain Res 1993; 53:155–165. 132. Arnsten A, Goldman-Rakic P. Selective prefrontal cortical projections to the region of the locus coeruleus and raphe nuclei in the rhesus monkey. Brain Res 1984; 306:9–18. 133. Rajkowsi J, Kubiak P, Ivanova S, Aston-Jones G. State-related activity and reactivity of the locus coeruleus neurons in behaving monkeys. Adv Pharmacol 1998; 42:740–744. 134. Gabriel M, Taylor C. Prenatal exposure to cocaine impairs neuronal coding of attention and discrimination learning. Ann NY Acad Sci 1998; 846:194–212.

224

Mayes and Pajulo

135. Romano AG, Harvey JA. Prenatal cocaine exposure: long term deficits in learning and motor performance. Ann NY Acad Sci 1998; 846:89–108. 136. Mayes LC, Bornstein MH, Chawarska K, Haynes OM, Granger RH. Impaired regulation of arousal in three-month-old infants exposed prenatally to cocaine and other drugs. Dev Psychopathol 1996; 8:29–42. 137. Henderson MG, McMillen BA. Effects of prenatal exposure to cocaine or related drugs on rat development and neurological indices. Brain Res Bull 1990; 24:207–212. 138. Peris J, Coleman HM, Millard WJ. Cocaine in-utero enhances the behavioral response to cocaine in adult rats. Pharmacol Biochem Behav 1992; 42:509–515. 139. Riley E, Foss J. The acquisition of passive avoidance, active avoidance, and spatial navigation tasks by animals prenatally exposed to cocaine. Neurotoxicol Teratol 1991; 13:559–564. 140. Riley EP, Foss JA. Exploratory behavior and locomotor activity: a failure to find effects in animals prenatally exposed to cocaine. Neurotoxicol Teratol 1991; 13:553–558. 141. Johns JM, Means MJ, Anderson DR, Means LW, McMillen BA. Prenatal exposure to cocaine II: effects on open-field activity and cognitive behavior in Sprague-Dawley rats. Neurotoxicol Teratol 1992; 14:343–349. 142. Spear LP, Kirstein CL, Bell J, et al. Effects of prenatal cocaine exposure on behavior during the early postnatal period. Neurotoxicol Teratol 1989; 11:57–63. 143. Hughes HE, Pringle GF, Scribani LA, et al. Cocaine treatment in neonatal rats affects the adult behavioral response to amphetamine. Neurotoxicol Teratol 1991; 13:335–339. 144. Cabib S, Castellano C, Cestari V, Filibeck U, Puglisi-Allegra S. D1 and D2 receptor antagonists differently affect cocaine-induced locomotor hyperactivity in the mouse. Psychopharmacology 1991; 105:335–339. 145. Drago J, Gerfen CR, Westphal H, Steiner H. D1 dopamine receptor-deficient mouse: cocaineinduced regulation of immediate-early gene and substance P expression in the striatum. Neuroscience 1996; 74:813–823. 146. Xu M, Xiu-Ti J, Cooper DC, et al. Elimination of cocaine-induced hyperactivity and dopamine-mediated neurophysiological effects in dopamine D1 receptor mutant mice. Cell 1994; 79:945–955. 147. Hooks MS, Juncos JL, Justice JB, et al. Individual locomotor response to novelty predicts selective alterations in D1 and D2 receptors and mRNA. J Neurosci 1994; 14:6144–6152. 148. Gilde AB. Prenatal cocaine exposure affects motor activity following acute quinpirole injection. Diss Abs Int B Sci Eng [0419-4217] 1994; 55:1702. 149. Heyser C, Chen W, Miller J, Spear N, Spear L. Prenatal cocaine exposure induces deficits in Pavlovian conditioning and sensory preconditioning among infant rat pups. Behav Neurosci 1990; 104:955–963. 150. Brunzell DH, Ayres JJB, Meyer JS. Effects of prenatal cocaine exposure on latent inhibition in 1-year-old female rats. Pharmacol Biochem Behav 2002; 72:795–802. 151. Brunzell DH, Coy AE, Ayres JJB, Meyer JS. Prenatal cocaine effects on fear conditioning: exaggeration of sex-dependent context extinction. Neurotoxicol Teratol 2002; 24:161–172. 152. Morrow BA, Elsworth JD, Roth RH. Prenatal cocaine exposure disrupts non-spatial, shortterm memory in adolescent and adult male rats. Behav Brain Res 2002; 129:217–223. 153. Smith RF, Mattran KM, Kurkjian MF, Kurtz SL. Alterations in offspring behavior induced by chronic prenatal cocaine dosing. Neurotoxicol Teratol 1989; 11:35–38. 154. Vorhees CV, Reed TM, Acuff-Smith KD, et al. Long-term learning deficits and changes in unlearned behaviors following in utero exposure to multiple daily doses of cocaine during different exposure periods and maternal plasma cocaine concentrations. Neurotoxicol Teratol 1995; 17:253–264. 155. Garavan H, Morgan RE, Mactutus CF, Levitsky DA, Booze RM. Prenatal cocaine exposure impairs selective attention: evidence from serial reversal and extradimensional shift tasks. Behav Neurosci 2000; 114:725–738. 156. Heyser CJ, McKinzie DL, Athalie F, Spear NE, Spear LP. Effects of prenatal exposure to cocaine on heart rate and nonassociative learning and retention in infant rats. Teratology 1994; 49:470–478.

NDS of Prenatal Cocaine Exposure

225

157. Chelonis JJ, Gillam MP, Paule MG. The effects of prenatal cocaine exposure on reversal learning using a simple visual discrimination task in rhesus monkeys. Neurotoxicol Teratol 2003; 25:437–446. 158. Gendle MH, White TL, Strawderman M, et al. Enduring effects of prenatal cocaine exposure on selective attention and reactivity to errors: evidence from an animal model. Behav Neurosci 2004; 118:290–297. 159. Morgan RE, Garavan HP, Mactutus CF. Enduring effects of prenatal cocaine exposure on attention and reaction to errors. Behav Neurosci 2002; 116:624–633. 160. Gabriel M, Taylor C, Burhans L. In utero cocaine, discriminative avoidance learning with low-salient stimuli and learning-related neuronal activity in rabbits (Oryctolagus cuniculus). Behav Neurosci 2003; 117:912–926. 161. Heyser C, Spear N, Spear L. The effects of prenatal exposure to cocaine on conditional discrimination learning in adult rats. Behav Neurosci 1992; 106:841–849. 162. Reinoso BS, Undie AS, Levitt P. Dopamine receptors mediate differential morphological effects on cerebral cortical neurons in vitro. J Neurosci Res 1996; 43:439–453. 163. Vogt BA, Gabriel M, eds. Neurobiology of Cingulate Cortex and Limbic Thalamus: A Comprehensive Handbook. Boston: Birkhausser, 1993. 164. Spear LP, Campbell J, Snyder K, Silveri M, Katovic N. Animal behavioral models. Increased sensitivity to stressors and other environmental experiences after prenatal cocaine exposure. Ann NY Acad Sci 1998; 846:76–88. 165. Dunn AJ, Kramarcy NR, eds. Neurochemical Responses in Stress: Relationship Between the Hypothalamic–Pituitary–Adrenal and Catecholamine Systems. In: Handbook of Psychopharmacology. New York: Plenum Publishing Corporation, 1984. 166. Hutchings DE, Fico TA, Dow-Edwards DL. Prenatal cocaine: maternal toxicity, fetal effects and locomotor activity in rat offspring. Neurotoxicol Teratol 1989; 11:65–69. 167. Spear L, Frambes N, Kirstein C. Fetal and maternal brain and plasma levels of cocaine and benzoylecgonine following chronic SC administration of cocaine during gestation in rats. Psychopharmacology 1989; 97:427–431. 168. Bilitzke PJ, Church MW. Prenatal cocaine and alcohol exposures affect rat behavior in a stress test (the Porsolt swim test). Neurotoxicol Teratol 1992; 14:359–364. 169. McMillen B, Johns J, Bass E, Means L. Learning and behavior of rats exposed to cocaine throughout gestation. Teratology 1991; 43:495. 170. Molina VA, Wagner JM, Spear LP. The behavioral response to stress is altered in adult rats exposed prenatally to cocaine. Physiol Behav 1994; 55:941–945. 171. Goodwin GA, Bliven T, Kuhn C, Francis R, Spear LP. Immediate early gene expression to examine neuronal activity following acute and chronic stressors in rat pups: examination of the neurophysiological alterations underlying behavioral consequences of prenatal cocaine exposure. Physiol Behav 1997; 61:895–902. 172. Wood RD, Molina VA, Wagner JM, Spear LP. Play behavior and stress responsivity in periadolescent offspring exposed prenatally to cocaine. Pharmacol Biochem Behav 1995; 52:367–374. 173. Wood RD, Spear LP. Prenatal cocaine alters social competition of infant, adolescent, and adult rats. Behav Neurosci 1998; 112:419–431. 174. Meaney MJ, Aitken DH, Viau V, Sharma S, Sarrieau S. Neonatal handling alters adrenocortical negative feedback sensitivity and hippocampal type II glucocorticoid receptor binding in the rat. Neuroendocrinology 1989; 50:597–604. 175. Bland-Stewart LM, Seymour HN, Beeghley M, Frank DA. Semantic development of African– American children prenatally exposed to cocaine. Sem Speech Lang 1998; 19:167–187. 176. Malakoff ME, Mayes LC, Schottenfeld R, Howell S. Language production in 24-month-old inner-city children of cocaine-and-other-drug-using mothers. J Appl Dev Psychol 1999; 20:159–180. 177. Lutiger B, Graham K, Einarson TR, Koren G. Relationship between gestational cocaine use and pregnancy outcome: a meta-analysis. Teratology 1991; 44:405–414. 178. Vogel G. Cocaine wrecks subtle damage on developing brains. Science 1997; 278:38–39.

226

Mayes and Pajulo

179. Held JR, Riggs ML, Dorman C. The effect of prenatal cocaine exposure on neurobehavioral outcome: a meta-analysis. Neurotoxicol Teratol 1999; 21:619–625. 180. LaGasse LL, Seifer R, Lester BM. Interpreting research on prenatal substance exposure in the context of multiple confounding factors. Clin Perinatol 1999; 26:39–54. 181. Baron R, Kenny D. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986; 51:1173–1182. 182. Black M, Schuler M, Nair P. Prenatal drug exposure: neurodevelopmental outcome and parenting environment. J Pediatr Psychol 1993; 18:605–620. 183. Singer L, Arendt R, Farkas K, Minnes S, Huang J, Yamashita T. Relationship of prenatal cocaine exposure and maternal postpartum psychological distress to child developmental outcome. Dev Psychopath 1997; 9:473–489. 184. Mayes LC. Substance abuse and parenting. In: Bornstein M, ed. The Handbook of Parenting. Hillsdale, NJ: Erlbaum, 1995:101–125. 185. Davidson Ward S, Schuetz S, Wachsman L, et al. Elevated plasma norepinephrine levels in infants of substance-abusing mothers. Am J Dis Child 1991; 145:44–48. 186. Mirochnick M, Meyer J, Cole J, Herren T, Zuckerman B. Circulating catecholamine concentrations in cocaine-exposed neonates: a pilot study. Pediatrics 1991; 88:481–485. 187. Needlman R, Zuckerman B, Anderson GM, Mirochnick M, Cohen DJ. Cerebrospinal fluid monoamine precursors and metabolites in human neonates following in utero cocaine exposure: a preliminary study. Pediatrics 1993; 92:55–60. 188. Mangano C, Gardner J, Karmel B. Differences in salivary cortisol levels in cocaine-exposed and noncocaine-exposed NICU infants. Dev Psychobiol 1992; 25:93–103. 189. Karmel BZ, Gardner JM, Magnano CL. Neurofunctional consequences of in utero cocaine exposure. NIDA Res Monogr 1991; 105:535–536. 190. Mayes LC, Molfese DL, Key AF. Event-related potentials in cocaine-exposed children during a Stroop Task. Neurotoxicology and Teratology 2005; 27:797–813. 191. Lidow MS. Consequences of prenatal cocaine exposure in nonhuman primates. Dev Brain Res 2003; 147:23–36. 192. Mayes LC, Bornstein MH, Chawarska K, Granger RH. Information processing and developmental assessments in 3-month-old infants exposed prenatally to cocaine. Pediatrics 1995; 95:539–545. 193. DiPietro JA, Suess PE, Wheeler JS, Smouse PH, Newlin DB. Reactivity and regulation in cocaine-exposed neonates. Infant Behav Develop 1995; 18:407–414. 194. Brown JV, Bakeman R, Coles CD, Sexson WR, Demi AS. Maternal drug use during pregnancy: are preterms and full-terms affected differently? Dev Psych 1998; 34:540–554. 195. Gingras JL, Feibel J, Dalley B, Muelenaer A. Maternal polydrug use including cocaine and postnatal infant sleep architecture: preliminary observations and implications for respiratory control and behavior. Early Hum Develop 1995; 43:197–204. 196. Karmel BZ, Gardner JM. Prenatal cocaine exposure effects on arousal-modulated attention during the neonatal period. Dev Psychobiol 1996; 29:463–480. 197. Alessandri SM, Bendersky M, Lewis M. Cognitive functioning in 8- to 18- month old drugexposed infants. Dev Psych 1998; 34:565–573. 198. Regaldo MG, Schechtman VL, Del Angel AP, Bean XD. Cardiac and respiratory patterns during sleep in cocaine-exposed neonates. Early Hum Develop 1996; 44:187–200. 199. Regaldo M, Schechtman V, Del Angel A, Bean X. Sleep disorganization in cocaine-exposed neonates. Infant Behav Develop 1995; 18:319–327. 200. Scher MS, Richardson GA, Day NL. Effects of prenatal cocaine/crack and other drug exposure on electroencephalographic sleep studies at birth and one year. Pediatrics 2000; 105:39–48. 201. Bendersky M, Lewis M. Arousal modulation in cocaine-exposed infants. Dev Psych 1998; 34:555–564. 202. Jacobson SW, Bihun JT, Chiodo LM. Effects of prenatal alcohol and cocaine exposure on infant cortisol levels. Dev Psychopath 1999; 11:195–208.

NDS of Prenatal Cocaine Exposure

227

203. Alessandri S, Sullivan M, Imaizumi S, Lewis M. Learning and emotional responsivity in cocaine-exposed infants. Dev Psych 1993; 29:989–997. 204. Azuma SD, Chasnoff IJ. Outcome of children prenatally exposed to cocaine and other drugs: a path analysis of three-year data. Pediatrics 1993; 92:396–402. 205. Griffith DR, Azuma SD, Chasnoff IJ. Three-year outcome of children exposed prenatally to drugs. Special Section: cocaine babies. J Am Acad Child Adolesc Psychiatry 1994; 33:20–27. 206. Jacobson SW, Jacobson JL, Sokol RJ, Martier SS, Chiodo LM. New evidence for neurobehavioral effects of in utero cocaine exposure. J Pediatr 1996; 129:581–590. 207. Struthers JM, Hansen RL. Visual recognition memory in drug-exposed infants. J Dev Behav Pediatr 1992; 13:108–111. 208. Hansen RL, Struthers JM, Gospe SM. Visual evoked potentials and visual processing in stimulant drug-exposed infants. Dev Med Child Neurol 1993; 35:798–805. 209. Metosky P, Vondra J. Prenatal drug exposure and play and coping in toddlers: a comparative study. Infant Behav Dev 1995; 18:15–25. 210. Coles CD, Bard KA, Platzman KA, Lynch ME. Attentional responses at eight weeks in prenatally drug-exposed and preterm infants. Neurotoxicol Teratol 1999; 21:527–537. 211. Singer L, Arendt R, Fagan J, et al. Neonatal visual information processing in cocaine-exposed and non-exposed infants. Infant Behav Dev 1999; 22:1–15. 212. Bendersky M, Gambini G, Lastella A, Bennett DS, Lewis M. Inhibitory motor control at five years as a function of prenatal cocaine exposure. J Dev Behav Pediatr 2003; 24:345–351. 213. Bendersky M, Gambini G, Lastella A, Bennett DS, Lewis MI. Inhibitory motor control at five years as a function of prenatal cocaine exposure. J Dev Behav Pediatr 2004; 24:345–351. 214. Savage J, Brodsky NL, Malmud E, Giannetta JM, Hurt H. Attentional functioning and impulse control in cocaine-exposed and control children at age ten years. J Dev Behav Pediatr 2005; 26:42–47. 215. Schroder MD, Snyder PJ, Sielski I, Mayes L. Impaired performance of children exposed in utero to cocaine on a novel test of visuospatial working memory. Brain Cognition 2004; 55:409–412. 216. Stanwood GD, Washington RA, Levitt P. Identification of a sensitive period of prenatal cocaine exposure that alters the development of the anterior cingulate cortex. Cereb Cortex 2001; 111:430–440. 217. Paule MG, Gillam MP, Allen RR, Chelonis JJ. Effects of chronic in utero exposure to cocaine on behavioral adaptability in rhesus monkey offspring when examined in adulthood. Ann NY Acad Sci 2000; 914:412–417. 218. Hurt H, Malmud E, Betancourt L, Braitman LE, Brodsky NL, Giannetta J. Children with in utero cocaine exposure do not differ from control subjects on intelligence testing. Arch Pediatr Adoles Med 1997; 151:1237–1241. 219. Phelps L, Cottone JW. Long-term developmental outcome of prenatal cocaine exposure. J Psychoeducational Assess 1999; 17:343–353. 220. Richardson G. Prenatal cocaine exposure: a longitudinal study of development. Ann NY Acad Sci 1998; 846:144–152. 221. Koren G, Nulman I, Rovet J, Greenbaum R, Loebstein M, Einarson T. Long-term neurodevelopmental risks in children exposed in utero to cocaine. The Toronto adoption study. Ann NY Acad Sci 1998; 846:306–313. 222. Lester BM, LaGasse LL, Seifer R. Cocaine exposure and children: the meaning of subtle effects. Science 1998; 282:633–634. 223. Delaney-Black V, Covington C, Templin T, et al. Expressive language development of children exposed to cocaine prenatally: literature review and report of a prospective cohort study. J Comm Dis 2000; 33:463–481. 224. van Baar A. Development of infants of drug dependent mothers. J Child Psychol Psychiatry 1990; 31:911–920. 225. van Baar AL, Soepatmi S, Gunning WB, Akkerhuis GW. Development after prenatal exposure to cocaine, heroin and methadone. Acta Paediatr Suppl 1994; 404:40–46.

228

Mayes and Pajulo

226. Angelilli ML, Fischer H, Delaney-Black V, Rubinstein M, Ager JW, Sokol RJ. History of in utero cocaine exposure in language-delayed children. Clin Pediatr 1994; 33:514–516. 227. Johnson JM, Seikel A, Madison CL, Foose SM, Rinard KD. Standardized test performance of children with a history of prenatal exposure to multiple drugs/cocaine. J Comm Dis 1997; 30:45–73. 228. Malakoff ME, Mayes LC, Schottenfeld RS. Language abilities of preschool-age children living with cocaine-using mothers. Am J Addict 1994; 3:346–354. 229. Madison CL, Johnson JM, Seikel A, Arnold M, Schultheis L. Comparative study of the phonology of preschool children prenatally exposed to cocaine and multiple drugs and nonexposed children. J Comm Dis 1998; 31:231–244. 230. Delaney-Black V, Covington C, Templin T, Ager J, Martier S, Sokol R. Prenatal cocaine exposure and child behavior. Pediatrics 1998; 102:945–950. 231. Nulman I, Rovet J, Altmann D, Bradley C, Einarson T, Koren G. Neurodevelopment of adopted children exposed in utero to cocaine. Can Med Assoc J 1994; 151:1591–1597. 232. Swanson MW, Streissguth AP, Sampson PD, Olson HC. Prenatal cocaine and neuromotor outcome at four months: effect of duration of exposure. J Dev Behav Pediatr 1999; 20:325–334. 233. Hawley TL, Disney ER. Crack’s children: the consequences of maternal cocaine abuse. Soc Policy Rep Res Child Dev 1992; 6:1–22. 234. Suffet F, Brotman R. Employment and social disability among opiate addicts. Am J Drug Alcohol Abuse 1976; 3:387–395. 235. Tucker MB. A descriptive and comparative analysis of the social support structure of heroin addicted women. Addicted Women: Family Dynamics, Self-Perceptions, and Support Systems. Washington, D.C.: U.S. Government Printing Office, 1979. 236. Regan D, Leifer B, Finnegan L. Generations at risk: violence in the lives of pregnant drug abusing women. Pediatr Res 1982; 16:91. 237. Boyd C, Mieczkowski T. Drug use, health, family, and social support in “crack”cocaine users. Addict Behav 1990; 15:481–485. 238. McMahon TJ, Rounsaville BJ. Substance abuse and fathering: some final comments on context and process. Addiction 2002; 97:1120–1122. 239. Burns WJ, Burns KA. Parenting dysfunction in chemically dependent women. In: Chasnoff I, ed. Drugs, Alcohol, Pregnancy, and Parenting. London: Kluwer Academic Publishers, 1988:159–171. 240. Lawson M, Wilson G. Addiction and pregnancy: two lives in crisis. Social Work in Health Care 1979; 4:445–457. 241. Fiks KB, Johnson HL, Rosen TS. Methadone-maintained mothers: 3-year-follow-up of parental functioning. Int J Addict 1985; 20:651–660. 242. Mondanaro JE. Women, pregnancy, children, and addiction. J Psychedelic Drugs 1977; 9:59–67. 243. Aron WS. Family background and personal trauma among drug addicts in the United States: implications for treatment. Brit J Addict 1975; 10:295–305. 244. Davis SK. Chemical dependency in women: a description of its effects and outcome on adequate parenting. J Subst Abuse Treat 1990; 7:225–232. 245. Black R, Mayer J. Parents with special problems: alcoholism and opiate addiction. Child Abuse Neglect 1980; 4:45–54. 246. Beeghly M, Frank DA, Rose-Jacobs R, Cabral H, Tronick EZ. Level of prenatal cocaine exposure and infant-caregiver attachment behavior. Neurotoxicol Teratol 2003; 25:23–38. 247. Eiden R. Maternal substance use and mother–infant feeding interactions. Infant Mental Health J 2001; 22:497–511. 248. Mayes L, Feldman R, Granger R, Haynes O, Bornstein MH, Schottenfeld R. The effects of polydrug use with and without cocaine on mother-infant interaction at 3 and 6 months. Infant Behav Develop 1997; 20:489–502.

NDS of Prenatal Cocaine Exposure

229

249. Johnson AL, Morrow CE, Accornero VH, Lihua X, Anthony JC, Bandstra ES. Maternal cocaine use: estimated effects on mother-child play interactions in the preschool period. J Dev Behav Pediatr 2002; 23:191–202. 250. Molitor M, Mayes L, Ward A. Emotion regulation behavior during a separation procedure in 18-month old children of mothers using cocaine and other drugs. Dev Psychopath 2003; 15:39–45. 251. Ukeje I, Bendersky M, Lewis M. Mother-infant interaction at 12 months in prenatally cocaine-exposed children. Am J Drug Alcohol Abuse 2001; 27:203–224. 252. Gottwald SR, Thurman SK. The effects of prenatal cocaine exposure on mother–infant interaction and infant arousal in the newborn period. Top Early Child Spec Educ 1994; 14:217–234. 253. Burn K, Chetnik L, Burns WJ, Clark R. Dyadic disturbances in cocaine-abusing mothers and their infants. J Clin Psychol 1991; 47:316–319. 254. LaGasse LL, Messenger D, Lester BM, et al. Prenatal drug exposure and maternal and infant feeding behaviour. Arch Dis Child 2003; 88:391–399. 255. Blackwell P, Kirkhart K, Schmitt D, Kaiser M. Cocaine/polydrug-affected dyads: implications for infant cognitive development and mother-infant interaction during the first six postnatal months. J Appl Dev Psych 1998; 19:235–248. 256. Burns KA, Chetnik L, Burns WJ, Clark R. The early relationship of drug abusing mothers and their infants: an assessment at eight to twelve months of age. J Clin Psych 1997; 53:279–287. 257. Eiden RD, Lewis A, Croff S, Young E. Maternal cocaine use and infant behavior. Infancy 2002; 3:77–96. 258. Kochanska G. Emotional development in children with different attachment histories: the first three years. Child Dev 2001; 72:474–490. 259. Belsky J, Pasco Fearon RM. Infant-mother attachment security, contextual risk, and early development: a moderational analysis. Dev Psychopath 2002; 14:293–310. 260. Carlson EA. A prospective longitudinal study of attachment disorganization/disorientation. Child Dev 1998; 69:1107–1128. 261. Hertsgaard L, Gunnar M, Erickson MF, Nachmias M. Adrenocortical responses to the Strange Situation in infants with disorganized/disoriented attachment relationships. Child Dev 1995; 66:1100–1106. 262. Shaw DS, Owens EB, Vondra JI, Keenan K. Early risk factors and pathways in the development of early disruptive behavior problems. Dev Psychopath 1996; 8:679–699. 263. Rodning C, Beckwith L, Howard J. Quality of attachment and home environments in children prenatally exposed to PCP and cocaine. Special issue: attachment and developmental psychopathology. Dev Psychopath 1991; 3:351–366. 264. Espinosa M, Beckwith L, Howard J, Tyler R, Swanson K. Maternal psychopathology and attachment in toddlers of heavy cocaine-using mothers. Infant Mental Health J 2001; 22:316–333. 265. Swanson K, Beckwith L, Howard J. Intrusive caregiving and quality of attachment in prenatally drug-exposed toddlers and their primary caregivers. Attach Hum Develop 2000; 2:130–148. 266. Fonagy P, Gergely G, Jurist E, Target M. Affect Regulation, Mentalization and the Development of the Self. New York: Other Press, 2001. 267. Levy DW, Truman S, Reflective functioning as mediator between drug use, parenting stress and child behavior. College Problems of Drug Dependence, Quebec City, Quebec.

SECTION THREE: EXPOSURE AND OUTCOME ASSESSMENTS

11 Immutable Elements—Variable Effects: Exposure Assessment for Neurotoxic Metals Alan H. Stern Division of Environmental and Occupational Health, University of Medicine and Dentistry of New Jersey, School of Public Health, Piscataway, New Jersey, U.S.A.

INTRODUCTION Among the metals that warrant consideration for their developmental neurotoxic potential, lead (Pb) (1) and mercury (Hg) (2) have clearly been the focus of the greatest attention. Other metals and metalloids, however, also pose at least the potential to cause developmental neurotoxicity, although developmental neurotoxicity may not be the most sensitive toxic endpoint in each case. These include manganese (Mn) (3), arsenic (As) (4), and to a lesser extent cadmium (Cd) (5). Recently, arsenic As has also been associated with reduced intellectual performance in children (6). As discussed below, the metals present unique issues for exposure assessment and pharmacokinetic modeling. Because the exposure assessment and pharmacokinetic considerations for methylmercury (MeHg) have been fairly well characterized, and because MeHg presents many of the exposure assessment complexities encountered among the various neurotoxic metals, this chapter includes a focus on MeHg as a case study from which generalizations applicable to the other developmental neurotoxic metals can be drawn.

GENERAL CONSIDERATIONS FOR METALS Metals are, by definition, elements. That means that metabolism, conjugation, and binding to various organic ligands notwithstanding, the parent metals are not degraded to simpler substances. With minor exceptions, metals are eliminated from the body in either feces or urine. The immutability of metals, can, therefore, be a major advantage in following the exposure or body burden because analysis for the element, itself, in urine, feces, or blood can integrate exposure regardless of the form in which the metal enters or leaves the body. On the other hand, some of the metals of concern exist in different valence states and/or as mixtures of inorganic and organo-metallic compounds (Hg, As, Mn) with different toxicities and pharmacokinetics. In such cases, analysis for the elemental form can obscure 231

232

Stern

the chemical-specific exposures and burdens. A notable exception is Pb, which largely exists in the body as the PbC2 ion and does not appear to undergo significant changes in valence state or to form organic ligands of toxicological significance in the body. These different valence states and compounds can have highly significant direct and indirect effects on developmental neurotoxicity. The form of the metal can significantly affect its absorption. MeHg is absorbed from the gastrointestinal tract with close to 100% efficiency, while inorganic Hg (HgC2) is, at most, moderately absorbed, and elemental Hg (Hg0) is essentially non-absorbed from the gastrointestinal tract. Although general pharmacologic principles would predict that ionized forms would be poorly absorbed across the gastrointestinal tract and across membranes in general, there does not appear to be a general rule as far as the metals are concerned, and absorption may have more to do with the essentiality of the metal or the ability of metal ions to mimic essential ions. A special case of this is MeHg, where the readily formed MeHg-cystein complex is recognized as the amino acid, methionine, and is actively transported across both the placenta and the blood–brain barrier. It is generally the case though, that absorption of metal salts is mediated by aqueous solubility, and essentially insoluble forms of a metal are not absorbed to any significant extent. Conversions and interconversions of the various forms of these metals can occur in various body compartments. Oxidation of MnC2 and Hg0 take place in the blood, and conversion of MeHg to HgC2, as well as the conversion of inorganic As to methylated As species, occurs largely in the liver. The different forms of a metal may not only display different pharmacokinetics, but they may also have different target organs either as a result of inherently different toxicologic mechanisms or secondary to pharmacokinetic factors. HgC2 is largely excluded from the brain by its relative inability to cross the blood–brain barrier. Thus, although HgC2 is inherently neurotoxic, it primarily affects the kidney where it tends to accumulate, while MeHg, which readily crossed the blood–brain barrier, primarily affects the central nervous system. The form of the metal may also influence its route of elimination. Soluble salts and soluble complexes are generally eliminated in the urine, while unabsorbed material and metals in lipid soluble forms are excreted in the feces. The form of the metal that is excreted may or may not be the form that is ingested or inhaled. Inorganic As is largely eliminated as organic forms, and MeHg is largely eliminated in the feces as HgC2. When exposures to more than one form of a metal occur more or less simultaneously, reconstruction of the individual exposures by analysis of either the elemental form, or of a specific species of the metal in excreta can produce confusing results. The various forms of a metal may also have different half-lives in the body as a whole and in various body compartments (e.g., blood, brain). Therefore, consideration of the relationships between exposure and effect, or attempts to reconstruct exposures from a sample of a biological tissue or fluid, need to consider the specific pharmacokinetics for each species in question. Organics tend to be retained in the body either because they bind strongly to proteins (e.g., serum albumin) or because they are lipophilic and sequester in fatty tissues. Some of the neurotoxic metals have significant protein binding. Pb binds strongly to hemoglobin, and Hg (particularly MeHg) binds strongly to the sulfhydral groups of many proteins including hemoglobin. However, the developmental neurotoxic metals are not particularly lipophilic even considering the organic forms of Mn, As, and Hg. In light of ongoing confusion, it is worth emphasizing here that MeHg is not lipophilic. It tends to distribute more or less uniformly to tissues, with kidney and liver having somewhat elevated levels. Contrary to the expectation for a lipophilic molecule, MeHg concentration is not enriched in human milk relative to inorganic Hg (7,8). The bioconcentration of MeHg in the body as well as its biomagnification in the food change is, rather, a function of the similarity of MeHg–cystein to methionine as well as its general strong proclivity for sulfhydral groups

Exposure Assessment for Neurotoxic Metals

233

including those in the bile, leading to enterohepatic circulation (9). For Pb, although binding to hemoglobin contributes to its retention in the body, the primary factor in the very long half-life of Pb is its sequestration in the bone due to the apparent similarity of the PbC2 ion to the CaC2 ion (1). Intake, retention, and systemic absorption of developmental neurotoxic metals via inhalation tend to be governed by general principles of particulate exposure (i.e., size, solubility, reactivity), or, in the case of Hg0, by general principles of vapor exposure.

GENERAL APPROACHES TO EXPOSURE ASSESSMENT FOR DEVELOPMENTAL NEUROTOXIC METALS The broad field of exposure assessment addresses the entire range of exposure from emissions of substances to the environment, to dispersion in the environment, to contact and uptake by receptors, to distribution within the body, and ultimately to the contact between the substance and a target organ or receptor. For the purposes of elucidating the potential effects of developmental neurotoxic metals, however, we are most directly concerned with those aspects of exposure from uptake to appearance at target organs or tissues. In this context, there are three general applications of exposure assessment, each of which addresses a somewhat different concern. On the clinical level, we may be concerned with determining whether an individual has experienced an exposure of sufficient magnitude to explain existing symptoms, or to require medical intervention. On the population level, we may ask about the characteristic levels of exposure, range of exposure or distribution of exposure in a population in order to determine the risk to the population relative to a standard or guideline, to identify populations appropriate for epidemiologic studies of effect, or to compare different populations in order to investigate whether differing aspects of their environments result in different levels of exposure. Finally, exposure assessment tools, including pharmacokinetic modeling, may be used to relate exposure to adverse health outcomes from epidemiologic studies for the purpose of deriving dose–response relationships that can support risk assessments and risk-based regulatory standards and guidelines. Exposure can be estimated by three types of approaches of increasing specificity with respect to exposure at the level of the individual. At the lowest level of specificity, generalized models are available that attempt to predict population exposure based on measured levels of a contaminant in the environment, and assumptions about rates of contact through various media. The usefulness of such approaches is limited by the validity of the models, by the uncertainty and variability in environmental measurements, and by the uncertainty and variability in the assumed contact rates across the population. Such an approach is best suited to broadly characterizing a population or for generating hypotheses about the impacts of specific contaminant sources in the environment. A more specific approach to estimating exposure of individuals or populations is to obtain data from the individual or population (or more likely from a sample of the population) on rates of contact with contaminated media such as drinking water (e.g., number of glasses of tap water consumed per day), soil (e.g., number of hours spent by children playing in areas with exposed contaminated soil), or diet (e.g., frequency and size of meals of contaminated food). Dietary data are most frequently derived from personal recall questionnaires, but can also be generated from food consumption diaries, and duplicate diet studies. Data on rates of contact can then be combined with estimates of the concentration of contaminants in particular media to estimate the daily intake of contaminants. Such an approach works best when the exposure to the contaminant comes from a single well-defined source such

234

Stern

as MeHg in fish, as in drinking water, or Pb in soil surrounding a smelter. Even in such cases, however, this approach is limited by uncertainty and variability in the estimated concentration of the contaminant in the media, and by uncertainty and bias in subjects’ recall of the normal and historic contact activity (e.g., How frequently is fish eaten per week?). This may be particularly the case in attempting to apply such an approach to children either through eliciting their direct recall, or by eliciting information from parents. From the standpoint of estimating individual exposure, the most specific approach is biomonitoring, that is, the measurement of contaminants in tissues or excreta of exposed individuals. In general, such measurements can be made with high accuracy. These measurements can be used with few additional considerations if the goal is simply to compare the relative exposure of two similar populations. However, if the goal is to estimate an intake dose (e.g., mg metal/kg body weight/day), then biomonitoring data must be considered in conjunction with a pharmacokinetic model that can relate the measured concentration in a tissue or excreta to intake (e.g., What dose of MeHg would result in a measured concentration of 58 mg of MeHg in blood?). This approach is limited by the accuracy and applicability of the pharmacokinetic model. The fraction of the population that most efficiently transfers an intake dose to the target tissue will be at greater risk from the same dose. Thus, when applied to populations, pharmacokinetic models should be able to estimate not only the average relationship between the intake dose and the measured biological concentration, but should also be able to describe the variability in that relationship in order to derive appropriately protective risk-based standards and guidelines. These considerations are discussed more fully in the National Research Council’s recent report on the toxicology of MeHg (10). Determination of total metal concentration in biological samples for developmental neurotoxic metals is generally a relatively straightforward process given the immutability of the elemental form. Historically, and still for most applications, one of several forms of atomic absorption specrophotemetry (AAS) has been used and provides practically useful detection limits for most concentrations corresponding to environmental exposure of concern. More recently, inductively coupled plasma atomic emission (ICP-AES) and inductively coupled plasma mass spectrometry (ICP-MS) have been successfully used. These methods all require sample extraction or ashing as well as additional sample preparation. For analysis of Hg in hair, X-ray fluorescence (XRF) has been used without extensive sample preparation (11). Analysis of single hair strands was able to elucidate the time course of Hg exposure in the Iraqi poisoning epidemic. Unfortunately, the practical detection limit for this method is larger than most current day exposure of interest. XRF has also been used to measure Pb deposited in bone in situ (1). This approach provides a window into historic and long-term chronic Pb exposures that, due to the turnover of most biologic material, is not available for the other neurotoxic metals. Where toxicity varies significantly for different species of the same metal, analyses yielding the concentration of the total metal can confound estimates of effect, dose–response or risk. However, determination of specific species of a metal is more difficult and generally requires compound-specific methods to separate the various species. These include column separation, and selective reactions with a given species. One common method for determination of MeHg involves separate determinations of inorganic and total Hg indirectly yielding MeHg by difference (12). With the recent elucidation of subtle neuro developmental effects from metals at lower levels of exposure than previously recognized, previously insignificant levels of background laboratory contamination by some of those metals can confound analyses of trace levels in biological material (13). This is particularly a concern for Pb and Mn.

Exposure Assessment for Neurotoxic Metals

235

The determination of which biological material to analyze as an appropriate measure of internal exposure can be complicated. Practical considerations pose the primary limitations for biomarkers of human exposure. Thus, although the concentration of these metals in the developing brain may be the most precise measure of exposures relating to neuro developmental effects, such tissue is rarely accessible. Other organs are likewise rarely available for analysis. Practical choices are generally limited to blood, hair, nails, milk, feces, and urine. Additionally, for Pb, in situ analysis of bone using XRF, and analysis of deciduous teeth provide useful measures of cumulative exposure based on retention of Pb in bone and dentin (14,15). Among these media, the selection of the most appropriate medium involves consideration of a variety of factors. Metal concentration in mother’s milk is clearly the most direct measure of lactational exposure, but is an indirect measure of either maternal or fetal body burden. Some metal species are preferentially eliminated in either urine or feces. MeHg is mostly eliminated in the feces, but HgC2 is almost entirely eliminated in the urine (2). It is, therefore, essential to understand the pharmacokinetics of distribution and elimination of the specific metal species of concern. In blood, the concentration following the initial distribution of an ingested dose among the various body compartments is a useful measure of available metal. However, for metals or species of metals that do not readily cross the blood–brain barrier or the placenta (e.g., HgC2), blood concentration may be an indirect measure, at best, of neuro developmental risk. Most of the other potential biologic media receive their burden of metal from the blood, and are subject to the same limitation. For metal species that do readily cross the blood–brain barrier and/or the placenta, the choice of medium depends, at least in part, on the persistence of a given dose of the metal in a body compartment relative to the duration of the period of vulnerability for the developmental endpoint in question. For chronic, essentially continuous, exposures, such as those that might occur from inhalation of a metal in ambient air from an ongoing source, or from a contaminated drinking water source, the concentration of the metal in any compartment will reach a steady-state, constant concentration in that compartment. In such a case, any compartment to which the metal is significantly distributed will likely provide a measure of ongoing individual exposure. Many exposures of concern, however, such as dietary exposures, occur episodically and with varying doses over time, resulting in peaks of exposure. For metals with relatively long half-lives in blood such as MeHg (approximately 50 days), and correspondingly long half-lives of elimination in urine or feces, concentrations of the metal in these media will tend to reflect average exposure over time. Peaks of exposure, including those that occur during critical windows of fetal development, may not be evident in these media, or may appear as modest increases over the course of repeated samples. On the other hand, for metals with short half-lives in blood such as As (40–60 hr), the blood concentration can reflect specific exposures, and may provide a reasonable indication of peak exposure, but samples need to be taken close to the exposures. This presents a serious practical difficulty. For such metals, blood or excreta do not preserve a useful record of exposure over time. In contrast, for those metals quantitatively retained in hair, hair provides a potential record of exposure over time. MeHg is incorporated into growing hair strands from matrix cells nourished by the blood, and is bound in the hair matrix. Thus, Hg concentration along the hair strand preserves a record of exposure during the period of hair growth at a rate of about 1 cm/mo. Concentration in hair has been used as a measure of As exposure, but the relationship between exposure and hair concentration does not appear to have been fully characterized (4). Hair is a less well characterized biomarker of exposure for Pb and Mn. A consideration with hair samples is that the portion of the hair strand reflecting current exposure requires a period estimated to be 20 days to 6 week to emerge above the scalp. Thus, maternal hair sampled at delivery actual reflects

236

Stern

exposures no later than the mid-third trimester, and, given the need to obtain a segment of reasonable length, probably reflects exposure extending into the second trimester. Hair also presents potential uncertainties due to both intra- and inter-individual variability in growth rate. This includes differences in growth rate among strands within the same hair sample, changes in morphology and growth rate during pregnancy, and the existence of a resting phase during which there is no active transfer of material from the blood (16–18). Additionally endogenous metals can be lost and exogenous metals (particularly those contained in shampoos and hair care products) can be added to hair strands (19). Considerations for the application of hair samples to exposure assessment in general and to assessment of in utero exposure assessment are discussed in greater detail in the NRC’s report on the toxicology of MeHg (10). Nails are biologically similar to hair, and have been used as a measure of MeHg exposure, but here too, the relationship between nail Hg concentration and exposure has not been well characterized (20). In general, researchers have access to these various media only at specific times such as medical contacts. Often maternal samples and cord blood samples are available only at, or close to, delivery. Depending on the half-life of the specific metal or metal species in blood or excreta, and on the applicability of hair samples, these samples of opportunity may reflect exposure only during a limited period of gestation. Such periods may include portions of neuro development that are vulnerable to the metal in question, but may also include relatively insensitive periods. They may also entirely miss the most vulnerable periods for the given metal. Thus, temporal mismatch between exposure measurements and vulnerable periods of development, a form of exposure misclassification, is always a concern with the use of biomarkers in investigations of in utero neurotoxicity. Furthermore, even if the concentrations of the metal in the available media do reflect exposures during vulnerable periods of development, the tests of neurologic function that are ultimately applied as the measures of effect may not measure the outcomes effected by the metal during the exposure period reflected by the biomarker. Underlying these problems are the relatively large gaps that remain in our knowledge of which neuro developmental functions are vulnerable during which stages of development. This is particularly the case for in utero development (21). More precise information about the temporal occurrence of specific neuro developmental vulnerabilities combined with a better understanding of the toxicological mechanisms of these metals that operate during development will lead to the use and development of more precise exposure assessment tools. There are myriad ways that an exposure measure, including a biomarker measurement, can be mismatched with an outcome, particularly in a relatively large study. It is thus unlikely that in such a study, a given set of exposure measurements will randomly associate with an outcome measurement, and very unlikely that such associations will be repeated across several different outcome measurements. This concept has been stated as “exposure misclassification biases toward the null.” The observation of several statistically meaningful associations, therefore, should be interpreted as an indication of a true association between the exposure reflected in the given exposure metric and the developmental outcome. In light of the current level of imprecision in our application of exposure metrics for neuro developmental toxic metals, the study designs most likely to find true associations between exposure and outcome are those that employ several different measures of exposure. Biomarkers of exposure are generally surrogates for the ultimate, but inaccessible exposure measure of interest. In the case of neuro developmental toxic metals, the ultimate exposure measure is generally either the concentration in the developing brain, or the retrospective maternal or childhood intake dose. When two or more valid biomarkers of exposure are available the choice of which to use should depend, at least in part, on the

Exposure Assessment for Neurotoxic Metals

237

pharmacokinetic relationship between the biomarkers and the exposure that is ultimately being estimated. For example, if the intent is to estimate the maternal intake dose corresponding to an exposure in utero putatively resulting in an observation of reduced neurologic performance, then maternal blood concentration of the metal at the time of fetal developmental vulnerability is pharmacokinetically closer to the maternal intake dose than is fetal cord blood concentration. On the other hand, if the intent is to estimate the concentration in the developing fetal brain, then fetal cord blood concentration of the metal is pharmacokinetically closer to the brain concentration than is maternal blood. In the latter case, however, other factors may also need to be considered. For instance, depending on the half-life of the metal in cord blood (and the fetus’ ability to metabolize and eliminate the metal), cord blood concentration may only reflect a portion of the third trimester of gestation. In such a case, if the developmental effect of concern results from a fetal vulnerability during the second or early third trimester of gestation, the connection between cord blood metal concentration and fetal brain concentration may be pharmacokinetically close, but temporally distant. Metal concentration in maternal hair (or nails) can be thought of as a spur off the pharmacokinetic main line connecting intake and brain. As a route of elimination from blood, it is at least two steps removed from both maternal intake and fetal brain concentration. It is, therefore, a pharmacokinetically more distant measure of either maternal intake or fetal brain concentration than is maternal or cord blood. This, deficit, may, however, be tempered by hair’s ability to retain a measure of exposure over time, providing a possible solution to temporal mismatches between exposure measures and the timing of developmental vulnerabilities.

APPLICATION OF PHARMACOKINETIC MODELS Biomarkers of exposure provide information in the form of concentration of the metal or metal species in a particular tissue or medium. For purposes of comparison of individual values to reference values, or for statistical comparison of the exposure of two or more populations, or for the testing of hypotheses of the relationship between exposure and effect, concentrations, by themselves, may be adequate. Although concentration-response relationships are related to the dose–response relationship, the concentration of a metal in a tissue or compartment is not, by itself, an estimate of dose. Thus, when the goal is to combine exposure data with outcome data to develop dose–response relationships to relate intake to risk, and/or to derive public health strategies, it is necessary to translate the measured metal concentrations to a dose (e.g. mg metal/kg body weight/day). This process of translation requires a pharmacokinetic model (sometimes referred to as a toxicokinetic model). Among their capabilities, these models predict the relationship of an administered dose to the concentration in one or more body compartments and media. Two basic types of pharmacokinetic models exist, physiologically-based pharmacokinetic (PBPK) models and steady-state models. PBPK models combine chemical-specific data on transfer coefficients between blood and individual tissue compartments, and blood flow rates for each compartment and employ simultaneous differential equations to construct a dynamic model of distribution and deposition in all specified compartments simultaneously. Because such models are dynamic, they can predict short-term occurrences such as the change in blood concentration of a metal as it is undergoing distribution among the various compartments. PBPK models can also predict changes in concentration in each compartment as body and organ weight changes occur with growth or pregnancy. Models have also been constructed that address the interconversion among the various metal species. Of particular note, PBPK models have been constructed to predict the changes in

238

Stern

concentration of metal in fetal and maternal compartments as the fetus grows. The model developed for MeHg by Clewell et al. (22) is a good example of the power of PBPK models. Other examples for As, and Pb are described in their respective ATSDR Toxicological Profiles (1,4). Despite their obvious theoretical strengths, PBPK models have practical limitations. They are, by their nature, data intensive. The data required to populate the model, particularly parameters for transfer coefficients for each body compartment, are often not available in terms of metal and tissue-specific values. These values must be estimated or extrapolated. This is compounded by the need, as discussed above, to produce not only estimates of the central tendency of the relationship between dose and biomarker concentrations, but to also address the population variability in those relationships. This requires that model parameters be not only specific to the metal or metal species of concern, but that they describe the distributional nature of their occurrence in the population. This range of data is rarely available for PBPK model parameters and distributional data as well as central tendency data must often be estimated. Variability estimates are difficult to validate, and it is often unclear how the estimates of variability produced on the basis of these multiple estimates relate to true population variability. The other basic type of pharmacokinetic model is this steady-state model. This approach starts with the limiting assumption that a compartment or set of linked compartments in the body is in kinetic steady-state. That is, that the rate of material entering the compartment (e.g., by ingestion) is equal to the rate of material leaving the system by metabolism, or transfer to another compartment. The condition of steady-state occurs when a substance enters a system at a constant rate and is eliminated at a rate that is a constant percent of the mass of the substance remaining in the system. Steady-state conditions are most applicable to constant or frequent intakes such as inhalation or drinking water exposures. Depending on the frequency of consumption, dietary intakes of the neuro developmental toxic metals may not result in full steady state conditions. The simplest form of the steady-state model is the one-compartment model. Blood is generally the compartment of choice. For metals or metal species with relatively long half-lives in the blood, such as MeHg, irregular or moderately infrequent dietary intake can still result in an approximation of steady state that allows the steady state model to be used with reasonable accuracy. The one-compartment model contains a relatively few parameters. Given the static nature of this model, fixing the value of these parameters, fixes the value of the dose. To estimate the intake dose that gave rise to a measured blood concentration, the model requires the measured concentration, the elimination rate constant (a function of the half-life in blood), the blood volume, the body weight, the fraction of the dose absorbed, and the fraction of the absorbed dose in the blood at steady state. For the neurotoxic metals these parameters are generally reasonably well characterized in terms of central tendency estimates, and existing information may also support development of data-specific distributions of population variability for the parameters. Data may also be available to estimate pregnancy-specific or childhood-specific model parameters that can lead to more developmentally specific estimates of the relationship between concentration and dose. The analysis of Stern (23) is an example of the use of distributional data in the one-compartment steady-state model applied to pregnancy. The one-compartment, steadystate model cannot directly predict relationships outside the modeled compartment. However, empirical ratios can be used to estimate the concentration in other compartments given the concentration in the modeled compartment. For instance, data on the maternal blood-cord blood Hg ratio can be used to estimate the cord blood Hg concentration given the model’s prediction of the maternal blood Hg concentration resulting from a given dose (24). The IEUBK (Integrated Exposure Uptake Biokinetic) model developed by the U.S. EPA is a special case among steady state model. It predicts children’s blood Pb

Exposure Assessment for Neurotoxic Metals

239

concentration starting with concentrations in environmental media. The biokinetic component of the overall model is a linked multiple compartment steady state model that is iteratively calculated over discrete time intervals (25). In general, the trade-off between PBPK models and steady-state models is between model flexibility on the one hand, and model transparency on the other.

MERCURY AS A CASE STUDY Exposure Assessment Approaches Dietary Assessment MeHg exposure is almost exclusively limited to fish consumption. The distinctiveness of fish as a specific food probably makes recall easier than for a food that is more common and eaten more casually. To some extent, this makes dietary assessment more straightforward than for other dietary contaminants such as chlorinated organics that are widely distributed in the diet. Thus, on the one hand, consumers may be able to recall the number, types, and size of fish meals over the course of an extended period such as a week. On the other hand, the relatively infrequent consumption of fish, averaging 1–2 meals per week for most fish consumers (26,27), can lead to recall inaccuracies and bias when attempting to estimate or reconstruct consumption over a longer period of time (27). MeHg exposure has been estimated by three different methods utilizing dietary data. In dietary recall studies, participants are asked to reconstruct their fish consumption over some specific period of time that can extend over several days, to a year (26,27). Shorter recall periods tend to give more accurate recall, but are subject to short-term variations in individuals’ consumption patterns, and tend to underestimate the consumption patterns of infrequent consumers (27,28). Diet diaries kept by participants are not subject to bias or uncertainties in recall. However, given average rates of fish consumption of 1–2 meals per week, reporting generally needs to be conducted over an extended period of time in order to obtain a reasonable estimate of intra-individual variability in consumption. This requires a commitment to persistence on the part of participants. Partial reporting or failure to complete the diaries on the part of some participants can result in biased data. The U.S. EPA (29) has used the two-day diary portion of the Continuing Study of Food Intake by Individuals (CSFII) of the U.S. to estimate Department of Agriculture to estimate fish consumption and MeHg intake in the U.S. population. The short reporting period in this database tends to ensure accurate reporting, but, as with short-term recall data, tends to largely reflect the patterns of frequent consumers. Diet diaries have often been used in conjunction with duplicate diet studies to estimate MeHg exposure (30,31). To estimate MeHg intake from either dietary recall or diet diaries, consumption data must be combined with estimates of the characteristic MeHg for each species of fish consumed. This leads to several different areas of uncertainty. Both sellers and consumers may misidentify the species of fish purchased and consumed. The notion of “characteristic” MeHg concentrations for each species assumes that the source of such information is both geographically precise, and up-to-date. Burger et al. (32) have reported that mean Hg concentrations in some commonly consumed species of commercial fish in New Jersey differ significantly from national U.S. data for the same species of fish reported by the U.S. FDA It is not clear whether these differences result from regional differences in the sources of the commercial fish supply and/or from the differences in the time periods represented by the two databases. In addition, the use of characteristic data on MeHg concentration implies that consumers consume fish with the characteristic or mean MeHg for that

240

Stern

species. Over time, regression-to-the-mean will tend to result in average exposure levels that do, indeed, reflect exposure to the mean concentration of MeHg in each species consumed. However, for species with a large variance in MeHg concentration, short-term exposures, including those during critical neuro developmental periods, can differ significantly from those estimated on the basis of characteristic concentrations. In duplicate diet studies, participants save a portion of food identical to the portion they consume. The duplicate portions are analyzed for the contaminant of concern, thus eliminating problems with species identification and characteristic concentrations. However, such studies place great demands on participants, particularly when food is consumed outside the house. Participants must be highly motivated and compliance wanes after a few days (33). Duplicate diet studies tend to be of relatively short duration and are thus, subject to the biases inherent in the other short duration studies of fish consumption. A more detailed discussion and comparison of dietary assessment of MeHg exposure can by found in the NRC (10) report on MeHg. Estimates of MeHg exposure based on dietary data have generally been useful for overall characterization of population exposures. They tend to be less used for quantifying individual exposures within a cohort. While there is generally acknowledged to be no “gold standard” in exposure metrics for MeHg, biomarkers of exposure, particularly those that can be temporally related to specific periods of development, are generally viewed as the most precise measures of individual exposure. Estimates of MeHg exposure from dietary data have generally been found to correlate moderately at best with biomarker-based estimates (34,35). Biomarkers of Exposure Blood Hg The mean half-life of MeHg in blood in adults is about 50 days (2). However, there is considerable population variability around this value (36). MeHg concentration in blood thus has the potential to be a practical measure of exposure over a period of weeks to months. Nonetheless, blood MeHg concentration has limitations that must be taken into account for it to be useful in neuro developmental toxicity exposure assessment. For consideration of exposure of the developing fetus in utero, concentration in both maternal blood and fetal cord blood can be measured. Total Hg concentration in blood is a simpler and cheaper analysis than is a MeHg-specific determination. For fish consumers with at least moderate consumption, the blood total Hg concentration will largely reflect MeHg, although the actual proportion will depend on the amount and frequency of fish consumption. In the U.S. population of women of childbearing age, the ratio of MeHg to total Hg was 0.64 comparing the 50th percentiles of exposure of both analyses. However, this ratio was 0.92 at the 90th percentile of exposure (37). For populations with little or no fish consumption, however, inorganic Hg, potentially including a significant contribution from dental amalgam fillings (38), can constitute a large proportion of the total blood Hg concentration. It is, therefore, advisable to consider the extent and variability of fish consumption in the population before using total Hg blood concentration as the sole measure of MeHg exposure. Individuals with relatively frequent fish consumption and little variability in the species consumed will approximate steady state with respect to MeHg concentration in blood, and blood concentrations should remain relatively stable as long major physiological parameters are also stable. For such individuals, the timing of a blood sample for Hg concentration is not a major consideration. Even with frequent consumption and a constant diet, however, pregnancy can alter maternal blood Hg concentrations because of changes in maternal blood volume and body weight, and because of transfer of MeHg from maternal blood to the fetus. For individuals who are

Exposure Assessment for Neurotoxic Metals

241

infrequent consumers, and/or for individuals with irregular consumption of species with high MeHg concentrations, blood Hg concentrations will vary in response to short-term dietary patterns. In the absence of accurate dietary information, blood Hg concentration is a “snapshot” of exposure at a point in time. From such information, it is not possible to determine whether an observed blood Hg concentration reflects average exposure, a shortterm elevation, or a declining concentration. If the goal is to obtain a cross-sectional picture of exposure in a population based on a reasonable sample size, then the “snapshot” nature of the individual observations of blood Hg concentration is not a drawback. However, if the goal is to relate exposure to a neuro developmental outcome or measure, then the “snapshot” nature of blood Hg concentration can lead to exposure misclassification in terms of a temporal mismatch between the time period reflected by the blood Hg concentration and the exposure during the sensitive developmental period reflected in subsequent neurological testing. This is particularly the case for neurotoxic exposure in utero. Although the mean half-life of MeHg in adult blood is about 50 days, there is some suggestion that during the third trimester of pregnancy, the half-life in maternal blood decreases significantly (23). Assuming a 50 day half-life, maternal blood sampled at delivery will reflect exposure from about the second half of the second trimester to the nearly the end of the third trimester. However, all other things being equal, Hg concentration will be most heavily influenced by exposure during the most recent halflife, corresponding to the last half of the third trimester. The situation is somewhat more complicated for the use of cord blood Hg concentration as a biomarker of exposure. MeHg appears to readily cross the placenta from maternal blood. However, the ratio of cord blood to maternal blood at or close to delivery is about 1.7 (24). The larger concentration on the fetal side of the placenta may result from simple differences in hematocrit between maternal and fetal blood, or may indicate differential binding. In addition, it appears that the fetus has little or no ability to metabolize MeHg to inorganic Hg (39). It is, therefore, possible that the half-life of MeHg in fetal blood is longer than in maternal blood. Cord blood Hg concentration may therefore preserve a record of exposure extending somewhat further backward in time compared to maternal blood. Neuro-developmental outcomes affected during the period represented by maternal or cord blood Hg may show meaningful statistical associations with blood Hg concentration. However, the strength of those associations, and the power to detect such associations will diminish the further the vulnerable gestational period is from the last half of the third trimester. In addition to the temporal uncertainty inherent in the exposure information contained in any measurement of blood Hg concentration, there is a general lack of information about the timing of the occurrence of periods of specific neuro developmental sensitivity to MeHg. Therefore, at the present time, the determination of whether either maternal or cord blood Hg is an appropriate biomarker of exposure for use in investigating associations between MeHg exposure and specific adverse neuro developmental outcomes can only be made empirically and retrospectively based on whether the blood Hg concentration predicts meaningful statistical associations with the specific measures of neurologic function or performance employed in a given study. The failure to detect an association may indicate the absence of a causal association, or may indicate that blood Hg was a temporally or pharmacokinetically imprecise exposure metric for that particular study. Hair Hg For determination of MeHg exposure, hair Hg concentration has some unique advantages. MeHg transfers from the blood to the growing hair follicles and binds strongly. This is less the case for inorganic Hg, and hair total Hg is generally 80–90% MeHg even in

242

Stern

populations with moderate to little fish consumption (34,40,41). Therefore, analysis of total Hg in hair generally provides a close estimate of MeHg concentration, and it is rarely necessary to speciate hair Hg. Sampling of hair is relatively non-invasive. Although the notion of giving a hair sample may initially put off subjects who envision unsightly gouges in their coiffures, hair analysis for Hg requires only about 50 mg of hair, about the width of a matchstick, that is generally sampled from the under layer at the nape of the neck. From the standpoint of MeHg exposure assessment, however, the most salient feature of hair is that it retains a record of MeHg exposure during the entire growth of the follicle. Hair grows at about 1 cm/mo (see Chapter ref. 10 for a detailed discussion of hair growth and its implications for MeHg exposure assessment). In theory then, hair contains a nearly realtime record of MeHg exposure whose time course can be related to the calendar. There are some complexities inherent in this relationship including variability in hair growth rate, and the time lag between incorporation of MeHg in the follicle and its emergence above the scalp (10). This time lag is about 20 days. For developmental effects potentially extending through childhood, this does not pose a significant concern; however, for assessment of MeHg exposure during gestation, this lag means that maternal hair sampled at delivery will not reflect fetal exposure during the end of the third trimester. Analysis of Hg in hair from neonates, and to a lesser extent from infant hair would seem to circumvent this problem and to provide a pharmacokinetically more direct measure of late gestational MeHg exposure, but this biomarker has not been well characterized and is not always available (42,43). If sufficiently long maternal hair strands can be obtained, exposure during most of gestation can be represented. If the entire strand is analyzed together, the resulting Hg concentration will reflect average exposure. To obtain more detailed information on exposure at specific times during gestation, however, requires that hair strands be analyzed in longitudinal portions corresponding to shorter intervals of time. Ideally, continuous determination of Hg concentration along the hair strand would give a real-time record of exposure. Such information could be extremely valuable in allowing investigation of the association of various parameters of exposure with developmental outcomes. In contrast to the standard association of average exposure over most of gestation given by analysis of 8–9 cm long strands, continuous analysis could allow investigations of associations between developmental outcomes and peak exposure, average peak exposures, peak exposure during a given trimester, etc. This could address the still thorny question of whether peak or “bolus” exposures to MeHg during gestation resulting from periodic or episodic meals of sea food with high Hg concentrations are more causally related to adverse developmental outcomes than moderately elevated, but continuous exposure from a constant sea food diet. Cox et al. (11) used XRF to recover essentially continuous information on the time course of exposure in the MeHg poisoning in Iraq in the early 1970s. An added feature of this analysis was that it could be conducted on a single strand of hair, thereby eliminating imprecision due to the misalignment of multiple hair strands. Unfortunately, the XRF methodology used in that study has a relatively large practical detection limit that was sufficient for the large exposures in the Iraqi poisoning, but too large for use with populations with more moderate levels of exposure. To date, more sensitive continuous methods for quantification of Hg in hair have not been applied to exposures assessment of MeHg exposed populations. However, techniques such as laser ablation (44) appear to hold promise for the application of continuous hair strand analysis of Hg to dose–response assessment of populations at current levels of exposure. More commonly, information about the time course of exposure to MeHg is recovered from hair by segmental analysis. In this approach, sections from a bunch of aligned hair strands are analyzed separately. Usually, 1–3 cm sections are cut, corresponding to 1 month’s to one trimester’s worth of exposure. By comparing the

Exposure Assessment for Neurotoxic Metals

243

Hg concentration among sections, this approach can provide information about the variability of exposure over time. However, while this may provide evidence of the existence of peak exposures, the actual magnitude of putative peak exposures resulting from individual high MeHg meal or several such meals over a short period of time cannot be determined. This results partly from the difficulty in accurately aligning hair strands within a bunch. However, the more fundamental problem is that analysis of hair segments even as short as 1 cm averages periods of increasing and declining concentration along with the corresponding peak concentration. Information on peak exposure is thus “smeared.” Analysis of segments shorter than 1 cm has generally not been practical. A more detailed analysis of this problem is presented in the NRC (10) report. One potential advantage of hair for MeHg exposure analysis is that it has the potential to provide some information on neonatal and infant exposure. Samples obtained from children less than one-year old could, depending on the length of their hair, reflect exposure at or close to birth. As with blood Hg, given the various uncertainties involved, the determination of whether hair Hg is the an appropriate choice of exposure metric for investigating associations between exposure and neuro developmental outcomes can only be made retrospectively within a given study. Milk Hg For assessment of lactational exposure to MeHg, the Hg concentration of breast milk is an obvious exposure metric. However, lactational exposure is a good example of the importance of addressing the forms of neuro developmental metals at the point of exposure. Milk is derived from maternal plasma. MeHg in the blood tends to bind preferentially to erythrocytes, while inorganic Hg is enriched in the plasma. The inorganic Hg in the plasma reflects not only sources of direct inorganic Hg exposure such as dental amalgams, but also the metabolism of MeHg. Thus, compared to whole blood, milk will tend to be enriched in inorganic Hg (7). In one study of fish consumers in Sweden, for instance, the mean organic Hg (essentially MeHg) concentration in maternal blood was six times the mean concentration in mother’s milk (45). Thus, breast feeding significantly increases infants’ exposure to MeHg, but there is only a moderate correlation of total Hg in maternal milk and total Hg in infants’ hair (reflecting largely MeHg body burden) (43). Overall, the extent of MeHg-specific exposure is considerably less than would be estimated based on measurement of total Hg concentration in either maternal blood or milk. Milk, like hair, can be thought of as a pathway of maternal MeHg elimination. Blood Hg concentrations in lactating women are characteristically lower than in non-lactating women. Therefore, it is important to identify lactating women when characterizing population-based exposure through hair or blood Hg concentration. Nail Hg Nails are biochemically closely related to hair. Nails similarly appear to retain a record of MeHg exposure, Toenails are generally preferred to fingernails probably because they are assumed to be less exposed to exogenous contamination, and because they are less frequently trimmed. Toenail Hg concentration has been shown to reflect fish consumption (20,46). With an average growth rate of about 0.1 cm/mo, toenails have the potential to integrate long term exposure in a single clipping. This telescoping of exposure information can be an advantage in studies of exposure over extended periods of development where long-term average exposure may be the most appropriate metric. However, it is likely to be a disadvantage for identifying exposures over shorter and specific time periods such as those occurring during windows of vulnerability in gestation. To date, toenails have not

244

Stern

been well characterized as a biomarker of MeHg exposure. There is no model for relating intake dose to toenail concentration, nor is there a clear relationship between toenail Hg concentration and the corresponding concentration in either hair or blood (47,48). Toenail Hg has been used as the measure of exposure in two large scale epidemiological studies of MeHg and cardiovascular disease in adult men (49,50), but it does not appear to have been employed in studies of the neuro developmental effects of MeHg. Further characterization of this biomarker could enable it as another useful tool for the exposure assessment of MeHg.

Recent Applications of MeHg Neuro-Developmental Exposure Assessment Population-Based Exposure Assessment The U.S. Centers for Disease Control and Prevention (CDC) has conducted an ongoing assessment of Hg exposure in the U.S. population of young children (1–5 years old) and women of childbearing age (16–49 years old) through its National Health and Nutrition Examination Survey (NHANES). The general purpose of this assessment is to describe and track the distribution of MeHg exposure in the sensitive population. A more specific focus of assessment has been to gauge the fraction of the sensitive population that exceeds the U.S. EPA Reference Dose (RfD) for MeHg of 0.1 mg/kg body weight/day. The RfD is the U.S. EPA’s estimate of the virtually safe dose for the most sensitive groups in the population (51). The Hg exposure assessment was originally embedded in CDC’s multichemical assessment of biomarkers, the first of which was based on NHANES data collected in 1999 (52), and the second of which used NHANES data from the period 1999–2000 (53). Subsequent data collection through NHANES focused more specifically on Hg and updated the assessment through 2002 (54). Both blood Hg concentration (52–55) and hair Hg concentration (52,56) were measured. As concentrations, neither measure is directly comparable to the RfD. Conclusions about the extent of exceedance of the RfD were based on the predictions of the one-compartment pharmacokinetic model that give a rough correspondence between the steady-state intake of MeHg at the RfD and blood and hair Hg concentrations of 5.8 mg/L and 1.0 mg/g respectively. Based on the blood Hg data 6–8% women of childbearing age exceeded the RfD. Based on the hair Hg data, about 10% exceeded the RfD. The differences in these estimates can be attributed to the different exposure information provided by these two metrics and the uncertainties inherent in estimating the intake dose corresponding to each concentration. These differences are instructive of the limitations in the use of these exposure metrics. The blood Hg data spans four years. The lower estimate of the fraction of the population exceeding the RfD comes from the most recent data. There is, however, no clear statistical trend. It should be kept in mind that these summary estimates of national exposure may obscure significant regional differences, particularly differences in fish consumption. Data on hair Hg from pregnant women and women of childbearing age in New Jersey suggests that coastal areas have higher rates of fish consumption and correspondingly higher exposures to MeHg than the national average (26,34). Because the MeHg RfD was derived on the basis of a dose–response relationship that is specific for exposure during gestation, it does not relate specifically to the virtual safe dose for post-natal exposure. Nonetheless, it has been used as a convenient benchmark for assessing the relative magnitude of childrens’ exposure. At the 95th percentile, the highest percentile of the population distribution reported, childrens’ blood Hg concentrations did not exceed the concentration corresponding to the RfD in any of the years of the CDC data. In general, children’s blood

Exposure Assessment for Neurotoxic Metals

245

Hg concentration was 20–35% that of women of childbearing age. Likewise, childrens’ hair Hg concentration did not exceed the value corresponding to the RfD. Children’s hair concentrations were about 30–60% that of women of childbearing age with most values in the upper end of this range. It is not entirely clear why the childrens’ hair Hg concentrations were generally closer to the women’s values than were the children’s blood Hg concentrations. It is possible that this apparent discrepancy results from the fact that analysis of hair tends to reflect average exposure whereas blood analysis reflects a more current picture of the exposures status. For infrequent consumers of fish, hair samples will be less influenced than blood samples by a period of non-consumption concurrent with the time of sampling. Therefore, given that in general, children tend to consume fish less often than adults, the comparison of hair concentrations would tend to show better agreement between children’s exposure and adult exposure. Nonetheless, this discrepancy again points out the limitations and uncertainties inherent in the use of a single biomarker of MeHg exposure. The Key Cohort Studies There have been three cohort studies that have played a key role in the development of the current RfD for MeHg. In each study, fetal exposure was estimated on the basis of sampling at, or close to, delivery. A detailed discussion of exposure assessment in these studies can be found in the section on Dose Estimation in the NRC (10) report. In the New Zealand study (57,58), 9 cm of maternal hair closest to the scalp was sampled. The analysis relating hair total Hg concentration and performance on neuro developmental tests at 4 years and again at 6 years of age was based on the average concentration of Hg in the entire 9 cm sample. Measures of post-natal exposure were not investigated. In addition, for a subset of the mothers, the full length hair sample was segmented into 91 cm pieces, and the segmental analysis was used to give a rough estimate of the variability of exposure during gestation. As discussed previously, 9 cm of hair sampled at delivery represents approximately 9 months of exposure, but does not include the last portion of the third trimester. It does, however, include at least several weeks of maternal exposure prior to conception that may be only indirectly related to fetal exposure. In the Seychelles Child Development Study (SCDS) (41), 9 cm of hair was likewise sampled from mothers at delivery, and a second sample intended to reflect the same period of exposure was obtained from most subjects 6 months later. It is not entirely clear how or whether these samples were combined. In addition, the 9 cm hair sample was segmented based on an estimate of the portions representing the trimesters of gestation. In the initial assessments of the relationship between exposure and developmental performance at 6 mo, 1.6 years, and 2.4 years of age, the measure of exposure was the average concentration of total Hg in 9 cm of maternal hair. Follow-up investigations of this cohort at 5.5 years and 9 years of age each sampled the 1 cm of hair closest to the child’s scalp from at least a portion of the subjects as an estimate of concurrent exposure (59–62). Concurrent and maternal hair samples were used in statistical analysis of developmental test scores to assess the influence of gestational and post-natal MeHg exposure on performance. In the studies conducted in the Faroe Islands, both maternal hair and cord blood total Hg concentrations were employed as measures of gestational MeHg exposure (63). Children’s hair total Hg was also collected at 1 years and 7 years of age, and used to investigate the potential influence of pre-natal and post-natal MeHg exposure on performance at 7 years of age (64). The maternal hair samples were not of uniform length and varied from 3–9 cm. Umbilical cord blood total Hg collection was uniform, and straightforward. These measures of exposure were employed to investigate the

246

Stern

relationship between gestational exposure and neuro developmental test performance at 7 years of age. Subsequent assessments of this cohort at 7 and 14 years old for neurophysiological function, and autonomic cardiovascular parameters potentially affected during development employed concurrently sampled hair total Hg as measures of postnatal MeHg exposure as well as the original cord blood total Hg concentrations (65– 67). MeHg exposure in a second Faroese cohort has also been assessed using maternal hair, and cord blood total Hg (68). One of the critical questions in assessing the influence of MeHg on neurologic development is the potential importance of peak, or bolus, exposures compared to average exposure over the course of gestation. It is, therefore, unfortunate that detailed dietary assessments of MeHg intake for the cohorts in these key studies were either not conducted or (in the case of the New Zealand study) not reported. Given the current inability of both the hair and blood measures of exposure to precisely quantify short-term peak exposures in utero, dietary data would have been useful in identifying those children who were subject to such exposures. Hair or Blood? There is an ongoing discussion in the literature as to the most informative biomarker for investigating the relationship between gestational MeHg exposure and neurodevelopmental effects. In part, this discussion is driven by speculation about the role of biomarker selection in the apparent discrepancy between the results of the SCDS and the Faroes study. As discussed previously, the more distant two pharmacokinetic parameters are, the more variable will be the relationship between them. Thus, maternal hair would be expected to be the most precise measure of the maternal intake dose, whereas fetal cord blood would be the most precise measure of the dose to the fetal brain during the third trimester. When exposure during other periods of gestation are considered, however, the picture is less clear. Cernichiari et al. (69) argued that, based on comparisons of the concentration of total Hg in autopsy samples from infants of unspecified age, the proximal 1 cm of maternal hair total Hg was a better statistical predictor of infant brain total Hg than either maternal blood total Hg, or infant blood total Hg. This is taken as evidence that maternal hair Hg is a more precise measure of fetal brain MeHg exposure than is fetal cord blood Hg. The interpretation of these data is, however, uncertain. First, it should be noted that, to the extent that we are interested in the developing fetus, infant brain, and infant blood, are, at best, surrogates for the tissues of primary interest. Differences in physiology and developmental status make quantitative comparisons between fetal tissue and infant tissue uncertain. Second, the statistical basis for identifying maternal hair Hg rather than infant blood Hg as the better predictor of infant brain Hg is ambiguous. The correlation coefficients for blood and hair Hg with infant brain Hg are, in fact, similar. The error of the regression slope estimate for hair is much larger (as would be expected on the basis of increased pharmacokinetic variability), while at the same time, the coefficient of variation is smaller (10,69). Another way to look at this question is to examine the ability of each measure to explain outcomes of interest in neuro development performance. As discussed previously, in studies such as these, it is more likely that an observed association between a measure of exposure and a measure of performance reflects a true relationship than that such an association occurs merely by chance. In the Faroes study, both maternal hair and cord blood Hg concentration were investigated as dependent variables in regression analyses of neuro developmental test performance. These measures showed a very similar pattern of association with test scores. In most, but not all cases, cord blood Hg gave slightly larger regression coefficients and smaller p-values (10). The overall similarity in the picture of association suggests that both measures were reflective of a valid underlying

Exposure Assessment for Neurotoxic Metals

247

association. On the other hand, the stronger association with cord blood for most of the outcomes suggests that for these outcomes, cord blood was the more direct measure of the effects of MeHg on the fetal brain. However, it is important to keep in mind that hair and cord blood Hg do not measure the same exposures either temporally or pharmacokinetically. Thus, it is significant that for some of the tests, maternal hair provided the stronger association. We may hypothesize that the tests that are more strongly associated with hair Hg concentration reflect aspects of fetal neurodevelopment that are, in one or more ways, different from those that are more strongly associated with cord blood Hg. Unfortunately, the Faroes study was the only major cohort study to employ both hair and cord blood measures of MeHg exposure. It would be premature to conclude that cord blood Hg is, in general, the best measure for investigating the neuro developmental effects of gestational exposure to MeHg. Furthermore, it may well be the case that the “best” exposure measure will depend on the specific endpoints investigated. Given that the interaction of MeHg with neurodevelopment is both uncertain and likely to be complex, the most robust study designs will be those that employ multiple measures of MeHg exposure potentially including maternal and children’s hair; maternal, children’s, and cord blood; and dietary assessment. Dose Reconstruction Risk assessment for the neuro developmental effects of MeHg has developed in a manner that requires exposure assessment to function in two different ways. The first of these involves the estimation of the relationship between a biomarker (e.g., hair Hg, blood Hg) and a neurodevelopmental outcome. Based on the data from the Faroes cohort, the lower 95% percent confidence interval on the Hg concentration in cord blood corresponding to a doubling of the population in the lowest 5% of performance was identified using the benchmark dose approach (10,70). This concentration (58 mg/L Hg in the NRC analysis; 10) then served as a point-of-departure for the derivation of a RfD. Having thus related biomarker concentration to effect, it is then necessary to relate the biomarker concentration to the maternal intake dose in order to structure risk-based information in a way that lends itself to public health assessments and public health strategies. This has been referred to as “dose reconstruction.” (10). As discussed previously, relating tissue concentration to dose requires a pharmacokinetic model. This relationship is inherently variable in a population. For a fixed cord blood Hg concentration (i.e., the point-of-departure), the intake dose giving rise to that concentration is realistically described as a probability distribution. The pharmacokinetic model that describes this relationship should, therefore, be able to predict this variability. In order to do this, the input parameters to the model should, themselves, be characterized as probability distributions. Estimates of the variability in the MeHg maternal intake dose giving rise to 58 mg Hg/L in cord blood from a PBPK model and two independent analyses of the one-compartment model showed reasonably good agreement (71). However, these analyses showed less agreement when the central tendency estimates of the intake dose from these analyses were compared. In part, this reflected the fact that these analyses were not necessarily specific in selecting input data that were specific either to pregnancy in general or more specifically, to the third trimester of pregnancy (as the period most closely related to the concentration of Hg in cord blood). In addition, these analyses did not address differences between Hg concentration in maternal and cord blood. A recent re-analysis using the one-compartment model (23) has addressed these issues. The extent of variability estimated by this analysis is instructive. The estimate of the maternal MeHg

248

Stern

intake dose giving rise to 58 mg Hg/L in cord blood for the population at the lowest 5% of the range of doses (i.e., the 95th percentile of the population) was 37% of the estimated median dose in the population. For the population in the lowest 1% of the range of doses (i.e., the 99th percentile) the estimated dose was only 25% of the median dose. Interestingly, the largest single component of this variability was the variability in the relationship between maternal blood Hg and cord blood Hg concentration (24). Thus, as discussed previously, when the model must account for an additional transfer of MeHg between pharmacokinetic compartments, with a resulting increase in the pharmacokinetic distance between the dose and the biomarker tissue concentration, the variability in the relationship between the dose and the biomarker concentration, likewise, increases.

REFERENCES 1. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Lead. Atlanta, GA: Public Health Service, U.S. Department of Health and Human Services, 1999. 2. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Mercury. Atlanta, GA: Public Health Service, U.S. Department of Health and Human Services, 1999. 3. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Manganese. Atlanta, GA: Public Health Service, U.S. Department of Health and Human Services, 2000. 4. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Arsenic. Atlanta, GA: Public Health Service, U.S. Department of Health and Human Services, 2000. 5. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Cadmium. Atlanta, GA: Public Health Service, U.S. Department of Health and Human Services, 1999. 6. Wasserman GA, Liu X, Parvez F, et al. Water arsenic exposure and children’s intellectual function in Araihazar, Bangladesh. Environ Health Persp 2004; 112:1329–1333. 7. Skerfving S. Mercury in women exposed to methylmercury through fish consumption, and in their newborn babies and breast milk. Bull Environ Contam Toxicol 1988; 41:475–482. 8. Oskarsson A, Palminger Hallen I, Sundberg J. Exposure to toxic elements via breast milk. Analyst 1995; 120:765–770. 9. Urano T, Iwasaki A, Himeno S, Naganuma A, Imura N. Absorption of methylmercury compounds from rat intestine. Toxicol Lett 1990; 50:159–164. 10. National Research Council. Toxicological Effects of Methylmercury. Washington, DC: National Academy Press, 2000. 11. Cox C, Clarkson TW, Marsh DO, Amin-Zaki L, Tikriti S, Myers GG. Dose-response analysis of infants prenatally exposed to methyl mercury: an application of a single compartment model to single-strand hair analysis. Environ Res 1989; 49:318–332. 12. Magos L, Clarkson TW. Atomic absorption determination of total, inorganic, and organic mercury in blood. J Assoc Off Anal Chem 1972; 55:966–971. 13. Grandjean P. Reference intervals for toxic metals: problems and prospects. Abraham J. Gitlitz memorial lecture. Ann Clin Lab Sci 1986; 16:67–74. 14. Fergusson JE, Purchase NG. The analysis and levels of lead in human teeth: a review. Environ Pollut 1987; 46:11–44. 15. Gulson B, Wilson D. History of lead exposure in children revealed from isotopic analyses of teeth. Arch Environ Health 1994; 49:279–283. 16. Hopps HC. The biologic bases for using hair and nail for analyses of trace elements. Sci Total Environ 1977; 7:71–89. 17. Katz S, Chatt A. In: Hair Analysis: Application in the Biomedical and Environmental Sciences. New York: VCH Publishers, 1988:6–12.

Exposure Assessment for Neurotoxic Metals

249

18. Pecoraro V, Barman JM, Astore I. The normal trichogram of pregnant women. In: Montagnad W, Dobson RL, eds. Advances in Biology of Skin, Vol. IX, Hair Growth, Proceedings of the University of Oregon Medical School Symposium on the Biology of Skin. Oxford: Pergamon Press, 1967. 19. Kempson IM, Skinner WM. ToF-SIMS analysis of elemental distributions in human hair. Sci Total Environ 2005; 338:213–227. 20. MacIntosh DL, Williams PL, Hunter DJ, et al. Evaluation of a food frequency questionnairefood composition approach for estimating dietary intake of inorganic arsenic and methylmercury. Cancer Epidemiol Biomarkers Prev 1997; 6:1043–1050. 21. Rice D, Barone S, Jr. Critical periods of vulnerability for the developing nervous system: evidence from humans and animal models. Environ Health Persp 2000; 108:511–533. 22. Clewell HJ, Gearhart JM, Gentry PR, et al. Evaluation of the uncertainty in an oral reference dose for methylmercury due to interindividual variability in pharmacokinetics. Risk Anal 1999; 19:547–558. 23. Stern AH. A revised probabilistic estimate of the maternal methyl mercury intake dose corresponding to a measured cord blood mercury concentration. Environ Health Persp 2005; 113:155–163. 24. Stern AH, Smith AE. An assessment of the cord blood:maternal blood methylmercury ratio: implications for risk assessment. Environ Health Persp 2003; 111:1465–1470. 25. White PD, Van Leeuwen P, Davis BD. The conceptual structure of the integrated exposure uptake biokinetic model for lead in children. Environ Health Persp 1998; 106:1513–1530. 26. Knobeloch L, Anderson HA, Imm P, Peters D, Smith A. Fish consumption, advisory awareness, and hair mercury levels among women of childbearing age. Environ Res 2005; 97:220–227. 27. Stern AH, Korn LR, Ruppel BE. Estimation of fish consumption and methylmercury intake in the New Jersey population. J Expo Anal Environ Epidemiol 1996; 6:503–525. 28. Whipple C, Levin L, Seigneur C. Sensitivity of mercury exposure calculations to physical and biological parameters of fish consumption. Presented at Fourth International Conference on Mercury as a Global Pollutant, Hamburg, August 4–8. 29. U.S. Environmental Protection Agency. Mercury Study Report to Congress. Vol. IV: An Assessment of Exposure to Mercury in the United States EPA-452/R-97-006. U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, and Office of Research and Development, 1997. 30. Williamson DM, Choury E, Hilsdon R, Taylor B. Improving data quality in community-based seafood consumption studies by use of two measurement tools. J Environ Health 2004; 67:9–13. 31. Sherlock JC, Lindsay DG, Hislop JE, Evans WH, Collier TR. Duplicate diet study on mercury intake by fish consumers in the United Kingdom. Arch Environ Health 1982; 37:271–278. 32. Burger J, Stern AH, Gochfeld M. Mercury in commercial fish: optimizing individual choices to reduce risk. Environ Health Persp 2005; 113:266–271. 33. Thomas KW, Sheldon LS, Pellizzari ED, Handy RW, Roberds JM, Berry MR. Testing duplicate diet sample collection methods for measuring personal dietary exposures to chemical contaminants. J Exp Anal Environ Epidemiol 1997; 7:17–36. 34. Stern AH, Gochfeld M, Weisel C, Burger J. Mercury and methylmercury exposure in the New Jersey pregnant population. Arch Environ Health 2001; 56:4–10. 35. Legrand M, Arp P, Ritchie C, Chan HM. Mercury exposure in two coastal communities of the Bay of Fundy, Canada. Environ Res 2005; 98:14–21. 36. Stern AH. Estimation of the interindividual variability in the one-compartment pharmacokinetic model for methylmercury: implications for the derivation of a reference dose. Regul Toxicol Pharmacol 1997; 25:277–288. 37. Mahaffey KR, Clickner RP, Bodurow CC. Blood organic mercury and dietary mercury intake: National Health and Nutrition Examination Survey, 1999 and 2000. Environ Health Persp 2004; 112:562–570.

250

Stern

38. Vahter M, Akesson A, Lind B, Bjors U, Schutz A, Berglund M. Longitudinal study of methylmercury and inorganic mercury in blood and urine of pregnant and lactating women, as well as in umbilical cord blood. Environ Res 2000; 84:186–194. 39. Grandjean P, Weihe P. Neurobehavioral effects of interuterine mercury exposure; potential sources of bias. Environ Res 1993; 61:176–183. 40. Lindberg A, Bjornberg KA, Vahter M, Berglund M. Exposure to methylmercury in non-fisheating people in Sweden. Environ Res. 2004; 96:28–33. 41. Cernichiari E, Toribara TY, Liang L, et al. The biological monitoring of mercury in the Seychelles study. Neurotoxicol 1995; 16:613–628. 42. Lindow SW, Knight R, Batty J, Haswell SJ. Maternal and neonatal hair mercury concentrations: the effect of dental amalgam. BJOG 2003; 110:287–291. 43. Boischio AA, Henshel DS. Linear regression models of methyl mercury exposure during prenatal and early postnatal life among riverside people along the upper Madeira river, Amazon. Environ Res 2000; 83:150–161. 44. Rodushkin I, Axelsson MD. Application of double focusing sector field ICP-MS for multielemental characterization of human hair and nails. Part III. Direct analysis by laser ablation. Sci Total Enviro 2003; 305:23–39. 45. Oskarsson A, Schultz A, Skerfving S, Hallen IP, Ohlin B, Lagerkvist BJ. Total and inorganic mercury in breast milk in relation to fish consumption and amalgam in lactating women. Arch Environ Health 1996; 51:234–241. 46. Joshi AA, Douglass CW, Kim HD, et al. The relationship between amalgam restorations and mercury levels in male dentists and nondental health professionals. J Public Health Dent 2003; 63:52–60. 47. Suzuki T, Watanabe S, Matsuo N. Comparison of hair with nail as index media for biological monitoring of mercury. Sangyo Igaku 1989; 31:235–238. 48. Ritchie KA, Burke FJ, Gilmour WH, et al. Mercury vapour levels in dental practices and body mercury levels of dentists and controls. Br Dent J 2004; 197:625–632. 49. Guallar E, Sanz-Gallardo MI, van’t Veer P, et al. Heavy Metals and Myocardial Infarction Study Group. Mercury, fish oils, and the risk of myocardial infarction. N Engl J Med 2002; 347:1747–1754. 50. Yoshizawa K, Rimm EB, Morris JS, et al. Mercury and the risk of coronary heart disease in men. N Engl J Med 2002; 347:1755–1760. 51. U.S.EPA (2005). IRIS Database for Risk Assessment (Accessed March 31, 2005, at http:// www.epa.gov/iriswebp/iris/index.html). 52. Centers for Disease Control and Prevention. Blood and hair mercury levels in young children and women of childbearing age—United States, 1999. MMWR 2001; 50:140–143. 53. U.S. Centers for Disease Control and Prevention. Second National Report on Human Exposure to Environmental Chemicals, 2993 (Accessed March 31, 2005, at http://www.cdc.gov/ exposurereport/2nd/pdf/secondner.pdf). 54. U.S. Centers for Disease Control and Prevention. Blood mercury levels in young children and childbearing-aged women–United States, 1999–2002. MMWR 2004; 53:1018–1020. 55. Schober SE, Sinks TH, Jones RL, et al. Blood mercury levels in US children and women of childbearing age, 1999–2000. JAMA 2003; 289:1667–1674. 56. McDowell MA, Dillon CF, Osterloh J, et al. Hair mercury levels in U.S. children and women of childbearing age: reference range data from NHANES 1999–2000. Environ Health Persp 2004; 12:1165–1171. 57. Kjellstro¨m T, Kennedy P, Wallis S, Mantell C. Physical and Mental Development of Children with Prenatal Exposure to Mercury from Fish. Stage I: Preliminary Tests at Age 4. National Swedish Environmental Protection Board Report 3080. Solna, Sweden, 1986. 58. Kjellstro¨m T, Kennedy P, Wallis S, et al. Physical and Mental Development of Children with Prenatal Exposure to Mercury from Fish. National Swedish Environmental Protection Board Report No. 1989, 3642

Exposure Assessment for Neurotoxic Metals

251

59. Davidson PW, Myers GJ, Cox C, et al. Effects of prenatal and postnatal methylmercury exposure from fish consumption on neurodevelopment: outcomes at 66 months of age in the Seychelles Child Development Study. JAMA 1998; 280:701–707. 60. Myers GJ, Davidson PW, Palumbo D, et al. Secondary analysis from the Seychelles Child Development Study: the child behavior checklist. Environ Res 2000; 84:12–19. 61. Davidson PW, Palumbo D, Myers GJ, et al. Neurodevelopmental outcomes of Seychellois children from the pilot cohort at 108 months following prenatal exposure to methylmercury from a maternal fish diet. Environ Res 2000; 84:1–11. 62. Myers GJ, Davidson PW, Cox C, et al. Prenatal methylmercury exposure from ocean fish consumption in the Seychelles child development study. Lancet 2003; 361:1686–1692. 63. Grandjean P, Weihe P, Jorgensen PJ, Clarkson T, Cernichiari E, Videro T. Impact of maternal seafood diet on fetal exposure to mercury, selenium, and lead. Arch Environ Health 1992; 47:185–195. 64. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 65. Murata K, Weihe P, Budtz-Jorgensen E, Jorgensen PJ, Grandjean P. Delayed brainstem auditory evoked potential latencies in 14-year-old children exposed to methylmercury. J Pediatr 2004; 144:177–183. 66. Sorensen N, Murata K, Budtz-Jorgensen E, Weihe P, Grandjean P. Prenatal methylmercury exposure as a cardiovascular risk factor at seven years of age. Epidemiol 1999; 10:370–375. 67. Grandjean P, Murata K, Budtz-Jorgensen E, Weihe P. Cardiac autonomic activity in methylmercury neurotoxicity: 14-year follow-up of a Faroese birth cohort. J Pediatr 2004; 144:169–176. 68. Steuerwald U, Weihe P, Jorgensen PJ, et al. Maternal seafood diet, methylmercury exposure, and neonatal neurologic function. J Pediatr 2000; 136:599–605. 69. Cernichiari E, Brewer R, Myers GJ, et al. Monitoring methylmercury during pregnancy: maternal hair predicts fetal brain exposure. Neurotoxicol 1995; 16:705–710. 70. Rice DC, Schoeny R, Mahaffey K. Methods and rationale for derivation of a Reference Dose for methylmercury by the U.S.A. EPA. Risk Anal 2003; 23:107–115. 71. Stern AH, Clewell HJ, Swartout J. An objective uncertainty factor adjustment for methylmercury pharmacokinetic variability. Human Ecolog Risk Assess 2002; 8:885–894.

12 Studying the Relation Between Pesticide Exposure and Human Development Dana B. Barr National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia, U.S.A.

Asa Bradman Center for Children’s Environmental Health, School of Public Health, University of California, Berkeley, California, U.S.A.

Natalie Freeman Center for Environmental and Human Toxicology, College of Veterinary Medicine and College of Public Health and Health Professions, University of Florida, Gainesville, Florida, U.S.A.

Robin M. Whyatt Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, U.S.A.

Richard Y. Wang National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, Georgia, U.S.A.

Luke Naeher College of Public Health, University of Georgia, Athens, Georgia, U.S.A.

Brenda Eskenazi Center for Children’s Environmental Health, School of Public Health, University of California, Berkeley, California, U.S.A.

INTRODUCTION Approximately 888 million pounds of conventional pesticides (excluding chlorine disinfectants, specialty biocides) are used annually in the United States (2). An additional 800 million pounds of wood preservatives are used annually (2). Approximately 78% are used in agriculture; 10% in homes and gardens; and the remainder in government, commercial or industrial applications. Herbicides constitute the bulk of conventional pesticides used (44%), and insecticides (10%), fungicides (6%) and other insecticides (40%) make up the remainder (2). In 2001, approximately 73 million pounds of organophosphorus (OP) insecticides were used annually (2). Although some agricultural 253

254

Barr et al.

uses of OP insecticides were restricted in 2001, overall use in this sector has not changed in recent years. As a percentage of total insecticide use, OP insecticide use increased from 58 to 70% between 1980 and 2001 (2). However, residential uses of chlorpyrifos and diazinon, two common OP insecticides, were eliminated in 2001 and 2002, respectively (3), although they remain widely used in agriculture. Malathion and acephate remain widely used residential OP insecticides. Carbamate insecticides also are used widely in both residential and agricultural applications. Synthetic pyrethroids, and to some extent nicotinyl insecticides, have largely replaced residential uses of OP insecticides and are now the dominant class of insecticides used in homes and gardens (4). Before the burgeoning use of OP, carbamate, and pyrethroid insecticides in the last three decades, organochlorine (OC) insecticides were used widely. Some OC insecticides used in the past include p,p 0 -dichlorodiphenyltrichloroethane (p,p 0 -DDT), dieldrin, and several isomers of hexachlorocyclohexane. Because of their detrimental effects on wildlife (5–10) and suspected harmful effects in humans (11), most OC insecticides were banned for use in the United States and many other developed countries by the 1980s. However, many of the OC insecticides persist for years in the environment and human tissue (12–16), and thus are still of concern. Only a few OC pesticides, including lindane and endosulfan, are used in U.S. agriculture. The current widespread use of pesticides and the persistent nature of pesticides used earlier have resulted in their widespread exposure to the U.S. population. For example, insecticide contamination in residential environments, including air, dust, and surfaces, has been documented in a variety of urban and rural environments (17–19). Detectable insecticide residues were found in approximately 47% of the fruit and vegetable samples tested as part of market-basket surveys by the USDA in 2002 (20). Finally, biomonitoring studies have shown that young children, pregnant women, and fetuses are directly exposed to insecticides on a regular basis (19,21–31), including OC, OP, and pyrethroid insecticides, all of which are neurotoxic chemicals. Developing fetuses and young children are a uniquely vulnerable population. Young children spend most of their time indoors (32), have extensive hand-to-mouth contact (33), and are often on dusty floors or in other microenvironments contaminated with pesticides. Young children also eat, drink, and breathe more air per unit of body weight than do adults (34). This combination of behavior and physiology results in higher exposures to young children than to adults in the same environment. The fetus can be exposed in utero by transplacental transfer (19). It is more susceptible to the neurotoxic effects of pesticide exposures (34). For example, levels of cholinesterase and paraoxonase, enzymes which modify the toxicity of OP insecticides, are much lower in umbilical cord blood (35–38), suggesting that very young children are more vulnerable to these exposures. Finally, fetuses and children develop rapidly, with critical developmental windows when they are particularly vulnerable to insecticide exposures. Studying the relation between pesticide exposure and disease outcomes in children is challenging, especially when research focuses on current-use pesticides that involve transient exposures by compounds that have short half-lives in the body (39–41). For example, organophosphate and pyrethroids are metabolized and excreted in the urine within hours to days after an exposure (4,42–47). Thus, in epidemiologic studies, exposures must be assessed frequently during the prenatal, perinatal, and early postnatal periods (39,41). In this chapter, we provide a brief overview of neurotoxic effects of insecticides and then focus on the exposure assessment issues that must be considered in epidemiologic studies of insecticides and neurobehavioral outcomes in young children. We focus on currently used insecticides (Table 1), primarily OP, carbamate, and pyrethroid insecticides; however, we also include information about some banned

Pesticide Exposure and Human Development Table 1

255

Selected Neurotoxic Insecticides and Their Primary Mode of Action

Mode of action

Insecticide

Modulates function of DDT and metabolites voltage-gated sodium channels (DDE, DDD) Permethrin Cypermethrin Deltamethrin Resmethrin Allethrin Bioallethrin Cyfluthrin Fenvalerate Esfenvalerate Sumithrin AChE inhibitor Azinphos methyl Chlorethoxyphos Chlorpyrifos Chlorpyrifos methyl Coumaphos Dichlorvos Diazinon Dicrotophos Dimethoate Disulfoton Ethion Fenitrothion Fenthion Isazaphos-methyl Malathion Methidathion Methyl parathion Naled Nitrofen Oxydemeton-methyl Parathion Phorate Phosmet Pirimiphos-methyl Sulfotepp Temephos Terbufos Tetrachlorviphos Carbaryl Propoxur Carbofuran Benfuracarb Carbosulfan Furathiocarb Pirimicarb Bendiocarb Aldicarb Methomyl

Class Organochlorine pesticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Pyrethroid insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Organophosphorus insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides Carbamate insecticides (Continued)

256

Barr et al.

Table 1 Selected Neurotoxic Insecticides and Their Primary Mode of Action (Continued) Mode of action Modified function of nicotinic ACh receptors Blocks GABA receptor

Insecticide Imidacloprid Acetamiprid Nitenpyram Fipronil

Class Nicotinyl insecticide Nicotinyl insecticide Nicotinyl insecticide Phenylpyrazole insecticide

Abbreviations: ACh, acetylcholine; AChE, acetylcholinesterase; GABA, gamma-aminobutyric acid.

OC insecticides (Table 1) because the potential for exposure to these insecticides or their degradates remains. TOXICOLOGY OP and Carbamate Insecticides The acute toxic effects of OP and carbamate insecticides result from the inhibition of acetyl cholinesterase (AChE) in the nervous system, causing an accumulation of the neurotransmitter acetylcholine at neuronal junctions that leads to continued stimulation then suppression of neurotransmission (43). Neither carbamate nor OP insecticides irreversibly inhibit AChE unless the bound moiety is aged by the loss of alkyl side chains; however, the duration of inhibition by OP insecticides is longer than that for carbamate insecticides (43). Among adults with exposure to OP and carbamate insecticides in occupational settings, the neurotoxic effects of exposure have been well documented (48–54). The neurotoxic effects consist of suppression of cholinesterase activity in the central nervous system and peripheral nervous system, the development of progressive peripheral neuropathies caused by the phosphorylation and inhibition of neurotoxic esterase, and finally respiratory paralysis and limb ataxia (49). Symptoms associated with acute, high-dose exposures include reduction in blood cholinesterase activity leading to muscarinic and nicotinic manifestations, which can be demonstrated as parasthesias, disturbances of visual function, cognitive difficulties, anxiety, muscle weakness, numbness, ataxia, paralysis, and respiratory distress. Long-term effects following acute exposure include muscular atrophy of the ataxic limbs (49,51), short-term memory loss, delayed polyneuropathies, and Parkinsonism (50,52,53). Responses to low-dose, chronic exposures tend to be more diffuse and nonspecific, including headaches, fatigue, anxiety, emotional reactivity, drowsiness, insomnia, loss of concentration and attention span, impaired memory, and visual function (50,52,54). During development, however, neurologic effects of OP and carbamate insecticide exposure, even at low levels, may be more detrimental (55). Modulation of acetylchloline levels in utero during nervous system development may be of particular concern. Neurotransmitters, including acetylcholine, play essential roles in the cellular and architectural development of the brain (56). Moreover, pregnancy is a time of increased risk because plasma AChE activity is already reduced during the first two trimesters (57,58). Depending on the stage of development, the same neurotransmitter can promote cell replication, elicit a switch from replication to differentiation, promote or arrest cell growth, evoke apoptosis or program genes that determine future responsiveness of the cell to external stimulation (56). Experimental data confirm that AChE inhibition disrupts these processes, including effects on cell replication and differentiation, axonogenesis and synaptogenesis (59–62). Finally, these processes can also be disrupted through

Pesticide Exposure and Human Development

257

noncholinergic mechanisms that involve alterations in the expression and function of nuclear transcription factors (63). Although immature animals recover more rapidly from AChE inhibition (64), tests on young rodents show greater susceptibility to OP insecticides that decreases with age (34,65–67). In several animal studies, the lethal dose in immature animals was up to two orders of magnitude lower than the adult dose (68–70). Young animals may be susceptible because of lower activity of detoxifying enzymes (paraoxonase or chlorpyrifos-oxonase) that deactivate OP metabolites (37,65,71–75). In one human study, Berkowitz et al. (76) reported an association between chlorpyrifos metabolites in maternal urine samples during pregnancy and infant head circumference at birth among infants of mothers with low paraxonase activity (76). In addition, polymorphisms in other metabolically important genes (e.g., various P450 enzymes, GSTM1, GSTP1) may be relevant to toxicities associated with early-life exposures to insecticides. Nervous system ontogeny can be divided into: (1) early brain development (i.e., first trimester in humans); and (2) the brain growth spurt (i.e., third trimester through the first two years of life in humans, when there is extensive axonal and dendritic growth, synaptogenesis, proliferation of glial cells, and myelinization) (56,61–64,77,78). Toxicants introduced during the early period can cause malformations, whereas those introduced in the later period can result in behavioral and cognitive changes, such as in learning and memory (79–81). In rodents, the cholinergic transmitter system undergoes rapid development during the later phase, the first 3–4 wk after birth (82). Given the diversity of mechanisms and target tissues, the developing brain is likely to be vulnerable from early embryonic life into childhood (55,56,63,83,84). Many OP insecticides are reasonably lipophilic and readily cross the placental barrier (19,85). Considerable evidence in animals links exposure to OP insecticides in utero or the early postnatal period with adverse neurodevelopment (86,87). Pyrethroid Insecticides and DDT Other major insecticide classes that are replacing home and some agricultural uses of OP and carbamate insecticides are natural pyrethrins and synthetic pyrethroids (4). Pyrethrins are naturally occurring insecticides found in a variety of plants such as chrysanthemums (4). Synthetic pyrethroids are human-made derivatives of pyrethrins, designed to be more chemically potent and environmentally stable (88). Like most other classes of insecticides, the pyrethroids are acute neurotoxicants (89–91), and although dissimilar in structure and environmental persistence to DDT, their modes of action are strikingly similar (89,92). Pyrethroid insecticides and DDT modulate the function of voltage-gated sodium channels (93). Specifically, they alter the permeability of excited nerve cells to sodium ions and cause repetitive nerve impulses which can vary between a few dozen for the less toxic noncyano substituted pyrethroids (Type I pyrethroids) to up to a thousand for the more toxic cyano substituted pyrethroids (Type II pyrethroids) (89–91). They also have other neurobiologic actions, including affects on central g-amino butyric acid (GABA), noradrenergic and dopaminergic or chlolinergic neurotransmission (94). In general, pyrethroids are considered among the lower toxicity insecticides, in part because mammals have higher levels of the enzymes that detoxify pyrethroids than do insects, and pyrethroids are rapidly metabolized and excreted in mammalian systems (95). However, detoxification enzymes (e.g., carboxyesterases) involved in metabolism of pyrethroids are much lower during fetal and early postnatal development than they are later in life (96,97), suggesting that young children may be more susceptible to adverse effects than adults.

258

Barr et al.

In humans, pyrethroid insecticides are metabolized by esterases, mainly in the liver, and the metabolites are renally eliminated with a half-life of about 6 hr (98). Acute high doses of pyrethroid insecticides can cause nervous system effects such as lack of coordination, tremors, vomiting, diarrhea, and irritability. Pyrethroid insecticides can also be skin and respiratory irritants (4). Experimental data in laboratory rodents have shown that the pyrethroid esfenvalerate can cause short-term behavioral effects in adult rodents (94). Other Neurotoxic Insecticides New insecticides are continually being developed to overcome many problems, such as mammalian toxicity and pest resistance, encountered with previous insecticides. Two insecticide classes introduced in the last decade include phenylpyrazole and nicotinyl insecticides (95). These insecticides are selectively more toxic to insects than to mammals. Phenylpyrazole insecticides, such as fipronil, have been shown to block insect GABA receptors which differ from those of vertebrates (99). Nicotinyl insecticides, such as imidacloprid, block the nicotinergic neuronal pathway that is more abundant in insects than in mammals (95). This blockage leads to the accumulation of acetylcholine, resulting in the insect’s paralysis, and eventually death. HUMAN STUDIES ON THE NEUROTOXIC EFFECTS OF PESTICIDES IN CHILDREN As reviewed above, numerous animal studies have demonstrated the potential for in utero or early insecticide exposure to impact human neurodevelopment (86). In addition, acute poisonings in children result in neurologic impairment (100). However, few studies have assessed the neurobehavioral functioning of infants and children who have been exposed to either low levels of OC (Table 2) or OP (Table 3) insecticides (101–106) and no studies that have examined the impact of other insecticide groups such as carbamates or pyrethroids. For the most part, the few studies that have been conducted are inconclusive because of methodologic limitations or the unknown long-term impact of early findings on later neurodevelopment or because they have not yet been replicated. OC Insecticides In an early investigation of a large North Carolina birth cohort with 1980s background exposure, Rogan et al. (101) demonstrated that p,p 0 -dichlorodiphenyldichloroethene (DDE) levels derived from measurements in cord serum and in breast milk were doserelated to hyporeflexia noted on the Brazelton Neonatal Behavioral Assessment Scale (BNAS). The infants were assessed once and most were assessed within the first week but up to one month of life. The long-term implications of abnormal reflexes on neurodevelopment are not known. In addition, these findings were not confirmed by Stewart et al. (107) in an Oswego birth cohort of neonates evaluated twice on the Brazelton scale in the first two days of life. However, this study was considerably smaller than the North Carolina study. Also, the Oswego infants likely had lower exposure to DDE given the later date of study (DDT was banned in the United States in 1972). In a follow-up study of the North Carolina birth cohort, DDE serum levels were associated with better performance on the Mental Developmental Index (MDI) of the Bayley Scales of Infant Assessment in the 6 mo olds but not when they were 12 mo old (108) and on the Psychomotor Developmental Index (PDI) at 18 and 24 mo (109). There was no

Rogan and Gladen. (1991)

Gladen et al. (1988)

Rogan et al. (1986)

Exposurea Outcome

670 children age 24 mo

676 children age 18 mo

Birth cohort from three health centers in North Carolina (same as above)

720 infants age 12 mo

DDE Placenta, maternal and cord Bayley Scales (BSID) serum and milk/colostrums combined into one measure/ woman MDI

Prenatal: placenta, maternal and Mental Development (MDI) cord serum, and milk at birth combined into one measure/ woman Postnatal: cumulative exposure at Psychomotor Development (PDI) time of evaluation based on concentration in milk and duration of breast-feeding

912 neonates %1 mo old DDE Placenta, maternal and cord Brazelton Neonatal Behavioral Assessment Scale (BNBAS) serum and milk/colostrums combined into one measure/ woman Birth cohort from three (Medianw2 mg/g milk lipid or health centers in North w50 mg/mL milk) Carolina Bayley Scales of Infant 786 infants age 6 mo DDE Development (BSID)

Population

Neurotoxic Effects of Organochlorine Insecticide Exposure in Children

Author, year

Table 2

[ PDI at 24 mo (Continued)

Postnatal DDE exposure: No associations at 6 or 12 mo with PDI or MDI No association at 18 mo with PDI or MDI

No association observed with PDI and transplacental exposure

Prenatal DDE exposure. [ MDI at 6 mo, but association disappears by 12-mo assessment

[Dose-related number of abnormal reflexes (hyporeflexia) whigher scores on regulation-of-states

Results

Pesticide Exposure and Human Development 259

Darvill et al. (2000)

Gladen and Rogan. (1991)

Author, year

Exposurea Outcome

219 at 12 mo old

PDI Birth cohort from three health centers in North Carolina (same as above) 712 children assessed at DDE Placenta, maternal and cord McCarthy scales of children’s abilities at 3, 4, or 5 yr 3, 4, or 5 yr serum and milk/colostrums combined into one measure/ woman Grades in English and math506 children up to age 10 1 ematics from report cards ⁄2 yr from grade 3 or later Birth cohort from three health centers in North Carolina (same as above) Fagan test of infant intelligence Same cohort as in Cord serum levels DDE, HCB, Stewart et al. Mirex; breast milk on a subgroup 230 at 6 mo old (The primary exposure was PCB)

Population

Table 2 Neurotoxic Effects of Organochlorine Insecticide Exposure in Children (Continued)

At 12 mo, inverse correlation, rZ-.14, pZ.03 (Authors report that there is no association)

At 6 mo, no association with DDE (HCB and Mirex not presented)

No association with grades

No association with McCarthy scores

No association with MDI at 24 mo

Results

260 Barr et al.

a

nZ92 infants 13 mo olds Cord serum DDE (medianZ 0.85 ng/mL serum or w145 mg/g lipid) and HCB Birth cohort living near an electrochemical factory in Spain

YGriffiths Scales: locomotor, personal-social, performance and eye-hand coordination Associations strongest if shorter period of breastfeeding HCB No association

2X dose of DDE/ Y3.5 points MDI Y 4.0 points PDI

Mental Development-MDI

Psychomotor DevelopmentPDI Griffiths Scales of Infant Development

p, p 0 , DDE YMDI and PDI-dose-related

Bayley scales of infant development

Conversions from whole weight to lipid adjusted values and vice versa were performed assuming an average milk lipid content of 4% and an average serum lipid content of 0.6%. The average milk lipid content is for mature milk; colostrum lipid content is lower.

Ribas-Fito´ et al. (2003)

Pesticide Exposure and Human Development 261

Farmworkers work in strawberries

Latino adolescents in Oregon

Occupation Farmworkers pre vs post season (6–36 days after pre-season evaluation)

Exposure based on:

Rohlman et al. nZ96 migrant farmworkers (2001)

nZ51 non farmworkers (NFW) 13–18 yr old (Farmworkers are significantly older and less educated and more males)

Exposure

Population

Neurotoxic Effects of Organophosphorus Insecticide Exposure in Children

Author, year

Table 3

Progressive ratio test Selective attention test Serial digit learning Continuous performance test

Simple reaction time Digit span

Behavioral assessment and Research System (BARS): Finger tapping Symbol-digit

Outcome

All showed improvement Authors state results likely confounded by cultural and educational differences in groups, and by practice effects

Pre vs. post season (66 retested)

Preseason vs NFW Yon cognitive tests (digit symbol, digit learning, digit span) Z motor tests Postseason vs. NFW Z on all tests

Pre or post season farmworkers vs. non farmworkers (once only)

Results

262 Barr et al.

Urinary p-nitro phenol (PNP) R100 ng/mL for one household member OR

Exposed in Mississippi:

PNPR300 ng/mL OR

Household wipe measures of methyl parathion (MP) R150 mg/100 cm2 Communities in Mississippi and Ohio where Exposed in Ohio: illegal household spraying for pest control was known to have occurred. Urinary PNP R100 ng/mL for one household member OR Household MP R132.9 mg/100 cm2 High exposure:

MZ6 yr old (rangeZ2–11)

nZ147 unexposed

Zeitz-Ruckart nZ132 exposed et al. (2004)

Y Motor skills (VABS)

Story Memory Immediate and Delayed (WRAML) Trail-Making test (O9 yo only) Verbal Cancellation Test Personality Inventory for Children

(Continued)

Problems with sample size due to attrition and age appropriate assessment

[ Undisicplined behavior and other behavioral problems Findings not consistent across sites

w Fine visuo-motor (purdue)

More of the exposed group improved at retest Z General intelligence, visual-motor integration, multistep processing (K-BIT, VMI and trail-making) Y Selective attention (verbal cancellation) Y Verbal delayed memory (story)

Z proportion classified as below expected

Purdue Pegboard

Kaufman Brief Intelligence Test (K-BIT)

Pediatric Environmental Neurobehavioral Test Battery (PENTB): Developmental Test of VisualMotor Integration (VMI)

Pesticide Exposure and Human Development 263

Young et al. (2005)

Author, year 2

MP R1000 mg/100 cm

Exposure

Unexposed from same communities with no exposure based on State records Maternal urinary dialkyl nZ381 ! 2 mo olds CHAMACOS birth phosphate metabolite cohort living in an agricultural community; levels during pregnancy mostly latino farmworker families in (MZ14 and 26 wk) and Salinas Valley California early postnatal (MZ 7 days) CHAMACOS birth cohort living in an agricultural community; mostly latino farmworker families in Salinas Valley California

Population

Outcome

BNBAS

Retested 1 yr later

Vineland Adaptive Behavior Scales

Table 3 Neurotoxic Effects of Organophosphorus Insecticide Exposure in Children (Continued)

No association with early postnatal exposure measures

[Dose-related number of abnormal reflexes (hyporeflexia) with in utero measure especially in infants O3 days at testing

Authors state results are not conclusive subacute exposure may be related to transient associations

Results

264 Barr et al.

Pesticide Exposure and Human Development

265

association of transplacental or breastfeeding exposure to DDE on the McCarthy Scales of Children’s Abilities at 3,4, or 5 years or school performance at 8 to 101⁄2 years (110). These findings from North Carolina differ somewhat from findings from other cohorts. In a follow-up study of a subset of the Oswego infants at 6 and 12 mo (102), the authors report no association at either time point of cord serum DDE levels with performance on the Fagan Test of Infant Intelligence (a novelty preference task with moderate predictive validity on later measures of psychometric intelligence), although they present a significant negative correlation between DDE levels and performance at 12 mo. This negative correlation is in line with a study of 92 infants from a community in Spain with high levels of atmospheric hexachlorobenzene (HCB) (103). Although no association existed with HCB, cord serum DDE levels were significantly negatively associated with cognitive, psychomotor, and social development of 13-mo-old children assessed on the Bayley and the Griffiths Scales of Infant Development. Similar findings were observed among 4- to 5-yr-old children participating in a small ecologic study in an indigenous community in Mexico (104). In that study, children living in the valley where a variety of OC, OP, and pyrethroid insecticides were used in farming were compared with those living on the foothills where farming was done without insecticides. Although no biomarkers of exposure were used in this study, a previous study documented high exposure to OC insecticides including DDE. Compared to the children from the foothills, the children living in the valley demonstrated poorer motor stamina and skill, eye-hand coordination, delayed memory, and drawing ability and had more aggressive behavior. OP Insecticides Recent studies using biomarkers of exposure have found that in utero exposure to OP insecticides may impact birth weight and length (29,31), length of gestation (111), and in some susceptible subpopulations, other measures of growth such as head circumference (76). These studies suggest that low levels of OP insecticide exposure have the potential to impact other aspects of development. However, few studies have been published that specifically assessed the neurodevelopment of children exposed to OP insecticides. In neonates from the CHAMACOS birth cohort study of primarily Latino agricultural families, Young et al. (112) reported that maternal dialkyl phosphate urinary metabolite levels during pregnancy were associated with an increase in number of abnormal reflexes on the BNAS. Effects were seen particularly in those who were assessed after three days following birth. These associations, which were dose-related, were similar to the findings of Rogan et al. (101) observed in relation to DDE exposure. In a study of older children (ages 2 to11 yr) living in communities in Mississippi and Ohio where methyl parathion was sprayed, Ruckart et al. (106) reported that in comparison with unexposed children, exposed children had decreased selective attention, delayed verbal memory, and poorer motor skills as well as behavioral problems. Exposure to methyl parathion was based on urinary measurements in household members and dust wipe samples. However, the results were not consistent in all analyses and at both sites; hence, the authors concluded that the results suggested transient associations but were inconclusive. In another inconclusive study of migrant farmworker adolescents, Rohlman et al. (105) showed that, in comparison to non- farmworking adolescents, farmworker adolescents performed poorly on a number of tests even prior to the season when exposure would have occurred. The farmworker adolescents showed substantial improvement when assessed post season, presumably after exposure. This study was impacted by educational, age, and sex differences in the groups and substantial practice effects and loss to follow-up.

266

Barr et al.

In summary, few studies in humans have examined the potential neurodevelopmental impact of chronic low-level insecticide exposure. The limited data, however, suggest some association of both OC insecticide exposure, namely DDE, and OP insecticide exposure with neonatal reflexes, although the clinical significance of this observation is unknown. Some data also suggest an impact on later neurodevelopment, but the data are too sparse to make any conclusions.

FACTORS AFFECTING FETAL, INFANT, AND EARLY CHILDHOOD EXPOSURE Factors Affecting Fetal Exposures For many compounds such as lead, a clear relation exists between maternal and fetal exposures (113). Pesticide researchers have assumed that maternal biomonitoring measurements during pregnancy indicate exposures to the fetus. Several recent studies that indicate direct exposures to the fetus tend to confirm this presumption (19,28,30). In utero pesticide exposures are a central focus of several current cohort studies (19,29,31,76,111,114) and will be part of the planned National Children’s Study (115). Thus, researchers need to understand physiologic changes during pregnancy that affect the distribution of pesticides in the body. For example, plasma volume, body water, cardiac output, and circulating hormones increase during pregnancy (116). These changes can alter the steady-state concentration of pesticides in the body or their toxicity. Chemicals absorbed into the body are distributed to highly perfused organs and extravascular tissues. The extent to which a chemical distributes to extravascular tissues and compartments depends on the body mass of the individual and on the chemical’s lipid solubility, ionic state, and protein-binding capacity. Lipophilic chemicals, such as OC insecticides, readily distribute to adipose tissue. These particular insecticides tend to accumulate in this compartment because they are slowly metabolized by the body, and this tissue is poorly perfused by blood. Total body water increases about 50% during pregnancy and contributes to the majority of the increase in body mass during this period (117). This causes increased plasma volume, which leads to lower plasma concentration of the chemical as pregnancy progresses and the exposure-dose remains constant or unchanged (118). This effect is most apparent with chemicals that are water soluble, which would apply to many of the pesticide metabolites. Another cause for lowered plasma chemical concentration during pregnancy is increased renal clearance. Renal blood flow increases by 25% to 50% at the end of the first trimester because of increased cardiac output. This promotes the elimination of chemicals in the urine by glomerular filtration and leads to a lower plasma concentration of the chemical (119). Thus, if a dose-dependent health effect is anticipated for a chemical, then the response is unlikely to be maintained during the gestational period if the exposure-dose remains constant. The distribution of chemicals across the placenta is primarily by passive diffusion, which is determined by the ionic state and concentration gradient of the chemical. Because fetal blood pH (w7.00–7.20) is slightly more acidic than maternal blood pH (7.40), pesticides that are weak bases will become ionized and trapped in the fetal circulation, leading to higher levels in cord blood than in maternal blood. The opposite effect would occur with chemicals that are weak acids (120). Most OP, carbamate, or pyrethroid insecticide metabolites are weak acids and are unlikely to become “ion trapped” in fetal circulation. This mechanism alone can contribute to variations in the levels of these metabolites measured in cord blood. In addition, the “ion trapping” of metabolites

Pesticide Exposure and Human Development

267

behaving as weak bases and with biologically toxic activity could potentially have detrimental effects on the fetus. Lipophilic chemicals in the non-ionized state can passively diffuse across the placenta and into the fetal circulation. Because many currentuse pesticides are somewhat lipophilic—for example, chlorpyifos has a logKow of about four (121)—this may be another process by which these chemicals may enter the fetal circulation. The amount of blood flow to the placenta is another critical variable determining the distribution of chemicals, particularly lipophilic chemicals, from maternal to fetal circulation (122). Some factors causing decreased placental blood flow include maternal hypotension, toxemia, and pre-eclampsia, and use of vasoconstricting drugs or chemicals (e.g., sympathomimetic agents) (123,124). Finally, the plasma protein-binding capacity of the pesticide can determine transfer across the placenta (118). The binding to plasma proteins by a chemical limits its ability to distribute to extravascular compartments, to exert its biological activity at the site of action, and to be eliminated by glomerular filtration. During pregnancy, two changes occur that concern plasma proteins: a decrease in plasma protein concentration caused by increased plasma volume and a decrease in protein binding sites because of saturation by increased hormone levels. The latter effect increases protein-unbound chemical levels (125). Chemicals that bind to plasma proteins with minimal affinity will cross the placenta and be bound by fetal protein and accumulate in that compartment, leading to levels higher than those measured in the maternal compartment. This effect is expected to be greatest after 37 wk of gestation because fetal plasma albumin levels are highest at that time. Chemicals with a high protein binding capacity (e.g., chlorpyrifos) are most likely to be affected by changes in plasma protein levels. Further research is needed to characterize the extent of protein binding affinity for currently used pesticides. Most current-use pesticides are rapidly metabolized in the body by the hepatic microsomal mixed-function oxidase system, although blood and other tissues may contribute to this process as well (126). For most of these chemicals, this process yields metabolites that are not biologically active. However, for the phosphorothionate insecticides (e.g., chlorpyrifos, parathion, fenthion, dimethoate, fenitrothion), the oxon analogue metabolite is much more biologically active at the cholinesterase site than the parent chemical (43). Similarly, carbamate metabolites that maintain the amide bond will have anticholinesterase activity like the parent chemical. For pyrethroids, geometric isomers are metabolized at different rates and have variable levels of biological activity. In general, the cis isomers are metabolized by oxidases, which is a slower process than the esterase-mediated hydrolysis of the trans isomers. In developing systems, chemical metabolism can be delayed because of immaturity. Feto-placental metabolism of chemicals can occur at either the placenta or fetal liver (127). Although fetal hepatic chemical metabolism begins within the first trimester of gestation, it is believed to have a small contribution to the biotransformation of chemicals because of the immature state of the liver. Thus, chemicals can persist for longer periods in the fetal circulation than in the maternal circulation. The significance of these findings to health assessment studies is that the biologically active chemical most proximate to the health effect of interest should be selected as the exposure biomarker; otherwise, the ability to identify such associations may be complicated by increased variability in response. Factors Affecting Childhood Exposures Efforts have been under way by the United States Environmental Protection Agency (U.S. EPA) to standardize childhood age-categories on the basis of behavioral and physiological characteristics for harmonizing exposure and risk assessments across all branches of

268

Barr et al.

the agency (128,129). Life stages can be identified for the development of physiologic systems and their changing sensitivity to toxicants (130). As noted above, sensitivity to neurotoxicants may be influenced not only by the stage of neurologic development, but also by metabolic processes that influence how the neurotoxicant is processed by the body. Similarly, behavioral life stages can be characterized by age-specific behavior patterns. Of particular interest for neurotoxicants are mouthing behaviors that peak in frequency and duration when infants are teething. These behaviors may, however, have their most significant influence on the ingestion of neurotoxicants during toddlerhood when the child is relatively independent of the caretaker, is mobile, and has not yet developed good hygiene habits. The mantra that children are not little adults is particularly true in interpreting and understanding exposure data. Identification of the factors that contribute to exposure requires understanding how children at different stages of development differ in activities and activity levels and sleeping and eating patterns and places they spend time (131–134). For approximately nine months of the year, school-aged children’s lives are dramatically influenced by their attendance in school. It is where they spend most of their waking hours during the work week and where they may have many of their insecticide exposures. For younger children, where the children spend time is driven by a variety of factors including whether the mother works, the family’s level of affluence, the family structure, and community characteristics (132). In the National Human Activity Pattern Survey (NHAPS) (133), exposure and activity pattern data were obtained for individuals of all ages in a population-based study. NHAPS found that household insecticide use was equally likely for all age groups (Table 4). At the same time children as a group spent more time on the floor than did adults, and among all children, time spent on floors decreased with age. The youngest children were least likely to wash hands and most likely to contact floors, dirt, and grass—potential sources of insecticide exposure. Children differ by age and by sex (Table 4). In an elaboration of NHAPS on dirt and grass contact potential, Wong et al. (135) studied 2- to 13-yr-olds and found both sex- and activity-driven differences in grass and dirt contact, with boys having greater contact than girls, and sports players having more contact than nonplayers of sports. The dirt/grass contact behaviors of children aged less than two years and teenagers did not appear to be Table 4 Age-Specific Changes in Potential Exposure Behaviors in Children: Percentage of Individuals Exhibiting Activity Age group (years) Activity Use of insecticides in homes to rid of pests Wash hands%twice yesterday Shower/bathe yesterday Spend time on floor Play outdoors Play in sand/dirt/grass Less than 30 min on dirt yesterday (doers) Less than 30 min on grass yesterday (doers) N NZnumber of participants. Source: From Ref. 133.

1–4

5–11

12–17

18–64

39.3% 29.3% 76.2% 86.6% 55.5% 43.3% 71.7% 41.2% 263

43.0% 19.0% 15.6% 44.1% 34.1% 28.4% 70.0% 31.5% 348

43.1% 16.0% 40.2% 9.5% 7.8% 7.0% 82.9% 39.0% 326

43.9% 4.6% 82.8% 7.7% 5.2% 3.9% 85.7% 38.4% 2972

Pesticide Exposure and Human Development

269

sports driven. Sex differences in outdoor activities including grass and dirt contact, also have been found in several insecticide exposure studies (23,136,137). Thus, researchers planning studies of pesticide exposure and human development must design sampling strategies that characterize exposure during rapidly changing life stages and by sex. ASSESSING EXPOSURE TO NEUROTOXIC PESTICIDES Historically, human exposure assessment to hazardous chemicals, including neurotoxic insecticides, have served two purposes: (1) estimating population level exposure for risk assessment purposes; and, (2) estimating individual-level exposures to classify exposure for epidemiologic studies. The ability of any individual exposure measure to classify exposure for epidemiologic studies is often unknown, unless they have been validated. The U.S. EPA developed a paradigm to describe the necessary parameters required to determine risks associated with exposures to environmental chemicals (138,139). A modified version of this paradigm (Fig. 1) can be applied to studies evaluating effects of neurotoxic insecticides. According to the paradigm, the exposure must be adequately described to positively establish the relation between neurotoxic effects and insecticide exposure. In fact, failure to link exposure to neurotoxic insecticides to neurotoxic effects in humans may result as much or more from poor exposure assessment than from a lack of association. Exposure assessment is a complex process for healthy adults but it becomes exponentially more complex when it involves fetuses and young children. Tables 5 and 6 (41) describe the exposure assessment data than can be realistically obtained during fetal and early childhood development. To most accurately evaluate neurotoxic effects of insecticide exposure, collecting the samples at or near critical timepoints in development is imperative. In general, three types of approaches (Fig. 1) have been used to describe exposure, each with its own advantages and limitations (40,41). EXPOSURE CHARACTERIZATION

RISK CHARACTERIZATION Survey Data Collection

SOURCE/STRESSOR FORMATION

•

Proximity to source Pesticide use data Questionnaires • Time-activity diaries • Food consumption diaries

NEUROTOXIC EFFECT

• •

DEPOSITION IN THE ENVIRONMENT

Modulating factors include: Genetics Lifestyle • Coexposures

re e) su ng po lu ex in, ith sk w ., ct .g ta (e on e C rfac te in

• •

Biological Measurements

Enviromental Measurements •

Air. dust and soil measurements Duplicate det measurements • Dermal dosimetry

EXPOSURE

DOSE

•

Modulating factors include: • Temporality • Magnitude • Frequency

Modulating factors include: • Bioavailabilty • Absorption • Toxicokinetics

•

Measurements in biological matrices such as urine, blood, saliva, meconium, amniotic fluid • Measurements of altered genes, proteins and metabolism

Figure 1 Exposure-health effects paradigm (adapted from U.S. EPA). The three general approaches for exposure assessment are shown.

270

Barr et al.

Table 5 Potential Preconception, Pregnancy, and Perinatal Sample Collection for Nonpersistent Insecticide Analysis Sample (1) Maternal urinea (1) Maternal blooda (1) Cord blooda Meconium (1) Colustrum/ Breastmilk (2) Maternal saliva (2) Dietary assessmentd (1) Home aire samplea (1) Home composite dust samplee,a (1) Questionnaire (1) Ecologic analysis (e.g., GIS)e

First Pre- conception tri-mester

Second trimester

Third trimester

Perinatal period

x

x

x xb,c

x

x

x

x x

x

x xc xc x x x

x x

x x

x x

x x x x

x x

Key: (1) indicates measures that have been used in Prior Epidemiologic studies. (2) indicates those that are more Experimental or Costly. a Media with existing laboratory methods for likely target insecticides (e.g., urine, dust, air, food). b Blood collection piggy-back on glucose tolerance test. Blood samples crucial for PON1 status, ACHe. c Blood collection that is normal part of medical care. Blood samples crucial for PON1 status, ACHe. d Duplicate diet sampling, food frequency questionnaire, or other method (see text). e For each home lived in. Source: From Ref. 41.

These approaches are collecting historical information, measuring concentrations in environmental media, and measuring concentrations in biological media. For the OC pesticides that persist in the environment and in people, such as DDT or its degradate/ metabolite DDE, biological measurements are generally sufficient for describing the exposure. However, exposure assessment for currently used pesticides, which tend to have short environmental and biological residence times, is more challenging (39–41). These exposures tend to be transient and often episodic; thus, one sampling or measurement would provide only a short-term dosimeter (41,46,140). Typically, exposure to these pesticides is best assessed by more than one approach and may require collection of multiple biologic and environmental samples to effectively depict exposure and exposure pathways (40). In addition, biological samples need to be collected at periods temporally relevant to the pesticide exposure or critical window of development (Tables 4–6). Survey Data Survey data collection can include ecologic and geographic data, questionnaire information, food diaries, time-activity diaries or videotapes. Methods to use ecologic data to classify insecticide exposure for health studies include simple classification of residential location in agricultural areas or use of detailed residential histories linked with nearby insecticide use histories (141,141–143). These methods in part are supported by studies in Washington showing that insecticide exposures may be higher in children living closer to fields (18). However, high variability in agricultural environments, such as row versus orchard crops, crop-specific insecticide use, application methods, and local meteorology may affect the relation of nearby insecticide use to residential exposures.

a,b

x

x

x x

x

3 yr

x x

x

4 yr

x

x x x

x x

x x

x x

x x

x x

b

Pediatric urine bag, cloth diaper or insert or disposable diaper gel matrix sample for nontoilet trained children. If not diapered, spot samples or multiple spots. Media with existing laboratory methods for likely target insecticides (e.g., urine, dust, air, food). c Blood collection at young age piggy-back on CDC recommended lead screen at 12 and 24 mo. Blood samples crucial for PON1 status, ACHe. d Choking hazard for saliva collection for children less than 3 yr with current protocol. e Duplicate diet sampling, food frequency questionnaire, or other method. f For each home lived in. Source: From Ref. 41.

a

x

x x

2 yr

x x

x

5 yr

X x

Each home/yr

x

x

18 mo

Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr

x

x

x

x x x

12 mo

Each home/yr

x

x

x

9 mo

Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr Each home/yr

x

6 mo

x

3 mo

Age of child

Early Child Sample Collection for Exposure Assessment of Currently Used Insecticides

(a) Urine (a) Bloodc,b (a) Breast milk (b) Salivad (b) Dietary assessmente (a) Home air samplef,b (a) Home dust samplef,b (a) Questionnaire (a) Ecologic analysis (e.g., GIS)f

Sample

Table 6

Pesticide Exposure and Human Development 271

272

Barr et al.

Additional validation studies including direct measures of pesticides in environmental or biologic samples are needed before ecologic data, such as agricultural pesticide use information, can be used to classify exposures for epidemiologic studies. Questionnaires traditionally have been used to provide semiquantitative information for exposure characterization (144). They are used to identify both sources and routes of exposure by collecting data on global issues, such as household and residential chemical use patterns (e.g., insecticide treatments within and around the home), presence of home vegetable/fruit gardens, parental occupation, proximity to external chemical sources, and home age/condition characteristics that may be associated with presence of neurotoxicants (145,146). Questionnaires can also be used to identify important milestones in child development that may affect exposure, such as the transition to solid foods at 4–6 mo of age. Such questionnaires typically are completed by the parent or head of household. When increasing level of detail is desired, time-activity diaries provide detailed macro-activity (where an individual spends time) and micro-activity (specific activities) data (133,147). Children older than 12 yr have little difficulty understanding how to complete these diaries but may have difficulty in compliance (148). For younger children, the primary caretaker has the responsibility as surrogate responder for the child (133,135). Use of surrogate responders works well for variables that are global such as whether a child spent time at school or played outdoors. Detailed information about how much time a child spent in specific locations or information about micro-activities (what a child touches or mouths) often is reported poorly or inaccurately by surrogate responders (137,149). To address this issue, some exposure assessors have adopted a variety of observational techniques to capture the micro-activities of the child (136,149–158). The most commonly reported method is videography in which exposure-related behaviors are quantified. In addition, the use of global position systems has also been shown as a useful approach to quantify the macroenvironment of children (159). Of particular interest for neurotoxicants are hand or mouth contacts with dusty surfaces, floors, or soil, hand-tomouth activities, hand contacts with food, and hand washing behaviors (23,151,160,161). Statistical associations between contact activities with substrates and insecticide hand loadings and/or biomarkers of exposure have been found in several studies (23). These associations have not been verified through empirical studies. A behavior of concern especially with young children is the consumption of food that has fallen on the floor. In laboratory studies, Rohrer et al. (162) found that many OP and pyrethroid insecticides adhered to bologna, cheese, and apple slices placed on contaminated hard surfaces. Observational techniques allow verification of parental reports and collection of quantifiable data on food handling and retrieval behavior that would otherwise be difficult to obtain. Ultimately, these types of studies will validate questionnaires and other instruments that can be used to categorize exposure in epidemiologic studies where individual videography is not possible. Measurements in Environmental Media Measurements of insecticides in environmental media can be used provide information about routes of exposure, and when coupled with time-activity and food consumption data, can provide powerful data on the frequency and magnitude of exposures. Many neurotoxic insecticides are semivolatile (e.g., OP, carbamate and OC insecticides) or nonvolatile (e.g., pyrethroids) (163); however, both semivolatile and nonvolatile insecticides can be resuspended into air on particles by human and pet activity (163) and are readily detectable in indoor and personal air samples (19,146,163–169). These studies have shown that inhalation exposure to semivolatile insecticides in indoor air can be substantial and may be

Pesticide Exposure and Human Development

273

a primary route of exposure after residential use of insecticides. For example, an aggregate exposure assessment of chlorpyrifos among adult residents from Baltimore, Maryland, found that inhalation exposures accounted for approximately 85% of total daily dose (164). Similarly, results from the U.S. EPA Non-Occupational Insecticides Exposure Study (NOPES) indicate that 85% of the total daily exposure of adults to airborne insecticides is from breathing air inside the home (146). However, exposures can vary considerably, particularly for children; thus, the primary route of exposure in children will depend on household pesticide use patterns, behavioral and dietary habits, and the volatility of the insecticides. Studies designed to characterize children’s exposure to insecticides indicate that the largest number of insecticides and the highest concentrations are found in household dust than in air, soil, and food (18,163). Most household dust samples contain one or more pesticides at detectable concentrations and the concentrations remain relatively constant over months to years (170–172). Because of hand-to-mouth activities, dust may be a significant medium of contaminant exposure for young children in the home environment. Dust sampling has several limitations: (1) the timing of application is not known; (2) levels in the dust may reflect use months to years before the sampling; and (3) depending upon the dust type sampled, dust concentrations may differ and the potential for exposure may differ, as well. For example, dust on hard surfaces may be readily available to transfer to children’s skin and result in nondietary ingestion or dermal exposures, whereas insecticide-contaminated dust lodged deeply in carpets may not be available to children. Carpet and other dusts may function as a reservoir for household insecticide contamination, recontaminating surfaces and air after cleaning depending on the physical and chemical properties (“fugacity”) of the specific compounds. Because of the variability in childhood activities and the potential exposures to dust, the inter-relations of environmental and personal exposure measures may be difficult to interpret. Diet is a potentially significant pathway of exposure to insecticides for children (34,173). Most food types contain some pesticide residues (174). Detectable insecticide residues were found in approximately 47% of the fruit and vegetable samples tested as part of the USDA’s market-basket surveys in 2002 (174). However, insecticide residues vary significantly across foods. Therefore, individual dietary exposures have been difficult to estimate reliably using food consumption questionnaires (175). Instead, studies generally have estimated dietary exposures by measuring insecticides in duplicate diet samples, in which study participants prepare and collect duplicate portions of all foods and beverages consumed. These studies are considered the gold standard; however, they are extremely costly, burdensome on the participant, and not feasible for large epidemiologic studies. Duplicate diet studies also may underestimate dietary exposure if study designs do not account for contamination of foods from indoor sources, such as handling of food by children who also contact contaminated surfaces or dust. As with the videography studies reviewed above, these types of studies likely will lead to validated questionnaires and other instruments that can be used in epidemiologic studies. Biomonitoring of Exposure Measurements of concentrations of neurotoxic pesticides in biological media directly indicate exposure and absorption of the pesticide (or degradate) into the body (39,176). The most common matrices employed for biomonitoring are blood and urine. Blood serum or plasma is the preferred medium for measurements of biologically persistent chemicals such as DDT and DDE (39). Because OC insecticides are highly lipophilic, they tend to sequester in the lipid stores in the body and remain at equilibrium. Thus, higher lipid

274

Barr et al.

concentrations in the blood would shift the equilibrium and increase the level of the OC insecticides in the blood. To correct for variation among blood lipids, these concentrations sometimes are normalized on lipid content, which then reduces the overall variability in measurements among persons (177). As with environmental measurements, biomonitoring of currently used insecticides is more complex because these chemicals typically are metabolized rapidly and excreted in urine (39,140). Metabolites of pyrethroid, OP, and carbamate insecticides have been measured frequently in urine samples collected from both occupational and nonoccupational exposures (27,178–183). These measurements are relatively easy to obtain; however, they often provide equivocal information on the parent pesticide. For example, 3,5,6-trichloropyridinol (TCPY) is a metabolite of both chlorpyrifos and its methyl analogue; thus, urinary TCPY measurements would not identify, unequivocally, the parent insecticide (184). Further complicating interpretation, TCPY also can be found in the environment from degradation or soil microbe and plant metabolism (185,186). Thus, urinary TCPY measurements also may reflect exposure to TCPY, which is not known to be neurotoxic. Consequently, measurements of TCPY in urine may result in misclassification of chlorpyrifos exposure and, therefore, obscure any link between exposure and adverse neurologic effects. Another limitation of urinary metabolite measurements is the variable dilution of spot, or convenience, urine samples. Creatinine adjustment of urinary metabolites is the standard method for correcting for urine dilution; however, recent studies suggest that creatinine adjustment in pregnant women and children may not be appropriate (152,187,188). The appropriateness of creatinine adjustment lies largely with the diversity of the population studied. Children excrete proportionately less creatinine (Fig. 2) than adolescents or adults, and women excrete less creatinine than men (188). Diet, disease, race/ethnicity, musculature, dietary supplements, and pregnancy also can influence creatinine excretion (187), thus increasing the variability in creatinine-adjusted data rather than eliminating variability from urine dilution.

70

%<30 Ug/dl creatinine

60 50 40 30 20 10 0 <13 months 25-36 months 49-60 months 13-24 months 37-48 months >60 months

Figure 2 Proportion of low creatinine levels by age of child (months) in a pesticide exposure study. Source: From Ref. 139.

Pesticide Exposure and Human Development

275

Finally, because of variability in exposure scenarios and timing and biological residence times of insecticides, spot urine samples may not reflect daily pesticide exposure and thus may misclassify exposure. One recently published study reported that first morning void samples were likely to more accurately represent total daily exposure in children for certain OP insecticides (189). Another recent study demonstrated that one spot urine sample was sufficient to accurately classify exposure tertiles in men enrolled in a sperm-quality study (190); however, the level of confidence in the classification rose along with the number of longitudinal urine samples collected. Few studies have investigated the ability of spot or 24-hr urine samples to accurately classify childhood exposures; however, several urinary validation studies are in progress. Despite these limitations, urinary insecticide metabolite measurements still provide meaningful data. In the absence of other reliable exposure data, they can help to more accurately classify exposure. When coupled with environmental measurements and certain historical information, a good understanding of the exposure can be derived (40). Furthermore, laboratory methods for measuring pesticides in urine are further developed than other biological exposure measurements; urine samples are also less invasive than other media to collect; and in many cases, reference data already exist (46). Currently used neurotoxic insecticides also have been measured in blood plasma or serum as the intact insecticide or its metabolite (19,31,191). In general, blood concentrations of pyrethroid, OP, and carbamate insecticides or their metabolites are about three orders of magnitude lower than their metabolite concentrations in urine (39,140). To accurately measure such low levels, relatively large amounts of blood (w10 mL blood resulting in 4 mL of serum) and highly sophisticated analytical instrumentation is required; thus the measurements are more complex and more expensive, and methodology is less readily available for most pesticides than urinary metabolite methods. However, blood measurements have the inherent advantage of easily allowing a dose estimate calculation using the known blood volume and distribution among tissues (39,140). Furthermore, the measurement of the parent chemical, rather than a metabolite, offers unequivocal identification of the source pesticide. Other matrices have been used for biomonitoring, including meconium, amniotic fluid, and saliva (28,30,192). Because meconium accumulates in the fetal bowels for 24–30 wk before elimination, measurements of pesticides in this medium may provide an estimate of cumulative fetal exposure during pregnancy (193). Saliva may be an easier and less costly matrix to collect, reducing costs in longitudinal sample collections; however, current data show that saliva levels of certain OP insecticides are only a fraction of the blood levels (194), so detection of these chemicals in saliva may be problematic.

RECOMMENDATIONS FOR FUTURE STUDIES Few would argue that exposures to neurotoxic insecticides may have adverse neurologic effects on susceptible populations, and potentially even healthy adults. However, few studies have evaluated these potential effects, and many studies that have been conducted are plagued by small sample size or poor exposure assessment. Designing studies with a good exposure assessment is a difficult process often hampered by cost and participant burden. Clearly, studies evaluating exposures to currently used pesticides require multiple measurements taken over critical time spans to adequately describe exposures. However, this design requires knowledge of the critical time windows of exposure required to develop certain effects, and in many cases, these time windows are not fully understood. Thus, to fully capture the exposure and relate it to

276

Barr et al.

some unknown time window, a limitless number of samples potentially could be measured; however, this is neither practical from a participant burden nor cost perspective. A better option would be to conduct smaller studies to evaluate variations in exposures among a particular population group and to determine the number of samples required to accurately classify their average exposure over a given period. Another complication in exposure assessment is the selectivity of the measurement for the exposure incurred. Studies are needed to identify and develop biomarkers more specific for a given insecticide and, preferably, a biologically active marker. For example, measurement of a urinary metabolite may not as accurately reflect the amount of an OP insecticide that was available for AChE inhibition, but current methods for measuring AChE inhibition lack the sensitivity to detect the low-level exposures that result in measurable urinary metabolites. Better methods for selectively measuring such inhibition—for example, direct measurement of adducted AChE—may provide more biologically relevant data. A further recommendation is that studies be conducted compatibly to allow metaanalyses of the data generated. In instances where few studies exist or where the technology has far surpassed the existing literature base, common characteristics still should be included to allow the use of older studies. Furthermore, negative findings also are important and should be reported with equal enthusiasm as positive findings. Despite the fact that insecticides have been used in the United States and other countries for over a century, few studies specifically have focused on evaluating their potential neurotoxic effects on humans. The limited data, however, suggest some association between DDE and OP insecticide exposure on neurodevelopment. Improved understanding and application of exposure assessment in future studies will allow a better determination of whether such neurotoxic effects result from low-level insecticide exposures.

ACKNOWLEDGMENTS This chapter in part draws on several authors’ (DB, AB, RW, LN) experiences sitting on the National Children’s Study (NCS) Chemical Exposures Working Group. We acknowledge the extensive contributions workgroup members made to our understanding of insecticide exposure issues during the development of a white paper on assessing exposures to chemical and biological agents for the NCS (1). The time Asa Bradman and Brenda Eskenazi devoted to this publication was supported by grant number P01ES009605 from the National Institute of Environmental Health Sciences (NIEHS), NIH, and grant number RD-831710 from the Environmental Protection Agency (EPA). The time Robin Whyett devoted to this publication was supported by NIEHS grant numbers P01 ES009600, R01 ES068977, R01ES11158, R01ES10165, R01ES12468; and EPA grant numbers R827027, 82860901, and RD-832141. This publication’s contents are solely the responsibility of the authors and do not necessarily represent the official views of NIEHS, NIH, and EPA, or CDC.

REFERENCES 1. Chemical Exposures Workgroup. Measurement and analysis of exposure to environmental pollutants and biological agents during the National Children’s Study. (Accessed 1 Mar, 2005, at http://www.nationalchildrensstudy.gov/research/methods_studies/.)

Pesticide Exposure and Human Development

277

2. Kiely T, Donaldson D, Grube A. Pesticide Industry Sales and Usage: 2000 and Market Estimates. Washington, D.C.: U.S. EPA, 2004. 3. U.S. EPA. Interim reregistration eligibility decision for chlorpyrifos. EPA 738-R-01-007, 2002. 4. ATSDR. Toxicological Profile for Pyrethrins and Pyrethroids. Atlanta: Agency for Toxic Substances and Disease Registry, 2003. 5. Allen JR, Barsotti DA, Carstens LA. Residual effects of polychlorinated biphenyls on adult nonhuman primates and their offspring. J Toxicol Environ Health 1980; 6:55–66. 6. Bulger WH, Kupfer D. Estrogenic action of DDT analogs. Am J Ind Med 1983; 4:163–173. 7. Berglind R, Dave G. Acute toxicity of chromate, DDT, PCP, TPBS, and zinc to Daphnia magna cultured in hard and soft water. Bull Environ Contam Toxicol 1984; 33:63–68. 8. O’Shea TJ, Fleming WJ, Cromartie E. DDT contamination at Wheeler National Wildlife Refuge. Science 1980; 209:509–560. 9. Allen JR, Barsotti DA, Lambrecht LK, Van Miller JP. Reproductive effects of halogenated aromatic hydrocarbons on nonhuman primates. Ann NY Acad Sci 1979; 320:419–425. 10. Allen JR, Hargraves WA, Hsia MT, Lin FS. Comparative toxicology of chlorinated compounds on mammalian species. Pharmacol Ther 1979; 7:513–547. 11. Berger S, Pardo B, Skorkowska-Zieleniewska J. Nutritional implications of pesticides in foods. Bibl Nutr Dieta 1980; 29:1–10. 12. Ferrando MD, Alarcon V, Fernandez-Casalderrey A, Gamon M, Andreu-Moliner E. Persistence of some pesticides in the aquatic environment. Bull Environ Contam Toxicol 1992; 48:747–755. 13. Mes J, Marchand L, Davies DJ. Organochlorine residues in adipose tissue of Canadians. Bull Environ Contam Toxicol 1990; 45:681–688. 14. Turusov V, Rakitsky V, Tomatis L. Dichlorodiphenyltrichloroethane (DDT): ubiquity, persistence, and risks. Environ Health Perspect 2002; 110:125–128. 15. Vieira ED, Torres JP, Malm O. DDT environmental persistence from its use in a vector control program: a case study. Environ Res 2001; 86:174–182. 16. Frank R, Rasper J, Smout MS, Braun HE. Organochlorine residues in adipose tissues, blood and milk from Ontario residents, 1976–1985. Can J Public Health 1988; 79:150–158. 17. Bradman MA, Harnly ME, Draper W, et al. Pesticide exposures to children from California’s Central Valley: results of a pilot study. J Expo Anal Environ Epidemiol 1997; 7:217–234. 18. Simcox NJ, Fenske RA, Wolz SA, Lee IC, Kalman DA. Pesticides in household dust and soil: exposure pathways for children of agricultural families. Environ Health Perspect 1995; 103:1126–1134. 19. Whyatt RM, Barr DB, Camann DE, et al. Contemporary-use pesticides in personal air samples during pregnancy and blood samples at delivery among urban minority mothers and newborns. Environ Health Perspect 2003; 111:749–756. 20. Barry JW. The USDA Forest Service pesticide spray behavior and application development program—an overview. J Am Mosq Control Assoc 1996; 12:342–352. 21. Adgate JL, Barr DB, Clayton CA, et al. Measurement of children’s exposure to pesticides: analysis of urinary metabolite levels in a probability-based sample. Environ Health Perspect 2001; 109:583–590. 22. Loewenherz C, Fenske RA, Simcox NJ, Bellamy G, Kalman D. Biological monitoring of organophosphorus pesticide exposure among children of agricultural workers in central Washington State. Environ Health Perspect 1997; 105:1344–1353. 23. Shalat SL, Donnelly KC, Freeman NC, et al. Nondietary ingestion of pesticides by children in an agricultural community on the US/Mexico border: preliminary results. J Expo Anal Environ Epidemiol 2003; 13:42–50. 24. Royster MO, Hilborn ED, Barr D, Carty CL, Rhoney S, Walsh D. A pilot study of global positioning system/geographical information system measurement of residential proximity to agricultural fields and urinary organophosphate metabolite concentrations in toddlers. J Expo Anal Environ Epidemiol 2002; 12:433–440.

278

Barr et al.

25. Fenske RA, Kissel JC, Lu C, et al. Biologically based pesticide dose estimates for children in an agricultural community. Environ Health Perspect 2000; 108:515–520. 26. Lu C, Fenske RA, Simcox NJ, Kalman D. Pesticide exposure of children in an agricultural community: evidence of household proximity to farmland and take home exposure pathways. Environ Res 2000; 84:290–302. 27. Barr DB, Bravo R, Weerasekera G, et al. Concentrations of dialkyl phosphate metabolites of organophosphorus pesticides in the U.S. population. Environ Health Perspect 2004; 112:186–200. 28. Bradman A, Barr DB, Claus Henn BG, Drumheller T, Curry C, Eskenazi B. Measurement of pesticides and other toxicants in amniotic fluid as a potential biomarker of prenatal exposure: a validation study. Environ Health Perspect 2003; 111:1779–1782. 29. Perera FP, Rauh V, Tsai WY, et al. Effects of transplacental exposure to environmental pollutants on birth outcomes in a multiethnic population. Environ Health Perspect 2003; 111:201–205. 30. Whyatt RM, Barr DB. Measurement of organophosphate metabolites in postpartum meconium as a potential biomarker of prenatal exposure: a validation study. Environ Health Perspect 2001; 109:417–420. 31. Whyatt RM, Rauh V, Barr DB, et al. Prenatal insecticide exposures and birth weight and length among an urban minority cohort. Environ Health Perspect 2004; 112:1125–1132. 32. Silvers A, Florence BT, Rourke DL, Lorimor RJ. How children spend their time: a sample survey for use in exposure and risk assessments. Risk Anal 1994; 14:931–944. 33. Zartarian VG, Ozkaynak H, Burke JM, Zufall MJ, Rigas ML, Furtaw EJ, Jr. A modeling framework for estimating children’s residential exposure and dose to chlorpyrifos via dermal residue contact and nondietary ingestion. Environ Health Perspect 2000; 108:505–514. 34. National Research Council. Committee on Pesticides in the Diets of Infants and Children. Pesticides in the Diets of Infants and Children. Washington, D.C.: National Academy Press, 1993. 35. Costa LG, Richter RJ, Li WF, Cole T, Guizzetti M, Furlong CE. Paraoxonase (PON 1) as a biomarker of susceptibility for organophosphate toxicity. Biomarkers 2003; 8:1–12. 36. Faustman EM, Silbernagel SM, Fenske RA, Burbacher TM, Ponce RA. Mechanisms underlying children’s susceptibility to environmental toxicants. Environ Health Perspect 2000; 108:13–21. 37. Costa LG, Li WF, Richter RJ, Shih DM, Lusis AJ, Furlong CE. Pon1 and organophosphate toxicity. In: Costa LG, Furlong CE, eds. Paraoxonase (Pon1) in Health and Disease: Basic and Clinical Aspects. Boston: Kluwer Academic Press, 2002. 38. Furlong CE, Holland N, Richter RJ, Bradman A, Eskenazi B. PON1 status of farmworker mothers and children as a predictor of organophosphate sensitivity Pharmacogenetics and Genomics, in press. 39. Barr DB, Wang RY, Needham LL. Biological monitoring of exposure to environmental chemicals throughout the life stages: requirements and issues for consideration for the National Children’s Study. Environ Health Perspect 2005; 113:1083–1091. 40. Needham LL, Whyatt RM, Ozkaynak H. et al. Exposure assessment in the National Children’s Study: Introduction. Environ Health Perspect 2005; 113:1076–1082. 41. Bradman A, Whyatt RM. Measures for characterizing exposures to nonpersistent pesticides during pregnancy and early childhood in the National Children’s Study: A review. Environ Health Perspect 2005; 113:1092–1099. 42. Garfitt SJ, Jones K, Mason HJ, Cocker J. Exposure to the organophosphate diazinon: data from a human volunteer study with oral and dermal doses. Toxicol Lett 2002; 134:105–113. 43. Karalliedde L, Feldman RJ, Henderson LO, Marrs T. Organophosphates and Health. London: Imperial College Press, 2001. 44. Griffin P, Mason H, Heywood K, Cocker J. Oral and dermal absorption of chlorpyrifos: a human volunteer study. Occup Environ Med 1999; 56:10–13.

Pesticide Exposure and Human Development

279

45. Hardt J, Angerer J. Biological monitoring of workers after the application of insecticidal pyrethroids. Int Arch Occup Environ Health 2003; 76:492–498. 46. Wessels D, Barr DB, Mendola P. Use of biomarkers to indicate exposure of children to organophosphate pesticides: implications for a longitudinal study of children’s environmental health. Environ Health Perspect 2003; 111:1939–1946. 47. Nolan RJ, Rick DL, Freshour NL, Saunders JH. Chlorpyrifos: pharmacokinetics in human volunteers. Toxicol Appl Pharmacol 1984; 73:8–15. 48. Kwong TC. Organophosphate pesticides: biochemistry and clinical toxicology. Ther Drug Monit 2002; 24:144–149. 49. Davies JE. Neurotoxic concerns of human pesticide exposures. Am J Ind Med 1990; 18:327–331. 50. Levin HS, Rodnitzky RL. Behavioral effects of organophosphate in man. Clin Toxicol 1976; 9:391–403. 51. Korsak RJ, Sato MM. Effects of chronic organophosphate pesticide exposure on the central nervous system. Clin Toxicol 1977; 11:83–95. 52. Soliman SA, Farmer J, Curley A. Is delayed neurotoxicity a property of all organophosphorus compounds? A study with a model compound: parathion Toxicology 1982; 23:267–279. 53. Cherniak MG. Organophosphos esters and polyneurophathy. Ann Inter Med 1986; 104:265–266. 54. Levin HS, Rodnitzky RL, Mick DL. Anxiety associated with exposure to organophosphate compounds. Arch Gen Psychiatry 1976; 33:225–228. 55. Moser VC, Chanda SM, Mortensen SR, Padilla S. Age- and gender-related differences in sensitivity to chlorpyrifos in the rat reflect developmental profiles of esterase activities. Toxicol Sci 1998; 46:211–222. 56. Slotkin TA. Developmental cholinotoxicants: nicotine and chlorpyrifos. Environ Health Perspect 1999; 107:71–80. 57. Evans RT, O’Callaghan J, Norman A. A longitudinal study of cholinesterase changes in pregnancy. Clin Chem 1988; 34:2249–2252. 58. Howard JK, East NJ, Chaney JL. Plasma cholinesterase activity in early pregnancy. Arch Environ Health 1978; 33:277–279. 59. Bigbee JW, Sharma KV, Chan EL, Bogler O. Evidence for the direct role of acetylcholinesterase in neurite outgrowth in primary dorsal root ganglion neurons. Brain Res 2000; 861:354–362. 60. Lauder JM, Schambra UB. Morphogenetic roles of acetylcholine. Environ Health Perspect 1999; 107:65–69. 61. Dam K, Seidler FJ, Slotkin TA. Chlorpyrifos releases norepinephrine from adult and neonatal rat brain synaptosomes. Brain Res Dev Brain Res 1999; 118:129–133. 62. Dam K, Seidler FJ, Slotkin TA. Developmental neurotoxicity of chlorpyrifos: delayed targeting of DNA synthesis after repeated administration. Brain Res Dev Brain Res 1998; 108:39–45. 63. Dam K, Seidler FJ, Slotkin TA. Transcriptional biomarkers distinguish between vulnerable periods for developmental neurotoxicity of chlorpyrifos: implications for toxicogenomics. Brain Res Bull 2003; 59:261–265. 64. Garcia SJ, Seidler FJ, Qiao D, Slotkin TA. Chlorpyrifos targets developing glia: effects on glial fibrillary acidic protein. Brain Res Dev Brain Res 2002; 133:151–161. 65. Benke GM, Murphy SD. The influence of age on the toxicity and metabolism of methyl parathion and parathion in male and female rats. Toxicol Appl Pharmacol 1975; 31:254–269. 66. Brodeur J, Dubois K. Comparison of acute toxicity of anticholinesterase insecticides to weanling and adult male rats. Proc Soc Exp Biol Med 1963; 114:509–511. 67. Moser VC. Dose-response and time-course of neurobehavioral changes following oral chlorpyrifos in rats of different ages. Neurotoxicol Teratol 2000; 22:713–723. 68. Pope CN, Chakraborti TK. Dose-related inhibition of brain and plasma cholinesterase in neonatal and adult rats following sublethal organophosphate exposures. Toxicology 1992; 73:35–43.

280

Barr et al.

69. Pope CN, Chakraborti TK, Chapman ML, Farrar JD, Arthun D. Comparison of in vivo cholinesterase inhibition in neonatal and adult rats by three organophosphorothioate insecticides. Toxicology 1991; 68:51–61. 70. Whitney KD, Seidler FJ, Slotkin TA. Developmental neurotoxicity of chlorpyrifos: cellular mechanisms. Toxicol Appl Pharmacol 1995; 143:53–62. 71. Atterberry TT, Burnett WT, Chambers JE, Atterberry TT, Burnett WT, Chambers JE. Agerelated differences in parathion and chlorpyrifos toxicity in male rats: target and nontarget esterase sensitivity and cytochrome p450-mediated metabolism. Toxicol Appl Pharmacol 1997; 147:411–418. 72. Li WF, Costa LG, Richter RJ, et al. Catalytic efficiency determines the in-vivo efficacy of pon1 for detoxifying organophosphorus compounds. Pharmacogenetics 2000; 10:767–779. 73. Mortensen SR, Chanda SM, Hooper MJ, Padilla S. Maturational differences in chlorpyrifosoxonase activity may contribute to age-related sensitivity to chlorpyrifos. J Biochem Toxicol 1996; 11:279–287. 74. Davies HG, Richter RJ, Keifer M, Broomfield CA, Sowalla J, Furlong CE. The effect of the human serum paraoxonase polymorphism is reversed with diazoxon, soman and sarin. Nat Genet 1996; 14:334–336. 75. Lassiter TL, Padilla S, Mortensen SR, Chanda SM, Moser VC, Barone S Jr. Gestational exposure to chlorpyrifos: apparent protection of the fetus? Toxicol Appl Pharmacol 1998; 152:56–65. 76. Berkowitz GS, Wetmur JG, Birman-Deych E, et al. In utero pesticide exposure, maternal paraoxonase activity, and head circumference. Environ Health Perspect 2004; 112:388–391. 77. Campbell CG, Seidler FJ, Slotkin TA. Chlorpyrifos interferes with cell development in rat brain regions. Brain Res Bull 1997; 43:179–189. 78. Song X, Violin JD, Seidler FJ, Slotkin TA. Modeling the developmental neurotoxicity of chlorpyrifos in vitro: macromolecule synthesis in pc12 cells. Toxicol Appl Pharmacol 1998; 151:182–191. 79. Drachman DA. Memory and cognitive function in man: does the cholinergic system have a specific role? Neurology 1977; 27:783–790. 80. Karczmar AG. Cholinergic influences on behavior. In: Waser PG, ed. Cholinergic Mechanisms. New York: Raven Press, 1975:37–47. 81. Eriksson P, Talts U. Neonatal exposure to neurotoxic pesticides increases adult susceptibility: a review of current findings. Neurotoxicology 2000; 21:37–47. 82. Coyle JT, Yamamura HI. Neurochemical aspects of the ontogenesis of cholinergic neurons in the rat brain. Brain Res 1976; 118:429–440. 83. Qiao D, Seidler FJ, Padilla S, Slotkin TA. Developmental neurotoxicity of chlorpyrifos: what is the vulnerable period? Environ Health Perspect 2002; 110:1097–1103. 84. Garcia SJ, Seidler FJ, Slotkin TA. Developmental neurotoxicity elicited by prenatal or postnatal chlorpyrifos exposure: effects on neurospecific proteins indicate changing vulnerabilities. Environ Health Perspect 2003; 111:297–304. 85. Richardson RJ. Assessment of the neurotoxic potential of chlorpyrifos relative to other organophosphorus compounds: a critical review of the literature. J Toxicol Environ Health 1995; 44:165. 86. Eskenazi B, Bradman A, Castorina R. Exposures of children to organophosphate pesticides and their potential adverse health effects. Environ Health Perspect 1999; 107:409–419. 87. Landrigan PJ, Claudio L, Markowitz SB, et al. Pesticides and inner-city children: exposures, risks, and prevention. Environ Health Perspect 1999; 107:431–437. 88. Soderlund DM, Clark JM, Sheets LP, et al. Mechanisms of pyrethroid neurotoxicity: implications for cumulative risk assessment. Toxicology 2002; 171:3–59. 89. Vijverberg HPM, van den Bercken J. Neurotoxicological effects and the mode of action of pyrethroid insecticides. Crit Rev Toxicol 1990; 21:105–126. 90. Aldridge WN. An assessment of the toxicological properties of pyrethroids and their neurotoxicity. Crit Rev Toxicol 1990; 21:89–104.

Pesticide Exposure and Human Development

281

91. Bradbury SP, Coats JR. Comparative toxicology of the pyrethroid insecticides. Rev Environ Contam Toxicol 1989; 108:177. 92. Vijverberg HP, van der Zalm JM, van der BJ. Similar mode of action of pyrethroids and DDT on sodium channel gating in myelinated nerves. Nature 1982; 295:601–603. 93. Vijverberg HP, van der Zalm JM, van der BJ. Similar mode of action of pyrethroids and DDT on sodium channel gating in myelinated nerves. Nature 1982; 295:601–603. 94. Mandhane SN, Chopde CT. Neurobehavioral effects of low level fenvalerate exposure in mice. Indian J Exp Biol 1997; 35:623–627. 95. Narahashi T. Neurophysiological effects of insecticides. In: Krieger RI, ed. Handbook of Pesticide Toxicology: Principles. San Diego: Academic Press, 2001:335–351. 96. Sheets L. A consideration of age-dependent differences in susceptibility to organophosphorus and pyrethroid insecticides. Neurotoxicology 2000; 21:57–63. 97. Cantalamessa F. Acute toxicity of two pyrethroids, permethrin, and cypermethrin in neonatal and adult rats. Arch Toxicol 1993; 67:513. 98. Leng G, Leng A, Kuhn KH, Lewalter J, Pauluhn J. Human dose-excretion studies with the pyrethroid insecticide cyfluthrin: urinary metabolite profile following inhalation. Xenobiotica 1997; 27:1273–1283. 99. Hainzl D, Cole LM, Casida JE. Mechanisms for selective toxicity of fipronil insecticide and its sulfone metabolite and desulfinyl photoproduct. Chem Res Toxicol 1998; 11:1529–1535. 100. Zwiener RJ, Ginsburg CM. Organophosphate and carbamate poisoning in infants and children. Pediatr 1988; 81:121–126 [published erratum appears in Pediatr 81 683]. 101. Rogan WJ, Gladen BC, McKinney JD, et al. Neonatal effects of transplacental exposure to PCBs and DDE. J Pediatr 1986; 109:335–341. 102. Darvill T, Lonky E, Reihman J, Stewart P, Pagano J. Prenatal exposure to PCBs and infant performance on the fagan test of infant intelligence. Neurotoxicology 2000; 21:1029–1038. 103. Ribas-Fito N, Cardo E, Sala M, et al. Breastfeeding, exposure to organochlorine compounds, and neurodevelopment in infants. Pediatr 2003; 111:e580–e585. 104. Guillette EA, Meza MM, Aquilar MG, Soto AD, Garcia IE. An anthropological approach to the evaluation of preschool children exposed to pesticides in Mexico. Environ Health Perspect 1998; 106:347–353. 105. Rohlman DS, Bailey SR, Anger WK, McCauley L. Assessment of neurobehavioral function with computerized tests in a population of hispanic adolescents working in agriculture. Environ Res 2001; 85:14–24. 106. Ruckart PZ, Kakolewski K, Bove FJ, Kaye WE. Long-term neurobehavioral health effects of methyl parathion exposure in children in Mississippi and Ohio. Environ Health Perspect 2004; 112:46–51. 107. Stewart P, Reihman J, Lonky E, Darvill T, Pagano J. Prenatal PCB exposure and neonatal behavioral assessment scale (NBAS) performance. Neurotoxicol Teratol 2000; 22:21–29. 108. Gladen BC, Rogan WJ, Hardy P, Thullen J, Tingelstad J, Tully M. Development after exposure to polychlorinated biphenyls and dichlorodiphenyl dichloroethene transplacentally and through human milk. J Pediatr 1988; 113:991–995. 109. Rogan WJ, Gladen BC. PCBs, DDE, and child development at 18 and 24 months. Ann Epidemiol 1991; 1:407–413. 110. Gladen BC, Rogan WJ. Effects of perinatal polychlorinated biphenyls and dichlorodiphenyl dichloroethene on later development. J Pediatr 1991; 119:58–63. 111. Eskenazi B, Harley K, Bradman A, et al. Association of in utero organophosphate pesticide exposure and fetal growth and length of gestation in an agricultural population. Environ Health Perspect 2004; 112:1116–1124. 112. Young JG, Eskenazi B, Gladstone EA, et al. Association between in utero organophosphate pesticide exposure and abnormal reflexes in neonates. Neurotoxicology 2005; 26:199–209. 113. Silbergeld EK. Maternally mediated exposure of the fetus: in utero exposure to lead and other toxins. Neurotoxicology 1986; 7:557–568. 114. Berkowitz GS, Obel J, Deych E, et al. Exposure to indoor pesticides during pregnancy in a multiethnic, urban cohort. Environ Health Perspect 2003; 111:79–84.

282

Barr et al.

115. The National Children’s Study. (Accessed 1 Mar, 2005, at http://www.nationalchildrensstudy. gov/.) 116. Guyton AC, Hall JE. 10th ed. Textbook of Medical Physiology. Philadelphia: W.B. Saunders Company, 2000. 117. Pirani BB, Campbell DM, MacGillivray I. Plasma volume in normal first pregnancy. J Obstet Gynaecol Br Commonw 1973; 80:884–887. 118. Heikkila A, Renkonen OV, Erkkola R. Pharmacokinetics and transplacental passage of imipenem during pregnancy. Antimicrob Agents Chemother 1992; 36:2652–2655. 119. Davison JM, Hytten FE. Glomerular filtration during and after pregnancy. J Obstet Gynaecol Br Commonw 1974; 81:588–595. 120. Levy G. Clinical pharmacokinetics of aspirin. Pediatrics 1978; 62:867–872. 121. Tomlin CDS. 11th ed. The Pesticide Manual. Farnham, U.K.: British Crop Protection Council, 1997. 122. Bassily M, Ghabrial H, Smallwood RA, Morgan DJ. Determinants of placental drug transfer: studies in the isolated perfused human placenta. J Pharm Sci 1995; 84:1054–1060. 123. Gaylard DG, Carson RJ, Reynolds F. Effect of umbilical perfusate pH and controlled maternal hypotension on placental drug transfer in the rabbit. Anesth Analg 1990; 71:42–48. 124. Adamson SL, Morrow RJ, Bull SB, Langille BL. Vasomotor responses of the umbilical circulation in fetal sheep. Am J Physiol 1989; 256:R1056–R1062. 125. Dean M, Stock B, Patterson RJ, Levy G. Serum protein binding of drugs during and after pregnancy in humans. Clin Pharmacol Ther 1980; 28:253–261. 126. Laws ER, Hayes WJ. Handbook of Pesticide Toxicology. San Diego, CA: Academic Press, 1991. 127. Pelkonen O. Drug metabolism in the human fetal liver. Relationship to fetal age. Arch Int Pharmacodyn Ther 1973; 202:281–287. 128. Selevan SG, Kimmel CA, Mendola P. Identifying critical windows of exposure for children’s health. Environ Health Perspect 2000; 108:451–455. 129. U.S. EPA. Guidance on selecting the appropriate age groups for assessing childhood exposures to environmental contaminants. EPA/630/P-03/003A, NECA-F-1449, 2003. 130. Guzelian PS, Henry CJ, Olin SS. Similarities and Differences Between Children and Adults: Implications for Risk Assessment. Washington, D.C.: ILSI Press, 1992. 131. Robinson JP. How Americans Use Their Time. New York: Praeger Publishers, 1977. 132. Timmer SG, Eccles J, O’Brien K. In: Juster FT, Stafford FP, eds. How children Use Time. In: Time, Goods, and Well-being. Ann Arbor, MI: University of Michigan, Institute for Social Research 1985:353–382. 133. Tsang AM, Klepeis NE. Descriptive statistics from a detailed analysis of the National Human Activity Pattern Survey (NHAPS) data. EPA/600/R-96/074, 1996. 134. Cohen Hubal EA, Sheldon LS, Burke JM, et al. Children’s exposure assessment: a review of factors influencing children’s exposure, and the data available to characterize and assess that exposure. Environ Health Perspect 2000; 108:475–486. 135. Wong EY, Shirai JH, Garlock TJ, Kissel JC. Adult proxy responses to a survey of children’s dermal soil contact activities. J Expo Anal Environ Epidemiol 2000; 10:509–517. 136. Freeman NC, Jimenez M, Reed KJ, et al. Quantitative analysis of children’s microactivity patterns: The Minnesota Children’s Pesticide Exposure Study. J Expo Anal Environ Epidemiol 2001; 11:501–509. 137. Black K, Shalat SL, Freeman NC, Jimenez M, Donnelly KC, Calvin JA. Children’s mouthing and food-handling behavior in an agricultural community on the U.S./Mexico border. J Expo Anal Environ 2004. 138. U.S. EPA. EPA’s Draft Report on the Environment 2003. U.S. Environmental Protection Agency: Available online: http://www.epa.gov/Envindicators/roe/pdf/tdHealth.pdf. 139. International Biomarker Workshop. Available online: http://www.ilsi.org/file/EnvironPublicHealthParadigm7-21-04.pdf. 140. Barr DB, Barr JR, Driskell WJ, et al. Strategies for biological monitoring of exposure for contemporary-use pesticides. Toxicol Ind Health 1999; 15:168–179.

Pesticide Exposure and Human Development

283

141. Ward M, Nuchols J, Weigel S, Maxwell S, Cantor K, Miller R. Geographic information systems. A new tool in environmental epidemiology. Ann Epidemiol 2000; 10:477. 142. Rull RP, Ritz B. Historical insecticide exposure in California using insecticide use reports and land-use surveys: an assessment of misclassification error and bias. Environ Health Perspect 2003; 111:1582–1589. 143. Gunier RB, Harnly ME, Reynolds P, Hertz A, Von Behren J. Agricultural insecticide use in California: insecticide prioritization, use densities, and population distributions for a childhood cancer study. Environ Health Perspect 2001; 109:1071–1078. 144. Lebowitz MD, Quackenboss JJ, Kollander M, Soczek ML, Colome S. The new standard environmental inventory questionnaire for estimation of indoor concentrations. JAPCA 1989; 39:1411–1419. 145. Savage EP, Keefe TJ, Tessari JD, et al. National study of chlorinated hydrocarbon insecticide residues in human milk, U.S.A. I. Geographic distribution of dieldrin, heptachlor, heptachlor epoxide, chlordane, oxychlordane, and mirex. Am J Epidemiol 1981; 113:413–422. 146. Whitmore R, Immerman FW, Camann DE, Bond AE, Lewis RG, Schaum JL. Nonoccupational exposures to pesticides for residents of two U.S. cities. Arch Environ Contam Toxicol 1994; 26:47–59. 147. Sexton K, Adgate JL, Eberly LE, et al. Predicting children’s short-term exposure to pesticides: results of a questionnaire screening approach. Environ Health Perspect 2003; 111:123–128. 148. Schwab M, Terblanche AP, Spengler JD. Self-reported exertion levels on time/activity diaries: application to exposure assessment. J Expo Anal Environ Epidemiol 1991; 1:339–356. 149. Reed KJ, Jimenez M, Freeman NC, Lioy PJ. Quantification of children’s hand and mouthing activities through a videotaping methodology. J Expo Anal Environ Epidemiol 1999; 9:513–520. 150. Zartarian VG, Streicker J, Rivera A, et al. A pilot study to collect micro-activity data of twoto four-year-old farm labor children in Salinas Valley, California. J Expo Anal Environ Epidemiol 1995; 5:21–34. 151. Akland GG, Pellizzari ED, Hu Y, et al. Factors influencing total dietary exposures of young children. J Expo Anal Environ Epidemiol 2000; 10:710–722. 152. O’Rourke MK, Lizardi PS, Rogan SP, Freeman NC, Aguirre A, Saint CG. Pesticide exposure and creatinine variation among young children. J Expo Anal Environ Epidemiol 2000; 10:672–681. 153. Tulve NS, Suggs JC, McCurdy T, Cohen Hubal EA, Moya J. Frequency of mouthing behavior in young children. J Expo Anal Environ Epidemiol 2002; 12:259–264. 154. Juberg DR, Alfano K, Coughlin RJ, Thompson KM. An observational study of object mouthing behavior by young children. Pediatrics 2001; 107:135–142. 155. Reed KJ. Quantification of Children’s Hand Mouthing Activities Through a Videotaping Methodology. New Jersey: Rutgers, the State University of New Jersey, 1997. 156. Zartarian VG, Ferguson AC, Ong CG, Leckie JO. Quantifying videotaped activity patterns: video translation software and training methodologies. J Expo Anal Environ Epidemiol 1997; 7:535–542. 157. Zartarian VG, Ferguson AC, Leckie JO. Quantified dermal activity data from a four-child pilot field study. J Expo Anal Environ Epidemiol 1997; 7:543–552. 158. Zartarian VG, Ferguson AC, Leckie JO. Quantified mouthing activity data from a four-child pilot field study. J Expo Anal Environ Epidemiol 1998; 8:543–553. 159. Elgethun K, Fenske RA, Yost MG, Palcisko GJ. Time-location analysis for exposure assessment studies of children using a novel global positioning system instrument. Environ Health Perspect 2003; 111:115–122. 160. Freeman NC, Ettinger A, Berry M, Rhoads G. Hygiene- and food-related behaviors associated with blood lead levels of young children from lead-contaminated homes. J Expo Anal Environ Epidemiol 1997; 7:103–118.

284

Barr et al.

161. Freeman NC, Sheldon L, Jimenez M, Melnyk L, Pellizzari E, Berry M. Contribution of children’s activities to lead contamination of food. J Expo Anal Environ Epidemiol 2001; 11:407–413. 162. Rohrer CA, Hieber TE, Melnyk LJ, Berry MR. Transfer efficiencies of pesticides from household flooring surfaces to foods. J Expo Anal Environ Epidemiol 2003; 13:454–464. 163. Lewis RG, Fortune CR, Blachard FT, Camann DE. Movement and desposition of two organophosphorus pesticides within a residence after interior and exterior applications. J Air Waste Manage Assoc 2001; 51:351. 164. Pang Y, MacIntosh DL, Camann DE, Ryan PB. Analysis of aggregate exposure to chlorpyrifos in the NHEXAS-Maryland investigation. Environ Health Perspect 2002; 110:235–240. 165. Esteban E, Rubin C, Hill R, Olson D, Pearce K. Association between indoor residential contamination with methyl parathion and urinary para-nitrophenol. J Expo Anal Environ Epidemiol 1996; 6:375–387. 166. Lewis RG, Fortmann RC, Camann DE. Evaluation of methods for monitoring the potential exposure of small children to pesticides in the residential environment. Arch Environ Contam Toxicol 1994; 26:37–46. 167. Clayton CA, Pellizzari ED, Whitmore RW, Quackenboss JJ, Adgate JL, Sexton K. Distributions, associations, and partial aggregate exposure of pesticides and polynuclear aromatic hydrocarbons in the Minnesota Children’s Pesticide Exposure Study (MNCPES). J Expo Anal Environ Epidemiol 2003; 13:100–111. 168. Fenske RA, Black KG, Elkner KP, Lee CL, Methner MM, Soto R. Potential exposure and health risks of infants following indoor residential pesticide applications. Am J Public Health 1990; 80:689–693. 169. Whyatt RM, Camann DE, Kinney PL, et al. Residential pesticide use during pregnancy among a cohort of urban minority women. Environ Health Perspect 2002; 110:507–514. 170. Butte W, Heinzow B. Pollutants in house dust as indicators of indoor contamination. Rev Environ Contam Toxicol 2002; 175:1–46. 171. Rudel RA, Brody JG, Spengler JD, et al. Identification of selected hormonally active agents and animal mammary carcinogens in commercial and residential air and dust samples. J Air Waste Manag Assoc 2001; 51:499–513. 172. Roinestad KS, Louis JB, Rosen JD. Determination of pesticides in indoor air and dust. J AOAC Int 1993; 76:1121–1126. 173. Fenske RA, Kedan G, Lu C, Fisker-Andersen JA, Curl CL. Assessment of organophosphorous pesticide exposures in the diets of preschool children in Washington State. J Expo Anal Environ Epidemiol 2002; 12:21–28. 174. USDA Agricultural Marketing Service. United States Department of Agriculture National Organic Program Final Rule. 7 CFR Part 205. Washington, D.C.: USDA, 2000. 175. MacIntosh DL, Kabiru C, Echols SL, Ryan PB. Dietary exposure to chlorpyrifos and levels of 3,5,6-trichloro-2-pyridinol in urine. J Expo Anal Environ Epidemiol 2001; 11:279–285. 176. Needham LL, Barr DB, Calafat AM. Characterizing children’s exposures: beyond NHANES. Neurotoxicology, 2005; 26(4):547–553. 177. Phillips DL, Pirkle JL, Burse VW, Bernert JT, Jr, Henderson LO, Needham LL. Chlorinated hydrocarbon levels in human serum: effects of fasting and feeding. Arch Environ Contam Toxicol 1989; 18:495–500. 178. CDC. Second National Report on Human Exposure to Environmental Chemicals. National Center for Environmental Health: Atlanta, GA. (Accessed 5 June, 2003 at www.cdc.gov/ exposurereport.) 179. Heudorf U, Angerer J. Metabolites of pyrethroid insecticides in urine specimens: current exposure in an urban population in Germany. Environ Health Perspect 2001; 109:213–217. 180. Castorina R, Bradman A, McKone TE, Barr DB, Harnly ME, Eskenazi B. Cumulative organophosphate pesticide exposure and risk assessment among pregnant women living in an agricultural community: a case study from the CHAMACOS Cohort. Environ Health Perspect 2003; 111:1640–1648.

Pesticide Exposure and Human Development

285

181. CDC. National Report on Human Exposure to Environmental Chemicals. National Center for Environmental Health: Atlanta, GA. (Accessed 21 Sep, 2002 at http://www.cdc.gov/nceh/dls/ report/.) 182. Bradman A, Eskenazi B, Barr DB, Bravo R, Castorina R, Chevrier J. et al. Organophosphate urinary metabolite levels during pregnancy and after delivery in women living in an agricultural community. Environ Health Perspect, 2005; 113:1802–1807. 183. Baker SE, Olsson AO, Barr DB. Isotope dilution high-performance liquid chromatographytandem mass spectrometry method for quantifying urinary metabolites of synthetic pyrethroid insecticides. Arch Environ Contam Toxicol 2004; 46:281–288. 184. Needham LL. Assessing exposure to organophosphorus pesticides by biomonitoring in epidemiologic studies of birth outcomes. Environ Health Perspect 2005; 113:494–498. 185. Morgan MK, Sheldon LS, Croghan CW, et al. Exposures of preschool children to chlorpyrifos and its degradation product 3,5,6-trichloro-2-pyridinol in their everyday environments. J Expo Anal Environ Epidemiol, 2005; 15:297–309. 186. Wilson NK, Chuang JC, Lyu C, Menton R, Morgan MK. Aggregate exposures of nine preschool children to persistent organic pollutants at day care and at home. J Expo Anal Environ Epidemiol 2003; 13:187–202. 187. Boeniger MF, Lowry LK, Rosenberg J. Interpretation of urine results used to assess chemical exposure with emphasis on creatinine adjustments: a review. Am Ind Hyg Assoc J 1993; 54:615–627. 188. Barr DB, Wilder LC, Caudill SP, Gonzalez AJ, Needham LL, Pirkle JL. Urinary creatinine concentrations in the U.S. population: implications for urinary biologic monitoring measurements. Environ Health Perspect 2005; 113:192–200. 189. Kissel JC, Curl CL, Kedan G, et al. Comparison of organophosphorus pesticide metabolite levels in single and multiple daily urine samples collected from preschool children in Washington State. J Expo Anal Environ Epidemiol 2005; 15:164–171. 190. Meeker JD, Barr DB, Ryan L, et al. Temporal variability of urinary levels of non-persistent insecticides in adult men. J Expo Anal Environ Epidemiol, 2005; 15:271–281. 191. Barr DB, Barr JR, Maggio VL, et al. A multi-analyte method for the quantification of contemporary pesticides in human serum and plasma using high-resolution mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2002; 778:99–111. 192. Denovan LA, Lu C, Hines CJ, Fenske RA. Saliva biomonitoring of atrazine exposure among herbicide applicators. Int Arch Occup Environ Health 2000; 73:457–462. 193. Bearer CF. Meconium as a biological marker of prenatal exposure. Ambul Pediatr 2003; 3:40–43. 194. Lu C, Anderson LC, Morgan MS, Fenske RA. Correspondence of salivary and plasma concentrations of atrazine in rats under variable salivary flow rate and plasma concentration. J Toxicol Environ Health 1997; 52:317–329.

13 Assessing In Utero Exposure to Cannabis and Cocaine Gale A. Richardson Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, U.S.A.

Marilyn A. Huestis Chemistry and Drug Metabolism, Intramural Research Program, National Institute on Drug Abuse, National Institutes of Health, Baltimore, Maryland, U.S.A.

Nancy L. Day Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania, U.S.A.

INTRODUCTION The purpose of this chapter is to describe two methods, self-report and biological assessment, for obtaining data on cannabis and cocaine use during pregnancy. The term cannabis will be used in this manuscript, except when a particular study focused exclusively on marijuana, as this is the term most commonly used globally and it encompasses exposure to all forms of the cannabis plant, including marijuana and hashish. The aim is to contrast these two general methodologies and to highlight the advantages and disadvantages of each in order to aid researchers in evaluating assessment options. We begin with a brief summary of the prevalence of cannabis and cocaine. Then, we discuss methods of interviewing women about their use and some of the biological techniques used to measure substance use. The chapter concludes with suggestions for future research. PREVALENCE Cannabis is the most commonly used illicit drug, both in the general population and among pregnant women. In the 2003 National Survey on Drug Use and Health (NSDUH) (1), 50.6% of women between the ages of 18 and 25 reported that they had ever used marijuana, 24% had used marijuana in the past year, and 13% reported use within the last month. For women 26 yr and older, the rates were lower: 35.8%, 4.7%, and 2.5% of the women reported lifetime, past year, and past month use, respectively. In the National Pregnancy and Health Survey (NPHS), a national survey of drug use in pregnant women, 2.9% reported that they had used marijuana during pregnancy (2). 287

288

Richardson et al.

Studies find different prevalence rates depending on the sociodemographic characteristics of the population that is studied, the time of interview, and the method of assessment. In our study of the effects of prenatal marijuana exposure, 30% of the original screening sample of 1360 low socioeconomic status women attending a prenatal clinic reported marijuana use during their first trimester of pregnancy (3). Tennes et al. (4) and Zuckerman et al. (5) reported similar prevalence rates, 32% and 27%, respectively, in comparable inner-city pregnant populations. Fried et al. (6) reported a much lower prevalence rate in their study of middle class women: 13% reported first trimester marijuana use. As with other illicit drugs, the prevalence of cannabis use in non-treatment samples decreases as pregnancy progresses. For example, in our marijuana cohort, rates of 54%, 28.6%, and 25% were reported in the first, second, and third trimesters, respectively (3). Women who use cannabis during pregnancy also differ from women who do not use cannabis in their sociodemographic characteristics, such as race, age, marital status, education, and income (3,4,6). The prevalence of cocaine use is lower than that of cannabis. The 2003 NSDUH (1) found that, among women aged 18 to 25, 12.4% had ever used cocaine, 5% had used in the past year, and 1.5% reported use within the past months. For women 26 yr and older, a similar percentage reported having ever used cocaine (12.2%), but the rates for use in the past year (1.1%) and the past months (0.4%) were lower. The NPHS (2) reported that 1.1% of women reported cocaine use during pregnancy. In our longitudinal study of prenatal cocaine exposure, 8% of the women attending a prenatal clinic reported first trimester cocaine or crack use (7). Rates of prenatal cocaine exposure in other research studies vary from 2.6% in a partially rural sample from Florida (8) to 18% in a sample from the inner city of Boston (5). As with cannabis, cocaine use decreases during pregnancy (2,8–10). Women who use cocaine during pregnancy also differ from women who do not use cocaine in age, marital status, income, and race (9–11). Thus, the rates of both cannabis and cocaine use during pregnancy differ by the population surveyed, the geographic locale (12), urban versus rural samples (8), method of ascertainment (13), sample characteristics such as maternal age, sociodemographic status, race/ethnicity (2), and timing during pregnancy (14). These differences must be considered when the researcher chooses the method of assessing prenatal drug exposure.

SELF-REPORT METHODS The most common self-report methods for measuring prenatal drug exposure are subjectcompleted or interviewer-administered questionnaires. There are trade-offs between these two types of questionnaires. Subject-completed instruments are more cost-effective, while interviews are more labor-intensive, and therefore, more expensive. Women may be more willing to report labeled behaviors on a self-completed questionnaire than on an interview, but self-completed questionnaires are also more prone to errors from reading problems, not understanding the questions, or lack of care. In addition, self-completed questionnaires require a simplified approach to data collection: Questions cannot be complex, the vocabulary must be at no more than a sixth grade reading level, and the flow of the instrument must be clear and direct. By contrast, interviewers can elicit more complex information, use more complicated flow patterns, and can make sure that the subject understands the questions and is capable of and willing to answer the questions. The interviewer can clarify the questions and any inconsistencies in the answers. However, interviewers can influence the responses of the subject through their attitudes or unconscious body language, so it is

In Utero Exposure to Cannabis and Cocaine

289

critical to carefully train and regularly monitor the interviewers to maintain the integrity of the interview process. When the investigator’s primary variable of interest is prenatal substance exposure, we feel that the interviewer-administered questionnaire provides superior data to the subject-completed instruments. Therefore, the remainder of this section will be concerned with interviewer-administered questionnaires. The importance of interviewer selection and training cannot be overemphasized. More accurate reporting of substance use is obtained when interviewers do not also provide medical care to the subject. In addition, subjects are more likely to participate and respond accurately when the interviewers are matched to subject characteristics. [The reader is referred to Babor et al. (15) and DelBoca and Noll (16) for more complete discussions of interviewer and task characteristics that might influence reporting.] Interviewers must be knowledgeable about the street names of drugs and their modes of use. For example, Richardson and Day (17) found that while women were hesitant to use the word “crack,” they would describe how they used the drug, providing an accurate description of the substance used. Morral et al. (18) also discussed the importance of using appropriate terms or jargon for drug use, particularly with adolescents. Because the use of many substances is illegal, pregnant women may avoid describing their use accurately for fear of reprisal, including the potential loss of their children. The assurance of confidentiality is critical to ensure honest and accurate reporting. Subjects must be assured that their information will not be released to hospital medical staff or in legal proceedings. The Department of Health and Human Services can issue a Confidentiality Certificate to the investigator in order to provide this protection. In addition to interviewer and confidentiality issues, the question format used in an interview is very important. Questions should be worded assuming that the behavior occurs; that is, “When you use cocaine, how much do you use?” will elicit more reported use than will “Do you use cocaine?” The use of memory aids is also helpful, e.g., using a calendar will trigger the memory of holidays and special events and will aid accurate recall of substance use (19). The length of the recall period also affects reporting accuracy. Women are more willing to report on past use than on current behavior (19,20). However, the period for recall must not be too long. For example, some test-retest correlations were lower for a 5-months than for a 3-months recall period in our cohort (21). In addition, assessment times should be standardized so that all women are interviewed at the same time periods. Using varying assessment points, such as the first prenatal visit, will yield non-comparable estimates of use, since the timing of that event can occur anytime from the first trimester to delivery. The pattern of substance use is described by the quantity (how much), frequency (how often), and duration (how long) of use. This information is particularly important in research on a teratogen, as the dose and timing of exposure during pregnancy are critical determinants of specific outcomes. The way in which questions are framed also affects the amount and accuracy of reported drug use. Studies often ask only about usual use, assuming that this variable will capture the totality of a woman’s substance use. However, questions about a woman’s usual quantity of use elicit what the subject believes is the socially acceptable response. In the Maternal Health Practices and Child Development (MHPCD) Project, we ask first about usual quantity, and then ask about the women’s maximum and minimum quantities. Data from our study of prenatal marijuana exposure showed that reports of the usual quantity of marijuana use represented only 37% of the total first trimester consumption, maximum quantity contributed an additional 45%, and minimum quantity an additional 17% (22). Similarly, in our study of prenatal cocaine exposure (9), the usual quantity of cocaine use represented 47% of the total first trimester consumption, maximum quantity contributed an additional 50%, and minimum quantity

290

Richardson et al.

an additional 3%. Thus, asking only about usual patterns will significantly under-estimate drug usage. Some studies have asked only about frequency of use, although it is important to know how much a woman uses during each episode. By measuring only frequency, it is not possible to separate women who use heavily on each occasion from those who use small amounts per occasion; clearly, their overall consumption will be substantially different. A scale that combines frequency and quantity allows the researcher to define patterns of use, for example, separating high frequency and high quantity users from high frequency and low quantity users. In addition, measuring duration allows evaluation of the time period when women used at a specific level or pattern, again a critical piece of data for teratologic studies. With the exception of tobacco, most women decrease drug use during pregnancy, usually during the first trimester (2,23). This decrease in substance use makes assessment more difficult, since the behavior the woman is trying to remember, and the researcher is trying to measure, is changing. In addition, women may not know that they are pregnant for a period of time after conception. Thus, they would not report drug use during this time as use during pregnancy. In our experience, the women conceptualize first trimester as the time from pregnancy recognition to the end of the third months. In the MHPCD project, data were collected for three intervals: the time from conception to pregnancy recognition, the period from pregnancy recognition to confirmation of the pregnancy, and the time from confirmation to the end of the first trimester. We and others have found that women do not have difficulty describing these time periods during pregnancy (14,19). In the prenatal marijuana study (3), from conception to recognition, 66% of the marijuana users reported that their marijuana use was similar to their use prior to pregnancy, rather than to what they had reported as their first trimester use. From recognition until confirmation of pregnancy, 33% of the women reported that their marijuana use was similar to prepregnancy rather than first trimester levels. Among women who smoked two or more joints per day prior to pregnancy, rates were even higher; 83% reported that from conception to recognition, their use was most similar to their reported prepregnancy rates, and 52% reported that from recognition to confirmation, they still smoked marijuana at their prepregnancy rates (19). In the prenatal cocaine study (9), in the sample of women who received prenatal care, from conception to recognition, 65% of cocaine users reported that their cocaine use was similar to their use prior to pregnancy, rather than to what they had reported as their first trimester use. From recognition until confirmation of pregnancy, 24% of women reported that they still used cocaine at their prepregnancy levels. In the sample of women with no prenatal care, rates were even higher—79% and 67%, respectively. Thus, in the absence of data for these relevant periods within the first trimester, prepregnancy drug use is a more accurate assessment of exposure during the early part of the first trimester and reported first trimester use more accurately describes use during the later part of the first trimester. However, if data on these time periods are collected, first trimester use is reported more accurately. There are several advantages to the use of interviews to obtain drug use information. They allow the researcher to differentiate patterns of substance use, particularly during early pregnancy, information that is critical for teratologic studies. Interviews also allow the researcher to measure use over longer time periods. Interviews are also the least expensive way to obtain information about the mode of drug use (e.g., sniffed powder cocaine versus smoked crack cocaine). Biological measures can document smoked crack use by identifying anhydroecgonine methyl ester in the urine, but this identification requires a chromatographic analysis of the urine which is not possible with onsite urine

In Utero Exposure to Cannabis and Cocaine

291

tests (24). These advantages of interviews can be optimized, as described above. However, none of the above methods can correct for the subject’s forgetting, inability to understand the questions, or deliberate misreporting. Because of these disadvantages, many studies employ biological tests to document substance use. These methods also have advantages and disadvantages.

BIOLOGICAL MEASURES Maternal Drug Testing Detection of in utero drug exposure traditionally meant testing the urine of the mother during pregnancy or at birth. The advantages of urine testing include the ready availability of both on-site and laboratory-based tests, proven accuracy and reliability, applicability to automation, reasonable pricing, well-established quality control programs, high drug concentrations compared to concentrations in other biological matrices, and a large scientific database for interpretation. Disadvantages include the need for private collection facilities and collectors of the appropriate sex, the ease of adulteration of the test, and most importantly, the short window of drug detection. Urine tests reflect drug use over only a few days (25,26), except in the case of heavy cannabis users. Depending on the frequency and randomness of the urine testing, identification of recent and heavier users is more likely. Richardson et al. (9) and Zuckerman et al. (5) found that among women who reported cocaine use on the interview, those who had a positive urine screen for cocaine reported significantly more cocaine use than did women with a negative screen. In addition, women who are selected for screening in a clinical setting are not a random group of pregnant women. They are a biased group identified on the basis of demographic and clinical indicators that are, in themselves, predictors of pregnancy outcome. Matera et al. (27) compared women who had and did not have urine screens for cocaine and found that women who were tested were less likely to be Caucasian and multiparous. In addition, women who have positive urine tests are more likely to be African American, older, single, and to use other drugs (9,27–29). Adulteration is also a major problem due to the ready availability of numerous products that can change the pH of the specimen and/or oxidize drugs and metabolites in urine. Simple dilution of the specimen by increasing water and other beverage consumption can lower drug concentrations below the cutoff of the assay. Another disadvantage of urine testing is that the urine may remain positive following chronic drug usage for an extended time, creating a situation where it is difficult to differentiate new drug use from residual drug excretion (30,31). These patterns can be differentiated by normalization of the urine drug concentration to urine creatinine concentration and comparison of consecutive specimens (30). There are alternatives to urinalysis for monitoring maternal drug use during pregnancy. Alternative matrices, such as oral fluid (saliva), sweat, or hair, can be selected based on the needs and design of the testing program. Each biological matrix offers unique information about the timing and amount of drug use and window of detection. The ease and invasiveness of collection methods vary, as does the stability of the analytes. Furthermore, parent drug, as well as metabolites, may be present in the specimen, facilitating test interpretation. There also are disadvantages associated with each matrix, including cost, turnaround time, collection device adsorption of drug, extensive specimen preparation procedures, and limited controlled drug administration data to aid in the interpretation of test results.

292

Richardson et al.

The demand for oral fluid testing is increasing due to interest in roadside detection of drug-impaired drivers. Commercial kits designed to test oral fluid are becoming more available and costs of the assays will be reasonable due to competition. Unfortunately, many manufacturers have simply tried to modify urine assays and adapt them to the new matrix. Drug metabolites are primarily found in urine, while oral fluid usually contains both parent drug and metabolites. Drug concentrations also may be much lower in oral fluid than in urine. In many cases, the assays have not been adequately modified to target different analyte profiles and at lower cutoffs. The advantages of these tests are that oral fluid is collected non-invasively and under direct observation without the need for specialized facilities. Generally, detection times of drugs in oral fluid are similar to the short detection times in blood, although lower cutoff concentrations can significantly increase detection windows. Concentrations in oral fluid may be useful for predicting active drug concentrations in plasma, although this varies considerably due to the different physiochemical characteristics of drugs. A short detection time may be useful if recent drug use is the focus of monitoring. It also may be more difficult to adulterate an oral fluid specimen due to direct observation during collection, although this has not been fully investigated. Some collection devices include citric acid or another chemical to stimulate saliva flow. However, stimulation of salivary flow increases the pH of oral fluid changing the distribution of drug between plasma and saliva. Stimulation of flow also increases the amount of fluid, diluting drug concentrations and reducing sensitivity. Many drugs, including the stimulant class, reduce saliva production making specimen collection difficult. On-site oral fluid tests are available for some drug classes and are being developed and improved for many other drugs. Although this is an active research area, much of the scientific information needed for interpretation of test results is not yet available. Most basic drugs diffuse more readily into oral fluid from blood due to its lower pH. Therefore, basic drugs usually are found in higher concentrations in oral fluid than in plasma. One major drawback to oral fluid testing is that testing for cannabinoids has not yet been optimized. The most abundant cannabinoid found in oral fluid is parent D9-tetrahydrocannabinol (THC), not the non-psychoactive 11-nor-9-carboxy-THC metabolite. THC in oral fluid primarily comes from contamination of the oral mucosa during smoking, not through diffusion of drug from the blood. Furthermore, there are indications that passive contamination of oral fluid from cannabis smoke in the environment can produce positive test results for up to an hour (32). Another major issue for cannabinoid testing in oral fluid is that the highly lipophilic THC may be tightly adsorbed to the specimen collection device, greatly reducing test sensitivity. Sweat testing is a more recent and less investigated method of monitoring drug use during gestation. The primary advantage is that the sweat patch is worn for one week, and captures drug use over the entire interval. Drugs in sweat can also reflect use as much as 24 to 48 hr prior to patch application. The sweat patch allows water to pass through the semi-permeable membrane and accumulates drug excreted throughout the week. Both parent drug and metabolites are excreted. Thrice-weekly urine and weekly sweat specimens from participants in a methadone maintenance treatment program were tested for opiates and cocaine (33,34). Weekly sweat testing was equivalent or better to thrice weekly urine tests in identifying cocaine and opiate drug use. Although the sweat patch accumulates drug excreted within a week, some portion of the drug also may be reabsorbed into the skin, degraded on the patch, and escape through the semi-permeable membrane. Another advantage of sweat testing is that it decreases the opportunity for adulteration. Each sweat patch has a unique identification number, will not adhere to the

In Utero Exposure to Cannabis and Cocaine

293

skin if removed, and punctures of the patch are readily visible. Sweat patch collectors are instructed to carefully look for any signs of tampering at the time of patch removal. Limitations to sweat testing include intra- and inter-subject variability in sweat production, low analyte concentrations, occasional skin sensitivity, and possible contamination of the patch from the environment and during specimen handling. In general, sweat patch testing is considered to be qualitative, not quantitative in nature. As is the case for oral fluid, this new technology lacks scientific data to guide the interpretation of test results. There does not appear to be a dose-concentration relationship and the question of residual excretion of drug in heavy chronic users long after last use has not been fully resolved. In addition, only a single laboratory is offering routine testing of the sweat patch and the cost has been estimated to be 1.7 times more than with urine testing for cocaine (35). Hair is another alternative matrix that may be used to evaluate drug use during pregnancy. The primary advantage of hair testing is the long window of drug detection, although that is dependent upon hair length. Several studies have found that hair analysis identifies more cocaine use than urine screens (13,36,37). Hair collection is less invasive than urine collection and is done under direct observation. Three centimeters of hair can be collected every three months to detect drug use efficiently in the previous three months. Although the cost of hair testing is much higher than urine testing, the number of specimens required and the amount of staff time to collect specimens is greatly reduced. Other advantages are the stability of drug analytes in hair at room temperature and the resultant ease of storage, handling, and shipping of specimens. If a repeat specimen is required, a new specimen that reflects the original time of sampling is easy to collect. An individual cannot abstain from drug use for a short period of time prior to hair collection and avoid detection, as can be the case with urine, oral fluid, and sweat. Further, adulteration of the hair specimen by bleaching, dyeing, or straightening is easily apparent. One limitation of hair testing is the differential incorporation of basic drugs into hair according to the melanin content of the hair (38–44). Darker hair contains more melanin than lighter colored hair and will most likely contain greater concentrations of basic drugs, like cocaine or methamphetamine, if exposed to the same amount of drug. This obviously complicates the interpretation of hair test results. In the future, we may find that normalization of basic drug concentrations to melanin concentrations in hair will reduce this apparent discrepancy. Neutral and acidic drugs, i.e., THC, do not appear to preferentially bind to melanin and may have less variable disposition into hair of different colors and with different melanin content. Hair testing is a sensitive technology to detect basic drugs, such as cocaine, but lacks sensitivity to detect cannabis use, as compared to urine tests. A second major limitation of hair testing is potential contamination by drugs in the environment (45–47). Washing procedures are designed to remove external drug contamination; however, all of the drug may not be removed. Whether it is possible to differentiate contamination from actual drug use remains a controversial subject. Hair is a highly complex matrix requiring extensive specimen preparation to release drug analytes from the hair fiber. Sensitive and specific procedures are required to eliminate interference from endogenous hair components and to identify and quantify low concentrations of drug analytes. Liquid chromatography-mass spectrometry (LC/MS), gas chromatography-mass spectrometry (GC/MS), and tandem mass spectrometry (MS/MS) techniques are necessary due to the picogram to nanogram concentrations of different drug analytes in hair. The cost of hair analysis is usually higher, turnaround times for test results may be longer, and fewer laboratories offer the testing. Another limitation of hair testing is that recent drug use is not detected. Approximately 10 days to two weeks are required for new

294

Richardson et al.

drug use to be incorporated into hair that is available for collection at the scalp surface. Other disadvantages include a high refusal rate (36) and a higher rate of lack of hair samples among African Americans (48). Neonatal Drug Testing Pregnant drug users may not present for prenatal care, permitting drug monitoring only at the time of delivery. Until recently, urine testing was the primary method of identifying prenatal drug exposure. As described above for maternal testing, urine has a short window of drug detection of a few days for most drugs, and a few weeks following heavy, chronic cannabis exposure. Furthermore, there are problems with collection of urine from newborns (49). The collection device frequently fails to properly adhere to newborn skin, with subsequent loss of urine into the diaper. The collection device also may irritate sensitive neonatal skin. Meconium testing is a useful means of detecting in utero drug exposure, offering an extensive window of detection for parent compounds and metabolites. Meconium begins to form in utero during the 12th to 13th weeks of gestation, accumulates throughout gestation, and serves as a reservoir for drugs and waste products. Meconium is easily and non-invasively collected from diapers for up to 72 hr from the time of birth until the appearance of milk stool, although excretion of meconium may be delayed in premature infants. Refrigeration of meconium is recommended after collection, although most drugs and metabolites appear to be stable at room temperature during shipment, with the exception of fatty acid ethyl esters for the analysis of alcohol exposure. Although meconium is one of the most promising matrices for monitoring in utero drug exposure, there are many issues that need to be resolved prior to optimal use of this biological fluid. Generally, the window of drug detection is considered to be the last 20 weeks of pregnancy, although in the Infant Development, Environment, and Lifestyle (IDEAL) study, a multi-center, longitudinal investigation of the effects of prenatal methamphetamine exposure, the greatest sensitivity was achieved with specimens collected from offspring of women who reported methamphetamine use in the third trimester (50). Casanova et al. (51) also found a bias toward detection of cocaine use within three weeks of delivery. Thus, an estimate of first trimester exposure may not be obtained with meconium. Meconium is a highly complex biological tissue with high bile acid and lipid content. These characteristics require extensive specimen preparation to maximize sensitivity, make it difficult to separate drugs of interest, and frequently reduce drug recovery from meconium. In addition, meconium is not homogeneous, requiring careful mixing of the specimen prior to analysis. Another disadvantage of meconium testing is that the scientific database supporting meconium testing is limited. Research is needed to determine the best analytes to include in testing for different drug classes. Although high concentrations of morphine, cocaine, methadone, methamphetamine, cannabinoids, and metabolites have been reported in meconium (51–61), screening and confirmatory tests do not always yield the same results. A common scenario is that meconium tests are positive by the immunoassay screening procedure, but are not confirmed by more specific confirmatory tests. This could be due to unique biomarkers produced by immature fetal metabolism, but there are other feasible explanations. Meconium may contain interfering cross-reacting substances and/or poor recovery may contribute to the lack of sensitivity of confirmatory procedures. Screening methods for meconium include enzyme-, fluorescent polarization-, and radio-immunoassays (61,62). In most cases, immunoassays developed for urine testing are applicable due to presence of both parent drug and metabolites in meconium. Unfortunately, unique

In Utero Exposure to Cannabis and Cocaine

295

metabolites, present in variable concentrations, may have poor cross-reactivity in assays designed for urine testing. Confirmation can be obtained by LC/MS, GC/MS, and MS/MS. Early procedures attempted to identify in utero cocaine exposure by monitoring meconium for cocaine, benzoylecgonine (BE), and cocaethylene (63–68). In the Maternal Lifestyle Study (MLS), 90% of meconium specimens were confirmed by analyzing for BE only, while addition of m-hydroxybenzoylecgonine (m-OH-BE) to the standard testing procedures for cocaine and BE increased the identification of cocaineexposed infants by 10% (12). Innovative approaches have documented that inclusion of other cocaine metabolites, such as norcocaine, benzoylnorecgonine, m- and p-hydroxy cocaine, and m- and p-OH-BE, improves the sensitivity of cocaine detection (65,69–73). Oyler et al. (70) also reported m-OH-BE to be the most abundant cocaine analyte in meconium following in utero cocaine exposure. More than 75% of positive screening tests for cocaine can be confirmed when cocaine, BE, and m-OH-BE are monitored. Most research on meconium detection of in utero drug exposure has focused on cocaine and opiate use, while few studies describe the detection of cannabinoids. Meconium specimens were also analyzed for cannabinoids in the MLS study, although many in utero cannabinoid exposures may have been missed because only parent THC and inactive 11-nor-9-carboxy-THC metabolite were measured. Only 7.2% of meconium specimens screened positive for THC and only 36.5% of these were confirmed (12). ElSohly and Feng (74) identified two additional metabolites, 11-hydroxy-THC and 8-beta11-dihydroxy-THC, in the meconium of infants exposed to cannabis. Addition of these two analytes to confirmation procedures improved the detection of gestational cannabis use. In the IDEAL study, for specimens that screened positive, only 20% of cannabis specimens were confirmed (50). It is apparent that significant additional research is needed to identify potential biomarkers in meconium for the detection of cannabinoid exposure. In addition, proper interpretation of meconium test results requires that the researcher determine if a history of drug exposure can be obtained through serial analyses of meconium specimens, as some have suggested (75). Resolution of this controversial idea requires determination of whether drugs deposited in the fetal gastrointestinal track diffuse within meconium and whether they can be re-absorbed later in gestation. Another important issue is the degree of contamination of meconium by neonatal urine and how this affects interpretation of meconium test results. Perhaps fetal metabolism and excretion are characteristic of different gestational ages, producing unique metabolites that could suggest the timing of in utero drug exposure. Another neonatal matrix for monitoring in utero drug exposure is hair. Hair can be collected non-invasively and provides a wide window of drug detection from the time of earliest hair growth, although one cannot obtain first trimester information. Obtaining permission to collect infant hair specimens is difficult because some cultures discourage the cutting of hair, and many new mothers are reluctant to permit collection of newborn hair. In addition, there are few research reports on neonatal hair tests to guide result interpretation. As with adults, recent drug use cannot be detected in neonatal hair specimens. Approximately 10 days to two weeks is necessary for drug recently incorporated into hair to reach the scalp surface. However, if drug use occurs close to the time of delivery, it is possible that the drug could be absorbed into the neonates’ hair from the surrounding amniotic fluid. In addition, the neonate’s hair reflects only drugs used by the mother during gestation, eliminating the issue of environmental contamination. As with adults, the problem of hair color bias remains a relevant issue for testing neonatal hair. Few data are available on the relationship of amount and frequency of maternal drug consumption and concentrations in neonatal hair. However, it has been shown that drugs

296

Richardson et al.

and metabolites are incorporated into this biological matrix following in utero drug exposure. Cocaine, BE, norcocaine, and cocaethylene can be detected in hair following cocaine use and sensitive procedures are available for monitoring drug concentrations (37,76,77). Exposure to cannabis smoke can result in the incorporation of the parent THC in hair. However, in workplace and forensic testing, demonstration of the acidic 11-nor9-carboxy-THC metabolite is required to document use of the drug (78,79).

COMPARISONS OF SELF-REPORT AND BIOLOGICAL MARKERS There have been numerous investigations of the correspondence between self-report instruments and biological markers. The agreement between positive self-reports and positive toxicology screens varies widely from 33% (29) to 81% (80), with a meta-analysis reporting a median kappa of 0.42 for high-risk samples (81). One important factor in determining the extent of agreement is the level of drug use. The agreement between selfreport assessments and biological markers is greater with higher levels of drug use (82). In addition, the measures will have better agreement when the time frame covered by the questions is the same as the biological window for ascertainment. Another factor influencing agreement is the type of drug. Marijuana was detected more often by selfreport than by biological measures, whereas the reverse was true for cocaine (20,48,83). Agreement is also influenced by the purpose of the biological measurements. For example, research from treatment samples has found that agreement between self-report and biological measures is higher at treatment intake, when motivation to report drug use is thought to be higher, than at follow-up (48,80). Disagreement between self-report and biological measures can occur in two ways. One, the self-report is positive and the biological measure is negative. There is substantial evidence that a high percentage of women who admit drug use on self-report have negative toxicology screens (9,20,29,84,85). In general, this results from drug use that occurred outside of the window of detection of the biological assay. Thus, if one were to rely on biological screens alone, many of the women who used would be missed. Two, the biological measure is positive and the self-report is negative. This may result from mistakes in reporting or deliberate misreporting. The extent of disagreement will vary depending on the nature of the sample, as well as the issues of instrument design, interviewer training, confidentiality, and assay sensitivity, as discussed above. For example, in our study of prenatal cocaine exposure, 100% of the women in the prenatal care sample who had a positive urine screen were identified by our study interviewer as users. Reporting was less accurate among women in the no prenatal care sample; 85% of the women who had positive urine screens admitted use on the interview. Thus, 15% were misreporters (9). The sociodemographic characteristics of women who misreport are not consistently different from those who report accurately, so it is not possible to construct a profile that could be used to identify such women. Richardson et al. (9) found no differences in demographic or drug use characteristics between those who denied use and those who admitted use, nor did Rutherford et al. (85) and Zuckerman et al. (86). Fendrich et al. (20) found no racial differences in misreporting of cocaine use, although they did find that African Americans were more likely to misreport marijuana use. By contrast, Grant et al. (87) found that women who misreported cocaine use were more likely to be African American, single, and multiparous. The issue of misreporting highlights the importance of considering the characteristics of the sample to be recruited.

In Utero Exposure to Cannabis and Cocaine

297

CONCLUSIONS AND RECOMMENDATIONS This review has highlighted advantages and disadvantages of self-report and biological methods for assessing prenatal cannabis and cocaine exposure. The aim of this chapter was to provide the researcher with information to evaluate the different types of measurement and to optimize the assessments used in their research. The decision about which method to use must consider the advantages and disadvantages of each method, as well as the purpose of the study, recruitment source, sample characteristics, and financial considerations. In teratologic studies, the pattern, duration, and timing of use during pregnancy are important parameters in determining the effects of substance exposure on the offspring. As we discussed, it is necessary to interview the women to identify these parameters of substance use. Further, interviews identify use that is below the cutoff of detection for biological assessment, or outside of the time frame of detection, and allow the investigator to assess use over a longer time period. To provide accurate data using an interview, it is important that the assessment is carefully designed and the administration carefully monitored. However, interviews are subject to errors in reporting whether from forgetting, deliberate misreporting, or from problems in the design and administration of the assessment. Biological measures are not subject to the problems that might occur with self-report instruments. Biological specimens provide more accurate assessments and a high degree of reliability. They are not, however, able to determine the pattern of use. The choice of which biological measure to employ should be determined primarily by the desired time frame of detection, as well as by the acceptability of the specimen collection and the reliability of the measure within the collection system. When employing biological assessments, it also is critical to include confirmation of all positive screening results. A percentage of negative tests should be confirmed as well to ensure that the specimens are correctly collected and analyzed (12,88). There is a need for further development of the instruments and methodologies used for collection of substance use data by interview. More exploration is needed to develop better methods to define use and to ask about substance use in a way that allows the subject to report accurately and honestly. Investigations concerning how women think about drugs and how to frame questions around their drug use experience would be important contributions. The section on biological measures has detailed the current state of biological assessment and outlined recommendations for improvement in biological measures. For example, controlled drug administration data in non-pregnant drug users and controlled therapeutic drug administration data (i.e., buprenorphine and methadone in pregnant opiate addicts) are needed to guide the interpretation of results for sweat, oral fluid, and meconium testing. While it is important to develop better methods within modalities, it is equally critical to develop ways to use biological and interview methods together to validate and enhance the collection of data that will accurately describe the pattern of use during pregnancy. Accurate assessment of drug effects on the fetus requires determining the type, amount, frequency, and timing of maternal drug use. To this end, it is a priority to develop measures that have equivalent time frames and that can address the specific type of drug and method of drug use. We also need to develop methods to combine the data from the two modalities of measurement so that the advantages of each can be maximized. For example, protocols should be developed that outline ways of combining the information on pattern of use with the laboratory data on the amount of use during a specific window of time. In addition, it is important to define the relationships between drug exposure, drug concentrations, and neonatal and maternal outcome measures. This would require

298

Richardson et al.

developing methods to monitor women throughout gestation using both interview and biological measures to link neonatal and developmental outcomes with in utero drug exposure. In summary, there are advantages and disadvantages of both interview and biological assessments. The optimal assessment strategy is the combined use of interview and biological measures, which will yield the most accurate information on in utero drug exposure for epidemiologic research. This strategy will also allow validation across modalities. It is important to begin to develop methods to maximize the utility of each and to combine the information from each separate modality, so that it is possible to interpret the data jointly rather than separately.

ACKNOWLEDGMENTS Portions of this work were supported by the National Institute on Drug Abuse (DA008916, G. Richardson, Principal Investigator: DA03874, N. Day, Principal Investigator).

REFERENCES 1. Substance Abuse and Mental Health Services Administration. Results from the 2003 National Survey on Drug Use and Health: National Findings. Office of Applied Studies, NSDUH Series H-25, DHHS Publication No. SMA 04-3964. Rockville, MD, 2004. 2. National Institute on Drug Abuse, National Pregnancy and Health Survey: Drug Use Among Women Delivering Livebirths: 1992 National Institutes of Health, Publication No. 96-3819, NIH, Rockville, MD, 1996. 3. Day NL, Sambamoorthi U, Taylor P, et al. Prenatal marijuana use and neonatal outcome. Neurotoxicol Teratol 1991; 13:329–334. 4. Tennes K, Avitable N, Blackard C, et al. Marijuana: prenatal and postnatal exposure in the human. In: Pinkert TM, ed. Consequences of Maternal Drug Abuse. In: NIDA Research Monograph No. 59, Rockville, MD, 1985:48–60. 5. Zuckerman B, Frank DA, Hingson R, et al. Effects of maternal marijuana and cocaine use on fetal growth. N Engl J Med 1989; 320:762–768. 6. Fried PA, Watkinson B, Grant A, et al. Changing patterns of soft drug use prior to and during pregnancy: a prospective study. Drug Alcohol Depend 1980; 6:323–343. 7. Richardson GA, Day NL, McGauhey PJ. The impact of prenatal marijuana and cocaine use on the infant and child. Clin Obstet Gynecol 1993; 36:302–318. 8. Eyler F, Behnke M, Conlon M, et al. Prenatal cocaine use: a comparison of neonates matched on maternal risk factors. Neurotoxicol Teratol 1994; 16:81–87. 9. Richardson GA, Hamel SC, Goldschmidt L, et al. Growth of infants prenatally exposed to cocaine/crack: comparison of a prenatal care and a no prenatal care sample. Pediatrics 1999; 104:e18. 10. Singer LT, Salvator A, Arendt R, et al. Effects of cocaine/polydrug exposure and maternal psychological distress on infant birth outcomes. Neurotoxicol Teratol 2002; 24:127–135. 11. Eyler FD, Behnke M, Conlon M, et al. Birth outcome from a prospective, matched study of prenatal crack/cocaine use: I. Interactive and dose effects on health and growth. Pediatrics 1998; 101:229–237. 12. Lester BM, ElSohly M, Wright LL, et al. The maternal lifestyle study: drug use by meconium toxicology and maternal self-report. Pediatrics 2001; 107:309–317. 13. Markovic N, Ness RB, Cefilli D, et al. Substance use measures among women in early pregnancy. Am J Obstet Gynecol 2000; 183:627–632.

In Utero Exposure to Cannabis and Cocaine

299

14. Hertz-Picciotto I, Pastore LM, Beaumont JJ. Timing and patterns of exposures during pregnancy and their implications for study methods. Am J Epidemiol 1996; 143:597–607. 15. Babor TF, Brown J, Del Boca FK. Validity of self-reports in applied research on addictive behaviors: fact or fiction? Behav Asses 1990; 12:5–31. 16. Del Boca FK, Noll JA. Truth or consequences: the validity of self-report data in health services research on addictions. Addiction 2000; 95:S347–S360. 17. Richardson GA, Day NL. Detrimental effects of prenatal cocaine exposure: illusion or reality? J Am Acad Child Adolesc Psychiatry 1994; 33:28–34. 18. Morral AR, McCaffrey DF, Chien S. Measurement of adolescent drug use. J Psychoactive Drugs 2003; 35:301–309. 19. Day NL, Wagener DK, Taylor PM. Measurement of substance use during pregnancy: methodological issues. In: Pinkert TM, ed. Current Research on the Consequences of Maternal Drug Abuse. In: NIDA Research Monograph No. 59. Washington, DC: US Government Printing Office, 1985:36–47. 20. Fendrich M, Johnson TP, Wislar JS, et al. The utility of drug testing in epidemiological research: results from a general population survey. Addiction 2004; 99:197–208. 21. Robles N, Day NL. Recall of alcohol consumption during pregnancy. J Stud Alcohol 1990; 51:403–407. 22. Day NL, Robles N. Methodological issues in the measurement of substance use. Ann NY Acad Sci 1989; 562:8–13. 23. Day NL, Richardson GA, Geva D, et al. Alcohol, marijuana and tobacco: effects of prenatal exposure on offspring growth and morphology at age six. Alcohol Clin Exp Res 1994; 18:786–794. 24. Jacob P, III, Lewis ER, Elias-Baker BA, et al. A pyrolysis product, anhydroecgonine methyl ester (methylecgonidine), is in the urine of cocaine smokers. J Anal Toxicol 1990; 14:353–357. 25. Julien R. 6th ed. A Primer of Drug Action. New York: WH Freeman Company, 1992. 26. Vereby K. Cocaine abuse detection by laboratory methods. In: Washton AM, Gold MS, eds. Cocaine: A Clinician’s Handbook. New York: Guilford Press, 1987:214–228. 27. Matera C, Warren WB, Moomjy M, et al. Prevalence of use of cocaine and other substances in an obstetric population. Am J Obstet Gynecol 1990; 163:797–801. 28. Lindsay MK, Carmichael S, Peterson H, et al. Correlation between self-reported cocaine use and urine toxicology in an inner-city prenatal population. J Natl Med Assoc 1997; 89:57–60. 29. Ashling K, Gross AH, Coghlin DT, et al. Prevalence of positive urine drug screens in a prenatal clinic: correlation with patients’ self-report of drug use. R I Med 1994; 77:371–373. 30. Huestis MA, Cone EJ. Differentiating new marijuana use from residual drug excretion in occasional marijuana users. J Anal Toxicol 1998; 22:445–454. 31. Preston KL, Silverman K, Schuster CR, et al. Assessment of cocaine use with quantitative urinalysis and estimation of new uses. Addiction 1997; 92:717–727. 32. Niedbala S, Kardos K, Salamone S, et al. Passive cannabis smoke exposure and oral fluid testing. J Anal Toxicol 2004; 28:546–552. 33. Preston KL, Huestis MA, Wong CJ, et al. Monitoring cocaine use in substance abuse treatment patients by sweat and urine testing. J Anal Toxicol 1999; 23:313–322. 34. Huestis MA, Cone EJ, Wong CJ, et al. Monitoring opiate use in substance abuse treatment patients with sweat and urine drug testing. J Anal Toxicol 2000; 24:509–521. 35. Winhusen TM, Somoza EC, Singal B, et al. Measuring outcome in cocaine clinical trials: a comparison of sweat patches with urine toxicology and participant self-report. Addiction 2003; 98:317–324. 36. Kline J, Ng SK, Schittini M, et al. Cocaine use during pregnancy: sensitive detection by hair assay. Am J Pub Health 1997; 87:352–358. 37. Kuhn L, Kline J, Ng S, et al. Cocaine use during pregnancy and intrauterine growth retardation: new insights based on maternal hair tests. Am J Epidemiol 2000; 152:112–119. 38. Potsch L, Skopp G, Moeller MR. Influence of pigmentation on the codeine content of hair fibers in guinea pigs. J Forensic Sci 1997; 42:1095–1098.

300

Richardson et al.

39. Potsch L, Skopp G, Rippin G. A comparison of 3H-cocaine binding on melanin granules and human hair in vitro. Int J Legal Med 1997; 110:55–62. 40. Kikura R, Nakahara Y. Studies on mechanism of drug incorporation into hair. Kokuritsu Iyakuhin Shokuhin Eisei Kenkyusho Hokoku 1998; 116:30–45. 41. Joseph RE, Jr Su TP, Cone EJ. In vitro binding studies of drugs to hair: influence of melanin and lipids on cocaine binding to Caucasoid and Africoid hair. J Anal Toxicol 1996; 20:338–344. 42. Cone EJ, Joseph RE, Jr. The potential for bias in hair testing for drugs of abuse. In: Kintz P, ed. Drug Testing in Hair. Boca Raton, FL: CRC Press, 1996:69–93. 43. Gygi SP, Joseph RE, Jr., Cone EJ, et al. Incorporation of codeine and metabolites into hair. Role of pigmentation. Drug Metab Dispos 1996; 24:495–501. 44. Kidwell DA, Blank DL. Mechanisms of incorporation of drugs into hair and the interpretation of hair analysis data. In: Cone EJ, Welch MJ, Babecki MBG, eds. Hair Testing for Drugs of Abuse: International Research on Standards and Technology. In: NIH Pub. No. 95-3727. Rockville: National Institute on Drug Abuse, 1995:19–90. 45. Blank DL, Kidwell DA. External contamination of hair by cocaine: an issue in forensic interpretation. Forensic Sci Int 1993; 63:145–156. 46. Wang WL, Cone EJ, Testing human hair for drugs of abuse. V. Attempted differentiation of environmental cocaine contamination from active cocaine use. Second International Meeting on Clinical and Forensic Aspects of Hair Analysis, Genoa, Italy, June 1994. 47. Wang WL, Cone EJ. Testing human hair for drugs of abuse. IV. Environmental cocaine contamination and washing effects. Forensic Sci Int 1995; 70:39–51. 48. Hindin R, McCusker J, Vickers-Lahti M, et al. Radioimmunoassay of hair for determination of cocaine, heroin, and marijuana exposure: comparison with self-report. Int J Addict 1994; 29:771–789. 49. Kwong TC, Ryan RM. Detection of intrauterine illicit drug exposure by newborn drug testing. Clin Chem 1997; 43:235–242. 50. Zhao Z, Liu J, LaGasse LL, et al. Methamphetamine and amphetamine concentrations in meconium of neonates of women enrolled in the IDEAL study of in utero methamphetamine exposure 2004 FBI/SOFT/TIAFT Meeting, Washington, DC, August 28–September 3, 2004. 51. Casanova OQ, Lombardero N, Behnke M, et al. Detection of cocaine exposure in the neonate. Analyses of urine, meconium, and amniotic fluid from mothers and infants exposed to cocaine. Arch Pathol Lab Med 1994; 118:988–993. 52. Clark GD, Rosenzweig IB, Raisys VA, et al. The analysis of cocaine and benzoylecgonine in meconium. J Anal Toxicol 1992; 16:261–263. 53. Wingert WE, Feldman MS, Kim MH, et al. A comparison of meconium, maternal urine and neonatal urine for detection of maternal drug use during pregnancy. J Forensic Sci 1994; 39:150–158. 54. Ostrea EM, Jr., Knapp DK, Romero AI, et al. Meconium analysis to assess fetal exposure to nicotine by active and passive maternal smoking. J Pediatr 1994; 124:471–476. 55. Ostrea EM, Jr., Brady M, Gause S, et al. Drug screening of newborns by meconium analysis: a large-scale, prospective, epidemiologic study. Pediatrics 1992; 89:107–113. 56. Ostrea EM, Jr., Romero A, Yee H. Adaptation of the meconium drug test for mass screening. J Pediatr 1993; 122:152–154. 57. Rosengren SS, Longobucco DB, Bernstein BA, et al. Meconium testing for cocaine metabolite: prevalence, perceptions, and pitfalls. Am J Obstet Gynecol 1993; 168:1449–1456. 58. Goosensen M, Stolk LML, Smit BJ. Analysis of 11-nor-delta-9-THC-carboxylic acid in meconium with immunoassay and HPLC diode-array detection. J Anal Toxicol 1995; 19:330. 59. ElSohly MA, Feng S, Murphy TP. Analysis of methadone and its metabolites in meconium by enzyme immunoassay (EMIT) and GC/MS. J Anal Toxicol 2001; 25:40–44. 60. Maynard EC, Amoruso LP, Oh W. Meconium for drug testing. Am J Dis Child 1991; 145:650–652.

In Utero Exposure to Cannabis and Cocaine

301

61. Ostrea EM, Jr., Brady MJ, Parks PM, et al. Drug screening of meconium in infants of drugdependent mothers: an alternative to urine testing. J Pediatr 1989; 115:474–477. 62. DiGregorio GJ, Ferko AP, Barbieri EJ, et al. Determination of cocaine usage in pregnant women by a urinary EMIT drug screen and GC-MS analyses. J Anal Toxicol 1994; 18:247–250. 63. Ostrea EM, Jr., Welch RA. Detection of prenatal drug exposure in the pregnant woman and her newborn infant. Clin Perinatol 1991; 18:629–645. 64. Callahan CM, Grant TM, Phipps P, et al. Measurement of gestational cocaine exposure: sensitivity of infants’ hair, meconium, and urine. J Pediatr 1992; 120:763–768. 65. Lewis DE, Moore CM, Leikin JB. Cocaethylene in meconium specimens. J Toxicol Clin Toxicol 1994; 32:697–703. 66. Koren G, Klein J, Forman R, et al. Biological markers of intrauterine exposure to cocaine and cigarette smoking. Dev Pharmacol Ther 1992; 18:228–236. 67. Lombardero N, Casanova O, Behnke M, et al. Measurement of cocaine and metabolites in urine, meconium and diapers by gas chromatography/mass spectrometry. Ann Clin Lab Sci 1993; 23:385–394. 68. Browne SP, Tebbett IR, Moore CM, et al. Analysis of meconium for cocaine in neonates. J Chromatogr 1992; 575:158–161. 69. Steele BW, Bandstra ES, Wu NC, et al. m-Hydroxybenzoylecgonine: an important contributor to the immunoreactivity in assays for benzoylecgonine in meconium. J Anal Toxicol 1993; 17:348–352. 70. Oyler J, Darwin WD, Preston KL, et al. Cocaine disposition in meconium from newborns of cocaine-abusing mothers and urine of adult drug users. J Anal Toxicol 1996; 20:453–462. 71. Lewis D, Moore C, Becker J, et al. Prevalence of meta-hydroxybenzoylecgonine (M-OH-BZE) in meconium samples. Bull Int Assoc Forensic Toxicol 1995; 25:33–36. 72. Murphey LJ, Olsen GD, Konkol RJ. Quantitation of benzoylnorecgonine and other cocaine metabolites in meconium by high-performance liquid chromatography. J Chromatogr 1993; 613:330–335. 73. Browne S, Moore C, Negrusz A, et al. Detection of cocaine, norcocaine, and cocaethylene in the meconium of premature neonates. J Forensic Sci 1994; 39:1515–1519. 74. ElSohly M, Feng S. Delta-9-THC metabolites in meconium: identification of 11-OH-delta9-THC, 8 beta, 11-diOH-delta-9-THC, and 11-nor-delta-9-THC-9-COOH as major metabolites of delta-9-THC. J Anal Toxicol 1998; 22:329–335. 75. Ostrea EM, Jr., Knapp D, Tannenbaum L. Serial meconium drug analysis can estimate the chronology and degree of the infant’s in utero drug exposure. Pediatr Res 1993; 33:229A. 76. Scheidweiler KB, Cone EJ, Moolchan ET, et al. Dose-related distribution of codeine, cocaine, and metabolites into human hair following controlled oral codeine and subcutaneous cocaine administration. J Pharmacol Exp Ther 2005; 313:909–915. 77. Bourland JA, Hayes EF, Kelly RC, et al. Quantitation of cocaine, benzoylecgonine, cocaethylene, methylcgonine, and norcocaine in human hair by positive ion chemical ionization (PICI) gas chromatography-tandem mass spectrometry. J Anal Toxicol 2000; 24:489–495. 78. Musshoff F, Junker HP, Lachenmeier DW, et al. Fully automated determination of cannabinoids in hair samples using headspace solid-phase microextraction and gas chromatograph–mass spectrometry. J Anal Toxicol 2002; 26:554–560. 79. Staub C. Chromatographic procedures for determination of cannabinoids in biological samples, with special attention to blood and alternative matrices like hair, saliva, sweat and meconium. J Chromatogr 1999; 733:119–126. 80. Buchan BJ, Dennis ML, Tims FM, et al. Cannabis use: consistency and validity of self-report, on-site urine testing and laboratory testing. Addiction 2002; 97:98–108. 81. Magura S, Kang SY. Validity of self-reported drug use in high risk populations: a metaanalytical review. Subst Use Misuse 1996; 31:1131–1153. 82. Morral AR, McCaffrey D, Iguchi MY. Hardcore drug users claim to be occasional users: drug use frequency underreporting. Drug Alcohol Depend 2000; 57:193–202.

302

Richardson et al.

83. Neale J, Robertson M. Comparisons of self-report data and oral fluid testing in detecting drug use amongst new treatment clients. Drug Alcohol Depend 2003; 71:57–64. 84. Horrigan TJ, Piazza N. The Substance Abuse Subtle Screening Inventory minimizes the need for toxicology screening of prenatal patients. J Subst Abuse Treat 1999; 17:243–247. 85. Rutherford MJ, Cacciola JS, Alterman AI, et al. Contrasts between admitters and deniers of drug use. J Subst Abuse Treat 2000; 18:343–348. 86. Zuckerman B, Amaro H, Cabral H. Validity of self-reporting of marijuana and cocaine use among pregnant adolescents. J Pediatr 1989; 115:812–815. 87. Grant T, Brown Z, Callahan C, et al. Cocaine exposure during pregnancy: improving assessment with radioimmunoassay of maternal hair. Obstet Gynecol 1994; 83:524–531. 88. Strano-Rossi S. Methods used to detect drug abuse in pregnancy: a brief review. Drug Alcohol Depend 1999; 53:257–271.

14 Neurobehavioral Test Batteries for Children Diane S. Rohlman and W. Kent Anger Center for Research on Occupational and Environmental Toxicology, Oregon Health and Science University, Portland, Oregon, U.S.A.

IDENTIFYING THE PROBLEM Growing List of Chemicals One of the many factors that affect the growth and development of children is chemicals in the environment. Children are exposed to them through the air they breathe, the food they eat and the water they drink (1). The majority of these chemicals are not evaluated for their potential toxicity, effects on development, or interactive effects with other chemicals prior to commercial introduction (2,3). Attention has been focused on the interaction of chemicals in the environment and neurodevelopment in children (4–6). Children from all cultures and backgrounds are at risk. However, minorities and children from low-income families are often at greater risk because of poor nutrition, an impoverished environment, and limited access to medical care (4,7–9). In order to evaluate the impact of environmental exposures on neurodevelopment it is necessary to have effective methods that will allow accurate conclusions to be drawn. The selection or development of appropriate methods to test children and quality control steps when using these methods are the topic of this chapter.

Children More Vulnerable The magnitude of the task of toxicity testing is made more daunting by our lack of knowledge of the target organ—the effect(s) that will occur at the lowest concentration— and the likelihood that the target organ will differ at different points in time as development unfolds. Children have greater exposure to toxicants than adults due to both behavior (e.g., increased time spent crawling on the floor and increased hand-to-mouth behavior) and their increased food:body-mass ratio (they consume a greater amount of food and drink compared to their body ratio than adults) (5,10,11). The developing brain and organ systems of infants and children, as well as their immature metabolism, also make them more vulnerable to environmental toxicants (5,12). Periods of rapid brain development and the absence of a blood-brain barrier in fetuses and young infants allow neurotoxicants access to the developing brain during critical times of development (13). In contrast, adult exposure to neurotoxicants occurs in a developed 303

304

Rohlman and Anger

brain and may lead to alterations in brain functions and structures that were at one time normal (5,13–15). Finally, children have more years to live than adults and therefore have more time to develop diseases produced by slow-acting early exposures (5,13,15,16). Neurobehavioral Measures Reveal Chemical Effects Neurotoxic effects of environmental chemicals can vary along a continuum from minor subclinical deficits in sensory memory, motor, or cognitive functioning to mental retardation and clinical diseases (17,18). The type of exposure, the developmental stage, and the pattern and duration of exposure can all influence the outcome of exposure (13,19). Although severe effects of neurotoxic exposure are readily apparent, subtle effects are not as easily detected—especially in a standard clinical exam. Neurobehavioral tests have been used to identify adverse health effects of toxins in both adults (20) and children (21). In children, neurobehavioral tests were first used systematically to identify subclinical deficits in children with low levels of lead exposure (22). Other early research used neurobehavioral tests to identify subtle cognitive deficits in children born from moderate-drinking mothers (23). Since that time the research with children has expanded to include assessment of a variety of toxicants including environmental exposures and drugs of abuse. The impact on neurodevelopment of three environmental toxicants, lead, methyl mercury, and polychlorinated biphenyls (PCBs), referred to as the “Big Three” by Reference 21, has been extensively studied (21,24). These studies have demonstrated impaired cognitive functioning in children and adults exposed to these toxicants. Studies examining these three chemicals have used a variety of methods—both cross-sectional and longitudinal—and a range of measures. Often a prospective longitudinal design is used to examine the impact of prenatal exposure to different environmental toxicants on children from birth to older ages (25). Although these studies are important and necessary for determining the developmental effects of toxicants on the nervous system, they require substantial resources to develop and then to follow a cohort over time. Cross-sectional studies are another, much less expensive approach that can be used to examine the effects exposure to a toxicant. They can establish the effects of exposures at one point in time, and may also be useful in guiding the design and focus of more comprehensive longitudinal studies (25).

PLANNING THE RESEARCH STUDY The following is directed to the reader who intends to embark on a research study to determine if a particular chemical or chemical mix is producing adverse effects in the nervous system of children. In addition, it is assumed that there is an exposed sample of children to be tested, and that a similar sample of unexposed children of the same socioeconomic background and age range is available to serve as a comparison group. This defines the classic cross-sectional study that has been reported in literally hundreds of publications that detected, and to an extent characterized, neurotoxic effects in adults (20). Test Selection The first step in planning a study is to determine how much is known about the chemical to which the children are exposed. The chemical is either an established neurotoxicant in children, in adults, in animals or it is not established, and each category suggests a different

Neurobehavioral Test Batteries

305

test selection strategy. However, even if the adverse effects are well established, it is quite likely that the effect occurring at the lowest exposure concentration, or following the shortest duration, is not known. Thus, using a broad range of tests is always encouraged. The principles of selection are: 1. If the adverse effects of the neurotoxicant are well characterized in children (e.g., one of the “big three”), the tests shown in published data to detect the established effects should form the base or core for the research study. Additional or supplemental tests are encouraged to expand the research base. 2. If there is research demonstrating adverse effects of the chemical in animals or human adults, tests reflecting effects of the same functions should certainly be employed, but a core series of tests that measure a broad array of functions is also needed since these effects were not established in developing children. 3. If the chemical has not been studied in children and offspring of pregnant women or if previous findings are negative, a core series of tests that measure a broad array of functions is required. As suggested in 3, the other factor that needs to be considered when selecting tests is the age of the child. Preschool children differ substantially from adolescents, and different abilities can be assessed at each developmental stage. Finally, pragmatic concerns about the design of the study will also impact test selection. These factors are detailed below.

Tests of Functional Effects: Global Versus Domain-Specific Measures Two approaches can be considered for selecting tests of functions that might be affected by a chemical of interest. These are domain-specific tests that assess a range of cognitive and motor functions (26) and global or apical measures of cognitive functioning and development. Global Measures Global or apical measures include standardized clinical tests, such as IQ tests or the Bayley Scales of Infant Development (27,28). These tests or test systems provide a single score that incorporates multiple elements of different functional domains such as cognitive, fine motor, attention, and motivation (26). Because the scores on these tests rely on performance in a variety of domains, they are sensitive to a broad range of impairments (26,29). These measures are standardized and often provide normative scores for evaluating performance on the test. Furthermore, they provide measures that are well known, so it is easier for people to understand the impact of deficits on these measures than for more specific functional tests (30,31). However, there are problems with using global measures. For example, while these tests may be sensitive to the effect of neurotoxicants, they are often thought to be nonspecific and at times insensitive (26,32). These endpoints do not assess specific functional domains. Therefore, children with the same IQ score may have a very different pattern of performance on different domains of cognition (33). It is important to realize that even small deficits on these global measures can indicate impairment. In studies examining effects from toxic exposures, individuals typically perform in the normal range; however, they may show deficits in other domains such as academic performance or reading and spelling skills (33). Finally, these global measures may have limited utility because they may not be standardized in the population being tested.

306

Rohlman and Anger

Domain-Specific Tests The alternative approach to test selection utilizes domain-specific tests that assess specific cognitive functions such as learning or memory. If chosen well, these tests may optimize the chance of detecting the specific effects of neurotoxicants (19). These tests may also provide insight into the specific brain-behavior relationship associated with a distinctive pattern of deficits associated with specific neurotoxicants (26,31). Identifying the functions to be measured is the first step in selecting groups of tests (usually referred to as batteries) to detect the effects of a toxicant exposure. Because of incomplete knowledge about neurotoxic effects from different toxicants in humans it is also important to include behavioral tests that have been shown to be sensitive to neurotoxic exposures in human research on other chemicals (34,35). An adequate neuropsychological assessment of a child should include assessment of a broad range of behaviors including measures of attention and vigilance, executive function and problem solving, learning and memory, language, sensory processing, motor performance, social skills, affective behavior, and academic performance (36). Tests can be selected from established neuropsychological measures of sensory and motor function, and from tests used in animals (35,37,38). One disadvantage of domain-specific tests is the absence of published normative data (29). In order to evaluate effects of neurotoxicants on an exposed population, it is necessary to compare their performance with that of a comparable control group or to examine changes from baseline performance within individuals. The selection of an adequate control group with similar ethnic, socioeconomic, and educational background, as well as similar age and gender, is paramount. Any differences between the groups may impact performance on these tests. Neurobehavioral Test Batteries for Children In recent years, due to the growing concern over the potential impact of environmental exposures on neurological function in children (39,40), interest in test methods appropriate for children has grown. Unlike adult neurobehavioral testing, in which a number of test batteries have been developed (20), there have been very few attempts to develop specific neurobehavioral batteries for children. The Pediatric Environmental Neurobehavioral Test Battery or (PENTB) (41) is a consensus test battery developed to assess possible neurotoxic effects in children living near hazardous waste sites. It combines observational measures and questionnaires for very young children with performance measures that are introduced for preschool children. Although the battery was proposed in 1996, the PENTB has thus far been used in only one study of children exposed to Methyl Parathion. That study did not reveal definitive effects (42). Summary Both global measures and domain-specific test batteries have advantages and disadvantages. The goals of the study and the interpretation of the data should dictate which type of measures or combinations of measures should be selected. The age of the child and the availability of measures will also influence the decision to include different types of measures. Because of the sensitivity of infant measures of global functioning (e.g., Bayley) these tests may be useful when information about the toxicant is unknown. However, it may also be useful to include tests that assess more specific functions in order to obtain more information about the characteristics of the impairment (43). In children who are preschool age or older, a broader range of measures may be used since the children are able to reliably complete more types of measures (21). Overall, the approach

Neurobehavioral Test Batteries

307

to test selection is similar to that used for adults, with one very important exception: There are a much larger number of known variables and confounders that affect the analysis and interpretation of the results in children, the most important of which is the developmental stage. Testing Children at Different Ages When testing children, developmental issues are the primary consideration in test selection and implementation. Tests that assess neurobehavioral or cognitive development have been developed for different age ranges. These developmental stages are important for both identifying exposure pathways and determining which behaviors and cognitive functions can be assessed. Infants Tests have been developed to assess neurodevelopment in the newborn and infant. Two of the most common measures are the Bayley Scales of Infant Development (27,28) and the Brazelton Neonatal Behavioral Assessment Scale (44). Infants are assessed to evaluate the impact of prenatal exposure to a toxicant on the central nervous system. It is important in prospective studies to identify deficits from prenatal exposure and determine if they are consistent as the child matures and develops or if there is a recovery of function (21). One advantage of assessing functioning in infants is that performance on these measures is thought to be less susceptible to socioeconomic confounders because they are administered early in the child’s lifetime (43,45). As the child grows older, parental influence and home environment impact performance to an ever greater extent (43,46). Although the Bayley and other measures are often used in studies examining prenatal exposure to neurotoxicants, their predictive validity, that is, their ability to identify children at risk of cognitive deficits later in childhood is often low (21,47). This low predictive validity is due in part to the rapid development of the infants and also to the restricted behaviors that can be assessed at this young age (33). These issues have led to the development of domain-specific tests that can be used with infants, including the Fagan Test of Infant Intelligence (48) and the Haith Visual Expectancy Paradigm (49). Instead of providing a global measure of cognitive functioning, these measures assess more specific domains. The Fagan test was developed to assess skills that are used by infants as well as by older children and adults. Because these skills are assessed as part of tests of cognitive functioning in older children, this measure tends to have better predictive ability than tests such as the Bayley (21,50). Furthermore, specific pattern of deficits on the Fagan are emerging for infants exposed to different toxicants (51). Although infant tests may be difficult to interpret because of the rapidly changing development of the child, they provide a good index of a newborn’s general health status at that point in time (33). As the child matures, more complex tests can be used to assess neurobehavioral functioning. Preschool Age Children At preschool ages the cognitive ability of a child is more easily assessed (26). Global measures of functioning are available [e.g., Wechsler Preschool and Primary Scale of Intelligence (52); Kaufman Assessment Battery for Children (53)] and neurobehavioral tests that assess specific functions can also be incorporated into a battery of tests. The use of domain-specific tests can identify specific types of impairment. As the child develops, more subtle deficits may become apparent. The specificity of these measures is important

308

Rohlman and Anger

because it can provide information to link toxicant exposure to specific brain regions (43). Furthermore, identification of specific domains of impairment will enable remediation programs to be efficiently directed at the impaired function. The ability of tests administered at this age to predict later performance is greater than that of tests administered to infants (21). When assessing preschool children there are several important things to consider: (1) maintaining the motivation and attention of the child during the test session, because children become easily frustrated and may not complete the tests (41,54); (2) the relationship between the examiner and the participant, because the child relies on the examiner to give task instructions and to provide feedback on performance; (3) standardized administration of tests. School-Age Children A period of rapid development in cognitive functioning occurs between the ages of five and seven as children mature and they are exposed to formal educational settings (26). During this time individual variations in performance make it difficult to distinguish changes associated with neurotoxic exposure from those that reflect normal development. By age seven and older, assessment is easier as the pace of development slows (26). For example, the time required to respond to a speeded task, reaction time, improves (viz., reaction time decreases) as age increases (55). Further age-related improvements are seen in understanding instructions (56,57). School-age children are able to complete a variety of tests, and motivation loss poses less of a problem (26). At this age children are often administered the same tests as adults, though perhaps with less challenging parameters (58–60). By adolescence, the same test and parameter settings used with adults can be employed (61). Maintaining Motivation in Children It is very important to maintain the motivation of the children during the entire test session. This can influence the reliability and validity of the tests (62). One method is to include the use of reinforcers throughout the test session [e.g., tokens (54) or money (38)]. Other methods for maintaining motivation on neurobehavioral tests include: (1) using intrinsically interesting tasks, (2) shorter tests, (3) adjusting the difficulty level of the test items to the subject’s ability level, and (4) emphasizing power not speed in the tests (62). The examiner plays an important role in the assessment of children, instructing the child on how to perform the test and motivating the child throughout the test session. It is important for the examiner to establish rapport with the child and provide feedback to help maintain motivation. However, this brings with it critical quality control issues to ensure that exposed and control groups receive the same level of attention by the examiner(s). Procedural measures are needed to ensure that, if there are multiple examiners, each examiner has an equal amount of time with participants from both the exposed and comparison groups. In addition, examiners must be blinded to group membership (exposed/control) during testing if they interact significantly with the children.

Confounders Factors that can alter performance and effectively mimic an effect of a toxicant must be measured to accurately evaluate and interpret test results. The impact of demographic variables such as age, education, gender, and culture on performance in adults

Neurobehavioral Test Batteries

309

occupationally exposed to toxicants is well established (20). These factors are important when assessing the impact of neurotoxicants on development as well as a range of other potential confounders (63). Important factors include individual characteristics of the child, such as age at the time of testing and gender. Other environmental characteristics include the socioeconomic status of the family, parental education and intelligence, stressors in the family, school quality, computer experience, and prenatal exposure to toxicants such as alcohol, lead, tobacco, and pesticides. Many of these factors are used to characterize the quality of parenting which can also maximize or minimize the effect of a toxicant on performance. The Home Observation for Measurement of the Environment (HOME) scale (64) was developed to evaluate quality of parental input to the child (e.g., intellectual stimulation and emotional responsiveness).

METHODS DEVELOPMENT Adapting Neurobehavioral Tests for Children In some cases, behavioral performance tests have been developed specifically for children. In other situations, the tests that have detected neurotoxic effects in adults are appealing choices for similar studies in children because they have a proven ability to detect chemical exposure effects. One advantage of adapting adult tests for children is that these may allow comparisons across ages. Some of the modifications to parameters for computer-based tests used with adults include: (1) shortening the test duration; (2) increasing the stimulus display time and interstimulus interval; (3) changing the stimuli used from letters to animal shapes; and (4) changing the presentation format from visual to auditory (54,59,60). When computerbased tests are used it is important to assess previous experience with computers and electronic gaming systems (56,59,65). Cross-Cultural Issues Most neurobehavioral tests used to study children have been developed with children from industrialized countries. The use of neurobehavioral tests to evaluate culturally diverse populations poses unique challenges, in part because this topic has not been extensively studied (66). Factors that impact performance on neurobehavioral tests, including social and cultural background, need to be accounted for and these effects need to be controlled either through study design or statistical analysis in order to correctly interpret the data. Some global measures of functioning have been translated into other languages (see Ref. 67 for a review of measures available to assess Spanish-speaking children). However, normative data developed in one culture cannot be used for children from another culture. Many of the domain-specific tests do not require language to perform the tests [e.g., reaction time, continuous performance, match-to-sample, finger tapping, pegboard; (40)]. However, the instructions on how to perform the tests do have to be translated.

Translating Instructions Since most tests have been developed in industrialized countries, their instructions often have to be translated to use them with non-mainstream populations. Test instructions should be translated prior to beginning the study. Translating “on the fly” during testing is never appropriate because it is does not provide a standardized administration of the test.

310

Rohlman and Anger

Furthermore, translators have varying abilities, rapport with the participant may be compromised, and subtleties can be missed. A multi-step process should be used to translate test instructions. This includes an initial translation into the new language, back translation of the instructions into the original language, and a resolution of any differences between the original version and the translated version (36). The final step is pilot testing the instructions in the target population. Throughout the process it is helpful to have people familiar with the study population, and the language they speak, review the translations and materials. The formality of the language as well as other semantic and structural rules can also impact the translations (8). For example, formal language used by well-educated people is a particular concern in Spanish-speaking countries (68,69). Formal language, often used in a word-for-word translation, may also distort the original meaning of the material and lead to confusion, especially in a population with minimal education. Low literacy can also limit a person’s ability to understand written material. It is best to use a “functional translation,” which employs terms that are easily comprehensible but convey the equivalent idea or concept. Another concern when translating tests into other languages is cultural variation. Stimuli or items used in tests may have more than one correct translation, especially when different local dialects are taken into account. This can cause problems interpreting responses and scoring the tests if naming the item is the correct response. Additionally, items or scenarios described in the test materials may be unfamiliar to the population being tested (36,54,70). Acculturation Issues Acculturation may impact the familiarity of testing situations and hence performance on tests (71). Acculturation is complex and consists of multiple variables, including language, values, belief, gender roles, attitudes, leisure activities, media preferences, and school experience. There is also an underlying assumption with most of the tests developed for North American and European populations that the participants have had some school experience (40). This “socialization experience” from classroom settings may impact the test-taking behavior of the participants. People who complete the educational system in the U.S. learn to work on tests, to perform his or her best, and to do it quickly. A person raised in a Hispanic culture may put a greater emphasis on rapport with the examiner than on the results of a test or completing it quickly (36).

Pilot-Revise-Pilot Process Throughout the development of the methods it is crucial to pilot test the methods in a sample selected from the planned study population. It is important to utilize a “pilotrevise-pilot” sequential process. This will allow any problems with the test administration or materials to be revealed before the start of the study. Only in observing the performance of children from the study population is it possible to determine if the instructions, the test parameters, and the stimuli used in the test are appropriate. Problems or concerns that are detected during piloting can thus be corrected before time and resources are spent on collecting worthless data. If significant modifications are made, it is necessary to repeat the pilot process again. These steps should be repeated as many times as necessary until the methods are suitable.

Neurobehavioral Test Batteries

311

QUALITY CONTROL Once the study design has been developed, including identifying the population of interest, selecting the tests to be administered, and defining the time and money resources available, it is important to consider quality control issues. These are important to ensure that there is not confounding due to the examiner(s) in the study results. The development of a standard protocol, training the examiner(s) on the protocol, and conducting quality control checks throughout the study will ensure that the data are valid and reproducible. Test Location The selection of the test location is an important decision. At a minimum it is necessary to have a quiet, comfortable location with adequate lighting and heating. The room should be free from distractions (e.g., toys, television) and be separate from the waiting room. The equipment should be placed on sturdy tables and chairs of varying heights made available to accommodate children of different sizes. The waiting area should also be comfortable and contain books and toys for the child and siblings to use while waiting. A separate area for interviews with the parent should be used to ensure confidentiality of the data. Protocol Manual The development of a detailed examiner’s protocol manual is required for the consistent administration of methods to every child in the study. Standardized instructions ensure each child is given the same instructions and that test scoring is constant (33). The specific information on how to administer the tests is included in the manual: (1) the order the tests are administered, (2) all procedures, including the exact wording used by the examiner, (3) responses to possible questions or problems, (4) information about recording responses and backing up data, and (5) scoring. The protocol manual should also contain a complete list of the equipment required for the test session, directions for setting up the equipment and the testing site, copies of data forms, troubleshooting guidelines, a list of regular and emergency phone numbers of study principals, and instructions for recording information in the study logbook. Examiner Training Examiners need to have sufficient training with the individual tests and their flow in the battery. Repeated practice is required until they are comfortable with the instructions and manipulating the test equipment. Practice is also required with children similar to the ones in the study population. Each examiner should be observed during training and then certified that they are able to consistently administer the tests in a standardized format. An additional part of the certification is concerned with the examiner’s presentation during the test session. These requirements come from PENTB (41), and include: wearing appropriate attire; having an appropriate demeanor with the child and parents; establishing good rapport with the child and maintaining it throughout the test session; pacing the session well (so that it is not rushed and not lagging); handling interruptions during the session in an appropriate manner (without losing focus or momentum); using praise properly; redirecting the child’s attention in a reinforcing manner; being aware of the child’s state and responding appropriately; maintaining a professional demeanor that is appropriate to the child’s age; avoiding patronizing, condescending, and authoritarian tones of voice; and ending the session and returning the child to the parent or teacher in

312

Rohlman and Anger

a professional and efficient manner. During the study, “mid-study” site visits, with quality control checks, are conducted to verify that examiners are following protocol and to provide mid-course corrections when needed. Such corrections are usually minor, but they are essential to ensure that an examiner has not adopted methods or instructions that veer away from the protocol, which could lead to exclusion of data they have collected. Even minor deviations can multiply over the course of the study and make the results uninterpretable.

AN EXAMPLE: DEVELOPING A TEST BATTERY FOR PRESCHOOL CHILDREN An example may bring this complex problem into perspective. A neurobehavioral test battery to assess functioning in Hispanic preschool children was developed (54,72). The parents of the children worked in agriculture and exposure to organophosphate pesticides (OP) was the concern. Children of agricultural workers are considered to have a higher risk of exposure to pesticides compared to the general populations because of the proximity of their homes to the fields where pesticides are applied and from take-home exposure (73,74). Test Selection Following the approach outlined above, the research examining neurobehavioral effects of pesticide exposure was reviewed. Different groups of people exposed to pesticides have shown neurobehavioral deficits. These include OP-poisoned populations; occupational groups, both adults and adolescents who work with pesticides; and preschool children. The findings from these studies are presented in Table 1. The goal was to develop a battery that included measures that had demonstrated sensitivity to OP exposure and, because of the unknown nature of effects of OPs in children, to assess a wide range of neurobehavioral functions. The battery was assembled by combining performance tests from the Behavioral Assessment and Research System (BARS) and other, non–computer-based tests. Methods Development A multi-step process was used to develop the test battery (54,72). Parameter settings for the computer-based tests (e.g., number of trials, difficulty level) that would be appropriate for young children were initially assessed in English-speaking preschool children. During this stage of test development we were interested in the strategies or methods the children were using to complete the tests and the percentage of children that were able to complete these tests. Based on the performance of these children, modifications were made to the original test battery and parameter settings (Table 2). Five BARS tests were originally administered: Symbol-Digit, Digit Span, Finger Tapping, Continuous Performance, and Matching to Sample. The Symbol-Digit test was dropped from the battery because the children had difficulty completing the assessment portion of the test. While they understood how to perform the test, they would become restless and distracted during the assessment and thus fail to complete the test. A switch from visual to spoken presentation during Digit Span increased the percentage of children completing the test, although many children still had difficulty with the reverse portion (digits backward) of the test. The alternating version of the Finger Tapping (tap sequentially with the left hand, then the right, then the left, etc.) was eliminated from the battery because the children were

Neurobehavioral Test Batteries Table 1

Summary of Neurobehavioral Findings of Pesticide Exposure

Function Motor speed and coordination

Test Finger Tapping Pursuit Aiming Santa Ana Pegboard

Drawing Information processing speed Complex visualmotor processingC executive function

Purdue Pegboard Hand–Eye Coordination Draw-a-Person Simple Reaction Time Syntactic Reason Digit-Symbol

Symbol-Digit Trail Making Verbal abstraction

Similarities

Attention/short term memory

Digit Span

Sustained attention Memory

313

Letter Cancellation Continuous Performance Benton Visual Retention Delayed Recall

Population(s)

Source

Poisoned adults Occupational groups Poisoned adults Occupational groups Poisoned adults Occupational groups Occupational groups Preschool children

(81–83)

Preschool children Occupational groups Adolescent farmworker Occupational groups Poisoned adults High exposed workers

(87) (61,88,89)

Occupational groups Adolescent farmworker High exposed workers Poisoned adults Poisoned adults High exposed workers Poisoned adults Occupational groups Adolescent farmworker High exposed workers High exposed workers Poisoned adults Adolescent farmworker Poisoned adults Occupational groups High exposed workers Preschool children

(84–86) (83,84,86) (81) (87)

(88) (84,90–94)

(61,88) (84,92,94) (92,94) (61,82–84,92,94,95)

(94) (61,90) (81,84,94)

(87)

Note: Tests in bold indicate tests selected for the current battery.

unable to respond fast enough to correctly meet the training criterion, causing the child to repeat the instructions and training. Unsuccessful performance and repeating the instructions multiple times would reduce the motivation of the child to continue testing. The practice criterion was modified in the Continuous Performance Test. In addition to the computer-based tests, four non–computer-based tests were also administered. Three tests from the PENTB were selected: the Story Memory subtest from the Wide Range Assessment of Memory and Learning (WRAML), the vocabulary and matrices subtest from the Kaufman Brief Intelligence Test (K-BIT), and the Purdue Pegboard test. A new test of recall and recognition, the Object Memory test (69) was also administered. The children had no difficulty completing these tests. The second step was to evaluate the modified battery for Spanish-speaking preschool children. Although the children had no trouble completing some of the computer-based tests (Finger Tapping, Match-to-Sample, Digit Span forward), there were

314 Table 2

Rohlman and Anger Steps Used in the Development of a Neurobehavioral Test Battery

Test Digit Span (BARS) Finger Tapping (BARS) Match-to-Sample (BARS) Continuous Performance (BARS) Symbol-Digit (BARS) Divided Attention BARS (PENTB) Story Memory— WRAML (PENTB) Vocabulary and Matrices— K-BIT (PENTB) Purdue Pegboard (PENTB) Visual Motor Integration (PENTB) Object Memory test

Step 3 final battery modifications from Step 2

Step 2 Spanishspeaking modifications from Step 1

Step 1 English-speaking X X

Oral presentation No alternating trials

Forward only OK

X

Change placement in battery Slower stimulus presentations

OK

Dropped

–

Added reinforcers/ reduced test duration OK

X

Dropped

Dropped

X

Dropped

Dropped

X

OK

OK

–

–

OK

X

OK

Replaced unfamiliar objects

X

X

Changed success criterion

Changed rhyme

Note: Tests in bold indicate tests selected for the current battery.

still problems with the Symbol-Digit test, the backwards trial of the Digit Span test, and completing the practice trial of the Continuous Performance Test. The Spanish-speaking children were also unfamiliar with the rhyme used for the Divided Attention Test. In this test the children are required to recite a nursery rhyme (“Jack and Jill” in the PENTB version) while simultaneously tapping. Because there is no Spanish equivalent of this rhyme, the “Itsy Bitsy Spider” or “La Aran˜a Pequen˜ita” was used. Similar to the English-speaking children, the Spanish-speaking children had no difficulty with the Purdue Pegboard test. However, they had great difficulty with the WRAML Story Memory and the K-BIT subtests. The scenarios described in the WRAML stories discussing birthday parties and going fishing with a cat were unfamiliar to this population. Also, for some of the items in the K-BIT there was more than one correct translation (i.e., feather, bridge, owl). Although an attempt was made to account for this problem by allowing multiple answers to be correct, the problem could not be eliminated entirely, which made the test scoring potentially unreliable. The final problem was that the children were unfamiliar with an item in a test (viz., a whistle used in Object Memory). A decision was made to eliminate the Symbol-Digit, the backwards trial of Digit Span, and the WRAML Story Memory and K-BIT subtests from the battery because of the cultural inappropriateness of the items and the difficulty the children were having

Neurobehavioral Test Batteries

315

completing the tests. The unfamiliar items in Object Memory were replaced and the Visual Motor Integration test from the PENTB was added to the battery. These steps demonstrate the pilot-revise-pilot process that is necessary to develop methods for testing children (54,72).

ANALYZING THE RESULTS Controlling for Confounders When examining associations between exposure and neurobehavioral performance, it is important to realize that there are a number of other variables or confounders that may account for a finding. These variables may provide an alternative causal explanation for an observed relationship between an exposure and an outcome (26,31,43). These variables also need to be considered during data analysis to assess their impact on the performance measures. Some studies have collected data on as many as 20 or more potential confounding variables (43). These variables can make it difficult to interpret the results of a study (75). Multivariate analyses can be used to control for these confounders. However, it is not reasonable to include them all in the analysis. A widely-used “rule of thumb” is to test 20 participants for each independent variable (76). Therefore, if you are interested in one exposure and 15 control variables you will need to have a minimum of 320 participants. This may not be feasible or practical (26). The large number of analyses will also increase the error term, making it more difficult to detect effects of a neurotoxic exposure. An alternative strategy is to prescreen the variables to determine which confounding variables are related to the outcome. One approach is to examine the correlation between the variables and the outcomes, and using a criterion (p!0.10 has been suggested) to determine which variables should be included in the analysis (43). Step-wise regression may also be used to determine which variables may be included in the model. Learning or Practice Effects In order to draw accurate conclusions from the data it is important that children complete the tests in the battery. At times a child may not understand the instructions and the examiner may decide, in order to prevent the child from becoming frustrated, to skip the test and move to another test in the battery. This information should be recorded in the study logbook and should be reviewed when examining the data from the study. At other times, the child may complete the test, but an examination of the data may reveal that they did not understand the instructions. Performance criteria that exclude data from children with sub-optimal performance have been used (59,65). Learning or practice effects may be seen on some measures. Studies examining chronic effects of neurotoxicants may incorporate a longitudinal study design. Two types of practice-related measurement error have been identified (77). Test-specific practice occurs when participants develop better test-taking strategies when given the same versions of a test or procedure several times. Item-specific practice occurs if participants are presented with the same stimuli during repeated administrations of tests evaluating memory. Although alternate test forms have been found to significantly reduce practice effects, test-specific practice effects may still occur, especially if the test involves a novel concept or procedure (77). The influence of demographic variables on learning may also make it difficult to interpret test results. Different populations may have different learning curves so that stable performance on a test may be achieved with different amounts of

316

Rohlman and Anger

practice (78,79). An additional concern when using a battery for repeat assessment is that repeated testing may cause a restriction of range of scores at the upper end (80).

CONCLUSIONS The goal of this chapter is to provide guidance on the selection of methods to assess neurotoxic effects in children. The steps to design a study can be summarized as follows: † † † † † † † † †

Test selection: base on research in humans or animals, if possible Select tests appropriate to the age of the child Use global and domain specific tests for infants or young children Test a broad range of functions in older children Adapt tests or procedures for cross-cultural uniqueness Pilot-revise-pilot new methods, instructions and translations Develop and follow scrupulously a detailed protocol manual Train and certify examiners Use rigorous quality control procedures including mid-study checks of protocol adherence † Include potential confounders in data analyses Converging evidence from studies in multiple populations and animals will allow valid conclusions to be drawn, but reliable and effective methods are needed to trust the integrity of your data.

ACKNOWLEDGMENTS This work was supported by R21 ES08707 from the National Institute of Environmental Health Sciences (NIEHS). Its contents are solely the responsibility of the authors and do not necessarily represent the views of NIEHS. The authors are developers of the Behavioral Assessment and Research System (BARS) mentioned in this chapter.

REFERENCES 1. CDC (Center for Disease Control). Second National Report on Human Exposure to Environmental Chemicals. Atlanta, GA: Centers for Disease Control and Prevention, 2003. 2. Goldman LR, Koduru S. Chemicals in the environment and developmental toxicity to children: a public health and policy perspective. Environ Health Perspect 2000; 108:443–448. 3. U.S. Environmental Protection Agency. Chemical Hazard Data Availability Study: What do We Really Know about the Safety of High Production Volume Chemicals? Washington, DC: U.S. Environmental Protection Agency, 1998. 4. Koger SM, Schettler T, Weiss B. Environmental toxicants and developmental disabilities. Am Psychol 2005; 60:243–255. 5. Landrigan PJ, Kimmel CA, Correa A, Eskenazi B. Children’s health and the environment: public health issues and challenges for risk assessment. Environ Health Perspect 2004; 112:257–265. 6. NRC (National Research Council). Scientific Frontiers in Developmental Toxicology and Risk Assessment. Washington, DC: National Academy Press, 2000. 7. Mott L. The disproportionate impact of environmental health threats on children of color. Environ Health Perspect 1995; 103:33–35.

Neurobehavioral Test Batteries

317

8. Rhodes RL, Kayser H, Hess RS. Neuropsychological differential diagnosis of Spanishspeaking preschool children. In: Fletcher-Janzen E, Strickland TL, Reynolds CR, eds. Handbook of Cross-Cultural Neuropsychology. New York: Kluwer Academic/Plenum Publishers, 2000:317–333. 9. Frumkin H, Walker ED, Friedman-Jime´nez G. Minority workers and communities. Occup Med 1999; 14:495–517. 10. Cohen Hubal EA, Sheldon LS, Burke JM, et al. Children’s exposure assessment: a review of factors influencing children’s exposure, and the data available to characterize and assess that exposure. Environ Health Perspect 2000; 108:475–486. 11. Reed KJ, Jimenez M, Freeman NC, Lioy PJ. Quantification of children’s hand and mouthing activities through a videotaping methodology. J Exp Anal Environ Epidemiol 1999; 9:513–520. 12. Claudio L, Kwa WC, Russell AL, Wallinga D. Testing methods for developmental neurotoxicity of environmental chemicals. Toxicol Appl Pharmacol 2000; 164:1–14. 13. Costa LG, Aschner M, Vitalone A, Syverson T, Soldin OP. Developmental neuropathology of environmental agents. Annu Rev Pharmacol Toxicol 2004; 44:87–110. 14. Trask CL, Kosofsky BE. Developmental considerations of neurotoxic exposures. Neurol Clin 2000; 18:541–562. 15. Godfrey KM, Barker DJ. Fetal programming and adult health. Public Health Nutr 2001; 4:611–624. 16. Reuhl KR. Delayed expression of neurotoxicity: the problem of silent damage. Neurotoxicology 1991; 12:341–346. 17. Mendola P, Selevan SG, Gutter S, Rice D. Environmental factors associated with a spectrum of neurodevelopmental deficits. Ment Retard Dev Disabil Res Rev 2002; 8:188–197. 18. Landrigan PJ. Children’s environmental health. Pediatr Clin N Am 2001; 48:1319–1330. 19. Rice D, Barone SJ. Critical periods of vulnerability for the developing nervous system: evidence from human and animal models. Environ Health Perspect 2000; 108:511–533. 20. Anger WK. Neurobehavioural tests and systems to assess neurotoxic exposures in the workplace and community. Occup Environ Med 2003; 60:531–538. 21. Dietrich KN, Bellinger D. The assessment of neurobehavioral development in studies of the effects of prenatal exposure to toxicants. In: Needleman HL, Bellinger D, eds. Prenatal Exposure to Toxicants: Developmental Consequences. Baltimore: John Hopkins University Press, 1994:57–85. 22. Needleman HL, Gunnoe CE, Leviton A, et al. Deficits in psychologic and classroom performance of children with elevated dentine lead levels. N Engl J Med 1979; 300:689–695. 23. Streissguth AP, Bookstein FL, Sampson PD, Barr HM. The Enduring Effects of Prenatal Alcohol Exposure on Child Development, Birth Through 7 Years: A Partial Least Squares Solution. Ann Arbor, MI: University of Michigan Press, 1993. 24. Weiss B, Landrigan PJ. The developing brain and the environment: an introduction. Environ Health Perspect 2000; 108:373–374. 25. Jacobson JL, Jacobson SW. Prospective, longitudinal assessment of developmental neurotoxicity. Environ Health Perspect 1996; 104:275–283. 26. Jacobson JL, Jacobson SW. Assessing neurotoxicity in children. In: Tilson HA, Harry GJ, eds. Neurotoxicology. London: Taylor & Francis, 1999:339–366. 27. Bayley N. Bayley Scales of Infant Development. New York: Psychological Corporation, 1969. 28. Bayley N. 2nd ed. Bayley Scales of Infant Development. San Antonio, TX: Psychological Corporation, 1993. 29. Fiedler N, Feldman RG, Jacobson J, Rahill A, Wetherell A. The assessment of neurobehavioral toxicity: SGOMSEC joint report. Environ Health Perspect 1996; 104:179–191. 30. Weiss B. Vulnerability of children and the developing brain to neurotoxic hazards. Environ Health Perspect 2000; 108:375–381. 31. Dietrich KN. Environmental neurotoxicants and psychological development. In: Taylor G, Yeates KO, eds. Pediatric Neuropsychology: Research, Theory and Practice. New York: Guilford Press, 1999:206–227.

318

Rohlman and Anger

32. Bernstein JH. Assessment of developmental toxicity: neuropsychological batteries. Environ Health Perspect 1994; 102:141–144. 33. Bellinger DC. Perspectives on incorporating human neurobehavioral end points in risk assessments. Risk Anal 2002; 22:487–498. 34. Anger WK, Rohlman DS, Storzbach D. Neurobehavioral testing in humans. Current Protocols in Toxicology. New York: John Wiley & Sons, Inc., 1999 pp. 11.7.1–11.7.17. 35. Davidson PW, Weiss B, Myers GJ, et al. Evaluation of techniques for assessing neurobehavioral development in children. NeuroToxicology 2000; 21:957–972. 36. Fletcher-Janzen E, Strickland TL, Reynolds CR. Handbook of Cross-Cultural Neuropsychology. New York: Kluwer Academic/Plenum Publishers, 2000. 37. Sharbaugh C, Viet SM, Fraser A, McMaster SB. Comparable measures of cognitive function in human infants and laboratory animals to identify environmental health risks to children. Environ Health Perspect 2003; 111:1630–1639. 38. Chelonis JJ, Daniels-Shaw JL, Blake DJ, Paule MG. Developmental aspects of delayed matching-to sample task performance in children. Neurotoxicol Teratol 2000; 22:683–694. 39. Schettler T, Stein J, Reich F, Valenti M, Wallinga D. Harm’s Way: Toxic Threats to Child Development. Cambridge, MA: Greater Bostons Physicians For Social Responsibility (GBPSR), 2000. 40. Krasnegor NA, Otto DA, Bernstein JH, et al. Neurobehavioral test strategies for environmental exposures in pediatric populations. Neurotoxicol Teratol 1994; 16:499–509. 41. Amler RW, Gibertini M, eds. Pediatric Environmental Neurobehavioral Test Battery. Atlanta: US Department of Health and Human Services, Agency for Toxic Substances and Disease Registry, 1996:277. 42. Ruckart PZ, Kakolewski K, Bove FJ, Kaye WE. Long-term neurobehavioral health effects of methyl parathion exposure in children in Mississippi and Ohio. Environ Health Perspect 2004; 112:46–51. 43. Jacobson JL, Jacobson SW. Methodological considerations in behavioral toxicology in infants and children. Dev Psychol 1996; 32:390–403. 44. Brazelton TB. 2nd ed. Neonatal Behavioral Assessment Scale. London: Blackwell Scientific Publications, 1984. 45. McCall RB. Nature-nurture and the two models of development: a proposed integration with respect to mental development. Child Dev 1981; 52:1–12. 46. Bradley RH, Caldwell BM, Rock SL, et al. Home environment and cognitive development in the first 3 years of life: a collaborative study involving six sites and three ethnic groups in North America. Dev Psychol 1989; 25:217–235. 47. McCall RB, Hogarty PS, Hurlburt N. Transitions in infant sensorimotor development and the prediction of childhood IQ. Am Psychol 1972; 27:728–748. 48. Fagan JF, Singer LT. Infant recognition memory as a measure of intelligence. In: Lipsett LP, ed. Advances in Infancy Research. Norwood, NJ: Ablex, 1983:31–78. 49. Haith MM, Hazan C, Goodman GS. Expectation and anticipation of dynamic visual events by 3.5-month-old babies. Child Dev 1988; 59:467–479. 50. McCall RB, Carriger MS. A meta-analysis of infant habituation and recognition memory performance as predictors of later IQ. Child Dev 1993; 64:57–69. 51. Jacobson SW. Specificity of neurobehavioral outcomes associated with prenatal alcohol exposure alcoholism. Clin Exp Res 1998; 22:313–320. 52. Wechsler D. Wechsler Preschool and Primary Scale of Intelligence—Revised. San Antonio, TX: Psychological Corporation, 1989. 53. Kaufman AS, Kaufman NL. Kaufman Assessment Battery for Children. Circle Pines, MN: American Guidance Service, 1983. 54. Rohlman DS, Anger WK, Tamulinas A, Phillips J, Bailey SR, McCauley L. Development of a neurobehavioral battery for children exposed to neurotoxic chemicals. Neurotoxicology 2001; 22:657–665.

Neurobehavioral Test Batteries

319

55. Kail R. Processing time declines exponentially during childhood and adolescence. Dev Psychol 1991; 27:259–266. 56. Despres C, Richer F, Roberge M-C, Lamoureux D, Beuter A. Standardization of quantitative tests for preclinical detection of neuromotor dysfunction in pediatric neurotoxicology. NeuroToxicology, 2005; 26:385–395. 57. Rival C, Olivier I, Ceyte H. Effects of temporal and/or spatial instructions on the speedaccuracy trade-off of pointing movements in children. Neurosci Lett 2003; 336:65–69. 58. Altmann L, Sveinsson K, Kramer U, Winneke G, Wiegand H. Assessment of neurophysiologic and neurobehavioral effects of environmental pollutants in 5- and 6-year-old children. Environ Res 1997; 73:125–131. 59. Dahl R, White RF, Weihe P, et al. Feasibility and validity of three computer-assisted neurobehavioral tests in 7-year-old children. Neurotoxicol Teratol 1996; 18:413–419. 60. Otto DA, Skalik I, House DE, Hudnell HK. Neurobehavioral Evaluation System (NES): comparative performance of 2nd-, 4th-, and 8th-grade Czech children. Neurotoxicol Teratol 1996; 18:421–428. 61. Rohlman DS, Bailey SR, Anger WK, McCauley L. Assessment of neurobehavioral function with computerized tests in a population of Hispanic adolescents working in agriculture. Environ Res 2001; 85:14–24. 62. Letz R. Continuing challenges for computer-based neuropsychological tests. Neurotoxicology 2003; 24:479–489. 63. Jacobson JL, Jacobson SJ, Humphrey HEB. Effects of in utero exposure to polychlorinated biphenyls on cognitive functioning in young children. J Pediatr 1990; 116:38–45. 64. Caldwell BM, Bradley RH. Home Observation for Measurementof the Environment. Little Rock: University of Arkansas Press, 1979. 65. Rohlman DS, Arcury TA, Quandt SA, et al. Neurobehavioral performance in preschool children from agricultural and non-agricultural communities in Oregon and North Carolina. NeuroToxicology, 2005; 26:589–598. 66. Helms JE. Why is there no study of cultural equivalence in standardized cognitive ability testing? Am Psychol 1992; 47:1083–1101. 67. Gonzalez JJ. In: Ponto´n MO, Leo´n-Carrio´n J, eds. Pediatric assessment. In: Neuropsychology and the Hispanic patient. Mahwah, NJ: Lawrence Erlbaum Associates, Inc., 2001. 68. Bernal EM. Increasing the interpretive validity and diagnostic utility of Hispanic children’s scores on tests of achievement and intelligence. Mental Health of Ethnic Minorities. New York: Praeger, 1990 pp. 108–138. 69. Mahurin RK, Espino DV, Holifield EB. Mental status testing in elderly Hispanic populations: special concerns. Psychopharmacol Bull 1992; 28:391–399. 70. Puente AE, Ardila A. Neuropsychological assessment of Hispanics. In: Fletcher-Janzen E, Strickland TL, Reynolds CR, eds. Handbook of Cross-Cultural Neuropsychology. New York: Kluwer Academic/Plenum Publishers, 2000:87–104. 71. Ponton MO, Leon-Carrion J. Neuropsychology and the Hispanic Patient: A Clinical Handbook. Mahwah, NJ: Lawrence Erlbaum Associates, 2001. 72. Rohlman DS, Gimenes LS, Ebbert CA, Anger WK, Bailey SR, McCauley L. Smiling faces and other rewards: using the Behavioral Assessment and Research System (BARS) with unique populations. Neurotoxicology 2000; 21:973–978. 73. Fenske RA, Lu C, Barr DB, Needham B. Children’s exposure to chlorpyrifos and parathion in an agricultural community in Central Washington State. Environ Health Perspect 2002; 110:549–553. 74. Loewenherz C, Fenske RA, Simcox NJ, Bellamy G, Kalman D. Biological monitoring of organophosphorus pesticide exposure among children of agricultural workers in central Washington state. Environ Health Perspect 1997; 105:1344–1353. 75. Mink PJ, Goodman M, Barraj LM, Imrey H, Kelsh MA, Yager J. Evaluation of uncontrolled confounding in studies of environmental exposures and neurobehavioral testing in children. Epidemiology 2004; 15:385–393. 76. Tabachnick BG, Fidell LS. Using Multivariate Statistics. New York: Harper & Row, 1983.

320

Rohlman and Anger

77. Benedict RHB, Zgaljardic DJ. Practice effects during repeated administrations of memory test with and without alternate forms. J Clin Exp Neuropsychol 1998; 20:339–352. 78. Lowe C, Rabbit P. Test/re-test reliability of the CANTAB and ISPOCD neuropsychological batteries: theoretical and practical issues. Neurotoxicology 1998; 36:915–923. 79. Rohlman DS, Bailey SR, Brown M, Blanock M, Anger WK, McCauley L. Establishing stable test performance in tests from the Behavioral Assessment and Research System (BARS). Neurotoxicology 2000; 21:715–724. 80. White RF, Gerr F, Cohen RF, et al. Criteria for progessive modification of neurobehavioral batteries. Neurotoxicol Teratol 1994; 16:511–524. 81. Reidy TJ, Bowler RM, Rauch SS, Pedroza GI. Pesticide exposure and neuropsychological impairment in migrant farm workers. Arch Clin Neuropsychol 1992; 7:85–95. 82. Nishiwaki Y, Maekawa K, Ogawa Y, Asukai N, Minami M, Omae K. The sarin health effects study group. Effects of sarin on the nervous system in rescue team staff members and police officers 3 years after the Tokyo subway sarin attack. Environ Health Perspect 2001; 109:1169–1173. 83. Kamel F, Rowland AS, Park LP, et al. Neurobehavioral performance and work experience in Florida farmworkers. Environ Health Perspect 2003; 111:1765–1772. 84. Group TPHES. Rosenstock L, Keifer M, Daniell WE, McConnell R, Claypoole K. Chronic central nervous system effects of acute organophosphate pesticide intoxication. Lancet 1991; 338:223–227. 85. Bazylewicz-Walczak B, Majczakowa W, Szymczak M. Behavioral effects of occupational exposure to organophosphorous pesticides in female greenhouse planting workers. Neurotoxicology 1999; 20:819–826. 86. London L, Myers JE, Nell V, Taylor T, Thompson ML. An investigation into neurologic and neurobehavioral effects of long-term agrichemical use among deciduous fruit farm workers in the western cape. S Af Environ Res 1997; 73:132–145. 87. Guillette EA, Meza MM, Aguilar MG, Soto AD, Garcia IE. An anthropological approach to the evaluation of preschool children exposed to pesticides in Mexico. Environ Health Perspect 1998; 106:347–353. 88. Stephens R, Spurgeon A, Calvert IA, et al. Neuropsychological effects of long-term exposure to organosphosphates in sheep dip Lancet 1995; 345:1135–1139. 89. Fiedler N, Kipen H, McNeil K, Fenske R. Long-term use of organophosphates and neuropsychological performance. Am J Ind Med 1997; 5:487–496. 90. Steenland K. Chronic neurological effects of organophosphate pesticides. BMJ 1996; 312:1312–1313. 91. Yokoyama K, Araki S, Murata K, et al. Chronic neurobehavioral effects of Tokyo subway sarin poioning in relation to posttraumatic stress disorder. Arch Environ Health 1998; 53:249–256. 92. Savage EP, Keefe TJ, Mounce LM, Heaton RK, Lewis JA, Burcar PJ. Chronic neurological squelae of acute organophosphate pesticide poisoning. Arch Environ Health 1988; 43:38–45. 93. Wesseling C, Keifer M, Ahlborn A, McConnell Moon JD, Rosenstock K, Hogstedt C. LongTerm neurobehavioral effects of mild poisoning with organophosphate and n-Methyl carbamate pesticides among banana workers. Int J Occup Environ Health 2002; 8:27–34. 94. Farahat TM, Abdelrasoul GM, Amr MM, Shebl MM, Farahat FM, Anger WK. Neurobehavioral effects among workers occupationally exposed to organophosphorous pesticides. Occup Environ Med 2003; 60:279–286. 95. Cole DC, Carpio F, Julian J, Leon N, Carbotte R, De Almedia H. Neurobehavioral outcomes among farm and nonfarm rural ecuadorians. Neurotoxicol Teratol 1997; 19:277–286.

15 Role of Neuroimaging Patricia A. Janulewicz Department of Environmental Health, Boston University School of Public Health and Department of Psychology, University of Massachusetts, Boston, Massachusetts, U.S.A.

Roberta F. White Department of Environmental Health, Boston University School of Public Health, Boston, Massachusetts, U.S.A.

Carole Palumbo Department of Neurology, Boston University School of Medicine, Aphasia Research Center, VA Boston Healthcare System, Boston, Massachusetts, U.S.A.

For many years, subtle alterations in brain function that are associated with exposure to neurotoxic substances have been a central focus of research interest. These alterations, occurring in the absence of a clinically identifiable intoxication or encephalopathy, are important because they are the first indicators that a substance is adversely affecting the brain. Research into subtle toxicant-induced brain damage using behavioral outcome measures has been critical in identifying these early neurotoxicant effects. The use of neuroimaging technology has played a role in identifying the structural effects of exposure to toxicants resulting in clinical manifestations of intoxication, especially in substances such as carbon monoxide (1–3) and mercury (4,5). However, until recently imaging technology had limited capability for identifying subtle alterations in brain structure and function. Recent advances in the sensitivity of imaging technology have allowed application of these techniques to the investigation of neurotoxicant effects on the brain. Although it has not yet been widely applied in neurotoxicologic research, its advantages are obvious. First and foremost, imaging techniques allow investigators to view the structure and function of the brain in living organisms. Furthermore, imaging allows identification of the specific structures and neural systems that are affected by toxicant exposure. Based on the existing literature linking decrements in performance on particular types and domains of cognitive and behavioral tests, evidence from behavioral studies can be used to guide hypotheses concerning brain structures that may be particularly vulnerable to specific toxicants. Imaging provides pictorial evidence of dysfunction that has been identified on behavioral and neurophysiological studies of toxicant-exposed populations. Such evidence may be more convincing than purely behavioral outcomes in demonstrating toxicant-induced brain damage to the scientific and lay communities. In addition, imaging and behavioral studies can be used in tandem when investigating subtle effects of 321

322

Janulewicz et al.

toxicants. Positive behavioral findings on a large group of toxicant-exposed individuals can be utilized to guide confirmatory imaging studies on a subgroup. Similarly, a small group of individuals can be investigated using the more time-consuming and expensive neuroimaging methodology, followed by behavioral studies of the larger population when imaging results indicate likely exposure-related brain damage or dysfunction. In addition, the techniques can be applied to animals, allowing further exploration of neurotoxicant effects following administration of toxicants or toxicant mixtures. Given the capacity of various imaging techniques to explore structural, functional, and neurochemical sequelae of exposure to neurotoxicants, their application to research in behavioral toxicology is an exciting development. In this chapter, we will first describe specific neuroimaging techniques, with a brief summary of their application to the study of child neurodevelopment in general. This will be followed by descriptions of the use of these techniques for the assessment of the central nervous system (CNS) effects of exposures to known neurotoxicants. Finally, data from a study that examined functional imaging in a cohort of children from the Faroe Islands with well characterized prenatal exposure to methylmercury (MeHg) and polychlorinated biphenyls (PCBs) will be described.

NEUROIMAGING TECHNIQUES Computerized Tomography Computerized tomography (CT), invented in 1972 by Hounsfield and Cormack, is a technique that capitalizes on the fact that x-rays reflect the relative densities of the tissue through which they pass (6). During brain CT imaging, many x-rays are taken of the brain at an angle of 15-degrees to the canthomeatal line. Computing and mathematical techniques create the visual images of the brain that provide the concrete evidence of CT scan results. Because the gray matter, which is rich in cells, and the white matter, which is dominated by axons, have different densities, they reflect onto the x-ray film differently. This allows visualization of different brain structures. CT methodology is relatively inexpensive and fast. In the child neurodevelopment literature, CT techniques are most commonly described for clinical use. CT has been widely applied to the diagnosis, localization, surgical planning and post-operative surveillance of brain tumors (7–10). It has also been used extensively to define the etiology and prognosis of cerebral palsy (11–15) and to diagnose and treat closed head injury (16–18) and shaken baby syndrome (14,15). Studies have also appeared linking CT-determined lesion sizes to IQ (19), brain anomalies and linguistic function in children who are HIVC (20), and CT findings in autism (21–25). Structural Magnetic Resonance Imaging Structural magnetic resonance imaging (MRI) creates an anatomical representation of the brain. It was first utilized to image a human being in 1976, although the theory underlying the method was postulated in 1946 (26). The MRI instrument consists of a flat table that slides into a circular magnet. The technique is based on the following physiological parameters. Hydrogen atoms in the body’s water and fat each spin at their own rate. When a body is placed inside the MRI machine, its atoms align with the magnetic field surrounding it. A resonance frequency (RF) is emitted, and the hydrogen atoms that are reactive to the emitted frequency spin in unison. When the pulse is discontinued, the atoms return to their normal spin, emitting energy proportional to that which was absorbed.

Role of Neuroimaging

323

The emitted energy is then used to create an image. The nuclei of different tissues have different frequencies and therefore react only to a specific RF pulse. This allows the discrimination of different brain regions. MRI is relatively expensive but non-invasive. It has good spatial resolution, which is continually improving (in fact, new analytic methods allow re-analysis of older scans for specific brain structures). There are several different pulse sequences that can be applied, such as T1, T2, and FLAIR techniques. Each of these sequences is designed for optimal visualization of anatomy and/or pathology. The machine can be confining for children and for adults who are large, agitated, or claustrophobic. In some cases, these problems can be ameliorated by allowing practice and accommodation in the scanner or a mock scanner before the actual imaging sequences are performed. MRI is commonly used clinically with children who have brain tumors or traumatic brain injuries (27–41). It has also been applied as a diagnostic tool to determine the etiology of seizure disorders (42–44). In the research arena, structural MRI has been utilized to follow the pattern of brain changes associated with normal child neurodevelopment (45–47) and to define the structural brain changes associated with disorders such as autism (48–50), attention deficit hyperactivity disorder (ADHD) (51–54), post-traumatic stress disorder (PTSD) (55–57), major depression (58), and premature birth (59,60). Structural MRI has also been used in developmental neurotoxicological research investigating alcohol, cocaine, methamphetamine, lead, and methylmercury. These studies will be discussed later in this chapter.

Functional Magnetic Resonance Imaging Functional magnetic resonance imaging (fMRI) is a technique that creates images of the brain’s neuronal activity while an individual or animal is carrying out a behavioral task (61). Completed in the MRI scanner, it measures changes in regional blood flow (perfusion fMRI), random movement of water molecules (diffusion weighted fMRI), or regional differences in oxygenated blood (BOLD). The most commonly used technique, blood oxygen level dependent fMRI, or BOLD, will be described here. This technique is based on the fact that active neurons require oxygen, increasing blood flow. BOLD relies on the different paramagnetic properties of oxy- and deoxy-hemoglobin. The signal that is detected and presented in imaging results represents oxygenated hemoglobin from cells in active brain regions. While in the scanner, the examinee carries out functional tasks designed to activate specific brain structures or neural systems. Signals detected in active brain regions are then superimposed on a structural MRI of the examinee, allowing identification of activated areas. Such results can be summed across examinees and presented as images representing group data. fMRI is relatively expensive but noninvasive. It has good spatial but limited temporal resolution. A limitation of the method is that it is time-consuming, which can be problematic when asking children to be confined in the MRI environment for an extended period of time. There are also limitations on the types of tasks that can be used in the MRI environment due to space constriction and the presence of the magnet. Often, tasks are presented by computer and projected onto space that can be visualized by the examinee within the scanner (for a description of some types of fMRI tasks, see the section on the Faroe Islands project below). fMRI is now being applied clinically as a preoperative tool in patients with seizure disorders in place of the WADA technique to guide safe surgical pathways (62–66). Research applications have included the investigation of normal neurodevelopment (67–73). It has also been utilized in studies of ADHD treatment interventions (74) and in the investigation of functional abnormalities in autism (49,75,76). A recent study that used

324

Janulewicz et al.

fMRI to examine the effects of in utero exposure to methylmercury will be described later in this chapter. Magnetic Resonance Spectroscopy Magnetic resonance spectroscopy (MRS) allows visualization of the biochemical composition and metabolism of the brain (77). It has promise as a technique in behavioral toxicology because of its capacity to detect chemicals and neurotransmitters. It utilizes the same machine and theory as structural MRI, recording resonance frequencies for compounds other than water. MRS can detect sodium, potassium, carbon, nitrogen, and fluorines. The N-acetylaspartate/creatine (NAA/Cr) ratio is commonly measured using MRS technology. Creatine is used in the ratio as a “hold” measure because levels of it tend to stay constant in the brain. Increased levels of NAA have been utilized to indicate specific types of neuronal damage: in gray matter NAA is located in the neuronal cell body, while in white matter it is confined to the axon. Therefore, its presence can be used to discriminate cell body versus axonal damage (78). In order to be detected by MRS, a compound must possess nuclear spin, unique RF, spin relaxation, and relaxation properties. The end result of MRS is not an image of the brain as in MRI and fMRI but a print-out of the different radio frequencies and intensities of the compounds measured. This technique has been refined to be capable of comparing white and gray matter metabolism (79). MRS has not been commonly used in clinical practice with children but is becoming a more popular technique in the research arena. There have been a few MRS studies done of children in normal, psychiatric, and developmentally delayed populations (64,80–91). Some work has also been done with children who were exposed to cocaine, methamphetamine, or lead. These studies will be discussed later in this chapter. Positron Emission Tomography Positron emission tomography (PET) creates a picture of the brain’s active metabolism (92). For this technique, the examinee ingests glucose labeled with short acting radioisotopes. The glucose is transported into the brain, where active areas require more glucose and therefore demonstrate more radioactivity. The examinee is asked to perform tasks between the time of administration of radiolabeled glucose and scanning, so those brain regions that are more active can be identified through imaging. PET is relatively expensive and is invasive. It is a useful tool in that the examiner can administer behavioral tests outside of the scanner environment. Therefore, it is not limited by factors that must be considered in fMRI, such as motion during the task. However, PET has poor temporal and spatial resolution. Because radioactive tracers must be administered to the examinee, it is not a useful technique for studies requiring multiple scans over a short time. PET has been applied both clinically and in research with child populations. It has been utilized to study developmental disorders, including but not limited to autism, Asperger’s, ADHD, and phenylketonuria (PKU) (76,93–99). We were unable to locate any existing studies using PET in the field of developmental neurotoxicology. Single Photon Emission Computerized Tomography Single photon emission computerized tomography (SPECT) is similar to PET but uses one rather than two photons (6). Since only one photon is emitted, a special lens called a collimater is required to locate the photon. This technique allows detection of the brain

Role of Neuroimaging

325

metabolism of compounds other than glucose. Because the collimator is required, the detection efficiency is less than with PET. SPECT has been useful in clinical populations, especially for localizing focal brain areas responsible for epilepsy in childhood prior to surgery (100) and for the diagnosis of viral encephalitis (101,102). SPECT has been applied to research efforts to understand the neurological underpinnings of disorders such as obsessive compulsive disorder (103,104), Tourette’s syndrome (105–107), autism (99,108–110), and ADHD (103,111,112). In the developmental neurotoxicology literature, there are a few studies describing the usefulness of SPECT in identifying the effects of prenatal exposure to alcohol and cocaine. These are described later in this chapter. IMAGING IN DEVELOPMENTAL NEUROTOXICOLOGY RESEARCH Drugs of Abuse Ethanol CNS effects of prenatal exposure to ethanol (ETOH) are classified under the umbrella of fetal alcohol spectrum disorders. These include fetal alcohol syndrome (FAS), fetal alcohol effects, and alcohol-related neurobehavioral disorders (113). The first case of FAS was described in the 1970s by Jones and Smith (114), when the only way to observe alcoholrelated neuropathological abnormalities was through autopsy cases. In the past decade, structural MRI, PET and SPECT have been employed to visualize ETOH-related brain abnormalities. A number of structural imaging studies carried out on children with prenatal ETOH exposure have shown abnormalities in the cerebellum, corpus callosum, and basal ganglia, in addition to overall microcephaly (115,116). These findings are consistent with autopsy studies (117). Some researchers have used more sophisticated analytical techniques for MRI that allow evaluation of the brain as a whole, showing that high prenatal ETOH exposure may be associated with displaced corpus callosum (113–116), altered gray matter densities in temporal lobe regions (116), or altered white matter (114,116). To date, it has been impossible to clearly link these structural abnormalities to any one particular neurocognitive deficit associated with FAS. One study utilizing PET demonstrated that individuals with FAS had decreased glucose metabolism in the cerebellum as compared with controls (118). Riikonen et al. (1999) (119), using SPECT, showed lower activation in the left parietoocciptial region in FAS patients than controls. This brain region is thought to functionally mediate arithmetic ability, which is a common area of difficulty for children with FAS (120). Cocaine In adults, cocaine use has been shown to cause vascular constriction. Konkol et al. (1994) (121) used SPECT to examine infants exposed in utero to cocaine. None of the infants in their study evidenced abnormal cerebral blood flow, suggesting that infants are spared the severe vascular changes seen in adults. However, the cocaine-exposed children presented with abnormalities on behavioral measures and electroencephalogram. MRS methodology was applied to the study of the effects of in utero exposure to cocaine by Smith et al. (2001) (122). While the structural MRIs completed on the children in their sample were all normal, the MRS data revealed significantly higher creatine levels in the frontal white matter in the exposed group. There was no significant difference in NAA levels between the exposed group and controls in the two brain regions studied (frontal white matter and striatum).

326

Janulewicz et al.

Methamphetamine Prenatal methamphetamine exposure has been shown to cause birth defects, including growth retardation, cardiac defects, and clefting (123). The effects on the developing CNS are still unclear, although this question was addressed by two recent studies. Smith et al. (2001) (123) used both MRI and MRS methods to visualize the brains of 12 children with prenatal methamphetamine exposure. No visible structural brain differences were observed between the exposed and control groups. However, a change in brain metabolism was observed in the exposed children, who had significantly higher levels of creatine in the striatum when compared to controls. The levels of N-acetylaspartate did not differ significantly between the two groups in the two brain regions assessed (striatum and frontal lobes), suggesting that there was no neuronal loss or damage in either region. A behavioral questionnaire was also filled out for all of these children and no behavioral issues were identified. Chang et al. (2004) (124) used MRI to evaluate global brain and regional brain structure volume in children with methamphetamine exposure. The children were also administered a battery of neuropsychological tests. When compared to controls, the methamphetamine exposed children showed smaller volumes in the putamen, globus pallidus, hippocampus, and caudate. The reduction in the volumes of these brain structures was associated with worse performance on tasks of sustained attention and delayed verbal memory. Environmental Exposures Lead A case study using imaging methodology (79,125) compared a lead-exposed 10 years-old boy to his unexposed first cousin. The boy was diagnosed with lead exposure at the age of 3 years, with blood lead levels of 51 mg/dL at 38 months and 44 mg/dL at 41 months. At age10 he showed deficits in reading, writing, arithmetic, language and attention consistent with the known cognitive effects of early lead exposure (126–129). A structural MRI was normal. The MRS data showed a marked difference in the levels of brain metabolites in the frontal lobes between the exposed boy and his unexposed cousin. The NAA/Cr ratio was lower in both frontal gray and white matter and the Mi/Cr ratio was decreased. Following on this finding, Troupe et al. (78) recruited 16 children with documented lead exposure for MRI and MRS studies. The age at exposure ranged from 10 months to 3 years–5 months, the highest recorded lead level ranged from 23–65 mg/dl, and the age at testing ranged from 4–21 years. Unexposed siblings and cousins of the lead exposed children were used as controls. Structural MRIs were normal for all of the children (exposed and controls). MRS studies focused on the frontal lobes, partitioning out the white and gray matter using the technique invented by Lopez-Villegas (125). NAA/Cr ratios were significantly lower in the frontal gray matter of the exposed children than controls. No differences were found in any of the ratios measured in the frontal white matter between the two groups. Results suggested that lead selectively disturbs the gray matter in the frontal lobes of children exposed at an early age. Methylmercury An early CT scan study examined 5 children diagnosed with Minamata disease who had prenatal MeHg exposure (Matsumoto et al., 1988) (5). This study revealed ventricular enlargement (5 children), cerebellar atrophy (4 children), cortical atrophy of the frontal and parietal lobes (3 children), and cortical atrophy of the temporal and parietal lobes

Role of Neuroimaging

327

(2 children). In 1969, a family in the United States was exposed to high levels of MeHg through contaminated pork. Two of the individuals, aged 20 and 13 years at the time of exposure, underwent MRI twenty two years after the exposure took place which revealed tissue loss in the calcarine and parietal cortices and cerebellar folia (130).

FAROESE IMAGING STUDY In this section of the chapter, we summarize the application of fMRI techniques to the evaluation of the CNS effects of prenatal exposure to two neurotoxicants, MeHg and PCBs, in adolescents. A paper has already been prepared for publication describing the results for two of the behavioral challenge paradigms used, photic stimulation and fingertapping, in adolescents with high- and low- mixed MeHg and PCB exposure (131). In the present summary, we present the overall study design and describe the four behavioral challenge methods used as an illustration of how neuroimaging methodology can be employed in behavioral toxicology research. We also present detailed tables summarizing findings for the two tests not presented in the prior paper, again summarizing results for the high- and low-mixed exposure groups. Methods Participants Participants were 12 adolescent boys aged 15–16 who are members of the cohort of Faroese children assembled in 1985–86 (132). The 12 children were selected to include the 3 boys from the cohort who best represented each of the following four prenatal exposure groups: highest MeHg exposure, highest PCB exposure, highest mixed exposure and lowest mixed exposure (see White et al., 2005, for a detailed description of the exposure groups). In the high mixed group, MeHg exposure ranged from 81.30–114.00 mcg/l in cord blood, while PCB exposure ranged from 4.60–6.91 ng/g in cord tissue. In the low mixed group, MeHg exposure ranged from 4.30–13.80 mcg/l in cord blood and PCBs from 0.40–0.53 ng/g in cord tissue. All of the participants were right-handed. The project PI in the Faroe Islands, Dr. Pal Weihe, met with the parents of the 12 adolescents selected on the basis of exposure. All parents and children agreed to participate in the study, and the boys traveled to Massachusetts with four Faroese chaperones in order to complete study methods. The study was approved by the Faroese community ethical review board and by the McLean Hospital IRB. Informed consent was provided by the boys’ parents. Imaging Methods All participants underwent a structural MRI followed by a series of functional MRI studies. MRI scans were performed at the McLean Brain Imaging Center in Belmont, Massachusetts. Structural Imaging. Conventional MR images were acquired on all subjects prior to functional MR imaging using a 1.5 T General Electric (Milwaukee, Wiscansin, U.S.A.) whole body imaging device operating at software level 4.8. A T1-weighted mid-sagittal localizer image was used to determine a plane perpendicular to the anterior commissure– posterior commissure line, followed by double echo spin–echo 3 mm axial slices of whole brain. The imaging parameters were a 256!256 matrix, TRZ3000 ms, TEZ30 and 8 ms, 24 cm field of view with two interleaved acquisitions. This resulted in 108 contiguous double echo slices. The voxel dimensions were 0.975 by 0.975 by 3 mm. A 3D Fourier

328

Janulewicz et al.

transform spoiled gradient-recalled acquisition (3 DFT SPGR) was also used to generate images with good contrast between gray matter and white matter (133). These data consisted of 124 1.5 mm thick coronal slices collected with the following imaging parameters: TR Z35 msec, TE Z5 msec with one repetition, flip angleZ45 degrees with a 256!256 matrix. Voxel dimensions were 0.975 by 0.975 by 1.5 mm. Reports on structural scans were sent to the Faroese PI, who conveyed results to the children’s parents and physicians. Functional Imaging. The fMRI study protocol consisted of performing two cognitive tasks, as well as a motor task and a passive visual stimulation paradigm while undergoing a functional MRI scan. These tasks were selected for the study because extensive prior research has shown that there is a significant relationship between higher exposure to MeHg and/or PCBs and lower performance on tasks assessing the same neuropsychological domains. All scans were performed with a high-speed imaging device (1.5 T; General Electric Signa, modified by Advanced NMR Systems) and a quadrature head coil receiver system using a modified echo-planar imaging technique. For each participant, we collected approximately 15 slices with a thickness of 6 mm. These slices were acquired in continuous succession. A series of 50 sequential images were obtained during the cognitive task. Images were collected every three seconds using a gradient echo pulse sequence (TEZ40 msec, flip angleZ75 degrees.) An image matrix of 64!128 was used with a 3 mm!3 mm in plane resolution. The fMRI protocols were sequential task activation paradigms that started with 30 seconds of resting baseline followed by alternating active and resting states in 30 seconds intervals resulting in total paradigm lengths of 150 seconds. Control paradigm 1: Photic stimulation. The participant was asked to open his eyes and lie quietly while data were obtained over a two-and-a half-minute period consisting of two 30 second cycles of darkness alternating with two 30 second cycles of binocular, 8 Hz patterned flash photic stimulation. The light stimulation was produced by goggles with light emitting diodes. This sequence was included because it requires only passive viewing and does not require movement, motivation or attention. Activation in response to visual stimulation in the absence of activation in response to our other paradigms assures us that cortical activation is present and detectable for a simple sensory task. Challenge paradigm 2: Hooper Visual Organization Test. The Hooper Visual Organization Test (HVOT) (134) consists of test stimuli that are pictures of disassembled, but readily recognizable, objects. The participant was asked to name the object formed if the pieces were put together appropriately. The objects are high frequency items and the test is not highly taxing with regard to naming abilities. The stimuli were presented visually from a Macintosh-controlled video display. During each activation period, each stimulus item was presented visually via video display. Participants were asked to name each stimulus item as soon as it was recognized. Each stimulus was presented for 10 seconds whether or not the subject named the stimulus in that time period. During the relaxation period, the participant was asked to relax and no stimuli were presented. Challenge paradigm 3: Boston Naming Test. The Boston Naming Test (BNT) (135) requires confrontation naming of objects depicted in line drawings. During standard administration of this test, if no correct response is produced within 20 seconds, semantic and/or phonemic cues are given. For use in the scanner, participants were asked only to name the object to confrontation. No cues were given if no response was given. Twenty stimuli were selected to represent an equal number of high and low frequency items. Each stimulus was presented for 3 seconds whether or not the participant named the stimulus in

Role of Neuroimaging

329

that time period. During the relaxation period, the participant was asked to relax and no stimuli were presented. Challenge paradigm 4 and 5: Motor task: Finger movement. Participants were asked to touch each finger to their thumb in a sequential, repetitive procedure during the activation periods. During the relaxation phase, the child was asked to relax and not move his fingers. The task was performed with the right hand (Right Finger Tapping, RFT) then repeated as a separate task with the left hand (Left Finger Tapping, LFT). This task was included because it is robust, has demonstrated clear activation sites, and should produce easily detectable signal changes for all participants. Data Analysis Statistical parametric maps (SPMs) were generated for each participant using SPM99 (136,137) (from the Wellcome Dept. of Cognitive Neurology, London, U.K.) implemented in Matlab (The Mathworks Inc., Sherborn, Massachusetts, U.S.A.), with an IDL interface. The functional data sets were motion corrected within SPM99 using the first image as the reference. After motion correction, the functional data sets were normalized to a standard template from Montreal Neurological Institute (MNI). The normalized data sets were first resampled to 2!2!2 mm within MNI space using sinc interpolation and then smoothed using a 4!4!4 mm Full Width at Half Maximum Gaussian smoothing kernel (138). A 150 sec box-car waveform, convolved with hemodynamic response function, was used as the reference paradigm. Using general linear model and the hemodynamically-corrected reference paradigm, the T-score values were calculated for each voxel and the SPMs were generated. In order to determine the differences in activation between the high mixed MeHg and PCB exposure versus low mixed exposure groups, a second-level paired t-test analysis was performed. The SPM group maps were generated using a fixed-effects model within SPM99 with the individual contrast maps (139). The resulting maps were thresholded at a TO3.5 (p!.0005) with a minimum cluster-size threshold set at 10 voxels. The location, size, and the strength of activation (T-score) were calculated for each activation cluster in SPM99. Anatomical regions were determined for activation maxima using the Talairach Daemon (140,141). The analysis was repeated for each paradigm (Photic Stimulation, HVOT, BNT, LFT and RFT).

Results Paradigm 1: Photic Stimulation Photic stimulation results have been described in detail elsewhere (131). To summarize, the high mixed exposure group showed greatest activation in primary visual cortex, while the low mixed exposure group showed the typical activation pattern seen in unexposed groups, i.e., greatest activation in visual association cortex.

Paradigm 2: Hooper Visual Organization Test Comparison of the high exposure group to the low exposure group revealed small but significant differences in the cerebellum bilaterally, the frontal, parietal and occipital lobes bilaterally and the right temporal lobe during the HVOT (Table 1, Fig. 1).

330

Janulewicz et al.

Table 1 Foci of Maximally Activated Brain Regions Surviving the Hooper Visual Organization Test for High MeHg and PCB Compared to Low MeHg and PCB Analysis Brodmann’s area Left cerebellum, posterior Right cerebellum, posterior Right occipital lobe Right temporal lobe, inferior Left frontal lobe, inferior Left occipital lobe Right frontal lobe, inferior Right parietal lobe, postcentral gyrus Left parietal lobe, precuneus

20

18 13

7

Coordinates

Number of activated voxels

X

Y

Z

Max T-score

Max Z-score

74

K50

K62

K20

7.96

7.56

87

40

K68

K18

7.72

7.36

108

16

K92

0

6.92

6.65

66

34

K10

K36

7.35

7.03

45

K46

34

6

6.91

6.64

275 32

K2 44

K74 24

20 8

6.76 6.49

6.52 6.27

37

40

K32

52

6.41

6.20

20

K28

K78

38

5.64

5.50

Figure 1 Activation during the Hooper Visual Organization Test for high mixed MeHg and PCB exposure compared to low mixed MeHg and low exposure. (A) Activation for the high exposure group and (B) the low exposure group during the Hooper Visual Organization test. Activation is similar for the two groups in the cerebellum, but there is greater activation in the high exposure group in the occipital and parietal lobes than is seen in the low exposure group, suggesting the need for the high exposure group to recruit additional brain structures to aid in function.

Role of Neuroimaging

331

Table 2 Foci of Maximally Activated Brain Regions Surviving the Boston Naming Test when High MeHg and PCB Are Compared to Low MeHg and PCB Analysis

Region of activation Left cerebellum, posterior Right frontal lobe, inferior temporal gyrus Right frontal lobe, middle frontal gyrus Left frontal lobe, middle frontal gyrus Left parietal lobe, precuneus Right cerebellum, posterior lobe Right temporal lobe, Inferior temporal gyrus Left limbic lobe, cingulate gyrus Left frontal lobe, medial frontal gyrus Left temporal lobe medial frontal gyrus Left temporal lobe, middle temporal gyrus Left frontal lobe, medial gyrus Left temporal lobe, middle temporal gyrus Right frontal lobe, superior frontal gyrus Left frontal lobe, paracentral lobule (sensory cortex) Right frontal lobe, subcortical white matter Left frontal lobe, inferior frontal gyrus Left Frontal lobe, subcortical white matter

Coordinates

Brodmann’s area

Number of activated aoxels

– 20

479 455

10

130

30

50

10

30

K30

Max T-score

Max Z-score

9.95 9.27

Infinity Infinity

K6

8.70

Infinity

60

22

8.43

Infinity

K34 K70

22

7.86

7.48

181

32 K80 K24

7.48

7.15

70

64 K26 K16

7.31

7.00

173

20

Y

Z

K38 K80 K32 34 K10 K38

208

K14 K10

36

7.09

6.81

73

K6 K22

62

6.56

6.33

73

K6 K22

62

6.56

6.33

43

K54 K66

20

6.26

6.06

73

K6 K22

62

6.56

6.33

43

K54 K66

20

6.26

6.06

16

48

5.97

5.79

K4 K38

60

5.73

5.57

47 5

X

23

18

158

30

30

0

5.58

5.44

21

K50

28

6

5.48

5.34

32

K14 K26

46

5.23

5.11

Paradigm III: Boston Naming Test The greatest differences in activation between the high and low exposure groups during the BNT occurred in the left cerebellum, right frontal lobe, and the left precuneus of the medial parietal lobe. In addition, other regions of the left and right frontal and temporal lobes were also significantly different between the two groups during this task but to a lesser extent. See z-scores in Table 2. According to visual inspection of the individual group activation patterns, it appears that the group with high MeHg levels but low PCB levels had increased

332

Janulewicz et al.

activation in the occipital region while the group with high levels of PCBs and low levels of MeHg had greater activation in frontal regions. Though these analyses were beyond the scope of this report, it is interesting to note that the different exposures may be contributing to the different activation levels in specific regions of the brain. Paradigms IV and V: Left and Right Finger Tapping Data from the finger tapping challenge tasks are detailed in another report (131). Summarized briefly, these results showed that the boys with low mixed exposure used fewer cerebral resources to carry out the task (i.e., highly focal ipsilateral brain areas were activated during the task), while the children with high mixed exposure employed more brain areas for task completion (activation of bilateral brain areas). These findings were particularly prominent for the non-dominant (left) hand.

DISCUSSION These results derive from data collected as a pilot project and are based on small numbers of participants. Therefore, the results must be interpreted with caution. However, the strikingand statistically significant-differences in brain activation patterns on fMRI between the high and low mixed exposure groups are tantalizing. They suggest that high exposure may be associated with losses in functional capacity in specific brain structures (e.g., visual association cortex) resulting in adaptive recruitment of other brain structures for task completion. In addition, results suggest that tasks normally associated with focal brain activation show activation in several brain areas, again suggesting adaptive recruitment of additional brain structures to aid in function. Data briefly mentioned in this summary also raise the possibility that exposures to MeHg and PCBs may be associated with different activation patterns in brain structures. Further analyses of these data are on-going. This small pilot also suggests that imaging technology may be a powerful means of further exploring and confirming exposure-outcome results revealed by behavioral test methodology.

ACKNOWLEDGMENTS The work presented in the section “Faroese Imaging Study” is the joint effort of a team of collaborators. In addition to two of the chapter authors (RFW, CP), the team included Dr. Pal Weihe, Director, Faroese Health System; Frodi Debes, University of Southern Denmark; Dr. Kristin Heaton, Boston University School of Public Health; Dr. Philippe Grandjean, University of Southern Denmark; and Deborah Yurgelun-Todd, Harvard University School of Medicine, McLean Hospital Functional Imaging Center.

REFERENCES 1. Devine SA, Kirkley SM, Palumbo CL, White RF. MRI and neuropsychological correlates of carbon monoxide exposure: a case report. Environ Health Perspect 2002; 110:1051–1055. 2. Pavese N, Napolitano A, De Iaco G, et al. Clinical outcome and magnetic resonance imaging of carbon monoxide intoxication. A long-term follow-up study. Ital J Neurol Sci 1999; 20:171–178.

Role of Neuroimaging

333

3. Gale SD, Hopkins RO, Weaver LK, Bigler ED, Booth EJ, Blatter DD. MRI, quantitative MRI, SPECT, and neuropsychological findings following carbon monoxide poisoning. Brain Inj 1999; 13:229–243. 4. Korogi Y, Takahashi M, Okajima T, Eto K. MR findings of Minamata disease—organic mercury poisoning. J Magn Reson Imaging 1998; 8:308–316. 5. Matsumoto SC, Okajima T, Inayoshi S, Ueno H. Minamata disease demonstrated by computed tomography. Neuroradiology 1988; 30:42–46. 6. Kolb B, Whisaw IQ. Fundamentals of Human Neuropsychology. 4th ed. New York, NY: W.H. Freeman and Company, 1996. 7. Temple MJ, Langer JC. Image-guided surgery for the pediatric patient: ultrasound, computerized tomography, and magnetic resonance imaging. Curr Opin Pediatr 2003; 15:256–261. 8. Desai KI, Nadkarni TD, Muzumdar DP, Goel A. Prognostic factors for cerebellar astrocytomas in children: a study of 102 cases. Pediatr Neurosurg 2001; 35:311–317. 9. Sutton LN, Cnaan A, Klatt L, et al. Postoperative surveillance imaging in children with cerebellar astrocytomas. J Neurosurg 1996; 84:721–725. 10. Steinbok P, Hentschel S, Cochrane DD, Kestle JR. Value of postoperative surveillance imaging in the management of children with some common brain tumors. J Neurosurg 1996; 84:726–732. 11. Piovesana AM, Moura-Ribeiro MV, Zanardi VD, Goncalves VM. Hemiparetic cerebral palsy: etiological risk factors and neuroimaging. Arq Neuropsiquiatr 2001; 59:29–34. 12. Tomberg TA, Ta’lvik TA, Iurisson MKH, Tulu E. Pathologic changes of the brain in children with various forms of infantile cerebral palsy (computerized-tomographic data). Pediatriia 1989; 9:50–54. 13. Koch B, Braillier D, Eng G, Binder H. Computerized tomography in cerebral-palsied children. Dev Med Child Neurol 1980; 22:595–607. 14. Morad Y, Avni I, Benton SA, et al. Normal computerized tomography of brain in children with shaken baby syndrome. J Aapos 2004; 8:445–450. 15. Zepp F, Bruhl K, Zimmer B, Schumacher R. Battered child syndrome: cerebral ultrasound and CT findings after vigorous shaking. Neuropediatrics 1992; 23:188–191. 16. Dahl-Grove DL, Chande VT, Barnoski A. Closed head injuries in children: is hospital admission always necessary? Pediatr Emerg Care 1995; 11:86–88. 17. Costeff H, Groswasser Z, Goldstein R. Long-term follow-up review of 31 children with severe closed head trauma. J Neurosurg 1990; 73:684–687. 18. Clifton GL, Grossman RG, Makela ME, Miner ME, Handel S, Sadhu V. Neurological course and correlated computerized tomography findings after severe closed head injury. J Neurosurg 1980; 52:611–624. 19. Kornhuber HH, Bechinger D, Jung H, Sauer E. A quantitative relationship between the extent of localized cerebral lesions and the intellectual and behavioural deficiency in children. Eur Arch Psychiatry Neurol Sci 1985; 235:129–133. 20. Rotta NT, Silva AR, Silva CL, et al. Follow-up of patients with vertically-acquired HIV infection who are more than 9 years old. J Trop Pediatr 2003; 49:253–255. 21. Balottin U, Bejor M, Cecchini A, Martelli A, Palazzi S, Lanzi G. Infantile autism and computerized tomography brain-scan findings: specific versus nonspecific abnormalities. J Autism Dev Disord 1989; 19:109–117. 22. Tsai LY, Jacoby CG, Stewart MA. Morphological cerebral asymmetries in autistic children. Biol Psychiatry 1983; 18:317–327. 23. Campbell M, Rosenbloom S, Perry R, et al. Computerized axial tomography in young autistic children. Am J Psychiatry 1982; 139:510–512. 24. Damasio H, Maurer RG, Damasio AR, Chui HC. Computerized tomographic scan findings in patients with autistic behavior. Arch Neurol 1980; 37:504–510. 25. Hier DB, LeMay M, Rosenberger PB. Autism and unfavorable left-right asymmetries of the brain. J Autism Dev Disord 1979; 9:153–159.

334

Janulewicz et al.

26. Hornak, J (2005, May 10). The Basics of MRI. Retrieved ‘January 2005’ from http://www.cis.rit.edu/htbooks/mri/index.html. 27. Hadley DM, Teasdale GM, Jenkins A, et al. Magnetic resonance imaging in acute head injury. Clin Radiol 1988; 39:131–139. 28. Brunel H, Raybaud C, Peretti-Viton P, et al. Craniopharyngioma in children: MRI study of 43 cases. Neurochirurgie 2002; 48:309–318. 29. Davis PC, Wichman RD, Takei Y, Hoffman JC, Jr. Primary cerebral neuroblastoma: CT and MR findings in 12 cases. Am J Roentgenol 1990; 154:831–836. 30. Im SH, Chung CK, Cho BK, et al. Intracranial ganglioglioma: preoperative characteristics and oncologic outcome after surgery. J Neurooncol 2002; 59:173–183. 31. Sprung C, Baerwald R, Henkes H, Schorner W. A comparative study of CT and MRI in midline tumors of childhood and adolescence. Child Nerv Syst 1989; 5:102–106. 32. Thiesse P, Turjman F, Kaemmerlen P, et al. Contribution of imaging in the diagnosis and follow-up of brain tumors. Pediatrie 1991; 46:7–22. 33. Korones DN, Butterfield R, Meyers SP, Constine LS. The role of surveillance magnetic resonance imaging (MRI) scanning in detecting recurrent brain tumors in asymptomatic children. J Neurooncol 2001; 53:33–38. 34. Bigler ED. Neuroimaging in pediatric traumatic head injury: diagnostic considerations and relationships to neurobehavioral outcome. J Head Trauma Rehabil 1999; 14:406–423. 35. Breysem L, Cossey V, Mussen E, Demaerel P, Van de Voorde W, Smet M. Fetal trauma: brain imaging in four neonates. Eur Radiol 2004; 14:1609–1614. 36. Khoshyomn S, Tranmer BI. Diagnosis and management of pediatric closed head injury. Semin Pediatr Surg 2004; 13:80–86. 37. Pierallini A, Pantano P, Fantozzi LM, et al. Correlation between MRI findings and long-term outcome in patients with severe brain trauma. Neuroradiology 2000; 42:860–867. 38. Poussaint TY, Moeller KK. Imaging of pediatric head trauma. Neuroimaging Clin N Am 2002; 12:271–294 see also page. ix. 39. Prasad MR, Ewing-Cobbs L, Swank PR, Kramer L. Predictors of outcome following traumatic brain injury in young children. Pediatr Neurosurg 2002; 36:64–74. 40. Verger K, Junque C, Levin HS, et al. Correlation of atrophy measures on MRI with neuropsychological sequelae in children and adolescents with traumatic brain injury. Brain Inj 2001; 15:211–221. 41. Wilde EA, Hunter JV, Newsome MR, et al. Frontal and temporal morphometric findings on MRI in children after moderate to severe traumatic brain injury. J Neurotrauma 2005; 22:333–344. 42. Major P, Decarie JC, Nadeau A, et al. Clinical significance of isolated hippocampal volume asymmetry in childhood epilepsy. Neurology 2004; 63:1503–1506. 43. Gaillard WD. Structural and functional imaging in children with partial epilepsy. Ment Retard Dev Disabil Res Rev 2000; 6:220–226. 44. Gupta AK, Sharma R, Sarma D. Imaging in epilepsy. Indian J Pediatr 2000; 67:S40–S60. 45. Kennedy DN, Haselgrove C, McInerney S. MRI-based morphometric of typical and atypical brain development. Ment Retard Dev Disabil Res Rev 2003; 9:155–160. 46. Sowell ER, Delis D, Stiles J, Jernigan TL. Improved memory functioning and frontal lobe maturation between childhood and adolescence: a structural MRI study. J Int Neuropsychol Soc 2001; 7:312–322. 47. Sowell ER, Thompson PM, Tessner KD, Toga AW. Mapping continued brain growth and gray matter density reduction in dorsal frontal cortex: Inverse relationships during postadolescent brain maturation. J Neurosci 2001; 21:8819–8829. 48. Brambilla P, Hardan A, di Nemi SU, Perez J, Soares JC, Barale F. Brain anatomy and development in autism: review of structural MRI studies. Brain Res Bull 2003; 61:557–569. 49. Brambilla P, Hardan AY, di Nemi SU, et al. The functional neuroanatomy of autism. Funct Neurol 2004; 19:9–17. 50. McAlonan GM, Cheung V, Cheung C. Mapping the brain in autism. A voxel-based MRI study of volumetric differences and intercorrelations in autism. Brain 2005; 128:268–276.

Role of Neuroimaging

335

51. Hynd GW, Lorys AR, Semrud-Clikeman M, Nieves N, Huettner MI, Lahey BB. Attention deficit disorder without hyperactivity: a distinct behavioral and neurocognitive syndrome. J Child Neurol 1991; 6:S37–S43. 52. Hynd GW, Semrud-Clikeman M, Lorys AR, Novey ES, Eliopulos D, Lyytinen H. Corpus callosum morphology in attention deficit-hyperactivity disorder: morphometric analysis of MRI. J Learn Disabil 1991; 24:141–146. 53. Bussing R, Grudnik J, Mason D, Wasiak M, Leonard C. ADHD and conduct disorder: an MRI study in a community sample. World J Biol Psychiatry 2002; 3:216–220. 54. Hardan AY, Jou RJ, Keshavan MS, Varma R, Minshew NJ. Increased frontal cortical folding in autism: a preliminary MRI study. Psychiatry Res 2004; 131:263–268. 55. De Bellis MD, Hall J, Boring AM, Frustaci K, Moritz G. A pilot longitudinal study of hippocampal volumes in pediatric maltreatment-related posttraumatic stress disorder. Biol Psychiatry 2001; 50:305–309. 56. Thomas LA, De Bellis MD. Pituitary volumes in pediatric maltreatment-related posttraumatic stress disorder. Biol Psychiatry 2004; 55:752–758. 57. Villarreal G, Hamilton DA, Graham DP, et al. Reduced area of the corpus callosum in posttraumatic stress disorder. Psychiatry Res 2004; 131:227–235. 58. Ehrlich S, Noam GG, Lyoo IK, Kwon BJ, Clark MA, Renshaw PF. White matter hyperintensities and their associations with suicidality in psychiatrically hospitalized children and adolescents. J Am Acad Child Adolesc Psychiatry 2004; 43:770–776. 59. Skranes JS, Vik T, Nilsen G, Smevik O, Andersson HW, Brubakk AM. Cerebral magnetic resonance imaging and mental and motor function of very low birth weight children at six years of age. Neuropediatrics 1997; 28:149–154. 60. Woodward LJ, Mogridge N, Wells SW, Inder TE. Can neurobehavioral examination predict the presence of cerebral injury in the very low birth weight infant? J Dev Behav Pediatr 2004; 25:326–334. 61. Papanicolaou AC. Fundamentals of Functional Brain Imaging. Netherlands: Swets and Zeitlinger, 1998. 62. Cross JH. Epilepsy surgery in childhood. Epilepsia 2002; 43:65–70. 63. Medina LS, Bernal B, Dunoyer C, et al. Seizure disorders: functional MR imaging for diagnostic evaluation and surgical treatment—prospective study. Radiology 2005; 236:247–253. 64. Ross AJ, Sachdev PS. Magnetic resonance spectroscopy in cognitive research. Brain Res Brain Res Rev 2004; 44:83–102. 65. Ries ML, Boop FA, Griebel ML, et al. Functional MRI and Wada determination of language lateralization: a case of crossed dominance. Epilepsia 2004; 45:85–89. 66. Cataltepe O, Turanli G, Yalnizoglu D, Topcu M, Akalan N. Surgical management of temporal lobe tumor-related epilepsy in children. J Neurosurg 2005; 102:280–287. 67. Kawashima R, Taira M, Okita K, et al. A functional MRI study of simple arithmetic—a comparison between children and adults. Brain Res Cogn Brain Res 2004; 18:227–233. 68. Kwon H, Reiss AL, Menon V. Neural basis of protracted developmental changes in visuospatial working memory. Proc Natl Acad Sci USA 2002; 99:13336–13341. 69. Moses P, Roe K, Buxton RB, Wong EC, Frank LR, Stiles J. Functional MRI of global and local processing in children. Neuroimage 2002; 16:415–424. 70. Klingberg T, Forssberg H, Westerberg H. Increased brain activity in frontal and parietal cortex underlies the development of visuospatial working memory capacity during childhood. J Cogn Neurosci 2002; 14:1–10. 71. Richter W, Richter M. The shape of the fMRI BOLD response in children and adults changes systematically with age. Neuroimage 2003; 20:1122–1131. 72. Billingsley-Marshall RL, Simos PG, Papanicolaou AC. Reliability and validity of functional neuroimaging techniques for identifying language-critical areas in children and adults. Dev Neuropsychol 2004; 26:541–563. 73. Wood AG, Harvey AS, Wellard RM, et al. Language cortex activation in normal children. Neurology 2004; 63:1035–1044.

336

Janulewicz et al.

74. Vaidya CJ, Austin G, Kirkorian G, et al. Selective effects of methylphenidate in attention deficit hyperactivity disorder: a functional magnetic resonance study. Proc Natl Acad Sci USA 1998; 95:14494–14499. 75. Boddaert N, Zilbovicius M. Functional neuroimaging and childhood autism. Pediatr Radiol 2002; 32:1–7. 76. Boddaert N, Chabane N, Belin P, et al. Perception of complex sounds in autism: abnormal auditory cortical processing in children. Am J Psychiatry 2004; 161:2117–2120. 77. Ross B, Bluml S. Magnetic resonance spectroscopy of the human brain. Anat Rec 2001; 265:54–84. 78. Trope I, Lopez-Villegas D, Cecil KM, Lenkinski RE. Exposure to lead appears to selectively alter metabolism of cortical gray matter. Pediatrics 2001; 107:1437–1442. 79. Lopez-Villegas D, Kimura H, Tunlayadechanont S, Lenkinski RE. High spatial resolution MRI and proton MRS of human frontal cortex. NMR Biomed 1996; 9:297–304. 80. Bartha AI, Foster-Barber A, Miller SP, et al. Neonatal encephalopathy: association of cytokines with MR spectroscopy and outcome. Pediatr Res 2004; 56:960–966. 81. Chang L, Ernst T, Witt MD, Ames N, Gaiefsky M, Miller E. Relationships among brain metabolites, cognitive function, and viral loads in antiretroviral-naive HIV patients. Neuroimage 2002; 17:1638–1648. 82. Courvoisie H, Hooper SR, Fine C, Kwock L, Castillo M. Neurometabolic functioning and neuropsychological correlates in children with ADHD-H: preliminary findings. J Neuropsychiatry Clin Neurosci 2004; 16:63–69. 83. de Bildt A, Sytema S, Ketelaars C, Kraijer D, Volkmar F, Minderaa R. Measuring pervasive developmental disorders in children and adolescents with mental retardation: a comparison of two screening instruments used in a study of the total mentally retarded population from a designated area. J Autism Dev Disord 2003; 33:595–605. 84. Deruelle C, Rondan C, Gepner B, Tardif C. Spatial frequency and face processing in children with autism and Asperger syndrome. J Autism Dev Disord 2004; 34:199–210. 85. Filippi CG, Ulug AM, Deck MD, Zimmerman RD, Heier LA. Developmental delay in children: assessment with proton MR spectroscopy. Am J Neuroradiol 2002; 23:882–888. 86. Hashimoto T, Tayama M, Miyazaki M, et al. Differences in brain metabolites between patients with autism and mental retardation as detected by in vivo localized proton magnetic resonance spectroscopy. J Child Neurol 1997; 12:91–96. 87. Horska A, Kaufmann WE, Brant LJ, Naidu S, Harris JC, Barker PB. In vivo quantitative proton MRSI study of brain development from childhood to adolescence. J Magn Reson Imaging 2002; 15:137–143. 88. Keller MA, Venkatraman TN, Thomas A, et al. Altered neurometabolite development in HIVinfected children: correlation with neuropsychological tests. Neurology 2004; 62:1810–1817. 89. Magne C, Schon D, Besson M. Prosodic and melodic processing in adults and children. Behavioral and electrophysiologic approaches. Ann NY Acad Sci 2003; 999:461–476. 90. Mirza Y, Tang J, Russell A, et al. Reduced anterior cingulate cortex glutamatergic concentrations in childhood major depression. J Am Acad Child Adolesc Psychiatry 2004; 43:341–348. 91. O’Neill J, Levitt J, Caplan R, et al. 1H MRSI evidence of metabolic abnormalities in childhood-onset schizophrenia. Neuroimage 2004; 21:1781–1789. 92. Walker RC, Purnell GL, Jones-Jackson LB, Thomas KL, Brito JA, Ferris EJ. Introduction to PET imaging with emphasis on biomedical research. Neurotoxicology 2004; 25:533–542. 93. Boddaert N, Chabane N, Barthelemy C, et al. Bitemporal lobe dysfunction in infantile autism: positron emission tomography study. J Radiol 2002; 83:1829–1833. 94. Buchsbaum MS, Hollander E, Haznedar MM, et al. Effect of fluoxetine on regional cerebral metabolism in autistic spectrum disorders: a pilot study. Int J Neuropsychopharmacol 2001; 4:119–125. 95. Ernst M, Zametkin AJ, Matochik JA, Pascualvaca D, Jons PH, Cohen RM. High midbrain (18F)DOPA accumulation in children with attention deficit hyperactivity disorder. Am J Psychiatry 1999; 156:1209–1215.

Role of Neuroimaging

337

96. Haznedar MM, Buchsbaum MS, Wei TC, et al. Limbic circuitry in patients with autism spectrum disorders studied with positron emission tomography and magnetic resonance imaging. Am J Psychiatry 2000; 157:1994–2000. 97. Weglage J, Pietsch M, Funders B, Koch HG, Ullrich K. Deficits in selective and sustained attention processes in early treated children with phenylketonuria—result of impaired frontal lobe functions? Eur J Pediatr 1996; 155:200–204. 98. Yanai K, Iinuma K, Matsuzawa T, et al. Cerebral glucose utilization in pediatric neurological disorders determined by positron emission tomography. Eur J Nucl Med 1987; 13:292–296. 99. Zilbovicius M, Boddaert N, Belin P, et al. Temporal lobe dysfunction in childhood autism: a PET study. Positron emission tomography. Am J Psychiatry 2000; 157:1988–1993. 100. Buchhalter JR, So EL. Advances in computer-assisted single-photon emission computed tomography (SPECT) for epilepsy surgery in children. Acta Paediatr Suppl 2004; 93:32–35; discussion 36–7. 101. Kao CH, Wang SJ, Mak SC, Shian WJ, Chi CS. Viral encephalitis in children: detection with technetium-99m HMPAO brain single-photon emission CT and its value in prediction of outcome. Am J Neuroradiol 1994; 15:1369–1373. 102. Kao CH, Wang SJ, Mak SC, Shian WJ, Chi CS. Tc-99m HMPAO brain SPECT brain findings in pediatric viral encephalitis. Clin Nucl Med 1994; 19:590–594. 103. Cheon KA, Ryu YH, Kim JW, et al. Dopamine transporter density of the basal ganglia assessed with [123I] IPT SPECT in drug-naive children with Tourette’s disorder. Psychiatry Res 2004; 130:85–95. 104. Busatto GF, Buchpiguel CA, Zamignani DR, et al. Regional cerebral blood flow abnormalities in early-onset obsessive-compulsive disorder: an exploratory SPECT study. J Am Acad Child Adolesc Psychiatry 2001; 40:347–354. 105. Chiu NT, Chang YC, Lee BF, Huang CC, Wang ST. Differences in 99mTc-HMPAO brain SPET perfusion imaging between Tourette’s syndrome and chronic tic disorder in children. Eur J Nucl Med 2001; 28:183–190. 106. Lampreave JL, Molina V, Mardomingo MJ, et al. Technetium-99m-HMPAO in Tourette’s syndrome on neuroleptic therapy and after withdrawal. J Nucl Med 1998; 39:624–628. 107. Muller-Vahl KR, Berding G, Kolbe H, et al. Dopamine D2 receptor imaging in Gilles de la Tourette syndrome. Acta Neurol Scand 2000; 101:165–171. 108. Kaya M, Karasalihoglu S, Ustun F, et al. The relationship between 99mTc-HMPAO brain SPECT and the scores of real life rating scale in autistic children. Brain Dev 2002; 24:77–81. 109. Zilbovicius M, Garreau B, Tzourio N, et al. Regional cerebral blood flow in childhood autism: a SPECT study. Am J Psychiatry 1992; 149:924–930. 110. Chiron C, Leboyer M, Leon F, Jambaque I, Nuttin C, Syrota A. SPECT of the brain in childhood autism: evidence for a lack of normal hemispheric asymmetry. Dev Med Child Neurol 1995; 37:849–860. 111. Vles JS, Feron FJ, Hendriksen JG, Jolles J, van Kroonenburgh MJ, Weber WE. Methylphenidate down-regulates the dopamine receptor and transporter system in children with attention deficit hyperkinetic disorder (ADHD). Neuropediatrics 2003; 34:77–80. 112. Amen DG, Carmichael BD. High-resolution brain SPECT imaging in ADHD. Ann Clin Psychiatry 1997; 9:81–86. 113. Autti-Ramo I, Autti T, Korkman M, Kettunen S, Salonen O, Valanne L. MRI findings in children with school problems who had been exposed prenatally to alcohol. Dev Med Child Neurol 2002; 44:98–106. 114. Archibald SL, Fennema-Notestine C, Gamst A, Riley EP, Mattson SN, Jernigan TL. Brain dysmorphology in individuals with severe prenatal alcohol exposure. Dev Med Child Neurol 2001; 43:148–154. 115. Riley EP, Mattson SN, Sowell ER, Jernigan TL, Sobel DF, Jones KL. Abnormalities of the corpus callosum in children prenatally exposed to alcohol. Alcohol Clin Exp Res 1995; 19:1198–1202.

338

Janulewicz et al.

116. Riley EP, McGee CL, Sowell ER. Teratogenic effects of alcohol: a decade of brain imaging. Am J Med Genet C Semin Med Genet 2004; 127:35–41. 117. Pfeiffer JMF, Fischbach H, Bierich J, Volk B. Alcohol embryo-and fetopathy. J Neurol Sci 1979; 41:125–137. 118. Hannigan J, Martier SS, Chugani HT, Sokol RJ. Brain metabolism in children with foetal alcohol syndrome (FAS). A positron tomography study. Alcohol Clin Exp Res 1995; 19 p. abstract. 119. Riikonen R, Salonen I, Partanen K, Verho S. Brain perfusion SPECT and MRI in foetal alcohol syndrome. Dev Med Child Neurol 1999; 41:652–659. 120. Streissguth A, Barr H, Olson H, Sampson P, Bookstein F, Burgess D. Drinking during pregnancy decreases word attack and arithmetic scores on standardized tests: adolescent data from population-based prospective study. Alcohol Clin Exp Res 1994; 18:248–254. 121. Konkol RJ, Tikofsky RS, Wells R, et al. Normal high-resolution cerebral 99mTc-HMPAO SPECT scans in symptomatic neonates exposed to cocaine. J Child Neurol 1994; 9:278–283. 122. Smith LM, Chang L, Yonekura ML, et al. Brain proton magnetic resonance spectroscopy and imaging in children exposed to cocaine in utero. Pediatrics 2001; 107:227–231. 123. Smith LM, Chang L, Yonekura ML, Grob C, Osborn D, Ernst T. Brain proton magnetic resonance spectroscopy in children exposed to methamphetamine in utero. Neurology 2001; 57:255–260. 124. Chang L, Smith LM, LoPresti C, et al. Smaller subcortical volumes and cognitive deficits in children with prenatal methamphetamine exposure. Psychiatry Res 2004; 132:95–106. 125. Trope I, Lopez-Villegas D, Lenkinski RE. Magnetic resonance imaging and spectroscopy of regional brain structure in a 10-years-old boy with elevated blood lead levels. Pediatrics 1998; 101:E7. 126. Needleman HL, Schell A, Bellinger D, Leviton A, Allred EN. The long-term effects of exposure to low doses of lead in childhood. An 11-years follow-up report. N Engl J Med 1990; 322:83–88. 127. Bellinger DC, Stiles KM, Needleman HL. Low-level lead exposure, intelligence and academic achievement: a long-term follow-up study. Pediatrics 1992; 90:855–861. 128. Leviton A, Bellinger D, Allred EN, Rabinowitz M, Needleman H, Schoenbaum S. Pre- and postnatal low-level lead exposure and children’s dysfunction in school. Environ Res 1993; 60:30–43. 129. Fergusson DM, Horwood LJ, Lynskey MT. Early dentine lead levels and subsequent cognitive and behavioural development. J Child Psychol Psychiatry 1993; 34:215–227. 130. Davis LE, Kornfeld M, Mooney HS, et al. Methylmercury poisoning: long-term clinical, radiological, toxicological, and pathological studies of an affected family. Ann Neurol 1994; 35:680–688. 131. White RF, Palumbo C, Weihe P, Debes F, Heaton KJ, Grandjean P, Yurgelun-Todd D. Functional imaging findings in adolescents with high prenatal mixed exposures to methylmercury and polychlorinated biphenyls. Unpublished manuscript, 2005. 132. Grandjean PWP, Jorgensen PJ, Clarkson T, Cernichiari E, Videro T. Impact of maternal seafood diet on fetal exposure to mercury, selenium, and lead. Arch Environ Health 1992; 47:185–195. 133. Shenton M, Kikinia R, Jolesz F, et al. Abnormalities of the left temporal lobe and thought disorder in schizophrenia. N Engl J Med 1992; 327:604–612. 134. Hooper H. Hooper Visual Organization Test (VOT). Los Angeles: Western Psychological Services, 1983. 135. Kaplan E, Goodglass H, Weintraub S. The Boston Naming Test. 2nd ed. Philadelphia: Lea and Febiger, 1983. 136. Friston K, Frith CD, Frackowiak RS, Turner R. Characterizing dynamic brain responses with fMRI: a multivariate approach. Neuroimage 1995; 2:166–172. 137. Friston K, Frith CD, Turner R, Frackowiak RS. Characterizing evoked hemodynamics with fMRI. Neuroimage 1995; 2:157–165.

Role of Neuroimaging

339

138. Friston K, Worsley KJ, Frackowiak RSJ, Mazziotta JC, Evans AC. Assessing the significance of focal activations using their spatial extent. Human Brain Mapping 1994; 1:214–220. 139. Holmes A, Friston K. Generalisability, random effects & population inference. Neuroimage 1998; 7:s754. 140. Lancaster JL, et al. Automated Talairach atlas labels for functional brain mapping 2000; 10:120–131. 141. Lancaster J, Summerln JL, Rainey L, Freita SCS, Fox PT. The Talairach Daemon, a database server for Talairach Atlas Labels. Neuroimage 1997; 5:633.

16 Early Life Environmental Exposures and Neurologic Outcomes in Adults Marc G. Weisskopf Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A.

Robert O. Wright Department of Pediatrics, Children’s Hospital and The Channing Laboratory, Brigham and Women’s Hospital, Harvard Medical School and Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, U.S.A.

Howard Hu Department of Environmental Health, Harvard School of Public Health and The Channing Laboratory, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, U.S.A.

INTRODUCTION The theory that early life exposures might influence health later in life has been around since the beginning of the last century (1). The concept is perhaps most widely known with respect to early life nutrition and subsequent cardiovascular disease as adults, a hypothesis put forth by Professor David Barker and colleagues (2–4) and generally referred to as “the Barker hypothesis.” While the Barker hypothesis generally refers to nutritional deficiencies in early life, the wider concept of environmental exposures in early life affecting risk of adult disease—often referred to as “Critical Developmental Windows for Exposure”—has also been pursued in relation to other health endpoints. In particular, there is a long history within the field of mental health of examining the association between in utero exposures and adult schizophrenia (5,6). Could early life exposures affect other aspects of neurologic functioning in adulthood? As the population of many countries is aging rapidly, these questions take on added importance. The percentage of Americans over the age of 65 years is expected to increase from 13% in 1997 to 20% in 2047 (7). Based on current trends, this increase will dramatically increase the number of persons with cognitive impairment as well as the number with specific neurologic diseases of the elderly such as Alzheimer’s disease (AD) (8). Cognitive declines and impairment are already observed frequently among the elderly and are a threat not only to the health of the affected individuals, but also to the quality of life 341

342

Weisskopf et al.

of both the affected individuals and their caregivers. Impaired cognition among adults is associated with functional decline in activities of daily living, increased risk of injury to self and others, associated demands on caregivers, and an increased risk of mortality (8–12). In addition, mild cognitive impairment is increasingly being recognized as a transitional state between normal aging and dementia (13,14). Potentially modifiable risk factors may play a role in the development of specific neurodegenerative diseases such as AD and Parkinson’s Disease (PD), as well as cognitive impairment. Studies that determine these risk factors would have substantial impact on public health. To date, however, the etiologies of most neurodegenerative conditions remain largely unknown. While with many neurodegenerative diseases like AD and PD some percentage may be accounted for by genetic factors, most studies show that this percentage is small and a large contribution from environmental factors is likely (15,16). Outside of age, however, non-genetic risk factors for neurodegenerative disease have been difficult to identify. For AD, the best epidemiologic evidence exists for low educational attainment and history of depression as risk factors, and for non-steroidal antiinflammatory agents and estrogen-replacement therapy as protective factors, but even these data are only somewhat consistent (17). For PD there is strong epidemiologic evidence for a protective effect of cigarette smoking (18,19) and caffeine intake (20,21), but beyond those factors the evidence is less consistent. Occupational exposures to solvents and toxic chemicals, acute events such as stroke and chronic diseases such as hypertension, pulmonary insufficiency, and cardiac insufficiency, have been found to account for a significant proportion of the variance in cognitive performance among the elderly. The bulk of research into specific neurodegenerative diseases and cognitive function among adults, however, has focused on exposures that are relatively recent and occur in adulthood. The role of early life exposures in cognitive function among the elderly and specific neurodegenerative diseases has received little attention, but may be an untapped modifiable etiologic risk factor for these outcomes.

CRITICAL PERIODS IN BRAIN DEVELOPMENT Within the context of childhood development, developmental “critical periods” exists for specific aspects of normal brain development, which as a result also are periods during which the child is more susceptible to toxic exposures. Thus, exposures during these periods are more likely to influence longterm neurologic outcome than similar exposures occurring later in life. A complete discussion of early childhood development and these critical periods is beyond the scope of this paper (22). However, in order to illustrate how childhood exposures might influence the risk of neurologic outcomes in the adult, some background and examples are useful. Development in the nervous system is marked by tremendous cellular plasticity as the highly intricate and specific connections within and between brain regions that provides the basis for intellectual and behavioral functioning are established. As such, development is a period of unique susceptibility to the toxic actions of environmental contaminants, as toxic insults may disrupt this plasticity to alter connections and subsequent functions toward maladaptive programming. These effects ultimately can impair functioning at the behavioral level, and may do so by disrupting the proper establishment of neuronal structure in the developing brain, or more immediate effects on function, or both. Mechanisms by which such contaminants can adversely affect the proper establishment of neuronal structure and function include the disruption of normal synaptic transmission between neurons and associated plasticity, intracellular

Early Life Environmental Exposures

343

signaling, gene expression, and the disruption of cell proliferation or induction of cell death. Such effects at the cellular level could lead, at the behavioral level, to deficits in a variety of neurologic functions, and because they occur during the formative period of brain development they may persist into adulthood. The inherited disease phenylketonuria provides an example of such a critical period—as well as the potential interplay between genetics and environment in neurodevelopment (23). Subjects with the disease cannot metabolize the amino acid phenylalanine. This causes an increase in toxic phenylketones that destroys neuronal tissue and causes mental retardation. By screening for this disease at birth, neonates with the genetic variant are detected and a low phenylalanine diet is instituted. If the diet is instituted later than the first few months of life, permanent neuronal damage occurs. After the age of six, the dietary restrictions can be loosened and some subjects do not continue the diet. Ingestion of phenylalanine at this later age does not lead to sustained neuronal damage. Thus, this gene-environment based toxic exposure causes permanent irreversible damage if it occurs early in the life phase. Exposures in later life, after the brain’s “critical period,” are often reversible and less toxic. Fetal alcohol syndrome provides another dramatic example. Excessive alcohol consumption during pregnancy can lead to dramatic and permanent effects on the child including microcephaly, distinct facial features, hyperreactivity, and mental retardation (24–26). The most pronounced period of vulnerability to these effects appears to be the third trimester of pregnancy. Another example of the greater susceptibility of the developing brain to toxic insult is lead poisoning, which is associated with adverse neurocognitive outcomes that are age at exposure dependent. The strongest association between adverse cognitive outcomes and lead exposures has been demonstrated with umbilical cord blood lead levels (27–29). Thus, in utero exposure appears to be particularly toxic. Nevertheless, exposures under the age of 3 years have been demonstrated to be associated with a loss of IQ points (27,29). A meta-analysis of this association has been performed and suggests that an increase in blood lead level from 10 to 20 mg/dl is associated with an approximately 3 point drop in the IQ score among school-age children (29). Among adults, as with phenylketonuria, lead appears to be less toxic. Adverse effects on cognitive performance on younger working age adults are difficult to measure with blood lead levels less than 40 mg/dl. Work-place regulations reflect this difference and while lead poisoning in children is generally defined as a blood lead level more than 10 mg/dl, lead poisoning in adults is defined by the Occupational Safety and Healthy Administration at a blood lead level of 40 mg/dl. In addition, longitudinal studies have demonstrated that some of the neurocognitive deficits from lead poisoning in adults are reversible (28). Less reversibility has been demonstrated in children. Animal studies also support the concept of a critical period for the toxic effects of environmental contaminants. Rats exposed to lead during gestation display deficits in maze performance well into adulthood, even when exposures are halted after birth (30). Adult rats exposed to similar doses of lead do not demonstrate deficits in maze performance after exposure is stopped (31). These studies support the concept that developmental lead exposures are more likely to cause permanent deficits in learning than adult exposures. In addition to increased toxicity in young animals there may also be an increased effective internal dose when external exposures are the same. Manganese concentrations in several brain regions have been found to increase significantly and much more dramatically when exposure was during the neonatal period compared with the same exposure in adulthood (32–35). Various other compounds have also been found to have a critical developmental window during which exposure can have dramatic and persistent effects while exposure outside the window are far less toxic, including organophosphates

344

Weisskopf et al.

such as chlorpyrifos, organochlorines such as DDT and its metabolites, pyrethroids such as deltamethrin and bioallethrin, and nicotine (36,37).

ARE EARLY LIFE EXPOSURES RELATED TO ADULT NEUROLOGIC OUTCOMES? The most extensive literature relating to early life exposures and adult neurologic outcome is that for schizophrenia. Much evidence suggests an involvement of in utero nutritional and infectious factors in the subsequent development of schizophrenia. Although the specific causal mechanism remains undetermined, the predominant view of the mechanism linking prenatal insult(s) to subsequent schizophrenia is that the insult(s) cause structural and functional damage to certain brain regions that predisposes the individual to schizophrenia. A review of this literature is beyond the scope of this chapter, but is reviewed elsewhere (5). Intriguingly though, a recent study is the first to suggest the involvement of early life exposure to chemical agents—in this case mid-gestational lead exposure—in adult schizophrenia (38). Little research effort has focused on the involvement of early life exposures on other neurologic outcomes, although different lines of evidence leave open the possibility that such exposures may be relevant.

Alzheimer’s Disease While alternative interpretations exist, the known association between limited education and occupational attainment with AD could support the hypothesis that risk factors for AD may be established early in life (39). This theory is often referred to as the “cognitive reserve” hypothesis (39), and may be best illustrated by the results of a regional cerebral blood flow study. Among AD subjects matched on clinical dementia severity, but not education level, those with higher education had a significantly greater parietotemporal perfusion deficit—a marker of AD pathology (40). This suggests that for a given level of clinical dementia, those with higher education have greater underlying AD pathology, and the interpretation is that their education (a relatively early life event) provided them with a cognitive reserve with which to resist the dementing effects of AD. A competing interpretation would be that persons with low education who develop AD are diagnosed earlier. One of the more striking studies supporting the “cognitive reserve” hypothesis is research among nuns mostly from the Milwaukee area (41). These nuns were required to write autobiographies prior to taking their religious vows in their early twenties, an event that occurred 60 years prior to the study. The biographies were available for analysis and a cohort of survivors underwent neuropsychological testing 60 years later. Blinded assessment of idea density in these written reports was strongly associated with cognitive function in later life, and was also associated with neurofibrillary tangles among the nuns who had died and had autopsies. Among 25 deceased nuns, 10 had neuropathologically confirmed AD. Low idea density was present in 90% of those with AD and in only 13% of those without AD, a difference significant at pZ0.001 (relative riskZ6.8, our calculation). These results support the hypothesis that early life experience resulting in increased idea density (presumably the type of thing for which education or occupation in other studies serve as proxies) provides a “cognitive reserve” against the dementing effects of AD. The environments of each nun would likely be fairly uniform after entering the convent so that direct effects of exposures after the essay was written should not account for differences in

Early Life Environmental Exposures

345

outcome. Importantly, the nuns would be expected to have had similar health care and diagnosis of AD would not occur preferentially in those with low idea density. While the “cognitive reserve” hypothesis suggests that increased cognitive ability in early life can protect against AD, a logical counterpart to this theory is that events in early life that decrease or limit cognitive ability could predispose to AD. While genetic factors may also explain associations between cognition, education, and AD, genetic markers of AD have not been validated, and less than 20% of all cases have an established familial pattern (42). Thus, early life environmental exposures should be an important avenue of research. Along these lines there are intriguing results from cases of childhood lead poisoning. In 3 separate cases of children who died subsequent to severe lead intoxication, post-mortem neuropathologic studies found neurofibrillary tangles, which are characteristic of AD (43,44). There is also evidence from animal studies that early life exposure to toxicants may be a risk factor for AD. In a study of rabbits, exposure to high doses of lead in diet led to the development of neurofibrillary tangles in the brain (45). Basha and colleagues (46) recently found an intriguing association between early life lead exposure and an overexpression of AD related proteins in the elderly animal. Rats were exposed to lead in drinking water between post-natal day 1 and 20 at levels that did not produce any gross morphological or nutritional disturbances. Compared to unexposed rats at 20 months of age (adult for a rat), those exposed to lead neonatally showed significantly increased expression of mRNA for the amyloid precursor protein (APP), snippets of which aggregate to form the beta-amyloid deposits characteristic of Alzheimer brains, and significant increases in APP itself as well as the amyloidogenic beta-amyloid products. These increases were not seen in rats exposed to lead as adults. Thus, it was specifically early-life exposure to lead that produced the increases in mRNA and proteins associated with AD. Parkinson’s Disease Although the etiology of PD remains largely obscure, there is compelling evidence for a role of environmental causes. Some of the most compelling evidence of this comes from the study of twins. Tanner et al. (16) studied 71 monozygotic twin pairs and 90 dizygotic twin pairs that had at least one pair member with well-established PD. Among the 16 pairs with diagnosis at or before age 50 years, the concordance of PD among the 4 monozygotic pairs was 1.0, whereas the concordance of PD among the 12 dizygotic pairs was 0.167. This incongruity supports PD before age 50 years as likely influenced mainly by genetic factors. However, among the 141 pairs with diagnosis above age 50 years, the concordance of PD among the 65 monozygotic pairs was 0.108, which was virtually the same as the concordance of PD among the 76 dizygotic pairs of 0.105. This similarity of concordance rates, which has also been reported in previous studies of twins (47–49), has often been cited to support the hypothesis that PD after age 50 years is likely influenced mainly by environmental factors. A similar finding was also recently reported in a large study of Swedish twins (50). These studies, however, may have further implications. Although the concordance rates between monozygotic and dizygotic twins are similar, another striking aspect of these results is that the concordance rates in twins from these studies may be 2–5 times higher than the percentage of PD among non-twin siblings of Parkinson’s cases (51–53). While siblings, twin or not, likely share many postnatal environmental exposures, only twins share in utero environmental exposures. Thus, these data may implicate in utero exposures in the development of PD. In utero and perinatal insult has been reported to lead to a childhood Parkinsonism syndrome with features similar to classical Parkinsonism, including responsiveness to

346

Weisskopf et al.

L-DOPA (54,55). Childhood poliomyelitis has also been reported to be a significant risk factor for the development of Parkinsonism (56), and it has been speculated that intrauterine influenza might cause PD (57). A study of 172 PD cases and 343 age- and sexmatched controls in England found no association between PD and several early life factors, including birth weight, aspects of socioeconomic status of the family at birth, birth order, and sibship size (58). Cases were more likely, however, to have reported suffering from croup or diphtheria in childhood than controls. These findings and others in humans should serve to spur inquiry into the possibility of prenatal influences on the development of PD (55). Evidence from animal studies may also implicate an early life influence on subsequent development of PD. Exposure to pesticides has long been suspected of increasing the risk for PD as a result of numerous studies suggesting that aspects of rural or agricultural life increase the risk of PD, as well as the structural similarity between some pesticides and 1-methyl-4-phenylpyridinium (MPPC)—the oxidative metabolite of 1-methyl,-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)—which rapidly induces a Parkinson’s-like state (59). Paraquat is one pesticide with structural similarity to MPPC and exposure to this pesticide, as well as to maneb and particularly the combination of the two has been shown to induce damage to the nigrostriatal system and accompanying behavioral effects in mice (60–62). Recently it has also been shown that neonatal exposure to paraquat and maneb in the mouse can also induce nigral dopaminergic cell loss and decreases in striatal dopamine in the adult (63). Furthermore, this early life exposure renders the adult mouse more susceptible to the dopaminergic damaging effects of exposure to the pesticides in the adult. Other research has found that in utero exposure to the bacteriotoxin lipopolysaccharide leads to a reduction in the number of dopamine neurons at birth in rats (64). This reduction appears to be persistent, increases with age in a slow, protracted manner, and is accompanied by increased production of pro-inflammatory cytokines, thus bearing striking similarity to the progressive pattern of cell loss and inflammation seen in human PD. Intriguingly, in humans, a common complication of pregnancy is bacterial vaginosis, which is known to produce increased levels of lipopolysaccharides and pro-inflammatory cytokines in the chorioamniotic environment of the fetus. These findings add further intrigue to the possibility that exposure to prenatal neurotoxins may play a role in the subsequent development of PD. General Neurologic Function There is ample evidence for an association between early life exposure to lead and impaired cognitive function in children (29,65,66). Such data has also been put forth for other environmental contaminants such as methylmercury (67–70) and PCBs (71–73). Whether developmental exposure to environmental contaminants leads to impaired cognition in adulthood is less clear. As discussed above, there is evidence from animal studies showing that developmental exposure to several different compounds can have neurobiological and behavioral effects in adult animals, but there is less evidence addressing this in humans. White et al. reported the results of neurocognitive testing among elderly adults with a history of severe childhood lead poisoning (74). These patients were identified by review of hospital charts of children with lead encephalopathy from the 1930s and 40s. Technological limitations of this era precluded the measurement of blood lead levels, and the diagnosis of lead poisoning was made clinically. Inclusion criteria included the presence of nerve palsy or encephalopathy in conjunction with a history of pica and

Early Life Environmental Exposures

347

ingestion of paint chips prior to hospitalization. These symptoms have not been described at blood lead levels less than 70 mg/dl and generally do not occur until blood lead levels are greater than 100 mg/dl. It can therefore be assumed that these children all had lead levels greater than 70 mg/dl. The current definition of lead poisoning in children is a blood lead level of 10 mg/dl. Nineteen survivors were tracked and examined 48 to 60 years following hospitalization. A group of age, race, neighborhood and sex-matched controls from the community were then recruited. These controls were comparable to cases in education level and pack years of smoking. Study subjects were tested using the Wechsler Adult Intelligence Scale-Revised (WAIS-R), the Wechsler Memory Scale (WMS), verbal fluency, and a test of non-verbal reasoning (Raven Progressive Matrix). The battery of tests performed generated 17 separate test scores. Among the 18 matched pairs, cases performed inferiorly to controls in all but 3 measures. These 3 measures were the information and orientation subtests of the WMS on which most subjects performed at the ceiling level, and the digit span subtest of the WAIS-R. Statistically significant differences were found in logical memory tests and picture completion. Of note, this type of global performance decrement is similar to patterns noted in childhood studies of lead poisoning. Among subjects with adult lead exposures, decrements typically occur primarily in domains of short-term memory and motor function. Thus, the pattern of cognitive deficits suggests a persistence of decrements acquired from childhood lead exposures.

DIFFICULTIES IN STUDYING THE EFFECT OF EARLY LIFE EXPOSURES ON ADULT NEUROLOGIC OUTCOMES If true, the hypothesis of fetal or early life origins of adult neurologic diseases and dysfunction would have important implications for public health and clinical practice. The problem is that the long temporal separation between exposure and outcome poses unique and significant challenges to the establishment of causality. Those studying the fetal origins of cardiovascular disease and schizophrenia, of course, face this problem also, and the limitations have been the source of several lines of criticism directed at these areas of research (6,75,76). These criticisms should serve as useful guides for future research based on similar hypotheses. Although the study of developmental influences on adult outcomes is subject to the same limitations as any epidemiologic study, there are specific limitations that are of greater concern when dealing with long time intervals between exposure and outcome. We address some of these below and discuss ways of minimizing them when considering early life exposures to environmental contaminants. Exposure Measurement One approach to studying an exposure-disease relation across such a long time span is to use previously collected or ecological data for exposures remote in time. Although sometimes necessitated by the nature of such study, this can lead to imprecision in the exposure measurement or the use of some measure that is only a proxy for the exposure of interest. The exposure measure most commonly used in research into the fetal origins of cardiovascular disease is birth weight. Although, the recording of birth weight may be accurate, in the analysis of its relation to cardiovascular disease it has usually been interpreted as a measure of nutritional status during pregnancy, which has been a controversial step (6,75,76). In schizophrenia research, data on whether a mother was in mid-gestation during an influenza epidemic has been used as a surrogate for influenza

348

Weisskopf et al.

exposure of the fetus, but whether a particular mother actually had influenza was not known (6). Unlike birth weight, for which reliable records are generally kept, or infectious diseases, for which there may be records of sporadic epidemics, data on varying exposures to environmental contaminants in the distant past are unlikely to be available outside of large scale ecologic studies. Some isolated opportunities may exist. For example, mercury exposure among the population of the Faroe Islands is almost exclusively the result of consuming pilot whale meat (personal communication, Philippe Grandjean & Pal Weihe). The fishing of pilot whales has a long tradition in the Faroes, occurs at specific times during the years, and excellent records are kept on the numbers killed every year in different parts of the islands, which is a reasonable surrogate for local consumption. Because there is a fair amount of variation in the numbers killed each year one could examine the outcomes of children whose gestational period coincided with different levels of available whale meat. Another option could be to make use of medical records, but they are not without problems. Records from the distant past are undoubtedly in storage in an off-site facility. Lists of subjects with admission and discharge diagnosis would have to be individually reviewed and records ordered from storage at considerable expense. There is the cohort described in the paper by White et al. (74) in which childhood lead poisoning had been documented as discussed above. While it is relatively small, there were other subjects described who were not matched who could be followed-up for adult outcomes. The small size of this cohort, however, precludes any study of such outcomes as PD as there would certainly not be enough cases to have adequate power. Some pilot work on more common or continuous measure outcomes might be possible. Such possible isolated opportunities notwithstanding, definitive studies will require more precise measures of exposure, preferably biomarkers within the individuals. In the study of environmental contaminants, the most precise exposure measures are often biomarkers. Measuring exposure to lead provides a good example of some of the strengths and limitations of working with biomarkers of exposure to environmental contaminants in the study of early exposures and adult outcomes. Very well-established techniques for measuring concentrations of lead in both blood and bone exist, and they can provide researchers with very precise measures of even very low exposures. Furthermore, bone lead, which can be measured accurately and non-invasively using x-ray fluorescence technology (77), provides an excellent marker of cumulative exposure because the halflife of bone lead is on the order of decades (78–80). Many researchers including our group have used these methods to identify associations between cumulative lead exposure and several neurologic outcomes, including cognitive function (81–86), psychiatric symptoms (87), and amyotrophic lateral sclerosis (88). Given this long half-life, bone lead in the elderly likely does contain some lead acquired during childhood. However, because the time frame of interest is a small fraction of the subject’s lifetime and because, until recently, lead was a common environmental adult exposure (largely the result of leaded gasoline), even bone lead levels would likely represent mainly adult exposures. The most promising method is likely to be the use of an existing cohort with banked biological samples. This would only work, however, for studies of exposures to compounds that would not break down over time in the stored sample. Many of the currently used pesticides, for example, would likely degrade depending on how the biological sample was stored. For other compounds this would be feasible. Lead, for example, is an element and thus will not break down: any existing blood samples will contain the same concentration of lead today, regardless of age of the sample. Dental samples may also be a particularly good measure of early childhood lead exposure. Lead is incorporated into the dental matrix, and shed deciduous teeth have been used in multiple

Early Life Environmental Exposures

349

studies of childhood lead poisoning. There is still the continued problem of time, as the number of large cohorts of elderly populations today with banked samples from the early life years of the cohort is unknown and likely to be small. Nonetheless, opportunities do exist to use samples banked during the early life of current elderly cohorts. In fact the recent finding of an association between in utero lead exposure and adult schizophrenia took advantage of just such an opportunity (38). Another opportunity is the National Collaborative Perinatal Project (89), a cohort of nearly 60,000 US women who gave birth in the 1950s and 1960s and provided biological samples that were stored for later use. Finally, looking forward, such opportunities can be created by considering a life-course approach when establishing new cohorts. Our group has been studying lead exposure in a cohort of approximately 1200 mother-infant pairs in Mexico (90–93). These children can be followed up for adult neurologic outcomes, although the same caveat regarding sample size and power as discussed above would apply. The national children’s study, a US based multi-agency initiative, is also taking a life course approach and is planning to follow 100,000 children from birth (http://www. nationalchildrensstudy.gov). Although the initial plans are to follow these children from birth until age 21, with some foresight, extending this follow-up should be a possibility. Selection Bias/Loss to Follow-Up In the context of studying early life exposures, a particularly difficult complication that can introduce bias is loss to follow-up of some initial cohort whether studied prospectively or retrospectively. In the Hertfordshire, England, cohort used to study the fetal origins of cardiovascular disease, of the original 15,664 eligible subjects, only 5,700 births could be traced and even fewer were available for physiologic measurements (2). With large amounts of loss to follow-up, the validity of the results become increasingly dependent on the loss being uninformative, that is, that the exposure-outcome relation is the same in those followed and those lost to follow-up. The strengths and limitations with respect to selection bias and loss to follow-up of any study of early life exposure to environmental contaminants and adult neurologic outcome are no different than those for any other study requiring a very long period of follow-up. The best approach is obviously to exert as much effort as possible to following the entire cohort, and attempting to track down the outcomes of those lost to follow-up. This, however, can quickly become expensive when loss to follow-up is large. Taking advantage of registries such as those in the Scandinavian countries or at least nationally collected mortality data when appropriate for the outcome of interest is another approach. When loss to follow-up is inevitable, there will always be the concern that the censoring is dependent on past exposures, in which case G-estimation methods, structural nested models, and marginal structural models are examples of analysis techniques that can be used to address the effects of loss to follow-up in a study when that loss may be dependent on prior exposures (94,95). This may be particularly relevant when exposure is associated with early mortality (lead poisoning may predispose to hypertension and early cardiovascular disease for example). Confounding In the study of the effect of early life exposures on adult outcomes, the issue of confounding has interesting implications. A confounder should not be an intermediate on the path from exposure to outcome nor an effect of the outcome (96,97). Therefore, in theory, adjustment need not be made for anything occurring after early childhood. If such

350

Weisskopf et al.

adjustment is made for a factor that is actually affected by the early life exposure, the resulting effect measure can be dramatically altered (98). This has been a criticism of the fetal origins of cardiovascular disease studies, which have often controlled for adult weight in assessing the association between birth weight and cardiovascular outcome (6,75,76). The list of potential confounders when considering the effect of early life or in utero exposures is somewhat limited as they should really be restricted to covariates that precede the exposure of interest. Some confounders that take on added importance in such studies are genetic influences and socioeconomic factors, as their influence on the parental generation or early life of the individual can impact the further development of the individual. In considering early life exposures to environmental contaminants, confounding by genetic influences may be less of a concern as one would need to postulate a mechanism by which a particular gene influences one’s exposure to the contaminant. This does not, of course, rule out the possibility of gene-environment interactions with early life exposures. Indeed, the example of phenylketonuria described above is an excellent example of such a situation. Socioeconomic factors, however, are likely to be more important, and, unfortunately, are often the hardest to accurately assess. Socioeconomic factors can be related to the likelihood of exposure to different contaminants (99–101) and may affect neurologic outcome, particularly aspects of cognitive function (102–107). There are reasons for adjusting for post-exposure variables in models, particularly when studying early life exposure and adult outcome. The effect of an early life exposure on an adult outcome may be hard to detect if this effect accounts for only a small portion of the overall variance in the outcome. There may be many other factors that occur later in life that also impact the outcome. Adjustment for such factors will reduce the overall variance and increase the precision of the estimate of effect of the early life exposure— provided they are independent of the main exposure of interest. If the later life factors for which adjustment is made are affected by the early life exposure, not only could some of the true effect of the early life exposure be adjusted away, but bias could also be introduced. Because independence from the early life exposure will often be hard to know with certainty, one should consider such adjustments carefully, and examine the results with and without the adjustment. If adjustment is made for a post-exposure factor and it is truly independent of the early life exposure, then the effect estimate of the early life exposure should not be altered beyond random variation from the unadjusted estimate, although the precision may improve. Alternatively, one could guard against the possibility that the factor being controlled is also affected by the early life exposure by using analysis methods that can account for that dependence such as G-estimation, structural nested models, marginal structural models, or structural equation models (94,95). It may be of interest to examine variables on the causal path in an effort to determine by which subsequent effects early life exposure has its overall effect. One statistical method that can be used to attempt to dissect out the contribution from different mechanisms in a single analysis is structural equation modeling (SEM), developed for use in economics and introduced in the environmental health literature in 1985 (99). SEM can partition the variance of interrelated outcomes between those caused by direct influence and those related through indirect pathways (i.e., mediated by at least one other variable in the model) and allow examination of relations between variables that are in the same causal pathways. For example, SEMs could be used to test the hypothesis that some contribution of birth weight to adult cardiovascular disease is mediated through an effect on adult birth weight. Additional potential intermediates can be considered simultaneously. The coefficients produced by SEM will describe the strength of these relationships and are similar to regression coefficients.

Early Life Environmental Exposures

351

HOW CAN THE EFFECT OF EARLY LIFE EXPOSURES ON ADULT NEUROLOGIC OUTCOMES BEST BE STUDIED? Consideration of the points raised above, and addressing them as best as possible, will be crucial to any study of the early life origins of adult neurologic outcome. Are there other factors that would improve the chances of identifying effects in such a study? Obviously the primary obstacle to meaningful results is the long time span between exposure and outcome, particularly when considering specific diseases such as PD, AD, or amyotrophic lateral sclerosis, which have relatively late age onset. If there existed an early subclinical marker of these diseases, then the study feasibility would be greatly improved. Several candidate markers exist, such as those provided by positron emission scanning, magnetic resonance imaging (MRI), and magnetic resonance spectroscopy (MRS). An exciting possibility with respect to AD is that beta-amyloid depositions in the eye—which lead to a particular type of cataract termed supranuclear—may be an early, and easily accessible, indicator of the disease as these deposits have been found in Alzheimer’s patients but not controls (108). The use of an early subclinical marker of disease could shorten the time lag between exposure and outcome by providing an intermediate marker that could be separately linked with exposure and disease. The potential to use MRS in studies of early lead exposure and AD provides an example of the utility of an early biomarker. The etiology of AD is a loss of neuronal function. Neuropathologic studies have demonstrated the presence of excess neurofibrillary tangles and decreased neuronal density in the brains of subjects with AD (42). The ability to detect neuronal density noninvasively in discrete anatomical regions in the human brain has recently been developed. MRS is a newly established technique using nonionizing radiation to identify compounds and metabolites in human tissue (109,110). The technique is similar to MRI, a nonionizing form of imaging routinely used in a variety of disease states. MRS is procedurally no different from MRI, but it employs computer software to translate the MRI signals into spectral data that provide an indication of neuronal density. Specifically, peaks for N-acetylaspartate (NAA), creatine, and choline appear after dampening of the water signal. NAA, abundant in neurons and proposed as a chemical marker of neuronal tissue (109,110), is an intraneuronal chemical whose function is not completely understood. NAA is not present in glial cells, however, and animal studies have demonstrated that its concentration declines following neuronal death (111). Choline is predominantly from components of membrane phospholipids. Changes in concentrations of choline appear to correlate with demyelinating diseases (109,110,112). Creatine is homogeneously distributed in the brain and appears to be relatively resistant to change. Because of these properties, creatine is used as an internal standard to compensate for variation in signal intensity from the scanner (109). Thus, NAA and choline concentration are typically expressed as a ratio of NAA or choline concentration to creatine concentration. MRS is currently being used clinically in the assessment of neurologic disease (109) including AD. Several studies have demonstrated that subjects with AD have lower NAA/ creatine ratios than control subjects (113,114). Furthermore, correlation between neurocognitive testing and NAA/creatine ratios has been demonstrated using the MiniMental State Examination, albeit in a sample of only 22 subjects (113). NAA/creatine appears to be a specific marker of AD in that choline/creatine ratios do not appear to correlate with cognitive testing and have not been demonstrated to differ between subjects with AD and controls. MRS has been shown to differentiate between subjects with AD and subjects with vascular dementia, and, because the testing requires obtaining an MRI, the presence of old strokes can be diagnosed as well (109,110). Our group has also used MRS to examine adult monozygotic twin brothers with a long history of occupational lead

352

Weisskopf et al.

exposure from painting (115). One brother, however, primarily did the sanding while the other primarily did the painting leading to markedly different lead levels between the two brothers, although they were both elevated in comparison to levels in the general population. We found lower NAA/creatine ratios and generally lower neurocognitive test scores in the twin with higher lead exposure. If childhood lead poisoning is associated with AD in later life, then one might expect to find decreased neuronal density with MRS after childhood lead poisoning. This question has not been widely studied, but a couple of reports do exist (116,117). A boy with a history of lead poisoning at the age of 3 years was compared with his nonlead-poisoned male cousin. MRS demonstrated a 35% decrease in the NAA/creatine ratio of the gray matter (region of neuron cell bodies) in the lead-poisoned child at age 10 compared to his cousin at age 9 (Table 1). Choline/creatine ratios were similar. Further, this patient demonstrated impaired performance on neuropsychiatric tests such as the Wechsler Intelligence Scale for Children, while the control subject performed well. This is a unique case because these children were matched on age, sex, and socioeconomic status. In addition, they grew up in the same household following the lead poisoning, and thus their environments were similar. A subsequent study of 16 children with elevated blood lead levels (blood lead levels from 23 to 65 mg/dL) and 5 children whose measured blood lead levels had never been above 10 mg/dL found that the children with elevated blood lead levels had statistically significantly lower NAA/Cr ratios in frontal gray matter. NAA/ Cr ratios were also lower for these children in frontal white matter, but this did not reach statistical significance (116). These reports, while preliminary, demonstrate that neuronal density changes as measured by MRS may be related to both elevated lead exposure and AD. If these changes, perhaps specific to particular brain regions, can be shown to be an early biomarker for AD, then implicating early life lead exposure in producing such changes—a study with a shorter time interval between exposure and outcome—could strengthen the hypothesis that childhood lead poisoning is a risk factor for AD. As with the study of diseases, a biomarker of adverse neural effects that occur prior to cognitive function being impaired to the point where that impairment is detectable with standard neuropsychological tests would be beneficial. The study of early life exposures and general cognitive function involves some different considerations from the study of particular neurologic diseases, some of which offer certain advantages. In particular, unlike a diagnosed disease, general cognitive function can be measured on a continuous, or at least ordinal, scale, mild cognitive impairments should be more prevalent than cases of specific neurologic diseases, and cognitive function can be measured throughout life. The first two of these aspects would likely improve statistical power. The ability to measure cognitive function at any age allows for shorter time intervals between exposure and outcome than those imposed by the study of a particular adult neurologic disease.

Table 1

Neurochemical Ratios in Gray Matter of a Lead-Exposed and Non-Lead-Exposed Child

Brain metabolite ratio NAA/CR Cho/CR a

Lead-exposed mean ratioa (STD)

Non-lead-exposed mean ratio (STD)

0.83 (0.16) 0.45 (0.04)

1.29 (0.08) 0.44 (0.04)

Numbers represent average spectra. Abbreviations: NAA/CR, N-acetylaspartate/creatine; Cho/CR, choline/creatine. Source: From Ref. 117.

Early Life Environmental Exposures

353

In addition, cognitive function can be measured repeatedly so that changes and patterns over time can be assessed within an individual. Studying cognitive function has its limitations as well. When measuring change in performance over time there is often a learning effect (118–120), such that improvements from early tests are expected simply on the basis of the subject having taken the test before, rather than an intrinsic improvement in cognitive function. Having multiple tests per person can help minimize this problem. Other problems with many tests of cognitive function include issues related to ceilings or floors of the cognitive test (121). Hypothetically, if a test has a floor of zero and there is an effect of exposure that is relatively monotonic with higher exposures leading to lower scores, then subjects with the highest exposures may all score a zero. Conceptually this may be considered a form of measurement error, as there may be variation in cognitive functioning among subjects who score at the floor (or ceiling) of the test that is missed by the restrictions of the test. The problem is that this measurement error is differential (in the example above affecting only those with high exposures), which introduces bias (122,123). Nevertheless, the measurement of cognitive function has been extensively studied and many sensitive tests of cognitive function are in wide use (124). Validated tests used in the diagnosis of different neurologic diseases also exist, such as the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) battery used to measure cognitive function in AD (125). Studies of both specific neurologic diseases and cognitive impairment could be nested within a single cohort, and this design may be the most efficacious.

SUMMARY Available evidence suggests that early-life, even in utero, environmental exposures can influence adult neurologic outcomes. While difficult and likely expensive, such association can be studied. The optimal study design would likely be a retrospective cohort study with banked biological samples and longitudinal follow-up of subjects or a prospective cohort study. The identification and use of biomarkers of early, subclinical disease—MRS being one possibility—may allow for more precise detection of neurologic outcomes and allow for a shortened time interval between exposure and outcome thereby reducing some of the impediments to identifying associations. A full consideration of the causal pathways of these exposures should be considered in the data analysis. Care must be taken not to control for an intermediate variable, and appropriate statistical techniques employed when attempting to account for loss to follow-up or determine the contribution of different effects of the early life exposure on the outcome. Although the epidemiologic terrain can be hazardous, the potential public health benefits from the study of early life environmental exposures and adult neurologic outcomes could be tremendous, and with the graying of many societies, the benefits are likely to only increase.

REFERENCES 1. Kuh D, Smith GD. When is mortality risk determined? Historical insights into a current debate Soc Hist Med 1993; 6:101–123. 2. Barker DJ, Winter PD, Osmond C, Margetts B, Simmonds SJ. Weight in infancy and death from ischaemic heart disease. Lancet 1989; 2:577–580. 3. Barker DJ, Osmond C, Golding J, Kuh D, Wadsworth ME. Growth in utero, blood pressure in childhood and adult life, and mortality from cardiovascular disease. BMJ 1989; 298:564–567.

354

Weisskopf et al.

4. Barker DJ, Osmond C. Infant mortality, childhood nutrition, and ischaemic heart disease in England and Wales. Lancet 1986; 1:1077–1081. 5. Susser ES, Brown AS, Gorman JM. Prenatal exposures in schizophrenia. In: Spiegel D, ed. Progress in Psychiatry. Washington, D.C: American Psychiatric Press, 1999. 6. Susser EB, Brown A, Matte TD. Prenatal factors and adult mental and physical health. Can J Psychiatry 1999; 44:326–334. 7. Day JC. Population Projections of the United States by Age, Sex and Hispanic Origin: 1995 to 2050. US Bureau of the Census, Current Population Reports, P25–1130, U.S. Government Printing Office, Washington, DC, 1996. 8. Brookmeyer R, Gray S, Kawas C. Projections of Alzheimer’s disease in the United States and the public health impact of delaying disease onset. Am J Public Health 1998; 88:1337–1342. 9. Bassuk SS, Wypij D, Berkman LF. Cognitive impairment and mortality in the communitydwelling elderly. Am J Epidemiol 2000; 151:676–688. 10. Bedard M, Pedlar D, Martin NJ, Malott O, Stones MJ. Burden in caregivers of cognitively impaired older adults living in the community: methodological issues and determinants. Int Psychogeriatr 2000; 12:307–332. 11. Hebert R, Brayne C, Spiegelhalter D. Factors associated with functional decline and improvement in a very elderly community-dwelling population. Am J Epidemiol 1999; 150:501–510. 12. Park DC, Nisbett R, Hedden T. Aging, culture, and cognition. J Gerontol B Psychol Sci Soc Sci 1999; 54:75–84. 13. Rowe JW, Kahn RL. Successful aging. Gerontologist 1997; 37:433–440. 14. Tabert MH, Albert SM, Borukhova-Milov L, et al. Functional deficits in patients with mild cognitive impairment: prediction of AD. Neurol 2002; 58:758–764. 15. Bergem AL, Engedal K, Kringlen E. The role of heredity in late-onset Alzheimer disease and vascular dementia. A twin study. Arch Gen Psychiatry 1997; 54:264–270. 16. Tanner CM, Ottman R, Goldman SM, et al. Parkinson disease in twins: an etiologic study. JAMA 1999; 281:341–346. 17. Hendrie HC. Epidemiology of dementia and Alzheimer’s disease. Am J Geriatr Psychiatry 1998; 6:S3–S18. 18. Grandinetti A, Morens DM, Reed D, MacEachern D. Prospective study of cigarette smoking and the risk of developing idiopathic Parkinson’s disease. Am J Epidemiol 1994; 139:1129–1138. 19. Hernan MA, Zhang SM, Rueda-deCastro AM, Colditz GA, Speizer FE, Ascherio A. Cigarette smoking and the incidence of Parkinson’s disease in two prospective studies. Ann Neurol 2001; 50:780–786. 20. Ross GW, Abbott RD, Petrovitch H, et al. Association of coffee and caffeine intake with the risk of Parkinson disease. JAMA 2000; 283:2674–2679. 21. Ascherio A, Zhang SM, Hernan MA, et al. Prospective study of caffeine consumption and risk of Parkinson’s disease in men and women. Ann Neurol 2001; 50:56–63. 22. Kandel ER, Schwartz JH, Jessell TM. Principles of Neuroscience. New York: McGraw-Hill, 2000. 23. Luzatto L, Mehta A. In: Stanbury J, Wyngaarden J, Fredrickson D, eds. Phenylketonuria in Metabolic Basis of Inherited Disease. New York: McGraw, 1993. 24. Danis RP, Newton N, Keith L. Pregnancy and alcohol. Curr Probl Obstet Gynecol 1981; 4:2–48. 25. Lewis DD, Woods SE. Fetal alcohol syndrome. Am Fam Physician 1994; 50:1025–1032. 1035–1036. 26. Olney JW, Wozniak DF, Farber NB, Jevtovic-Todorovic V, Bittigau P, Ikonomidou C. The enigma of fetal alcohol neurotoxicity. Ann Med 2002; 34:109–119. 27. Dietrich KN, Succop PA, Bornschein RL, et al. Lead exposure and neurobehavioral development in later infancy. Environ Health Perspect 1990; 89:13–19. 28. ATSDR. The nature and extent of lead poisoning in children in the United States: a report to congress. Atlanta, GA: US Department of Health and Human Services, 1988.

Early Life Environmental Exposures

355

29. Schwartz J. Low-level lead exposure and children’s IQ: a meta-analysis and search for a threshold. Environ Res 1994; 65:42–55. 30. Kuhlmann AC, McGlothan JL, Guilarte TR. Developmental lead exposure causes spatial learning deficits in adult rats. Neurosci Lett 1997; 233:101–104. 31. Jett DA, Kuhlmann AC, Farmer SJ, Guilarte TR. Age-dependent effects of developmental lead exposure on performance in the Morris water maze. Pharmacol Biochem Behav 1997; 57:271–279. 32. Dorman DC, Struve MF, Vitarella D, Byerly FL, Goetz J, Miller R. Neurotoxicity of manganese chloride in neonatal and adult CD rats following subchronic (21-day) high-dose oral exposure. J Appl Toxicol 2000; 20:179–187. 33. Takeda A, Ishiwatari S, Okada S. Manganese uptake into rat brain during development and aging. J Neurosci Res 1999; 56:93–98. 34. Vojtisek M, Knotkova J, Svandova E, Svihalkova J, Paduanova Z. Maganese retention in rat brain. Rev Environ Health 1999; 14:251–256. 35. Webster WS, Valois AA. Reproductive toxicology of manganese in rodents, including exposure during the postnatal period. Neurotoxicology 1987; 8:437–444. 36. Eriksson P. Developmental neurotoxicity of environmental agents in the neonate. Neurotoxicology 1997; 18:719–726. 37. Qiao D, Seidler FJ, Padilla S, Slotkin TA. Developmental neurotoxicity of chlorpyrifos: what is the vulnerable period? Environ Health Perspect 2002; 110:1097–1103. 38. Opler MGA, Brown AS, Graziano J, et al. Prenatal lead exposure, delta-aminolevulinic acid, and schizophrenia. Environ Health Perspect 2004; 112:548–552. 39. Stern Y, Gurland B, Tatemichi TK, Tang MX, Wilder D, Mayeux R. Influence of education and occupation on the incidence of Alzheimer’s disease. JAMA 1994; 271:1004–1010. 40. Stern Y, Alexander GE, Prohovnik I, Mayeux R. Inverse relationship between education and parietotemporal perfusion deficit in Alzheimer’s disease. Ann Neurol 1992; 32:371–375. 41. Snowdon DA, Kemper SJ, Mortimer JA, Greiner LH, Wekstein DR, Markesbery WR. Linguistic ability in early life and cognitive function and Alzheimer’s disease in late life. Findings from the Nun Study. JAMA 1996; 275:528–532. 42. Bird TD, Miller BL. Alzheimer’s disease and other dementias. In: Kasper DL, Braunwald E, Fauci AS, Hauser SL, Longo DL, Jameson JL, eds. Harrison’s Principles of Internal Medicine. New York: McGraw-Hill Inc, 2005:2393–2406. 43. Niklowitz WJ, Mandybur TI. Neurofibrillary changes following childhood lead encephalopathy. J Neuropathol Exp Neurol 1975; 34:445–455. 44. Wisniewski K, Jervis GA, Moretz RC, Wisniewski HM. Alzheimer neurofibrillary tangles in diseases other than senile and presenile dementia. Ann Neurol 1979; 5:288–294. 45. Niklowitz WJ. Neurofibrillary changes after acute experimental lead poisoning. Neurology 1975; 25:927–934. 46. Basha MR, Wei W, Bakheet SA, et al. The fetal basis of amyloidogenesis: exposure to lead and latent overexpression of amyloid precursor protein and beta-amyloid in the aging brain. J Neurosci 2005; 25:823–829. 47. Eldridge R, Ince SE. The low concordance rate for Parkinson’s disease in twins: a possible explanation. Neurology 1984; 34:1354–1356. 48. Marsden CD. Parkinson’s disease in twins. J Neurol Neurosurg Psychiatry 1987; 50:105–106. 49. Ward CD, Duvoisin RC, Ince SE, Nutt JD, Eldridge R, Calne DB. Parkinson’s disease in 65 pairs of twins and in a set of quadruplets. Neurology 1983; 33:815–824. 50. Wirdefeldt K, Gatz M, Pawitan Y, Pedersen NL. Risk and protective factors for Parkinson’s disease: a study in Swedish twins. Ann Neurol 2005; 57:27–33. 51. Maher NE, Golbe LI, Lazzarini AM, et al. Epidemiologic study of 203 sibling pairs with Parkinson’s disease: the GenePD study. Neurology 2002; 58:79–84. 52. Marder K, Levy G, Louis ED, et al. Familial aggregation of early- and late-onset Parkinson’s disease. Ann Neurol 2003; 54:507–513. 53. Payami H, Zareparsi S, James D, Nutt J. Familial aggregation of Parkinson disease: a comparative study of early-onset and late-onset disease. Arch Neurol 2002; 59:848–850.

356

Weisskopf et al.

54. Widhalm S. Current status of conservative therapy of extrapyramidal movement disorders in childhood and adolescence. Wien Med Wochenschr Suppl 1985; 88:1–23. 55. Riederer P, Foley P. Mini-review: multiple developmental forms of parkinsonism. The basis for further research as to the pathogenesis of parkinsonism. J Neural Transm 2002; 109:1469–1475. 56. Nielsen NM, Rostgaard K, Hjalgrim H, Aaby P, Askgaard D. Poliomyelitis and Parkinson disease. JAMA 2002; 287:1650–1651. 57. Mattock C, Marmot M, Stern G. Could Parkinson’s disease follow intra-uterine influenza? A speculative hypothesis J Neurol Neurosurg Psychiatry 1988; 51:753–756. 58. Martyn CN, Osmond C. Parkinson’s disease and the environment in early life. J Neurol Sci 1995; 132:201–206. 59. Checkoway H, Nelson LM. Epidemiologic approaches to the study of Parkinson’s disease etiology. Epidemiology 1999; 10:327–336. 60. Thiruchelvam M, Richfield EK, Baggs RB, Tank AW, Cory-Slechta DA. The nigrostriatal dopaminergic system as a preferential target of repeated exposures to combined paraquat and maneb: implications for Parkinson’s disease. J Neurosci 2000; 20:9207–9214. 61. Brooks AI, Chadwick CA, Gelbard HA, Cory-Slechta DA, Federoff HJ. Paraquat elicited neurobehavioral syndrome caused by dopaminergic neuron loss. Brain Res 1999; 823:1–10. 62. McCormack AL, Thiruchelvam M, Manning-Bog AB, et al. Environmental risk factors and Parkinson’s disease: selective degeneration of nigral dopaminergic neurons caused by the herbicide paraquat. Neurobiol Dis 2002; 10:119–127. 63. Thiruchelvam M, Richfield EK, Goodman BM, Baggs RB, Cory-Slechta DA. Developmental exposure to the pesticides paraquat and maneb and the Parkinson’s disease phenotype. Neurotoxicology 2002; 23:621–633. 64. Carvey PM, Chang Q, Lipton JW, Ling Z. Prenatal exposure to the bacteriotoxin lipopolysaccharide leads to long-term losses of dopamine neurons in offspring: a potential, new model of Parkinson’s disease. Front Biosci 2003; 8:s826–s837. 65. Banks EC, Ferretti LE, Shucard DW. Effects of low level lead exposure on cognitive function in children: a review of behavioral, neuropsychological and biological evidence. Neurotoxicology 1997; 18:237–281. 66. Needleman HL, Gatsonis CA. Low-level lead exposure and the IQ of children. A metaanalysis of modern studies. JAMA 1990; 263:673–678. 67. Budtz-Jorgensen E, Keiding N, Grandjean P, Weihe P. Estimation of health effects of prenatal methylmercury exposure using structural equation models. Environ Health 2002; 1:2. 68. Weihe P, Hansen JC, Murata K, et al. Neurobehavioral performance of Inuit children with increased prenatal exposure to methylmercury. Int J Circumpolar Health 2002; 61:41–49. 69. Grandjean P, Weihe P, White RF, Debes F. Cognitive performance of children prenatally exposed to “safe” levels of methylmercury. Environ Res 1998; 77:165–172. 70. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-years-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 71. Grandjean P, Weihe P, Burse VW, et al. Neurobehavioral deficits associated with PCB in 7-years-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001; 23:305–317. 72. Jacobson SW, Fein GG, Jacobson JL, Schwartz PM, Dowler JK. The effect of intrauterine PCB exposure on visual recognition memory. Child Dev 1985; 56:853–860. 73. Jacobson JL, Jacobson SW, Humphrey HE. Effects of in utero exposure to polychlorinated biphenyls and related contaminants on cognitive functioning in young children. J Pediatr 1990; 116:38–45. 74. White RF, Diamond R, Proctor S, Morey C, Hu H. Residual cognitive deficits 50 years after lead poisoning during childhood. Br J Ind Med 1993; 50:613–622. 75. Paneth N, Ahmed F, Stein AD. Early nutritional origins of hypertension: a hypothesis still lacking support. J Hypertens Suppl 1996; 14:S121–S129. 76. Joseph KS, Kramer MS. Review of the evidence on fetal and early childhood antecedents of adult chronic disease. Epidemiol Rev 1996; 18:158–174.

Early Life Environmental Exposures

357

77. Aro ACA, Todd AC, Amarasiriwardena C, Hu H. Improvements in the calibration of 109Cd K x-ray fluorescence systems for measuring bone lead in vivo. Phys Med Biol 1994; 39:2263–2271. 78. Armstrong R, Chettle DR, Scott MC, Somervaille LJ, Pendlington M. Repeated measurements of tibia lead concentrations by in vivo x ray fluorescence in occupational exposure. Br J Ind Med 1992; 49:14–16. 79. Gerhardsson L, Attewell R, Chettle DR, et al. In vivo measurements of lead in bone in longterm exposed lead smelter workers. Arch Environ Health 1993; 48:147–156. 80. Hu H, Milder FL, Burger DE. X-ray fluorescence: issues surrounding the application of a new tool for measuring burden of lead. Environ Res 1989; 49:295–317. 81. Balbus-Kornfeld JM, Stewart W, Bolla KI, Schwartz BS. Cumulative exposure to inorganic lead and neurobehavioural test performance in adults: an epidemiological review. Occup Environ Med 1995; 52:2–12. 82. Muldoon SB, Cauley JA, Kuller LH, et al. Effects of blood lead levels on cognitive function of older women. Neuroepidemiology 1996; 15:62–72. 83. Payton M, Riggs KM, Spiro A, III, Weiss ST, Hu H. Relations of bone and blood lead to cognitive function: the VA Normative Aging Study. Neurotoxicol Teratol 1998; 20:19–27. 84. Schwartz BS, Stewart WF, Bolla KI, et al. Past adult lead exposure is associated with longitudinal decline in cognitive function. Neurology 2000; 55:1144–1150. 85. Weisskopf MG, Wright RO, Schwartzm J, et al. Cumulative lead exposure and prospective change in cognition among elderly men: the VA Normative Aging Study. Am J Epidemiol 2004; 160:1184–1193. 86. Wright RO, Tsaih SW, Schwartz J, et al. Lead exposure biomarkers and mini-mental status exam scores in older men. Epidemiology 2003; 14:713–718. 87. Rhodes D, Spiro A, 3rd, Aro A, Hu H. Relationship of bone and blood lead levels to psychiatric symptoms: the normative aging study. J Occup Environ Med 2003; 45:1144–1151. 88. Kamel F, Umbach DM, Munsat TL, Shefner JM, Hu H, Sandler DP. Lead exposure and amyotrophic lateral sclerosis. Epidemiology 2002; 13:311–319. 89. DHEW. The Collaborative Perinatal Study of the National Institute of Neurological Diseases and Stroke: The Women and Their Pregnancies: U.S Department of Health Education and Welfare, Public Health Service, National Institutes of Health, 1972. 90. Ettinger AS, Tellez-Rojo MM, Amarasiriwardena C, et al. Levels of lead in breast milk and their relation to maternal blood and bone lead levels at one month postpartum. Environ Health Perspect 2004; 112:926–931. 91. Hernandez-Avila M, Gonzalez-Cossio T, Hernandez-Avila JE, et al. Dietary calcium supplements to lower blood lead levels in lactating women: a randomized placebo-controlled trial. Epidemiology 2003; 14:206–212. 92. Gomaa A, Hu H, Bellinger D, et al. Maternal bone lead as an independent risk factor for fetal neurotoxicity: a prospective study. Pediatrics 2002; 110:110–118. 93. Hernandez-Avila M, Gonzalez-Cossio T, Palazuelos E, et al. Dietary and environmental determinants of blood and bone lead levels in lactating postpartum women living in Mexico City. Environ Health Perspect 1996; 104:1076–1082. 94. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology 2000; 11:550–560. 95. Witteman JC, D’Agostino RB, Stijnen T, et al. G-estimation of causal effects: isolated systolic hypertension and cardiovascular death in the Framingham Heart Study. Am J Epidemiol 1998; 148:390–401. 96. Robins JM, Greenland S. Identifiability and exchangeability for direct and indirect effects. Epidemiology 1992; 3:143–155. 97. Miettinen O. Confounding and effect-modification. Am J Epidemiol 1974; 100:350–353. 98. Hernan MA, Hernandez-Diaz S, Werler MM, Mitchell AA. Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. Am J Epidemiol 2002; 155:176–184.

358

Weisskopf et al.

99. Bornschein RL, Succop P, Dietrich KN, Clark CS, Que Hee S, Hammond PB. The influence of social and environmental factors on dust lead, hand lead, and blood lead levels in young children. Environ Res 1985; 38:108–118. 100. Elreedy S, Krieger N, Ryan PB, Sparrow D, Weiss ST, Hu H. Relations between individual and neighborhood-based measures of socioeconomic position and bone lead concentrations among community- exposed men: the Normative Aging Study. Am J Epidemiol 1999; 150:129–141. 101. Hu H, Payton M, Korrick S, et al. Determinants of bone and blood lead levels among community-exposed middle-aged to elderly men. The normative aging study. Am J Epidemiol 1996; 144:749–759. 102. Farmer ME, Kittner SJ, Rae DS, Bartko JJ, Regier DA. Education and change in cognitive function. The Epidemiologic Catchment Area Study. Ann Epidemiol 1995; 5:1–7. 103. Gallacher JE, Elwood PC, Hopkinson C, et al. Cognitive function in the Caerphilly study: associations with age social class, education and mood. Europ J Epidemiol 1999; 15:161–169. 104. Lee S, Kawachi I, Berkman LF, Grodstein F. Education, other socioeconomic indicators, and cognitive function. Am J Epidemiol 2003; 157:712–720. 105. Li CY, Wu SC, Sung FC. Lifetime principal occupation and risk of cognitive impairment among the elderly. Indust Health 2002; 40:7–13. 106. Lynch J, Kaplan G, Shema SJ. Cumulative impact of sustained economic hardship on physical, cognitive, psychological, and social functioning. N Engl J Med 1997; 337:1889–1895. 107. Turrell G, Lynch JW, Kaplan GA, et al. Socioeconomic position across the lifecourse and cognitive function in late middle age. J Gerontol B Psychol Sciences Soc Sci 2002; 57:S43–S51. 108. Goldstein LE, Muffat JA, Cherny RA, et al. Cytosolic beta-amyloid deposition and supranuclear cataracts in lenses from people with Alzheimer’s disease. Lancet 2003; 361:1258–1265. 109. Arnold DL, De Stefano N. Magnetic resonance spectroscopy in vivo: applications in neurological disorders. Ital J Neurol Sci 1997; 18:321–329. 110. van der Knaap MS, van der Grond J, Luyten PR, den Hollander JA, Nauta JJ, Valk J. 1H and 31P magnetic resonance spectroscopy of the brain in degenerative cerebral disorders. Ann Neurol 1992; 31:202–211. 111. Koller KJ, Zaczek R, Coyle JT. N-acetyl-aspartyl-glutamate: regional levels in rat brain and the effects of brain lesions as determined by a new HPLC method. J Neurochem 1984; 43:1136–1142. 112. Korenke GC, Pouwels PJ, Frahm J, et al. Arrested cerebral adrenoleukodystrophy: a clinical and proton magnetic resonance spectroscopy study in three patients. Pediatr Neurol 1996; 15:103–107. 113. Heun R, Schlegel S, Graf-Morgenstern M, Tintera J, Gawehn J, Stoeter P. Proton magnetic resonance spectroscopy in dementia of Alzheimer type. Int J Geriatr Psychiatry 1997; 12:349–358. 114. Shiino A, Matsuda M, Morikawa S, Inubushi T, Akiguchi I, Handa J. Proton magnetic resonance spectroscopy with dementia. Surg Neurol 1993; 39:143–147. 115. Weisskopf MG, Hu H, Mulkern RV, et al. Cognitive deficits and magnetic resonance spectroscopy in adult monozygotic twins with lead poisoning. Environ Health Perspect 2004; 112:620–625. 116. Trope I, Lopez-Villegas D, Cecil KM, Lenkinski RE. Exposure to lead appears to selectively alter metabolism of cortical gray matter. Pediatrics 2001; 107:1437–1442. 117. Trope I, Lopez-Villegas D, Lenkinski RE. Magnetic resonance imaging and spectroscopy of regional brain structure in a 10-years-old boy with elevated blood lead levels. Pediatrics 1998; 101:1066–1067. 118. Galasko D, Abramson I, Corey-Bloom J, Thal LJ. Repeated exposure to the Mini-Mental State Examination and the Information-Memory-Concentration Test results in a practice effect in Alzheimer’s disease. Neurology 1993; 43:1559–1563.

Early Life Environmental Exposures

359

119. Jacqmin-Gadda H, Fabrigoule C, Commenges D, Dartigues JF. A 5-years longitudinal study of the Mini-Mental State Examination in normal aging. Am J Epidemiol 1997; 145:498–506. 120. Mantere P, Hanninen H, Hernberg S, Luukkonen R. A prospective follow-up study on psychological effects in workers exposed to low levels of lead. Scand J Work Environ Health 1984; 10:43–50. 121. Morris MC, Evans DA, Hebert LE, Bienias JL. Methodological issues in the study of cognitive decline. Am J Epidemiol 1999; 149:789–793. 122. Weuve J, Glymour M, Berkman LF, Hu H, Spiro A, Robins JM. Identifying predictors of cognitive function and cognitive change when the measurement of cognitive function has a ceiling [Abstract]. Epidemiology 2004; 159:214. 123. Rothman KJ, Greenland S. Precision and validity in epidemiologic studies. In: Rothman KJ, Greenland S, eds. Modern Epidemiology. Philadelphia, PA: Lippincott-Raven Publishers, 1998:115–134. 124. Lezak M. Neuropsychological Assessment. New York: Oxford University Press, 1998. 125. Morris JC, Heyman A, Mohs RC, et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assessment of Alzheimer’s disease. Neurology 1989; 39:1159–1165.

SECTION FOUR: ANALYSIS AND INTERPRETATION

17 The Critical Concept of Control in Human Neurobehavioral Toxicology Studies: We Can, and Must, Do Better Paul Stewart Department of Psychology, State University of New York at Oswego, Oswego, New York, U.S.A.

INTRODUCTION In the late 1980s and early 1990s, a number of observational and epidemiological studies were conducted to examine potential cardioprotective effects of hormone replacement therapy (HRT) in women. In general, several multivariate correlational studies reported a 40% to 50% reduced risk for coronary heart disease associated with HRT (1–4), even after control for covariates such as race, height, weight, parity, smoking, and education (1). However, in the mid-1990s Matthews et al. (5) carefully and meticulously measured demographic and health characteristics in women prior to their initiation in HRT programs. Results indicated that in comparison to non-HRT users, there were multiple important psychosocial and health related variables that conferred a better cardiovascular risk profile in HRT users prior to their use of HRT. Women who went on to get HRT were generally better educated, had higher levels of high density lipoprotein (HDL) cholesterol, more leisure activity, lower blood pressure and lower body weight. A subsequent clinical trial was conducted, employing systematic manipulation of HRT in 16,608 randomly assigned patients (8506 HRT and 8102 placebo) across 40 US clinical centers. On May 31, 2002, after approximately 5 yr of follow-up, a report in the Journal of the American Medical Association indicated the trial was halted because significantly increased risk for coronary heart disease and related illnesses in the HRT group (6). What is the relevance of this story to human neurotoxicology studies? Quite plainly, that a number of highly consistent multivariate observational studies can be so blatantly and thoroughly contradicted by carefully controlled experimentation should give us all great pause. For the very tools and analyses used in these discredited studies are commonly used throughout most, if not all, of the multivariate correlational studies—including those in human neurotoxicology. Thus, asking ourselves “how can we do better” in studies of the effects of environmental contaminants in children is warranted. Many people outside the aforementioned field, both scientists and laypeople alike, are skeptical or uncertain 361

362

Stewart

concerning the results of these studies. As someone “inside” the field, I sometimes share this skepticism. Part of this skepticism lay in the almost reflexive doubt one experiences when bombarded by the media with each latest study showing how eating food X, or being exposed to substance Y, increases the risk for cancer subtype Z (but only in males!). And then only later we find studies that contradict these results, and so forth. It is hard not to be concerned about what is happening in non-experimental sciences when one simply stands back and notes the vast array of data that are generated which are often unparsimonious, atheoretical, highly disclaimed, and often contradictory. Part of this may be the fault of the media, who as Richard Doll noted, often report “half-baked and preliminary findings, without adequate allowance for the vagaries of chance, bias in reporting, and the complexity involved in the way different social and environmental factors are interrelated” (7). But even in apparently well-designed and respected studies, such as the Faroe and Seychelles MeHg studies, we find significant apparent contradictions. The former study reports several adverse associations with MeHg (8,9). The latter reports null and even many beneficial associations (10–14). Thus the problem is not just a perception, or just the media, but it may also lie in a more fundamental problem in nonexperimental studies of the exact kind discussed in this text. Like most if not all of the other authors in this text, I have had the good fortune to review countless manuscripts and grant proposals, and have been exposed to a myriad of different studies, including their different questions, approaches, methods, designs and results. These studies are not always or even usually limited to my own narrow research interests (PCB exposure). Many are studies asking similar questions about different contaminants. While they differ in technical aspects, these studies all share the goal of gathering data that will permit valid inferences to be drawn about the likelihood of a causal link between exposure to a given contaminant and neurobehavioral outcomes. They also share the common link that they are non-experimental. No subjects can ethically be randomly assigned to a PCB exposure or Pb exposure group. The data must be collected as it exists outside the laboratory, with all the strengths of external validity and pitfalls of confounding that go with it. The real question is, given these limitations, can one place confidence in the results and conclusions of these studies? The current answer, in my view, is only “sometimes,” when it should simply be an unqualified “yes.” In this chapter I intend to suggest that there are weaknesses in the field which must be addressed. It is also my intention to argue that not only are these weaknesses addressable, but that they should be addressed sooner rather than later. If human neurotoxicology studies of health effects of low-level environmental exposures are to maintain, let alone gain, scientific credibility, there must evolve a mature state of standards and practices which has not yet been realized. Experimental vs. Non-Experimental Designs Experiments are traditionally defined by (a) random assignment of subjects to conditions and (b) systematic manipulation of an independent variable (15). These characteristics strengthen our ability to infer causality between the independent and dependent variable because it minimizes the chance of confounding. As a trained experimentalist, I once subscribed to the notion that there is a qualitative distinction between experiments and non-experiments in terms of the ability to prove causality. I now believe this is a fundamental myth. The difference is a quantitative one, if defined in terms of the degree of confidence with which one may infer a causal relation. Random assignment is a means to attempt to minimize the chance the independent variable will be correlated with a confounding variable. It is therefore likely that the degree of confounding between

Human Neurobehavioral Toxicology Studies

363

the independent variable and other variables is less than in a non-experimental design. This is a quantitative, but not qualitative, difference from non-experiments. “Systematic manipulation” of the independent variable is not unique to experiments. Assuming determinism, all variables have been “manipulated” by something. In the true experiments this is done by the researcher, in non-experimental studies this is done by uncontrolled variables in the environment. Both cases can involve causal relations, but it is not the case that manipulation of the variable by an experimenter is necessary for causality. Causality exists outside the laboratory. The difference is that in an experiment, causality may be inferred with greater confidence because the design ensures that other factors or alternative explanations have been controlled and/or held constant. Causality is thus inferred either because the correlation between the other factors and outcome has been set close to zero (by minimizing variability of the confounder), or that the correlation between the other factors and the independent variable has been set close to zero (by random assignment of subjects to treatment group, counterbalancing or randomization of extraneous factors). Regardless of whether this is achieved through experimental manipulation or covariate control—it is the result that is important. And that result is the scenario whereby the independent variable predicts variability in the outcome variable that cannot be predicted by other variables. This is the definition of an excellently controlled experiment. But it is also the definition of an excellently controlled non-experiment. Control Is the Fundamental Issue Of course, the fact remains today that most non-experimental studies cannot infer causal relations nearly to the degree that true experiments can. The major problem for non-experimental studies is a disquieting lack of consensus regarding what defines acceptable control. Despite appearances, there is indeed a “control condition” in correlational (non-experimental) studies to which the “treatment” groups or “treated” individuals are compared. In a study which employs covariate control, the comparison is, quite literally, the difference between the actual outcome scores generated by the subjects and their “predicted” scores based upon the covariate, or control, equation model. Each individual subject has a specific predicted score; the score a given subject “should” have based upon his or her unique combination of covariate values that are entered into the multiple regression equation. Thus the predicted score itself represents an estimate of the score the subject should obtain if the null hypothesis is correct. In statistical parlance the difference between the predicted and actual scores can be quantified by recording the “residuals.” Residuals are the positive or negative numbers which reflect the degree to which the score fell above or below, respectively, the predicted score based upon the covariate equation (null hypothesis). It is therefore the mean of the differences between the predicted and observed scores in different exposure groups that is analogous to the mean difference between the scores of the control and experimental groups in a true experiment. Since predicted scores are the fundamental “control” numbers to which observed scores are compared, all results obtained from that study are based upon them. They are as fundamental to any non-experimental study as a control group is to any experimental study. Consequently, the structure of the control variable equation which generates these predicted (control) scores, and the covariates which enter that equation, are perhaps the most fundamentally important consideration in the study. Given the above, it is quite remarkable that there has been no universally accepted method for the construction of covariate models, nor what kinds of covariates should be included/excluded in these models, nor how these covariates should be measured in the first instance. In a very real sense, an analogous situation might be found if

364

Stewart

experimentalists were unable to agree to what defines the most basic elements of a control group. One can only imagine the state of experimental science were this to be the case. Yet this it is not too far from the truth to say that this condition characterizes non-experimental investigation, across multiple fields of study, including those described, and not described, in this text. The quantity and qualitative characteristics of covariates measured across the different studies in human neurotoxicology vary considerably. Weaknesses in Current Methods If the field of human neurotoxicology is to evolve and improve from a standpoint of control, we must identify the areas where the weaknesses are. Only then is improvement possible. I submit there are at least 3 areas in which there are weaknesses. These are (1) a lack of reasonable consensus about the number and types of control variables chosen in these studies, (2) an absence of standards for how control variables are measured, and (3) a lack of reasonable consensus for the statistical treatment of control variables. The Choice and Number of Control Variables Differs Widely Across Studies In the design of any human neurotoxicology study, the investigators are faced with the daunting task of choosing a set of covariates and then choosing the most valid and reliable instruments to measure them. There are a limited number of major covariates that most human neurotoxicology studies typically measure and control for. Some of most common include variables such as parental IQ, socioeconomic status (SES), the home environment (HOME), parental smoking, parental alcohol use, and child sex. Outside of these, there is much less consistency and standardization of control variables across studies in the literature. The absolute number of covariates measured varies widely across studies. For instance, in papers reporting PCB effects on early childhood cognitive development, the number of covariates considered vary 5-fold; from approximately 10 (16), to 14 (17), to 17 (18), to 24 (19) to more than 50 (20). How many covariates are enough? This is difficult to answer, but probably more than most studies are measuring today. The importance of assessing a very wide range of potential confounds is illustrated in Table 1. The correlation matrix shows the bivariate relationships between a large number of covariates and child global cognitive performance, and CPT (continuous performance test) performance, at 3 yr, 41⁄2 yr, and 9 yr of age. Out of the 46 covariates listed, there are a very large set of covariates which predict performance, about 20–30 for each outcome measure, even using a very restrictive (p!.05) criteria for association. It is perhaps tempting to assume that if one measures a smaller number of “major covariates” (e.g., parental IQ, SES, HOME, cigarette smoking, alcohol use, and child sex), then these will account for most, if not all, of the variance predicted by those 20–30 covariates in Table 1. Indeed, as shown in Table 2, these common covariates are robust predictors of children’s cognitive and behavioral development. Nevertheless, commonly used covariates such as those shown in Table 2 do not account for all the variability in the outcome measures in our own data set (The Oswego PCB study). Table 3 shows that several variables still remain significant predictors of outcome, even after adjustment of the outcome scores by the 6 common covariates in Table 2. In terms of global cognition (McCarthy and WISC scores), the two most consistent and important additional covariates appear to be maternal depression and maternal sustained attention. To my knowledge, these are not routinely measured or controlled for the human neurotoxicology literature.

Human Neurobehavioral Toxicology Studies Table 1

365

Predictors of Behavioral Development in Children

Demographic Maternal education Paternal education Parity of child SES (Hollingshead 2-Factor Index) Maternal IQ (PPVT & K-BIT) Maternal sustained attention (CPT) Maternal depression Maternal age Maternal height Paternal age Paternal height Paternal weight HOME 12 mo HOME 54 mo HOME 8 yr #Yrs Same address #Yrs near grt lakes Married (y/n) Day care Home care Health/nutrition Pre-pregnancy weight Weight gain during preg. Stress before preg. Stress-1st half preg. Stress-2nd half preg. Maternal illness history Obstetric optimality Vitamins during preg. Prescrip. meds. during preg. Non-prescrip. meds. during preg. Nutrition scale Infant characteristics Child sex Birthweight (grams) Head circumference Ballard:neuromuscular Ballard:physical Gestational age at birth EP (cord blood) Substance use Cigarettes/day 2nd hand smoke (hrs/day) Alcohol (# drinks/day) Herbal tea (drinks/mo) Decaf. coffee (drinks/mo) Diet soda (drinks/mo) Decaf. soda (drinks/mo) Caffein. beverages (drinks/mo)

McCarthy (3 yr)

McCarthy (41⁄2 yr)

WISC-III (9 yr)

Impulsive responding [4,8 & 10 yr (mean)]

.32*** .32*** K.15** .31*** .39*** .28*** K.31*** .26*** .01 .26*** .11* .01 .21*** .47*** .47*** K.01 .09 K.24*** .00 .02

.34*** .30*** K.03 29*** .44*** .35*** K.17** .29*** .04 .20*** .06 .11* .08 .43*** .44*** K.05 .11* K.24*** .07 .02

.36*** .30*** K.15** .38*** .52*** .33*** K.31*** .13* .02 .09* .16** .10* .15** .38*** .44*** .00 .01 K.12* .11* .07

K.19** K.20*** K.02 K.30*** K.20*** K.30*** .18** K.15** K.11* K.19** K.18** K.12* K.21*** K.31*** K.30*** K.08 .00 .18** K.09 K.08

K.04 .07 K.14** K.03 .00 K.15** .08 .15** .00 K.05 K.07

.03 .07 K.15** .03 .10* K.09 .05 .14** K.04 .01 K.13*

K.03 .06 K.08 .12* .07 K.11* .04 .06 K.02 .05 K.03

.16** K.06 .14* .01 K.05 .10* K.15** K.15** .12* .10* .11*

.23*** .18** .07 .03 .05 .08 K.14**

.24*** .24*** .10* .02 .02 .02 K.10*

K.01 .22*** .25*** .08 .10* .03 K.09

K.35*** K.16** K.15** K.21*** K.28*** K.19** .20***

K.22*** K.24*** .09 K.01 .06 .00 K.03 K.18**

K.17** K.22*** .10* .09* .10* K.13* K.10* K.11*

K.17** K.24*** .05 .10* .03 .02 .00 K.10*

.04 .32*** .02 K.05 K.14* .00 .00 .16**

*p!.20, **p!.05, ***p!.01. Abbreviations: SES, hollingshead 2-factor index; GCI, mcCarthy general cognitive index; PPVT, peabody picture vocabulary test; K-BIT, kaufman brief intelligence test; PPVT & KBIT, average of 2 IQ metrics; CPT, continuous performance test, d-prime score; HOME, home observation measure of the environment; EP, cord blood erythroprotoporphyrin.

366 Table 2

Stewart Common Covariates as Predictors of Child Cognitive Performance

Common covariates Maternal IQ HOME Socioeconomic status (SES) Maternal cigarette smoking Maternal alcohol use SEX

McCarthy 3 yr

McCarthy 41⁄2 yr

WISC 9 yr

Impulsivity (CPT) 4, 8 & 10 yr mean

.39*** .47*** .31*** K.22** .08 .23***

.44*** .43*** .29*** K.17** .10 .24***

.51*** .44*** .38*** K.17* .05 K.01

K.20** K.28*** K.30*** .04 .01 K.35***

***p!.005, **p!.05, *p!.10.

The inadequacy of reliance upon the “commonly used” covariates is even more salient for domain-specific endpoints such as CPT performance (Table 3). Even after adjusting the CPT performance data for all 6 major covariates, we find an additional 11 that predict residual variability. The significance of these latter associations range from “trends” (p!.10) to highly convincing (p!.001). Note that there would be several more than 11 if a larger (p!.20) criteria for covariate inclusion is adopted. These data raise two important points. First, we find that inclusion of a small number of major covariates does not sufficiently control for important, significant predictors of outcome, whether that endpoint is global (IQ) or domain-specific (impulsive responding on CPT tasks). Secondly, the set of covariates that predict IQ is very different than those Table 3 Additional Predictors of Child Performance After Adjustment for Common Covariates (Maternal IQ, the HOME, SES, Smoking, Alcohol Use and Child Sex) McCarthy 3 yr Maternal depression Maternal stress Maternal attention

K.18*** K.16** .15**

WISC-III 9 yr Maternal attention Maternal depression Maternal stress Birth weight

.21*** K.17** .15** .11*

*** p!.005, ** p!.05, * p!.10.

McCarthy 41⁄2 yr Maternal attention Maternal stress Maternal diet soda use Birth weight Maternal age Maternal nutrition

.20*** K.17** K.17** .16** .15** K.13*

Impulsivity (CPT) 4, 8 & 10 yr mean Ballard (physical) Maternal attention Ballard (neuromuscular) Pre-pregnancy weight Stress before pregnancy Birth weight Gestational age Infant head circumference 2nd-hand smoke exposure Maternal depression Cord blood EP

K.26*** K.22*** K.21** C.21** C.19** K.17** K.16** K.14* C.14* C.14* C.13*

Human Neurobehavioral Toxicology Studies

367

which predict impulsive responding. Thus, it is difficult to predict how many, or what combination, of covariates will be important predictors of children’s performance. These observations underscore the need to measure a very large array of covariates. It is possible that any number of these variables could relate to the exposure metric in a given study, and thus may be bona-fide confounders. Failure to measure and control for them would clearly bias the exposure-effect estimate. Second, even if these variables were unrelated to the exposure metric in a given study, statistical control for these variables could increase statistical power by reducing the error term of the regression equation (21). Failure to measure a large number of covariates means that one or both forms of control in many studies is potentially less than it could, or should, be. Unfortunately, I do not believe the majority of human neurotoxicology studies pay nearly enough attention to this issue.

The Measurement of Control Variables Differ by Study Measurement of a sufficient number of potential confounds is not the only area in which human neurotoxicology studies may improve. Even when the same control variables have been measured across studies, they have not all been measured in the same manner. Although Leon (22) reviews the dangers of “over-aggregation” of covariates, covariate data are routinely aggregated in the literature. In the case of maternal smoking during and/or prior to pregnancy, some studies have measured it in a continuous fashion (20), while others have aggregated the variable in a dichotomous fashion (8,18). Education of the parent has also been measured in both continuous (20) and dichotomous fashions (8). Arguments have been made for both approaches, with those measuring the data in a continuous fashion arguing that this most accurately reflects the natural variability of the variable, and those measuring the data in a dichotomous fashion raising concerns about the skewed distributions of some of these variables and resulting violations of statistical assumptions (23). The choice between these approaches is not necessarily trivial, since using continuous or dichotomous data will affect the control equation. This is demonstrated in Table 4. The choice to convert maternal smoking and maternal education, for example, from continuous (average cigarettes per day; maximum grade obtained) to dichotomous (smoke y/n; high school degree/college degree) variables almost universally reduces their predictive power on several endpoints in our dataset. The strength of association between maternal education and WISC-III performance, for example, dropped from rZ.36 to rZ.23 when transforming the data from a continuous to Table 4 Smoking and Education as Continuous or Dichotomous Variables in Relation to Multiple Behavioral Endpoints in Children Variable

McCarthy 3 yr

McCarthy 41⁄2 yr

WISC 9 yr

CPT-impulsive 4, 8 & 10 yr

Smoking (continous)a Smoking (dichotomous)b Education (continuous)c Education (dichotomous)d

K.22** K.18* .32*** .19***

K.17* K.09 .34*** .23***

K.17* K.13* .36*** .23***

.04 .12* K.19** K.12*

a

cigarettes/day, *p!.20 cigarettes (y/n), **p!.05 c highest grade achieved, ***p!.01 d high school grad/college grad. b

368

Stewart

a dichotomous measure. In terms of r-squared, this is more than a 50% reduction in predictive power. Thus, dichotomizing these variables results in a significant underestimation of their impact. This is not to say that converting continuous data into ordinal data is never warranted or useful. Sometimes, data simply are not distributed in a manner remotely approximating a normal distribution, and this cannot be corrected by simple log transformations. In these cases transforming data into an ordinal set may be required (23,40). However, the point is that the decision can influence the outcomes, and no agreed-upon framework has been adopted. What degree of skewness or kurtosis justifies transforming continuous data into an ordinal set? And if this is done, what assurance can be made that the “true” association between the covariate and the outcome is not artifactually diminished, as suggested in Table 4? We currently have no agreed-upon answers to these questions. In many papers I have reviewed over the past 10 yr, covariate data appear in dichotomized, trichotomized, or otherwise in aggregated form without explanation. Reliability and Validity of Covariates Industry-funded critiques of human neurotoxicology studies (24,25) have criticized the reliability of our exposure and outcome measures, in an attempt to argue that results of human neurotoxicology studies are spurious (e.g., replete with Type I errors). These critiques miss the mark in fundamental ways—problems with reliability in either the exposure or outcome data increase Type II errors, not Type I errors. This is due to the statistical principle that random measurement error (including problems with reliability) make it less likely that systematic relationships in the data can be found. If, however, there are concerns with the reliability, or validity, of covariates, then we indeed do have a problem with potentially reporting spurious findings. My concern is that the same standards for reliability and validity, so important in the development and construction of analytical, biological, and behavioral measurement tools, are not often raised for many covariate measures. This is particularly troublesome because the putative effects of low-level teratogens are often small. In this case, lessons may be gleaned from other fields which wrestle with the same issues. Relman & Angell (26), in their criticism of the literature concerning putative effects of psychosocial interventions on disease, stated “the reason weak effects are so difficult to study is that they are easily swamped by effects of confounding variables, either known ones or unknown..it would be essential to adjust very, very finely for [important confounds]...yet, failure to describe exactly how confounders were handled was a major problem in nearly all these studies.” Are we adjusting “very, very finely” for all important potential confounds? I believe we may not in some cases. While reliable, valid, and standardized measures are available for exposure assessment (e.g., analytical PCB, Pb, or MeHg methodologies), and outcome assessment (e.g., Full Scale IQ) exist, they only exist for a handful of control variables (Parental IQ, SES, HOME, Maternal Depression). Consider, for example, how we treat exposure data (i.e., PCB, Pb or MeHg levels) relative to how we treat data for potential confounders. As a field, we accept that the reliability and validity of exposure data warrant considerable scrutiny and attention. Thus accepted analytical standards and practices for their direct quantification in biological tissues have been developed. Further, when human neurotoxicology studies are published, it is almost a universal requirement that we publish at least one paper on the analytical methods alone. In contrast, equally rigorous standards have not been universally developed for more than a small handful of major covariates. On one hand we seem to have excellent tools for assessing maternal IQ (PPVT, K-BIT, WAIS), socioeconomic status (Hollingshead), and the home environment (H.O.M.E.

Human Neurobehavioral Toxicology Studies

369

scales). On the other hand we also accept self-report, retrospective recall measures for many potentially confounding exposures (i.e., smoking, alcohol use). Further, as mentioned earlier in this chapter, these variables are often expressed in crude categories. In addition, reliability data are not always presented, or even expected it seems, for many covariates. Finally, unlike the case for exposure data, we not seem to require stand-alone papers carefully detailing and describing the quality assurance/quality control for measuring every single potential confound. Why is this permissible? The implication is that either these are unimportant variables, or that measuring control variables with less precision and care than the exposure variable is “good enough” for the “purposes of control.” Neither position is really defensible. This is not an attempt to dismiss self-report measures, nor an attempt to suggest that human neurotoxicology studies cannot be conducted unless control variables are measured with perfect fidelity. The question I am raising is: what are the implications in a data set where the predictor variable is measured with great reliability, but the covariates are measured with moderate or even poor reliability? The answer should be clear even at the introductory statistics level. When correlated predictors are measured with differing reliability coefficients, the outcome will be biased in favor of the more reliably measured predictor. This is simply due to the fact that measurement error (i.e., lower reliability) interferes with the ability of one variable to predict another. This is a sobering thought when considering the difference in the amount of time and resources directed towards developing precise and reliable measures of chemical levels in human blood, compared to the amount of time and resources directed toward developing (and using!) precise and reliable measures of nutritional history, second hand smoke exposure, perinatal medication use, alcohol consumption, and numerous others. This is not to say that such measures are not available, only that if they are available, they have not been universally adopted. Since we do not have agreed-upon, standardized measurement tools for the majority of covariates, with known validity and reliability, we do not even know the extent of this problem. Differing Criteria for Entry of Covariates The lack of standardization in the number of covariates, the manner in which they are measured, and their psychometric (reliability and validity) properties is a multi-layered barrier to the effective interpretation and comparison of studies in human neurotoxicology. Unfortunately, this problem is even further compounded by significant discrepancies in the statistical treatment of covariates. In particular, the decision criteria for whether or not covariates enter an equation vary considerably across studies and disciplines. There appear to be at least two broad, conceptual approaches: empirical (p-value and/or change-inestimate approaches) and theoretically based approaches. Using an empirical approach, in the form of significance levels (p-values) of the covariate, has been a common method. However, there is considerable variability in the literature in regards to which p-values are used, and how they are employed. Inclusion criteria range from p!.20 in relation to outcome only (20,27), to p!.10 in relation to outcome (28,29), to p!.25 in relation to both exposure AND outcome (30). In other cases, p-values are employed differently depending on whether they are determined in relation to exposure or outcome [e.g., a study of prenatal cocaine exposure in children considered covariates if they were correlated with outcome at p!.20, but with the further restriction that the study groups differ on the covariate at a very restrictive p!.05 (31)]. Some investigators retain all variables related to outcome at a given alpha level from univariate analyses (20,27), and others employ multi-step analyses where variables may be removed or added based on additional criteria (18,32). Unfortunately, few if any studies cite empirical

370

Stewart

evidence to support the effectiveness at confounder control of the particular p-value or approach chosen. This is especially unfortunate given that there are indeed published data in the literature showing that not all approaches are equally efficacious (33,34). A distinctively different approach from the empirical approaches described above are theoretical criteria where covariates are entered, based not upon empirical relations with exposure and/or outcome, but rather presumed theoretical relevance to the outcome measure under test. This last approach presents potentially unique problems and will be discussed later in this chapter. Importantly, an approach that may be used adjunctively with all the previously described approaches is the “change-in-estimate” approach (33,34), where additional covariates are selected if they produce more than a 10% change in the neurotoxicant association with outcome. This approach is particularly effective to guard against bias in the original covariate model, since each and every covariate that did not meet the initial selection criteria is checked to see if it changes the exposure effect estimate (beta). This approach has been used as an adjunct to the p-value approach in the Pb literature (30), and very recently in our own work (35). Unfortunately, as with all the other methods, its use is inconsistent in the human neurotoxicology literature. Should There Be Standards for Covariate Entry Criteria? With so much variability in the approaches to covariate control, it is logical to ask whether the discrepant approaches play a role in discrepant outcomes across studies. Moreover, if these different approaches are producing discrepant outcomes, which approach is most valid? These are both potentially answerable questions. In regards to whether the different approaches may yield different outcomes, the answer may be, in some cases, yes. Using our own data set, I first experimented with 4 different statistical criteria for covariate entry. I regressed postnatal lead (Pb) exposure on commission errors on a continuous performance task in 41⁄2 yr-old children (nZ131 children with valid postnatal Pb exposure data). Four different criteria (p-values) for covariate inclusion were employed. For Method 1, I used the rule that a covariate must be related to both exposure AND outcome to be considered a “true” confounder (36). I also used a very restrictive alpha level (p!.10) for this analysis. For Method 2, I used the same definition of a confounder but used a larger (and thus more inclusive) alpha level for inclusion (p!.20) of covariates. Method 3 used a broader definition of a confounder, specifically, that any and all variables related to outcome are included, at a narrow alpha (p!.10). Method 4 used the broadest definition and inclusion criteria for confounders of all the methods: any and all variables related to outcome are entered, at a large alpha (p!.20). Out of all the 4 methods, this last method has been empirically demonstrated to be effective at confounder control in Monte-Carlo studies (33,34). Following these analyses, the impact of these covariate selection criteria was examined by re-running each of the regressions using a change-in-estimate criteria (described earlier) in the second step. This approach allows the analyst to determine whether the initial regression failed to account for all important confounders, because any unselected covariates which even weakly (O10%) change the exposure effect estimate (beta) are added. This approach has also been shown effective to control confounding bias in Monte-Carlo studies (33,34). Table 5 demonstrates the results of these analyses. The differing statistical criteria for covariate entry did indeed affect outcome. Briefly, the broader the definition of potential confounders, and the larger the alpha level for inclusion, the more variance in the outcome measure was accounted for, and the stronger the Pb effect became. In addition,

Human Neurobehavioral Toxicology Studies

371

Table 5 Postnatal Lead Exposure and Commission Errors on the CPT: Results Using Four Different Criteria for Covariate Inclusion (1) p!.10 in relation to exposure AND outcome

(3) p!.10 in relation to outcome

HOME 4 yr Secondhand smoke exposure Marital status

HOME 4 yr Secondhand smoke exposure Marital status HOME 12 mo Maternal pre-pregnancy weight Maternal sustained attention Obstetric optimality Sex Ballard (physical)

Covariate r2Z.10, p!.003 Adjusted r2Z.08 Lead outcome: BZC.09, pZ.25 D Lead outcome: BZC.12, pZ.14

Covariate r2Z.26, p!.001 Adjusted r2Z.20 Lead outcome: BZC.12, pZ.14 D Lead outcome: BZC.15, pZ.06

(2) p!.20 in relation to exposure AND outcome

(4) p!.20 in relation to outcome

HOME 4 yr Secondhand smoke exposure Marital status Socioeconomic status

HOME 4 yr Secondhand smoke exposure Marital status Socioeconomic status (SES) Number of children in household HOME 12 mo Maternal pre-pregnancy weight Maternal sustained attention Obstetric optimality Sex Ballard (physical) Paternal age

Covariate r2Z.11,p!.004 Adjusted r2Z.08 Lead outcome: Beta ZC.10, pZ.23 D Lead outcome: Beta ZC.13, pZ.13

Covariate r2Z.30,p!.001 Adjusted r2Z.23 Lead outcome: BZC.16, pZ.04** D Lead outcome: BZC.16, pZ.05**

**Association with Pb exposure statistically significant after control for the widest array of confounders. D, Results after Change-in-Estimate approach.

Table 5 shows that when the change-in-estimate approach was added to examine the impact of the confounder selection strategy, the effect estimates of Methods 1–3 were notably changed. Only the results of Method 4 were not significantly changed by this test. These results should not be surprising. Monte-Carlo studies demonstrate that either the change-in-estimate criteria, or a liberal p-value criteria (p!.20), perform acceptably as covariate criteria in covariate control simulations (33,34). Lower p-values have been shown to perform unacceptably (33,34). In short, the results from Method 4 are the most valid on the basis of the literature (33,34), and the fact that a check for bias using the change-inestimate supported the Method 4’s results. These observations argue for using either Method 4, the change-in-estimate procedure, or the two in combination. A strength of the combination approach is that it minimizes bias (every covariate is checked to see if it changes the effect estimate, using the change-in-estimate approach), and it maximized

372

Stewart

precision [covariates which are related to outcome are entered, regardless of their effect on the exposure-effect estimate, in order to reduce the error term of the equation (37)]. This has been used to excellent effect in the Pb literature (30), and recently in the PCB literature (35). Given these observations, and the fact that there is statistical evidence in the literature demonstrating the superior effectiveness of liberal p-values (p!.20) and change-in-estimate approaches to confounder control, it is not really sensible that we have failed to adopt such practices routinely. We do have an agreed standard for p-values in virtually all sciences for hypothesis testing of the predictor and outcome measure (pZ.05), which reflect a general consensus that minimizing Type I errors is more important than minimizing Type II errors. Most scientists would agree, however, that when controlling for covariates, these concerns are reversed. For in the effort to be sure that we enter all the variables which may otherwise serve to confound our results, it is the minimizing of Type II errors that is paramount. Entering a covariate which is not a true confounder or not “truly” related to outcome (analogous to a Type I error) may be a small mis-step, but failing to recognize, and thus enter, a truly important confounder (analogous to a Type II error) is a serious problem. This is why I believe conservative alpha levels should NOT be applied when performing analysis for covariate inclusion, and Monte-Carlo simulations support my view (33,34). By failing to control for an important confound because it did not achieve a highly conservative alpha (pZ.05 or pZ.10) in relation to outcome, we may be “falsely retaining the null for confounders” (thus excluding them) at very high rates. This makes the final analysis of the exposure and outcome variable less valid. Until there is adoption of a universal, evidence-based standard, the differing methodological practices will only serve to potentially confuse and obscure the actual relations among the data, whether null or not. As long as these discrepancies exist, methodological and statistical factors will remain a potentially simple rival explanation to the thoughtful, but more complex, hypotheses of effect modification put forth to explain discrepancies across studies (38).

ADDITIONAL (THEORETICAL) CRITERIA FOR COVARIATES The empirical criteria for covariate entry discussed above (i.e., entry based on p-values or change-in-estimate) are quite common in the majority of large, well-conducted human neurotoxicology studies. However, sometimes investigators employ additional criteria for covariate entry. In particular, one of these is a criteria based upon “theoretical relevance” of the covariate to the outcome measure. This occurs occasionally in major human neurotoxicology studies, but occurs with extreme regularity in studies I have reviewed outside of the Pb and PCB fields. I believe theoretical criteria merit focused discussion because they potentially introduce a serious bias problem in covariate analysis. The use of theoretical criteria for entry of covariates into an equation appears perfectly sensible. Variables which, based upon the literature, should be predictors of an outcome measure should be entered as control variables in the equation. Variables for which there is no basis or evidence for their relation to outcome should probably be ignored. There can be little disagreement that this practice is defensible when it comes to well-known variables such as parental IQ, SES, or the HOME and children’s IQ. There is a wealth of literature supporting the relations between the former 3 and the latter (20,30,19,37,18,17). Yet there are far more variables that may contribute to performance beyond “the big 3” and many more potential outcome variables than children’s IQ

Human Neurobehavioral Toxicology Studies

373

(Tables 1–3). It is in these latter scenarios that fundamental problems arise for theoretical criteria. Where is there consensus for, say, exposure to second hand smoke and performance on a continuous performance test? Is there a theoretical basis for a relation between consumption of caffeinated beverages during pregnancy and visual recognition memory in infants at 1 yr of age? The answers to these questions are less important than their implications: (a) there is a nearly limitless number of possible covariate-outcome combinations for which no literature are available and (b) there are a very limited number of covariate-outcome combinations for which there is general scientific consensus regarding the purported relation between the two. Thus, the incidences when the application of a theoretical criteria for covariate inclusion is truly valid are necessarily limited to a small number of covariate-outcome combinations. As a result, investigators employing theoretical criteria are required to use their own judgment and their own reading of the literature when developing theoretical justifications for inclusion/exclusion of particular covariates. In my view this makes the use of theoretical criteria for entry and/or exclusion of covariates most problematic and indeed, suspect. For it permits the judgments, opinions and even subtle biases of the investigator to determine which variables enter into the control equation and which do not. By extension, then, this allows investigator’s judgment (or bias) to determine the very nature of the control comparison. This is not desirable, and is doubly a concern given that, under many circumstances, the investigator may be working with both predictor and outcome data simultaneously, and is blind to neither. This may allow bias, whether conscious or unconscious, to influence the results. This can be easily illustrated in Table 6. Table 6 shows relationships between prenatal PCB exposure and children’s’ performance on the McCarthy Perceptual Performance Scale at 54 months of age. Column 1 shows the results as we reported them in Stewart

Table 6 Prenatal PCB Exposure and Perceptual Performance, Results With and Without Theoretical Criterion (1) Purely empirical criterion (p!.20) Maternal education Paternal education Socioeconomic status Maternal age Paternal age HOME (54 mo) Stress before pregnancy Stress 2nd half of pregnancy Nutritional history Birth weight Cigarette smoking Secondhand smoke Maternal illnesses Maternal decaffeineated coffee use Maternal IQ

PCB outcome: FZ2.24, pZ.137

(2) Empirical C Theoretical criterion Maternal education Paternal education Socioeconomic status Maternal age Paternal age HOME (54 Mo) Stress before pregnancy Stress 2nd half of pregnancy Nutritional history Birth weight Cigarette smoking Secondhand smoke Maternal illnesses Maternal decaffeinated coffee use Maternal IQ Maternal pre-pregnancy weight Maternal prescription medication use Maternal over-the-counter medication use PCB outcome: FZ4.39, pZ.038

374

Stewart

et al. (20), using an empirical criteria for covariate entry of p!.20, and we show no significant association between PCB exposure and McCarthy performance at 54 months of age (pZ.133). However, the 2nd column illustrates a hypothetical scenario whereby a researcher could take these same data, but add 3 additional covariates based upon supposed “theoretical relevance” to the outcome measure. These variables are maternal body weight just before pregnancy, maternal prescription medication use during pregnancy, and maternal over-the-counter medication use during pregnancy. Who could argue against the potential theoretical relevance of these variables to a newborn’s health? Probably no one. Using these 3 additional covariates, now a significant association (pZ.038) between PCB and Perceptual Performance is shown. But before getting too eager to dismiss our earlier negative results, consider I could also argue that the following covariates are also “theoretically relevant”: prenatal DDE exposure, prenatal MeHg exposure, stress during the 1st half of pregnancy, and the number of years living at current address (i.e. an index of home stability). Upon entering these variables, the relationship between prenatal PCB exposure and Perceptual Performance once again becomes nonsignificant (FZ2.52, pZ.114). The point here is that employing the “theoretical relevance” criteria allows the analyst a frightening degree of control over the outcome of the analysis. This is extremely problematic especially given the fact that there is really no objective criterion for “theoretical relevance”— it is up to the judgment of the analyst. A reviewer can never be sure why the covariates selected were chosen, and why others were left out. I fear the human ability to rationalize may allow an investigator, perhaps unconsciously, to choose the variables which allow for a confirmation of his/her hypothesis. One can always then rationalize, post-hoc, why the variables chosen were relevant, and why the variables left out were not. It is for this reason we do not use a theoretical criteria in the Oswego study and why I submit they should never be used in epidemiology. This is not to say investigators are unethical. But like anyone else, we are fallible. My argument, that theoretical criteria never be employed, questions the integrity of researchers no more than the requirement for blindedness does. It is simply an effective means of ensuring biasfree analysis. The use of a strictly empirical criteria for covariate inclusion, in my view, offers the only potentially bias-free approach. An empirical criteria for covariate inclusion makes no assumptions about the anticipated relationships between covariates and outcome, but rather sets an arbitrary, but objective, criteria for their inclusion. For example, in the Oswego study, we now enter any and all variables which relate to an outcome measure with a liberal alpha level of p!.20, followed by an additional change-in-estimate step to eliminate any residual bias. These components of this approach have strong support from Monte-Carlo studies (33,34). A potential criticism of a “purely” empirical approach is that by ignoring literature and/or theory in favor of an arbitrary p-value or change-in-estimate cutoff, the investigators may (1) fail to enter and important covariate that is clearly relevant based on the literature, but which failed to meet the arbitrary p-value cutoff, and (2) an empirical and liberal criteria of (such as p!.20) will result in the inclusion of a large number of covariates just by chance (1 out of 5, in the case of p!.20), generating significant noise in the model. Several points need to be made in this regard. Regarding point #1, it is highly unlikely that a variable that is “an important covariate that is clearly relevant based on the literature” would not be related to outcome at a high (liberal) alpha level (e.g., p!.20), and/or would not change the outcome of the results with a change-in-estimate analysis. If it did neither, one might question the literature base rather than the covariate model. Second, it is quite true that using a liberal alpha level will result in a greater number of covariates

Human Neurobehavioral Toxicology Studies

375

entered based upon chance relationships. Nevertheless, use of a liberal alpha for covariate inclusion has not prevented the Oswego study from routinely detecting subtle effects of PCB exposure in exposure children (39,40,20,27,35). Important Questions for the Discipline As the scientists who literally define “state” of our “art” in our fields, it is our responsibility to engage in a collective conversation about minimally acceptable standards and practices for human neurotoxicology studies. I have outlined 4 areas which I believe could benefit from discussion and debate: 1. Minimum standards for the number and kind of covariates that impact neurocognitive development 2. Minimum standards for validity and reliability of said covariates 3. Development of universal, objective standards for covariate inclusion and exclusion criteria 4. Development of universal standards and practices for the statistical modeling of control data. The reader will note that no particular approach is recommended. I certainly cannot claim to have all the answers. But the conversation needs to begin. The various approaches to each of these areas need to be put on the table. Then, empirical and theoretical arguments need to be brought to bear in order to determine the most powerful and effective approaches. I also recognize that the issue of covariate control is not always a simple one. A variable that is a control for one endpoint could be an effect mediator, moderator, or even an outcome measure for another endpoint. However, excessive concern over these situations obfuscates the issue. I am not dismissing these cases; rather, the issues I outline above are applicable to what are, quite frankly, the majority of analyses conducted today: where there is a desire to infer a putative cause-effect relation between two variables while attempting to control for the influence of a larger number potentially confounding variables. These recommendations above do not delve into the more complex issues of effect mediation or moderation, but neither do they depend on their resolution. For undoubtedly a set of standards and practices must be developed for those analyses as well. I would not be entirely surprised if critics of the neurotoxicology literature used this chapter as part of their arguments to convince the scientific community to dismiss studies in our field. Therefore I should make it clear that the criticisms of problems raised in this chapter are most decidedly not unique to neurotoxicology. If anything, the standards and practices to control in other non-experimental fields appears even less developed— especially in extremely large epidemiological studies where the resources to measure covariates with high fidelity are scarce. Nonetheless, I am focused on my own field because it is the one in which I work. The goal of this chapter is to improve our discipline, not to malign it. I also would not be surprised if some scientists dismissed some or all of the ideas outlined in this chapter as naı¨ve at best and a form of “scientific Orwellianism” at worst, the concern being that creative thinking, flexibility, innovation and new breakthroughs with respect to conceptual, empirical, or statistical treatment of control variables would be stifled in the context of a set of “rigid” standards and practices. In my view this concern is without merit and, in fact, pushes the field backward and not forward. The hallmark of a mature science is the development of common standards and practices. If a new conceptual or methodological approach were truly innovative and meritorious (and, assuming progress in the field, such advances should occur), then it is the responsibility of

376

Stewart

the investigator to empirically demonstrate this. If the evidence is convincing enough that a particular change in the standards and practices demonstrably improves the explanatory power or bias control, then no doubt it will be adopted. This will undoubtedly occur in due course if we develop some basic standards upon which approaches to control can be judged. Of course, it will take somewhat longer for a particular practice or approach to be accepted relative to the current “wild west” of non-standard, diverse and idiosyncratic approaches. However, the alternative is far worse. Imagine, if you will, this discussion taking place in the context of analytical chemistry or medicine. How trustworthy would our exposure data be if analytical MeHg, Pb or PCB chemists resisted standards and practices for the precise measurement of these toxicants in favor of an argument for an environment that encourages “freedom and creativity?” And imagine a similar situation in the case of treatments of specific diseases. The point here is that there exist disciplines in science that have achieved much more rigor and credibility than others. The goal of this chapter is to begin encouraging human neurotoxicology to become that kind of discipline.

REFERENCES 1. Ettinger B, Friedman G, Bush T, Quesenberry C. Reduced mortality associated with long-term postmenpausal estrogen therapy. Obstet Gynecol 1996; 87:6–12. 2. Stampfer M, Colditz G, Willett W, Manson J, Rosner B, Speizer F. Postmenopausal estrogen therapy and cardiovascular disease: Ten-year follow-up from the Nurses’ Health Study. N Engl J Med 1991; 325:756–762. 3. Grady D, Rubin S, Petitti D. Hormone therapy to prevent disease and prolong life in postmenopausal women. Ann Intern Med 1992; 117:1016–1037. 4. Rijpkema A, Van der Sanden A, Ruijs A. Effects of post-menopausal oestrogen-progestogen replacement therapy on serum lipids and lipoproteins: a review. Maturitas 1990; 2:259–285. 5. Matthews K, Kuller L, Wing R, Meilahn E, Plantinga P. Prior to use of estrogen replacement therapy, are users healthier than nonusers? Am J Epidemiol 1996; 143:971–984. 6. Rossouw J, Anderson G, Prentice R, et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women’s Health Initiative randomized controlled trial. Writing Group for the Women’s Health Initiative Investigators. July 17, 2002; 288:321–333. 7. Hennekens CH, Buring JE. Epidemiology in Medicine. 1st ed. Boston: Little, Brown & Co., 1987. 8. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 9. Grandjean P, Weihe P, Burse VW, et al. Neurobehavioral deficits associated with PCB in 7 yr-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001; 23:305–317. 10. Myers GJ, Davidson PW, Cox C, et al. Summary of the Seychelles child development study on the relationship of fetal methylmercury exposure to neurodevelopment. Neurotoxicol 1995; 16:711–716. 11. Davidson PW, Myers GJ, Cox C, et al. Longitudinal neurodevelopment study of Seychellois children following in utero exposure to methylmercury from maternal fish ingestion: outcomes at 19 and 29 mo. Neurotoxicol 1995; 16:677–688. 12. Davidson PW, Myers GJ, Cox C, et al. Effects of prenatal and postnatal methylmercury exposure from fish consumption on neurodevelopment: outcomes at 66 mo of age in the Seychelles Child Development Study. JAMA 1998; 280:701–707. 13. Davidson PW, Palumbo D, Myers GJ, et al. Neurodevelopmental outcomes of Seychellois children from the pilot cohort at 108 mo following prenatal exposure to methylmercury from a maternal fish diet. Environ Res 2000; 84:1–11.

Human Neurobehavioral Toxicology Studies

377

14. Davidson PW, Myers GJ, Shamlaye C, Cox C, Wilding GE. Prenatal exposure to methylmercury and child development: influence of social factors. Neurotoxicol Teratol 2004; 26:553–559. 15. Heiman GW. Research Methods in Psychology. Houghton Mifflin Co., 1995. 16. Gladen BC, Rogan WJ. Effects of perinatal polychlorinated biphenyls and dichlorodiphenyl dichloroethene on later development. J Pediatr 1991; 119:58–63. 17. Patadin S, Lanting C, Mulder P, Boersma ER, Sauer PJ, Weisglas-Kuperus N. Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in Dutch children at 42 mo of age. J Pediatr 1999; 134:33–41. 18. Walkowiak J, Wiener JA, Fastabend A, et al. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on pscychodevelopment in early childhood. Lancet 2001;358–1602–1607. 19. Jacobson JL, Jacobson SW, Humphrey HE. Effects of in utero exposure to polychlorinated biphenyls and related contaminants on cognitive functioning in young children. J Pediatr 1990; 116:38–45. 20. Stewart PW, Reihman J, Lonky EL, Darvill TJ, Pagano J. Cognitive development in preschool children prenatally exposed to PCBs and MeHg. Neurotoxicol Teratol 2003; 25:11–22. 21. Kleinbaum D, Kupper L, Muller K. Applied regression analysis and other multivariable methods. 2nd ed. Boston: PWS-Kent, 1988. 22. Leon DA. Failed or misleading adjustment for confounding. Lancet 1993; 342:479–481. 23. Winneke G, Walkowiak J, Kramer U. Strong opinions are no substitute for balance arguments: Comments on Cicchetti. Kaufman, and Sparrow’s critical appraisal of PCB cohort studies. Psychol Schools 2004; 41:655–659. 24. Cicchetti D, Kaufman A, Sparrow S. The relationship between prenatal and postnatal exposure to polychlorinated biphenyls (PCBs) and cognitive, neuropsychological, and behavioral deficits: A critical appraisal. Psychol Schools 2004; 41:589–636. 25. Kaufman A. Do low levels of lead produce IQ loss in children? A careful examination of the literature Arch Clin Neuropsychol 2001; 16:303–341. 26. Relman AS, Angell M. Resolved: Psychosocial interventions can improve clinical outcomes in organic disease. Psychosom Med. 2002; 64:558–563. 27. Stewart PW, Fitzgerald S, Reihman J, et al. exposure, the corpus callosum, and response inhibition. Environ Health Perspect 2003; 111:1670–1677. 28. Chiodo LM. Neurodevelopmental effects of postnatal elad exposure at very low levels. Neurotoxicol Teratol 2004; 26:359–371. 29. Jacobson JL, Jacobson SW. Prenatal exposure to polychlorinated biphenyls and attention at school age. J Pediatr 2003; 143:780–788. 30. Bellinger DC, Stiles KM, Needleman HL. Low-level exposure, intelligence and academic achievement: a long-term follow-up study. Pediatrics 1992; 90:855–861. 31. Lewis BA, Singer LT, Short EJ, et al. Four-year language outcomes of children exposed to cocaine in utero. Neurotoxicol Teratol 2004; 26:617–627. 32. Schantz SL, Gasior DM, Polverejan E, et al. Impairments of memory and learning in older adults exposed to polychlorinated biphenyls via consumption of Great Lakes fish. Environ Health Perspect 2001; 109:605–611. 33. Maldonado G, Greenland S. Simulation study of confounder-selection strategies. Am J Epidemiol 1993; 138:923–936. 34. Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol 1989; 129:125–137. 35. Stewart P, Reihman J, Gump B, Lonky E, Darvill T, Pagano J. Response inhibition at 8 and 91⁄2 yr of age in children prenatally exposed to PCBs. Neurotoxicol Teratol 2005; 6:771–780. 36. Schlesselman J. Case-Control Studies: Design, Conduct, Analysis. New York: Oxford University Press, 1982. 37. Jacobson J, Jacobson S. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. N Engl J Med 1996; 335:783–789.

378

Stewart

38. Bellinger DC. Interpreting the literature on lead and child development: the neglected role of the “experimental system”. Neurotoxicol Teratol 1995; 17:201–212. 39. Darvill T, Lonky E, Reihman J, Stewart P, Pagano J. Prental exposure to PCBs and infant performance on the Fagan Test of Infant Intelligence. Neurotoxicol 2000; 21:1029–1038. 40. Stewart PW, Reihman J, Lonky E, Darvill T, Pagano. Prenatal PCB exposure and Neonatal Behavioral Assessment Scale (NBAS) performance. Neurotoxicol Teratol 2000; 22:21–29.

18 Environmental Risk Assessment with Structural Equation Models Esben Budtz-Jørgensen Department of Biostatistics, University of Copenhagen, Copenhagen and Institute of Public Health, University of Southern Denmark, Odense, Denmark and Department of Biostatistics, Harvard School of Public Health, Boston, Massachusetts, U.S.A.

SUMMARY Environmental risk assessment relies on statistical methods for estimation of the dose-response relationship and for subsequent calculation of exposure limits. Interspecies extrapolations are avoided when human epidemiological data is available. However, observational studies always involve concerns regarding confounding, measurement error, missing data, and multiple comparisons. Standard statistical techniques, such as multiple regression analysis, are poorly suited for such data and will lead to biased and inefficient effect estimation and safe dose calculation. This paper explores the potential of structural equation models for improving current statistical methods in environmental risk assessment. The concreter case of neurobehavioral effects in Faroese children prenatally exposed to methylmercury is used for illustration.

INTRODUCTION Biostatistical methodologies are crucial in the analysis of environmental epidemiology data. However, standard methods are too simple for common data in this field, and may hinder identification and characterization of hazardous environmental factors. For example, it is usually not possible to obtain an error-free measurement of the causative exposure. Available exposure markers are affected by ordinary measurement errors in the laboratory, as well as biological fluctuations. This uncertainty is typically ignored in multiple regression analysis which assumes that predictor variables are measured with-out error. As a consequence, these methods will lead to biased effect estimation. Often exposure-associated effects are biased toward zero, while effects of potential confounders can be biased in either direction (1). Adjustment methods usually depend on the size of the total (analytical C biological) error. From standard quality control data, it may be possible to estimate the laboratory error, but the size of the 379

380

Budtz-Jørgensen

preanalytical biological error component is often not known and corrected effect estimation is seldom attempted. Statistical analysis is further complicated by the fact that prospective cohort studies often include a large number of disease endpoints. Prior information about the adverse effect of the agents being studied is often weak, which makes it difficult to rule out effects in advance. While inclusion of many outcome variables will reduce the risk of overlooking important health effects, it also increases the risk of chance findings; especially if each endpoint is analyzed separately without any clear a priori hypothesis. Therefore, to obtain a correct assessment of the “overall” significance of the observed exposure effect, it is often necessary to conduct some sort of correction for multiple testing. The Bonferroni method is the standard technique for this purpose. However, with many correlated outcomes, this method is known to be very conservative. This is a critical weakness, which further reduces the power to detect the relatively weak effects of low-level exposure to chemical substances. This power may already be low, due to the relatively high degree of imprecision commonly present in environmental epidemiology outcome and exposure variables. Thus, important health effects associated with the substance in question may remain undetected, if the statistical analysis is based on these crude methods. This paper shows that a more sophisticated class of statistical models, known as structural equations, can be useful in the analysis of environmental epidemiology data. These models were originally developed in the social sciences, where they have been used extensively. However, so far, very few applications exist in environmental epidemiology. In these models, observed variables are considered manifestations of underlying latent variables which can be linearly related. As we shall see, when more than one exposure variable is available, this structure allows estimation of the total imprecision in each exposure variable. Based on these estimates, the naive regression coefficients can be adjusted to give an unbiased reflection of the exposure-related effect. Furthermore, multiple testing problems can be addressed by also viewing the outcome variables as indicators of a limited number of latent response variables. The advantages of structural equations are illustrated in the case of neurobehavioral effects in children with methylmercury exposure. This chemical accumulates in fish and seafood, and can cause serious damage to the nervous system; especially when the exposure occurs prenatally (i.e., due to the mother’s diet during pregnancy). The National Academy of Sciences (NAS) (2) recently identified an epidemiological cohort study conducted in the Faroe Islands as the critical study for calculations of a safe exposure limit for mercury. The question, therefore, emerged as how to extract the best possible information from these results using modern biostatistical methods.

THE FAROESE MERCURY STUDY A birth cohort of 1022 Faroese children was established, and the intrauterine methylmercury exposure was determined by analysis of umbilical cord blood and maternal hair (3). In the absence of frank methylmercury poisoning, as seen in serious episodes in Japan and Iraq (2), the objective of the study was to identify and characterize any neurobehavioral changes and their relationship to mercury exposure. Thus, at age 7 years, 917 (90%) of the cohort members participated in a thorough clinical examination with focus on nervous system function (3). Neuropsychological tests were chosen to include tasks that would be affected by the neuropathological abnormalities described in congenital methylmercury poisoning and the functional deficits seen in children with early-life exposure to other neurotoxicants.

Environmental Risk Assessment

381

THE STANDARD APPROACH TO EFFECT ESTIMATION Standard multiple regression analysis showed strong mercury effects, especially for the Boston Naming Test (BNT) and for the reaction time score on the Continuous Performance Test (4). Adjustment for an additional covariate (Town7), indicating whether the child was living in a town, generally reduced mercury effects (Table 1); and for some of the outcomes mercury coefficients were clearly insignificant. Thus, the regression analysis produced a large number of coefficients with a wide range of p-values. Based on these results, the “overall” significance of the observed effect of the exposure is difficult to assess. Even if the exposure had no effect, some coefficients would be expected to be significant simply by chance. Not only is the regression result difficult to interpret, this analysis is also inefficient, because it fails to exploit that some test scores measure the same underlying function in the child. Finally, the standard analysis ignores exposure imprecision, and exposure effects are therefore likely to be underestimated. Inclusion of potential confounders with a strong

Table 1 Estimated Effects of a 10-Fold Increase in Mercury Exposure Using the Cord Blood Mercury Concentration and the Mercury Concentration in Maternal Hair, Respectively, as the Exposure Indicator Cord Blood Hg Response NES2 Finger tapping Preferred hand (FT1) Non preferred hand (FT2) Both hand (FT3) NES2 Hand-Eye Coordination Error scorea (HEC) NES2 Continuous Performance Test Ln total misseda Reaction timea Wechsler Intelligence Scale Digit spans (DS) Similarities Sqrt. block designs Bender Visual Gestalt Test Errors on copyinga Reproduction Boston Naming Test No cues (BNT1) With cues (BNT2) California Verbal Learning Test Learning (CVLT1) Short-term repro. (CVLT2) Long-term repro. (CVLT3) Recognition (CVLT4)

Maternal Hair Hg

Naive

Adjusted

p

Naive

Adjusted

p

K1.01 K0.55 K1.90

K1.17 K0.64 K2.20

0.08 0.31 0.10

K1.03 K0.91 K2.74

K1.51 K1.34 K4.02

0.08 0.11 0.02

0.03

0.04

0.27

0.05

0.07

0.10

0.22 34.6

0.25 40.0

0.07 0.002

0.08 16.2

0.12 23.8

0.52 0.13

K0.21 K0.003 K0.11

K0.23 K0.004 K0.13

0.14 0.99 0.31

K0.17 K0.23 K0.06

K0.25 K0.33 K0.09

0.24 0.57 0.59

0.33 K0.10

0.38 K0.12

0.49 0.54

0.33 0.07

0.48 0.10

0.51 0.68

K1.61 K1.70

K1.86 K1.96

0.002 0.001

K1.10 K1.12

K1.62 K1.64

0.04 0.03

K1.00 K0.46 K0.46 K0.26

K1.16 K0.53 K0.53 K0.30

0.23 0.06 0.10 0.21

K0.97 K0.41 K0.42 K0.19

K1.42 K0.60 K0.62 K0.28

0.27 0.11 0.15 0.38

Note: Effects are estimated in standard regression analysis before and after adjustment for exposure error. The p-value is obtained from the naive test ignoring exposure error. Furthermore, the covariate Town7 was included in addition to the confounders identified by Grandjean et al. (4). a Higher scores indicate an adverse effect.

382

Budtz-Jørgensen

association to the exposure may lead to further attenuation, even if these variables have no real effect on the outcome. In fact, the attenuation seen in the mercury effects after adjustment for Town7 may, at least in part, be a result of exposure error and high correlation between the exposure and this potential confounder (5). STRUCTURAL EQUATION MODELS In structural equation models (6,7), the general aim is to describe the conditional distribution of the dependent variables (Yi1,.,Yip) given the covariates (Zi1,.Ziq) of subject i, iZ1,.,n. This is achieved by specifying a measurement model and a structural model. In the measurement model, the dependent variables are assumed to be linearly related to a set of latent variables (hi1,.,him) and a random error: Yi1 Z v1 C l11 ,hi1 C . C l1m ,him C 3i1 , , ,

(1)

Yip Z vp C lp1 ,hi1 C . C lpm ,him C 3ip Typically, the number of latent variables (m) is much smaller than the number of observed variables (p). Furthermore, each of the dependent variables is usually considered a measure only of one latent variable. In this case, only one of the so-called “factor loadings” (lij) in each row will be different from 0. The measurement errors (3i1,.,3ip) follow a normal distribution with a mean of zero and a variance matrix of U. The structural part of the model describes how the latent variables are related to each other and to the covariates: hi1 Z a1 C Sjs1 b1j ,hij C Sj g1j ,Zij C zi1 , , ,

(2)

him Z am C Sjsm bmj ,hij C Sj gmj ,Zij C zim Thus, each of the latent variables may depend on the covariates and other latent variables. Effects of latent variables are described by b-parameters, while g- parameters give the effects of the covariates. The residuals (zi1,.,zim) are assumed to be independent of the measurement errors (3i1,.,3ip), while following a normal distribution with a mean of zero and a variance matrix of J. Extensions The set of structural equation models has been extended in several ways that may be beneficial for applications in environmental epidemiology. First, the model assumes a multivariate normal distribution for the dependent variables given the covariates. Even after transformations, this requirement is not likely to be satisfied when a large number outcomes are considered. This problem can addressed by using robust inferential techniques (8) or by applying more sophisticated structural equation models that also allow for categorical dependent variables (7,9). Second, in a so-called multiple group analysis, the parameters of the model can vary across a categorical group variable (7). This is

Environmental Risk Assessment

383

an important improvement, because it allows modeling of effect modification. For example, the effect of a latent exposure variable may depend on the sex of the subject. Lastly, the analysis is not necessarily restricted to complete cases. Under the assumption that data are missing at random (10), likelihood methods that exploit information from incomplete cases have been developed and made available in user friendly software (11).

ESTIMATION OF EXPOSURE MEASUREMENT ERROR This section describes how error variances in exposure variables can be estimated using structural equation models. This approach requires that more than one exposure variable are available. In the Faroese data, mercury concentrations were measured in cord blood (B-Hg) and maternal hair (H-Hg). After a logarithmic transformation, the relation between these variables is approximately linear (12). This observation leads to the following measurement model for the joint distribution of the exposure variables: logðB-HgÞ Z logðh1 Þ C 3BKHg

(3)

logðH-HgÞ Z vHKHg C lHKHg ,logðh1 Þ C 3HKHg

Thus, the observed mercury concentrations depend on the true exposure represented by the latent variable log (h1)and a random measurement error (3B-Hg or 3H-Hg). Scales of latent variables cannot be estimated from data, but must be defined before the analysis. Here, the true exposure is expressed on the scale of the cord blood concentrations, in the sense that a one unit increase in log (h1), on average, leads to a one unit increase in log (B-Hg). The factor loading lH-Hg allows for the fact that a one unit increase in log (B-Hg) may not correspond to a one unit increase in log (H-Hg). Similarly, the intercept (vH-Hg) allows for differences in means. The parameters of main interest here are the variances of the error terms 3B-Hg and 3H-Hg. In the structural part of the model, the mother’s average number of pilot whale dinners per month during pregnancy (Whale) was included as a predictor of the true exposure: logðh1 Þ Z a C g,logðWhale C 1Þ C z1

(4)

Here z1 is a normally distributed residual term with a mean of zero. Inclusion of whale meat consumption is important because it allows identification of the main parameters. In the model (3), based only on the two biomakers, the number of free parameters exceeds the number of sufficient statistics and measurement error variances cannot be estimated (5). The estimated standard deviations of the measurement error terms were 0.32 and 0.44 in the log-transformed mercury concentrations in cord blood and maternal hair, respectively (12). These variances are on different scales and cannot be directly compared. Instead, meaningful comparisons of the biomarker precisions can be based on the estimated correlation of each biomarker to the true exposure (Table 2). The analysis Table 2 Factor Loadings, Standard Deviations of Error Terms and Correlations to the True Exposure for Mercury Concentrations in Cord Blood and Maternal Hair Indicator

Loading (l)

Error standard deviation

Correlation to truth

log(B-Hg) log(H-Hg)

1 0.84

0.32 0.44

0.93 0.84

384

Budtz-Jørgensen

showed that the cord blood concentration was more precise than the hair concentration (pZ0.007). This result is in agreement with prior expectations and the results of Table 1, showing that the cord blood concentration was generally a stronger predictor of childhood test performance. The current comparison of biomarker precisions is independent of outcome information, but it rests on the assumptions of the structural equation model. Here we assumed that the two mercury concentrations and maternal pilot whale consumption are related to the same underlying exposure variable. Total measurement errors in the biomarkers are identified by assuming that the three variables are independent given the true exposure. The robustness of the results were assessed in sensitivity analysis. First, we fitted models with fixed degrees of correlation between measurement errors in the two biomarkers. For example, modest correlations could arise if the mercury load during the period (about 6 weeks pre-term) where the two biomarkers are both relevant deviated grossly form the true exposure. These analyses showed that the estimated measurement error variances increased as a function of the measurement error correlation, but the cord blood concentration remained the best indicator (5). In further analysis, we allowed for the possibility that blood and hair concentrations are measuring two different (but correlated) components of mercury exposure. In this alternative model, the hair concentration would measure its component with a smaller error than that given in Table 2, and it could even be a better measure of the hair component that the blood concentration. However, the empirical observation that the hair concentration is an inferior predictor of neurobehavioral deficits at 7 years must then be interpreted as an indication that an underlying hair component is of lesser toxicological relevance.

Adjusted Regression Coefficients Based on the estimated exposure imprecisions, the naive regression coefficients can be adjusted. Under the additive model of equation (3), the attenuation in the cord blood coefficient is given by the factor r Z ½varflogðB-HgÞjZgKvarð3BKHg Þ=varflogðB-HgÞjZg

(5)

Where ZZ(Z1,.,Zq)t is the vector of potential confounders. Equation (5) shows that, in studies where a higher degree of the exposure variation can be explained by the confounders, the results are more susceptible to measurement error problems. The conditional variance var[log(B-Hg)jZ] is estimated as the residual variance in the regression of log (B-Hg) on Z. Together with the measurement error variance, this value is then inserted into equation (5) to get an estimate of the attenuation factor. Attenuations of 0.865 and 0.681 were obtained for (log-transformed) mercury concentrations in cord blood and maternal hair, respectively. Consistent exposure effect estimates were derived by multiplying the naive coefficients by the inverse of the estimated attenuation factor (Table 1). Adjusted mercury effects are deattenuated, but at the same time the adjustment also increased the statistical uncertainty. In fact, the naive test of no exposure related effect is efficient in the regression models considered here (13). Thus, the exposure error adjustment does not increase the power to detect effects of the exposure. However, compared to the univariate approach taken in this section, power can be gained in structural equation models combining information from different outcome variables. This analysis will also address the multiple testing problem which clearly remains after the adjustment for exposure error.

Environmental Risk Assessment

385

EFFECT ESTIMATION IN STRUCTURAL EQUATION MODELS The effects of prenatal mercury exposure on childhood neuropsychological test performance has been estimated in structural equation models that include both exposure and outcome variables (14,15). The structural equation approach is most powerful when observed variables can be combined into fewer latent variables. Therefore, the neuropsychological response variables were sorted into major nervous system functions. Attention was restricted to outcomes in a group consisting of motor functions and another group encompassing cognitive function with a verbal component. A path diagram, illustrating the initial model, is given in Fig. 1. Here, observed variables are in boxes, while latent variables are in ovals. The model for the exposure variables, described in the previous section, is located in the left hand side of the diagram. Thus, mercury concentrations in cord blood and maternal hair are considered error prone indicators of the underlying true exposure, which is affected by maternal whale meat consumption. The neuropsychological test scores are incorporated to the right in the diagram. Thus, it was assumed that three NES Finger Tapping scores (FT1, FT2, FT3) and the score on NES Hand Eye Coordination (HEC) were manifestations of an underlying motor function (h3), while the two scores on the Boston Naming Test (BNT1, BNT2), the four scores on the California Verbal Learning Test (CVLT1,., CVLT4), and the Digit Spans (DS) test score all reflect a verbally-mediated function (h4). Measurement errors (3’s) in different outcomes were assumed to be independent. In the structural part of the model, true mercury exposure is hypothesized to affect the two latent outcome functions, after adjustment for effects of covariates: h3 Z a3 C b31 ,logðh1 Þ C g31 ,Z1 C/C g3q ,Zq C 23

(6)

h4 Z a4 C b41 ,logðh1 Þ C g41 ,Z1 C/C g4q ,Zq C 24

HEC

∋

FT1

∋

FT1

FT2

∋

FT2

FT3

∋

FT3

BNT1

∋

BNT2

∋

Confounders

HEC

h3 z3

∋

z1 B–Hg

log(B-Hg)

z4

∋

h4

BNT1 BNT2

DS

DS

CVLT1

CVLT1

CVLT2

CVLT2

CVLT3

CVLT3

CVLT4

∋

log(h1)

∋

log(H-Hg)

∋

H–Hg

∋

log(Whale +1)

∋

z2

CVLT4

Figure 1 Path diagram for the association between mercury exposure indicators and childhood neurobehavioral test scores.

386

Budtz-Jørgensen

Here b31 and b41 denote the effect of mercury exposure on motor function and the verbally-mediated function, respectively. These parameters are of main interest in the analysis. Potential confounders of the relation between prenatal mercury exposure and childhood test performance were included in the model as covariates. Therefore, these variables were assumed to be measured without error. The confounders are allowed to be correlated with the true exposure, and are assumed to affect the latent response functions in the child. Model Modifications The initial model fitted the data poorly (14). This was, in part, because the proposed correlation structure of the neurobehavioral scores was too simple. The model assumed that a child’s test scores are independent given the latent ability levels. However, tests reflecting the same latent function can be collected further into subgroups in which the tasks resemble each other more. Thus, three of the motor scores are from the finger tapping task (FT), two of the verbal scores are from the BNT and four others are from the CVLT. Scores in the same subgroup are likely to be correlated given the latent variable they are assumed to measure. Indicators with this property are sometimes described as being locally dependent. To take this into account, three new latent variables (h5, h6, h7) were included. Local dependence between the FT-scores were then modeled by letting h5 affect these scores only. Similarly, the BNT scores and the CVLT scores were assumed to depend on h6 and h7, respectively. In the initial model, covariates were assumed to affect the test scores only indirectly through the latent response functions. This restricts the effects of covariates on test scores reflecting the same latent response. Thus, a given covariate is assumed to affect different outcomes in the same way except for scale differences. For example, the ratio between expected increases in two verbal scores caused by a one unit increase in a covariate must be equal to the ratio of the factor loadings of these outcomes on the latent verbal function. A more flexible model for the covariate effects was obtained by allowing (a few) covariates to affect the outcomes also directly. Thus, the measurement model of equation (1) was extended to include effects of covariates on the observed dependent variables, i.e., YikZnkCSjlkj$hijCSjkkj$ZijC3ik. In a backward elimination procedure, six significant direct effect parameters were identified. Four of those allowed the test scores in boys and girls to vary (disproportionately) across indicators of the same latent variable (14). Results of the Structural Equation Analysis The extended model fitted the data nicely (14). Despite the strong improvement in model fit, estimated values for the main parameters changed only slightly as a result of the model modifications. Table 3 gives the estimated mercury effects in the final model. A highly

Table 3 Estimated Effect of 10-Fold Increase in Exposure on Motor Function and Verbally Mediated Function. Motor Function Is Expressed on the Scale of FT1, Verbal Function Is Expressed on the Scale of BNT2, While True Mercury Exposure Is on the B-Hg-Scale

Motor Function Verbal Function

b^

c s:e:

P

K1.03 K1.62

0.49 0.52

0.034 0.002

Environmental Risk Assessment

387

significant mercury effect was seen for the verbal function, while the effect on the motor function was weaker, but still significant at the 5% level. Thus, compared to the regression results this analysis gives a much simpler representation of the overall trends in complex Faroese data. The scales of the two latent neurobehavioral functions were defined by fixing the factor loadings of the responses FT1 and BNT2 at one. Therefore, it is estimated that the effect of a 10-fold increase in the true mercury exposure reduces the child’s motor ability by approximately one finger-tap with the preferred hand. The effect on the verbal function corresponds to a loss of 1.6 points on the cued BNT. These results are in agreement with the naive regression coefficients of Table 1. Thus, the structural equation results are similar to the strongest mercury coefficients obtained for individual scores of motor function and verbal skills, respectively. Finally, it is interesting to note that inclusion of the neurobehavioral outcomes and the confounders did not affect the measurement model for the exposures. The cord blood concentration still appeared to be more precise.

THE BENCHMARK APPROACH TO SETTING SAFETY STANDARDS Having identified the Faroese study as the critical study for calculation of an exposure limit for methylmercury, the NAS committee further recommended that this limit should be derived based on the so-called benchmark approach (2). This methodology was first developed for standardized experimental dichotomous (normal/abnormal) responses (16). The benchmark dose (BMD) is the dose that increases the risk of an abnormal response from P0 in unexposed subjects to P0CBMR for subjects at the BMD. BMR is a constant to be specified by the regulatory authorities. The statistical uncertainty is taken into account by basing the exposure limit calculation on the 95% lower (one-sided) confidence limit of the BMD, the so-called BMDL. Buldtz-Jørgensen et al. (17) described how the BMD approach can be applied to epidemiological data. Unlike experimental studies, observational studies usually do not include a control group completely free of exposure, the exposures are distributed continuously, and the response variable is influenced by confounders in addition to the exposure of interest. Therefore, let Y(X, Z) denote the response of a subject with exposure X and confounder values Z Z (Z1,.,Zq)t. We assume that the dependence of the response on the exposure and the confounders can be modeled using a regression model YðX; ZÞ Z b0 C bx gðXÞ C btz Z C 3;

(7)

where 3 w N(0,s2), and the dose-response function g is known and increasing with g(0) Z 0. Without loss of generality, we consider large responses to be disadvantageous. An adverse exposure effect therefore corresponds to a positive value of bx. In this general set-up, the definition of the BMD is less obvious. Still the main idea is that exposures corresponding to the BMD should increase the risk of an abnormal response by a prespecified BMR. Thus, the level of an abnormal response t0 must be specified. If t0 is the same for all subjects then the unexposed risk P{Y (0,Z) O t0} will depend on the confounders, and so will the BMD, which is not practical. This problem can be avoided by letting t0 depend on the confounders such that all unexposed subjects have the same prespecified risk P0 (i.e., P {Y (0,Z) Ot0} Z P0 for all Z). In agreement with the original definition for dichotomous data, the BMD is the dose that the increases the risk by BMR,

388

Budtz-Jørgensen

i.e., P{Y(BMD, Z) Ot0} Z P0CBMR. In the regression model of equation (7), this dose is independent of the confounders, and given by

BMD Z

8 > > > > <

if b x % 0

N 0

1

Us > > gK1 @ A > > : bx

(8) if b x O 0

where UZFK1(1KP0)KFK1(1KP0KBMR) (17). Often, P0 is fixed at 5%, while the BMR is either 5% or 10% (2). Furthermore, Budtz-Jørgensen et al. (17) derived a closed form approximation to the lower 95% confidence limit 8 > > ) > c b^ x Þ 1 C ðt2 Ku2 Þ=2df : b^ x C u95 SEð

c b^ x Þ%Ku95 if t Z b^ x =SEð c b^ x ÞOKu95 if t Z b^ x =SEð

95

(9) where u95z1.645 is the 95th percentile in the standard normal distribution, df is the c b^ x Þ is the estimated standard deviation of b^ x : number of degrees of freedom, and SEð

STANDARD APPLICATION OF THE BENCHMARK APPROACH In applications of the benchmark method, the choice of dose-response function g is critical. Previous analyses in similar data [see for instance (18)] used the so-called K-power model. This model consists of power functions g(x)ZxK, where the power K is to be estimated under the restriction K R 1. Thus, this model includes only convex functions. As the dose-effect could also be concave within some doseffi ranges, we also considered a pffiffiffiffiffiffiffiffiffiffi linear model [g(x)Zx], a square root model ½gðxÞZ xC 1 K1 and a logarithmic model [g(x)Zlog(xC1)]. Despite the fact that the models fitted the data almost equally well, a substantial variation in BMD results across the models was seen (19). The logarithmic model yielded the lowest BMDLs, while the linear model and the K-power model provided the highest. The strong model dependence of the BMDLs may seem in conflict with the concept put forward by Crump (16), that by estimating risks at moderate levels, such as 5% or 10%, the BMD should be relatively robust to model specification. However, as explained by Bødtz-Jørgensen et al. (17), when no data from unexposed controls are available, this is not true, as the method depends on extrapolations to zero dose. In the calculation of an exposure limit, the NAS committee ruled out the square root and the logarithmic dose-response functions as being biologically implausible (2). Furthermore, P0 and the BMR were fixed at 5%, so that exposure to the BMD will double the unexposed risk (i.e., from 5% to 10%). Even with these definitions, the BMDLs varied strongly between the Faroese outcomes (19). In agreement with the precautionary principle, the committee decided to disregard multiple comparisons concerns and based the final calculation on the outcome with the strongest mercury effect (i.e., the cued BNT score). This procedure yielded a BMDL of 58 mg/l in cord blood.

Environmental Risk Assessment

389

SOPHISTICATED BENCHMARK ANALYSIS The described benchmark calculation was based on regression analysis and it therefore suffers from the shortcomings of this method. The result depends on the outcome considered, the analysis is not efficient (important when calculating BMDL), and exposure error is ignored. In this section, it is described how these problems can be solved. Adjustment for Exposure Error First, we consider the effect of ignoring exposure error in the benchmark calculation. In the model (7), the BMD is given by gK1(Us/bx). The naive analysis (ignoring imprecision) underestimates bx, while the opposite is true for s (1). Thus, the BMD is overestimated. The BMDL also decreases as a function of bx and increases as a function of s (17). Therefore, failure to adjust for measurement error will lead to upward bias in the BMDL and, as a consequence, calculated exposure limits may not be as safe as intended. Based on the imprecisions estimated in the section, Estimation of exposure measurement error Budtz-Jørgensen et al. (20) calculated measurement error adjusted BMDLs. This was done using maximum likelihood estimation in a so-called structural approach (1). The likelihood function is the conditional density f(YjW,Z) of the response (Y), given the observed exposure (W) and the covariates (Z). This function determined by integrating over the true exposure ð ð f ðYjW; ZÞ Z f ðYjX; W; ZÞf ðXjW; ZÞdX Z c f ðYjX; ZÞf ðWjX; ZÞf ðXjZÞdX; (10) where we exploited that Y and W are independent given X, Z and c is a constant. In the latter expression, the distribution of Y given X and Z [f(YjX, Z)] is defined by the doseresponse model ½Y Z b0 C bx gðXÞC btz Z C 3: The distribution of W given X and Z [f(WjX,Z)] is given by the error model. Here we assumed that the additive error model was appropriate after transformation with a known monotone function, i.e., h(W)Zh(X)CU. For example, this model was applied in the section Estimation of exposure measurement error where we used an additive error model for log-transformed mercury concentrations. Calculation of the likelihood function further requires specification of the distribution of f(XjZ), the true exposure given the covariates. For this we assumed that h(X) followed a normal distribution conditional on Z. The complexity of the analysis depends on the relation between h and the doseresponse function g. In the simple case where gZh, f(YjW,Z) is a normal distribution and a closed form expression for the adjusted BMDL was obtained as ( qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi) BMDL Z gK1 r

gðBMDLnaive Þ2 KU2 varðUÞ=r ;

(11)

where BMDLnaive is the unadjusted result and r is the attenuation factor (5). As expected, the adjustment decreases the BMDL. This decrease becomes stronger for lower values of r. Thus, in studies with limited exposure variation or high exposure imprecision, the naive analysis will be more biased. Closed form solutions are not available in the general case where gsh. In the Faroese data, the NAS committee recommended a linear response model [g(x)Zx], while data suggested that an additive error model was appropriate after a log-transformation of the mercury concentrations [h(x)Zlog(x)]. In this case, calculation of the likelihood function (10) involves numerical integration, which may be accomplished using the SAS

390

Budtz-Jørgensen

procedure NLMIXED. Based on the exposure imprecision estimated in Estimation of exposure measurement error, we obtained an adjusted BMDL of 45.5 mg/l for the BNT score. This result indicates that the (unadjusted) BMDL recommended by the NAScommittee was overestimated by about 27%. On the other hand, by relying solely on the most significant dose-response relation, some downward bias may have been introduced in the NAS approach. This is further explored in the next section. Adjustment for Measurement Error and Multiple Comparisons Imprecisions in the response variable will generally not lead to bias in the regression coefficients, but the BMD analysis is sensitive to this type of error. This effect can be illustrated by assuming an additive error in the relationship between the true response (Y) K1 ~ i.e. Y~ Z Y C U:p and the measured response ðYÞ; The true BMD ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi is g (Us/bx), while the K1 naive analysis estimates the larger value g ðU s2 C varðUÞ=bx Þ: Thus, failure to adjust for errors in the response will lead to further overestimation of the BMD. Neurobehavioral test scores in children are likely to have a large component of unexplained variation and may therefore produce strongly biased BMD results. The structural equation approach allows for measurement error both in the exposure and in the response. The structural part of the model estimates the relation between true exposure and true response. Thus, BMD calculations based on this relation avoid problems caused by measurement errors. In addition, multiple comparisons are not an issue and the BMD is more efficiently estimated. In a model similar to the one developed in the section Effect estimation in structural equation models, Budtz-Jørgensen et al. (21) calculated joint benchmark results for the verbal outcomes. This analysis yielded a BMDL below the lowest of the individual BMDL results. The BMDL was 40.7 mg/l, as expressed in the cord blood scale. Note that, the result of the joint analysis is even lower than the BMDL obtained after exposure error adjustment of the strongest multiple regression dose-response relationship. DISCUSSION Public health research must identify environmental risk factors and then determine safety standards to limit human exposures at acceptably safe levels. This important challenge requires investigators to use the most sensitive and accurate methods to determine exposure levels, outcome variables, and potential confounders. Still, problems with imprecision in the exposure assessment, missing data, and multiple endpoints cannot be avoided in observational studies. This paper has emphasized the importance of taking these problems into account in the risk assessment. Standard statistical methods, such as multiple regression, ignore many of the problematic features, and will lead to biased and inefficient results. Failure to adjust for measurement error in the exposures and in the outcomes may lead to underestimation of the risk, while multiple comparisons are likely to have opposite effect. Structural equation models were shown to be useful in the process of environmental risk assessment. These models consist of linear relations between observed and latent variables. This structure allows both estimation of error size in observed variables and subsequent adjustment of the dose-response relationship. In the analysis of the Faroese data, exposure errors were first estimated in a simple structural equation model using only exposure variables. Naive regression coefficients and benchmark doses were then adjusted, based on these degrees of imprecision. This approach required limited modeling, but the multiple comparisons problem was not addressed and this analysis was as

Environmental Risk Assessment

391

inefficient as standard regression analysis. A more powerful and parsimonious representation of the exposure effects was obtained by also viewing the neurobehavioral outcomes as reflections of latent variables. The results of this more sophisticated analysis were similar to the most significant regression results. In fact, the joint BMDL was somewhat lower than that obtained using standard calculations based on the BNT score. Thus, in situations where the structural equation approach has not been applied, risk assessment based on the most significant effect may not be overly protective. The definition of the BMD depends on the response level in unexposed subjects. Unfortunately, epidemiological data usually lack an unexposed control group, and extrapolations to zero are therefore necessary. As a consequence, the results of this approach have been shown to be sensitive to the choice of dose-response model (17). Despite this problem, BMD results are still considered the best basis for identification of exposure standards (2). This paper has pointed to further weaknesses in the benchmark approach. It was shown that if measurement error in the exposure or in the response is ignored, then the BMD and the BMDL may be overestimated. Thus, application of less precise study variables will lead to higher and less protective exposure limits. This bias is counter to the precautionary principle, which would require that weaker knowledge must lead to more stringent standards (22,24). Adjusted results can be obtained in structural equation analysis. In cases where this analysis is not conducted, standard benchmark calculations should be supplemented by sensitivity analysis exploring the robustness of the result to imprecisions in the study variables. However, in the longer term, methods that handle uncertainty more appropriately must be developed. One possibility may be to base the calculations on hockey stick models, which assume no effect of the exposure below an unknown threshold. In the presence of an exposure imprecision, the threshold is typically underestimated (23,24). Therefore, contrary to the benchmark approach, the naive analysis that ignores the measurement error will produce lower hockey stick thresholds and thereby more protective standards. Unfortunately, other sources of uncertainty are not dealt with in this precautionary way. Thus, a limited sample size or an imprecise response variable will bias the estimated threshold upward (24).

ACKNOWLEDGMENTS I am grateful to Dr. Pal Weihe for allowing me to use the data from the Faroese cohort study and to Drs. Philippe Grandjean and Niels Keiding for inspiring discussions. This study was supported by grants from the National Institute of Environmental Health Sciences (ES06112 and ES09797), the U.S. Environmental Protection Agency (9W-0262NAEX), the European Commission (Environment Research Programme), the Danish Medical Research Council, and the Danish Health Insurance Foundation. The contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of the NIEHS, NIH, or any other funding agency.

REFERENCES 1. Carroll RJ, Ruppert D, Stefanski LA. Measurement Error in Nonlinear Models. London: Chapman & Hall, 1995. 2. National Academy of Sciences (NAS). Toxicological Effects of Methylmercury. Washington: National Academy Press, 2000.

392

Budtz-Jørgensen

3. Grandjean P, Weihe P, Jørgensen PJ, Clarkson T, Cernichiari E, Viderø T. Impact of maternal seafood diet on fetal exposure to mercury selenium, and lead. Arch Environ Health 1992; 47:185–195. 4. Grandjean P, Weihe P, White RF, et al. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997; 19:417–428. 5. Budtz-Jørgensen E, Keiding N, Grandjean P, Weihe P, White RF. Consequences of exposure measurement error for confounder identification in environmental epidemiology. Stat Med 2003; 22:3089–3100. 6. Bollen KA. Structural Equations with Latent Variables. New York: John Wiley and Sons, 1989. 7. Muthen B. A general structural model with dichotomous, ordered categorical and continuous latent variable indicators. Psychometrika 1984; 49:115–132. 8. Satorra A, Bentler PM. Corrections to test statistics and standard errors in covariance structure analysis. In: von Eye A, Clogg CC, eds. Latent Variable Analysis: Applications to Developmental Research. Newbury Park: Sage Publications, 1994:399–419. 9. Sammel DP, Ryan L, Legler JM. Latent variable models for mixed and discrete and continuous outcomes. J R Stat Soc B 1997; 59:667–678. 10. Little RJA, Rubin DB. Statistical Analysis with Missing Data. New York: Wiley, 1987. 11. Muthe´n LK, Muthe´n B. Mplus. The comprehensive modeling program for applied researchers. User’s Guide. Los Angeles: Muthe´n & Muthe´n, 1998. 12. Budtz-Jørgensen E, Grandjean P, Jørgensen PJ, Weihe P, Keiding N. Association between mercury concentrations in blood and hair in methylmecury-exposed subjects at different ages. Environmental Research 2004; 95:385–393. 13. Tosteson T, Tsiatis A. The asymptotic relative efficiency of score tests in a generalized linear model with surrogate covariates. Biometrika 1998; 75:507–514. 14. Budtz-Jørgensen E, Keiding N, Grandjean P, Weihe P. Estimation of health effects of prenatal mercury exposure using structural equation models. Environ Health 2002; 1:2. 15. Budtz-Jørgensen E, Keiding N, Grandjean P, Weihe P, White RF. Statistical methods for the evaluation of health effects of prenatal mercury exposure. Environmetrics 2003; 13:105–120. 16. Crump K. A New Method for Determining Allowable Daily Intakes. Fundam Appl Toxicol 1984; 4:854–871. 17. Budtz-Jørgensen E, Keiding N, Grandjean P. Benchmark Dose Calculation from Epidemiological Data. Biometrics 2001; 57:698–706. 18. Crump K, Van Landingham C, Shamlaye C, et al. Benchmark concentrations for methylmercury obtained from the Seychelles Child Development Study. Environ Health Perspect 2000; 108:257–263. 19. Budtz-Jørgensen E, Keiding N, Grandjean P. Benchmark modeling of the Faroese methylmercury data. Final Report to U.S.EPA, 1999. 20. Budtz-Jørgensen E, Keiding N, Grandjean P. Effects of exposure imprecision on estimation of the benchmark dose. Risk Analysis 2004; 24:1689–1696. 21. Budtz-Jørgensen E, Keiding N, Grandjean P. Application of structural equation models for evaluating epidemiological data and for calculation of the benchmark dose. Proceedings of the ISI International Conference on Environmental Statistics and Health at Santiago de Compostela, July. 2003:183-194. 22. Barnett V, O’Hagan A. Setting Environmental Standards: The Statistical Approach to Handling Uncertainty and Variation. London: Chapman & Hall, 1997. 23. Kuchenhoff H, Carroll RJ. Segmented regression with errors in predictors: Semi-parametric and parametric methods. Statatistics in Medicine 1997; 16:169–188. 24. Keiding N, Budtz-Jørgensen E. The precautionary principle and statistical approaches to uncertainly. Eur J Oncol Lib 2003; 2:185–191.

19 Incorporating the Social-Ecological Perspective into Studies of Developmental Neurotoxicity Virginia A. Rauh Heilbrum Department for Population and Familiy Health, Mailman School of Public Health, Columbia University, New York, New York, U.S.A.

Frederica Perera Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, New York, U.S.A.

Jennifer F. Culhane Department of Obstetrics and Gynecology, Drexel University College of Medicine, Philadelphia, Pennsylvania, U.S.A.

INTRODUCTION The focus of much environmental health science is to identify and quantify associations between potentially toxic exposures and disease outcomes. Consistent with a medical model, the introduction of biomarkers has moved the field forward by validating the degree of individual exposure and improving the precision of effect size estimates. This, in turn, has lead to increased technological and programmatic responses to the reduction of environmental pollution, including some positive changes in public health policy. However, little attention has been paid by environmental health scientists to the social conditions underlying (and often determining) the distribution of such toxic exposures, and even less attention to the processes whereby some social conditions may alter susceptibility to the toxicants, or vice versa. This chapter is divided into two sections, each addressing a set of methodologic issues raised by the social-ecological perspective in environmental science. The first section describes the co-occurrence of chemical and social exposures, providing an epidemiological rationale for a contextual approach to risk assessment in studies of developmental neurotoxicity. Virtually all human exposure scenarios involve multiple substances, whereby some exposures occur against a background of others, each at different levels of concentration. This background, or context, includes not only other chemical and physical agents but also the many social factors with direct or indirect influences on health and development. The distributions of both social and chemical risks are disproportionate, such that disadvantaged populations are more likely to experience adverse social and chemical exposures, as compared to more advantaged groups and 393

394

Rauh et al.

communities (1,2). Furthermore, chemical and social exposures operate at many different scales of influence, posing methodological problems for assessing risk. Although environmental scientists have developed complex dispersion models to quantify concentrations of ambient chemical pollutants, for example, they have seldom attended to the spatial aspects of social contextual influences on health and development (3). Taken together, the distribution and covariation of social and chemical hazards provide a complex epidemiologic backdrop for risk assessment. The second part of the chapter addresses the different types of possible relationships between chemical and social exposures, and how these relationships affect the calculation of risk estimates. On the one hand, confounding as a result of unmeasured or inadequately measured co-exposures can result in biased estimates of neurotoxic effects. On the other hand, chemical and social exposures may interact with each other, leading to risk reduction, risk enhancement, or other modulation of susceptibility at the individual level, the population level, or both. The presence of interaction may also depend on length of exposure and developmental stage of the individual, so that interactions with time (or age) will need to be considered. These various contingencies will have profound effects on individual differences in risk as well as population health disparities. The framework underlying this chapter is not original. In addition to the many individuals who have proposed social-contextual approaches to understanding environmental health (4), a number of excellent reports dealing with methods for studying environmental health risks have appeared in the last few years. These include reports from the Centers for Disease Control [Racial and Ethnic Approaches to Community Health (REACH) 2010: Addressing Disparities in Health, 2005] (5), the Environmental Protection Agency (Framework for Cumulative Risk Assessment, May, 2003) (6), the Committee on Integrating the Science of Early Childhood Development (From Neurons to Neighborhoods, 2000) (7), the Institute of Medicine (The Role of Environmental Hazards in Premature Birth: Workshop Summary, 2003) (8), the National Scientific Council on the Developing Child (Young Children Develop in an Environment of Relationships, 2004) (9) and the National Environmental Justice Advisory Council (Ensuring Risk Reduction in Communities with Multiple Stressors: Environmental Justice and Cumulative Risks/ Impact, 2004) (10). Building on these ideas, the present chapter combines current thinking about the joint distributions of chemical and social risks, methods for testing different scales of influence in risk assessment research, the meaning of statistical interaction between chemical and social risks, and possible biological mechanisms underlying these interactions.

THE CONFLUENCE OF SOCIAL AND CHEMICAL RISKS As multiple exposures to environmental toxicants are increasingly documented in urban (11) as well as rural settings (12), environmental scientists have been making the shift from single hazard risk assessment to cumulative risk assessment (13–15). In practice, the accumulation of toxic exposures has been shown to be a more powerful determinant of child health and development than any specific exposure (16–18). As advocated by the EPA (Framework for Cumulative Risk Assessment, 2003) (6), risk assessment should include both social and chemical hazards, and should be done with the focus on a population or a community, rather than studying the risk of one exposure on the individual. The convergence of adverse social and chemical exposures poses particular methodologic problems for etiological studies of neurodevelopment, because both kinds of exposures can be toxic to children and families, independently and in combination.

Incorporating the Social-Ecological Perspective

395

In addition, each correlate or component of social adversity may carry its own developmental risk, aside from that portion of social-related risk that can be attributed to chemical contaminants. For example, the effects of poverty may be transmitted or mediated by dietary practices, substance abuse, health service utilization, the quality of the home environment, and/or psychosocial stress (19–21). Taken together, these multiple risk factors account for considerable variability in child neurobehavioral functioning (22). There is a long history of studies documenting associations between social conditions and health outcomes. These social conditions include income (23–25), poverty (26,27), education (28), marital status and family structure (29,30), social support/stress/resources (31–36), inequality (37,38), and community quality (39) to name some relevant areas of research. Taken together, the effects of poverty-related factors on child health and neurodevelopment are frequently stronger than the effects of the chemical neurotoxicant(s) under study (40). Many chemical toxicant exposures are concentrated among low-income populations and in socio-economically disadvantaged communities (16,17,41). The convergence of toxic exposures and social adversities represents a type of environmental injustice, in which the greatest toxic burden is carried by those who can least afford the adverse health consequences (42). For example, exposures to environmental tobacco smoke (ETS) and polycyclic aromatic hydrocarbons (PAH) are high among low-income, urban, and minority populations, both because of the location of outdoor pollution sources and current socio-demographic patterns of tobacco use (43–48). Elevated cotinine levels have been reported in 70–80% of inner city children (49,50). In addition, many low-income urban areas are home to industrial and waste-related land uses and lack environmental amenities as a result of land use policy that may have been based on exclusionary zoning practices and residential segregation (51–53). For example, there is a lack of parks and open spaces in historically segregated communities like Northern Manhattan (54). During the major era of parks construction in New York City, only one of 255 playgrounds was built in Harlem (55). In the United States, 60% of Hispanics and 50% of African Americans, compared to 33% of Caucasians, live in areas failing to meet two or more of the national ambient air quality standards (48,56). These same minorities are also more likely than others to experience poverty and the many adversities that accompany poverty, including substandard housing, poor nutrition, and inadequate health care (57,58). The specific environmental exposures created and mediated by aspects of the built environment include air pollution such as fine particulate matter, other criteria pollutants (namely nitrogen oxides, sulfur dioxide, carbon monoxide), and construction debris and air toxics (e.g., volatile organic chemicals). Other environmental exposures linked with the built environment include noise and stress (59). A small but growing body of knowledge has looked specifically at stressors related to the physical environment, including personal safety, crime, and violence in the neighborhood (60). In addition, the outdoor environment is inextricably linked to the indoor environment, as pollutants, such as PM2.5, mercury, and other air toxics, can enter and accumulate in the home, especially in poor quality housing without proper ventilation. Poor housing is an example of a poverty-related condition that can be a source of toxic exposures and a pre-eminent stressor in the lives of many low-income, minority families (61,62). In New York City, for example, it is estimated that 74% of the children in Central Harlem and 60% in Washington Heights (both low income communities) live in fair to poor quality housing, as compared to 38% city-wide (63). The construct of housing includes physical/chemical characteristics (e.g., location, density, building height, maintenance, air quality, sanitation, pests, pesticides, etc.), social aspects (e.g., crowding,

396

Rauh et al.

threats to safety, segregation, noise, social networks, cost), and psychological components (interpersonal conflict, dislocation). The importance of adequate housing for the maintenance of health and positive development has long been a topic of scientific and public health policy discussion (64), yet measures of housing adequacy have not been well-used in studies of environmental toxicants (61). Using a measure of residential housing disrepair (e.g. holes in ceilings or walls, peeling paint, water damage) developed by Rauh et al. (65), Whyatt et al. (66) found a significant association between degree of disrepair and pesticide exposure, after controlling for ethnicity and neighborhood of residence. Specifically, substandard housing conditions (a major concern in this population) were found to be positively associated with pest infestations and high use of pest control measures. Furthermore, disrepair was significantly associated with demoralization (a measure of nonspecific psychological distress), suggesting that housing problems are psychologically stressful for urban families. At the population level, neurodevelopmental risk may also arise as a result of the inadequate supply of affordable housing for low-income persons and the increasing spatial segregation of housing by income, race/ethnicity, and social class into unsafe neighborhoods (51). Increasing concentration of poverty leads to housing disinvestments, physical deterioration, and the disintegration of informal social networks (67). When affordable housing is unavailable, families are forced to divert the resources needed for food, clothing, and other necessities to housing costs. Residential instability results, as families are forced to move frequently, move in with other families in overcrowded, unsanitary conditions, or experience periods of homelessness. Residential instability has been identified as one of the most important predictors of individual and community health, and constitutes a threat to the development of informal local friendship networks, kinship bonds, and local organizational ties (68). The psychological stress associated with homelessness is extreme, as families are deprived of a home as an object of attachment and a source of identity (69). Finally, social or economic deprivation need not be absolute in order to affect health and development. Inequality or relative deprivation may be at least as important an influence on child health and developmental outcomes as absolute deprivation (70–72). One argument views relative poverty as a form of social exclusion, and suggests that the racial discrimination underlying this exclusion (including residential segregation) contributes directly to general health (73) as well as child health outcomes. Indeed, there is a substantial literature on the negative health effects of perceived discrimination (74), but it is not yet clear whether these risks operate primarily at the level of individual experiences, or whether additional risk is conferred by exposure to institutionalized racism (75,76). Specific aspects of the wider social environment, including the degree of residential segregation (57), have been demonstrated to be more closely associated with neurodevelopmental outcomes than racial group membership at the individual level (77). One possible explanation for this finding that relative disadvantage is associated with poor health relates to the psychosocial stress experienced by those who have less advantage or power in the social structure as compared to others. Although inequality is greatest in situations where extreme poverty is found, there is evidence to show that the inequality itself confers risk independent of the concentration of poverty (78). Furthermore, because of the association between low income and pollution, areas with greater income equality may also experience greater exposure inequality. Geronimus has suggested that individuals become worn by the years of disadvantage and stress that arise from relative deprivation, and has termed the erosion of health associated with disadvantage “weathering” (79). A critical point is that we are not necessarily talking about extreme conditions, but rather the chronic stresses of overcrowding, inadequate garbage

Incorporating the Social-Ecological Perspective

397

removal, housing location near busy transportation routes, poor ventilation, etc.— conditions that are part of the everyday lives of the residents of many urban communities.

Methodologic Issues in Modeling Social Context The inclusion of social conditions in models of cumulative risk depends upon our ability to validly measure the various components of social context at appropriate scales of influence. Social context is inherently multi-level, comprising both micro and macro social conditions, and, it can be argued, everything in between. In their simplest form, multi-level studies typically include assessments at the individual and community-level, using some standard administrative unit to define community (e.g., health area, zip code, census tract, block, etc.). For example, exposure to poverty or substandard housing may be measured at the individual level (personal income, number of homeless episodes) and the community level (average income in the census tract, amount of concentrated poverty, proportion of imminently dangerous buildings, etc.). Similarly, exposure to chemical toxicants can be measured at the individual biomarker level or at the area level (e.g., concentration of pollutants in the indoor or outdoor air). Advances in statistical techniques that facilitate the modeling of multi-level influences and the growing interest in the use of geographic information systems have made analyses of community- and regional-level variation more feasible (80–83). A recent review of articles published prior to 1998 of the effect of local area social characteristics on various individual health outcomes in developed countries, controlling for individual socioeconomic status, found that all but two of the 25 reviewed studies reported a statistically significant association between at least one measure of the social environment and a health outcome, after adjusting for individual level socioeconomic status, despite heterogeneity in study designs, substitution of local area measures for neighborhood measures, and probable measurement error (84). Although multi-level studies in environmental health science are still relatively rare compared to individual-level or ecologicallevel investigations, the results of these studies nevertheless point to the potential importance of residential context on health. Multi-level analysis has the capacity to simultaneously assess the effects of individual-level and group-level exposures on individual outcomes (85). Such studies can address the question of whether local area characteristics have a measurable effect on outcome over and above individual exposures, or whether the apparent associations between aggregate measures and outcomes simply reflect individual-level characteristics of area residents. For example, does the mean income level in some defined community have an effect on individual outcome beyond the effect of individual income, and is the contextual effect independent of a given individual’s income (78). If individual and contextual factors both influence neurodevelopmental outcomes, then models that exclude one or the other set of risk factors are likely to be poorly specified and lead to misinterpretation of the effects of both individual and contextual-level factors. At the ecological level, the high correlation among various community characteristics poses problems in estimating the effects of distinct community characteristics. One way to get around this problem is to develop indices of related community-level constructs, but such indices may obscure the role of their distinct components. This could underestimate the association between a chemical exposure of interest and neurobehavioral outcome, if part of the chemical effect is carried by the association between the community quality index and the neurobehavior. Bellinger suggests using more differentiated measures of complex social constructs in order to

398

Rauh et al.

control for only those aspects of community quality that do not reflect exposure opportunities (86). Despite progress in this area, problems persist with respect to the conceptualization of and the distinctions between micro- and macro-level social phenomena. Link and Phelan have proposed the notion of “fundamental social causes” of disease (87) to explain how resources (such as knowledge, power, money, prestige, and social connections) are linked to disease outcomes through multiple (often shifting) risk-factor mechanisms. The notion is that inequalities in health are a function of persistent social inequalities, regardless of the intervening individual-level exposures, many of which can be shown to vary over time. To date, few studies have managed to successfully differentiate macro from micro factors, especially with respect to statistical analysis (80,88). So-called mixed models and multi-level approaches are now being employed in epidemiological studies, but further refinement is needed. In fact, it seems obvious that the classification of “levels” into two or even three categories of influence is somewhat crude, since the definition of the higher level “units” and the borders between the various levels are frequently unclear. One rationale for inclusion of multi-level assessment in environmental health studies is that the main effects operating at the individual level cannot be adequately understood without reference to group-level data. As will be discussed in the second section of this chapter, multi-level modeling permits the testing of cross-level questions in which the impact of individual exposure to toxicants may depend upon or be conditioned by community-level conditions or social processes. What Are the Components of Social Context that Might Affect Health Outcomes? Robert (89) classifies community-level factors into three distinct, yet related components: social environment, service environment, and physical environment. In fact, most studies that discuss why social context should matter for health outcomes focus on similar set of factors, all of which fall under these three general categories. Community social environment typically refers to social relationships, level of neighborhood cohesion or disorganization, norms of reciprocity, civic participation, level of crime, and related attributes. These characteristics are hypothesized to influence health and development through a number of potential pathways, including social support, coping strategies, and exposure to chronic stress (90–95). Others suggest that social characteristics of neighborhoods, perhaps through shared cultural norms and values, may influence health behaviors (89,93–97). Neighborhood social cohesion and residential stability in turn can affect the overall quality of neighborhoods more generally. Neighborhood social context may be related to the degree of political organization, influencing residents’ ability to demand public services, such as sanitation, fire, police, and emergency services (51,89,92,98). Poor quality of such services in turn may have a direct impact on child development by increasing exposure to stressful circumstances and by making residents more vulnerable to the effects of intentional and unintentional injuries. Availability and quality of other goods and services, such as grocery stores, recreational opportunities, health care facilities, and retail stores, are also likely to vary by neighborhood characteristics with socially and economically disadvantaged communities having fewer services than wealthier areas (94,99–102). One study of the distribution of food stores found significantly fewer (three to four times) supermarkets in poor and black communities as compared to more affluent white communities (54). The quality of the physical environment, including exposure to toxicants, noise and air pollution, and quality of the housing stock and public space, can

Incorporating the Social-Ecological Perspective

399

also have direct effects on individuals (90,93). Finally, residential location can also be related to availability of job and educational opportunities and thus influence health through family income and attained social status (89). Neighborhood context may be particularly important in explaining observed racial group differences in levels of toxic exposures. Racism and residential segregation has meant that black and white women, even with the same individual characteristics, are not exposed to equivalent everyday living conditions or access to resources. For example, African Americans are more likely than whites to live in neighborhoods with fewer highquality services, substandard housing, high rates of crime and violence, inadequate recreation areas, and greater numbers of “toxic” sites, such as diesel bus depots, waster transfer stations, and incinerators (51,103,104).

How Should Social Contextual Units Be Defined? Most studies that examine area-level variation or the effects of community context on health outcomes have used administrative or political boundaries to characterize neighborhoods and communities. In the United States, aggregate-level studies typically investigate area-level variation in health indicators at the level of the state (71,105), county, or metropolitan areas and central cities (57,106). Most, although not all, studies that integrate both individual and community-level characteristics in the analysis use U.S. census tracts as the neighborhood unit (90,92,97,100,107). Census block groups and alternative neighborhood boundaries are less frequently employed (94,108). Similarly, in European studies, political or administrative boundaries are typically used to define communities and neighborhoods (84). The choice of these units appears to be driven more by data availability and convenience than the fact that these areas are thought to represent neighborhoods as experienced by their residents. If geographic areas do not correspond to the actual geographical distribution of the causal factors linking social environments to health, these neighborhood units provide only a rough ecological profile and may not adequately capture the potential effects of community context on outcomes of interest (80,84). Surprisingly few studies have examined whether the way in which the neighborhood is defined matters. Elo et al. show that the way in which the neighborhood is defined (e.g., as a census block group, a tract, or a larger unit corresponding to two distinct Philadelphia neighborhood configurations) influences the extent to which racial differences in birth weight can be explained by neighborhood context (108). Using fixed effects estimates, block groups explain more of racial differences in birthweight than either census tracts or the alternative neighborhood configurations. Reijneveld et al. examined the same issue in Amsterdam using three different neighborhood definitions (neighborhoods, postal service code areas, and boroughs) to examine the effect of deprivation on adult health outcomes (109). The authors concluded that although the clustering of poor health was most pronounced at the neighborhood level, the choice of geographic classification had no effect on the magnitude of the deprivation effect. Despite their convenience, political and administrative units may not be the most appropriate way to delineate neighborhood boundaries. Furthermore, it may well be that different types of contextual influences operate at different levels of aggregation. Larger catchment areas, for example, may be appropriate for measuring availability of goods and services, such as health care facilities, recreational opportunities, and grocery stores, while smaller geographic units may be more appropriate for assessing the quality of housing stock, crime, and social characteristics of neighborhoods. In fact, changing the size of the

400

Rauh et al.

aggregation area can actually alter study results by masking heterogeneity within the area units. Environmental health studies that pay explicit attention to these issues are needed to advance our understanding of contextual influences on health outcomes, over and above the toxic contribution to risk. There are at least two types of data that can be used to characterize community context. Most commonly, aggregate-level variables are based on characteristics of individuals residing in the area, and are sometimes called derived variables. Secondly, structural or integral variables, such as the availability of goods and services, can be used. These variables do not have equivalent measures at the individual-level (80,84). Because of the availability of census and administrative data on individuals, most multi-level studies include only aggregate-level variables derived from individual-level data. These data represent means, medians, and distributions of individual-level characteristics as household income, poverty status, educational attainment, occupation, etc. Although such group level characteristics can influence social environments, social norms, health behaviors, and characteristics of neighborhood-based social networks, these characteristics may also serve as proxies for structural or integral variables such as service availability, housing quality, etc. (80,110). A growing number of sociological and public health studies have included these latter types of community-level measures (102,107,111,112). Regardless of how neighborhoods are defined, multilevel research has helped to draw attention to the role of social structures in health, particularly with respect to the problem of persistent health disparities. However, as recently reviewed by Oakes, very few such studies have attended to the question of causal inference, or recognized that neighborhood effects may not be truly independent effects (113). Oakes suggests that future multi-level studies make use of social experimental designs to better understand the underlying causal processes and to use this understanding to design interventions that more effectively reduce risk and improve public health. Given the complex relationships between toxic exposures and social conditions, this type of methodology might be very productively applied in the field of environmental health science.

Analytic Implications of Co-exposure: Covariate or Confounder? Control for social covariates in studies of environmental risk is important because it may reduce error or “noise” in the model, but not necessarily because such covariates actually bias (or confound) the association between the chemical exposure and the outcome of interest. If confounding does occur, adjustment for those social factors corrects a bias when estimating the chemical exposure-disease relationship. Inclusion and accurate measurement of all possible social-contextual factors become very important in constructing models of risk assessment when confounding is suspected. Evidence from the lead literature in particular argues for the careful assessment of social and behavioral covariates in studies of neurodevelopment, and demonstrates that social covariates can be stronger predictors of neurodevelopment than the toxicant of interest. For example, Wasserman et al. (40) found that the quality of the HOME environment; maternal age, intelligence, education, and language; and birthweight and gender all showed significant associations with 4-year I.Q., accounting for 42.7% of the variance in the general cognitive index. Bellinger et al. (114) found that social factors (such as social class, maternal I.Q., marital status, preschool attendance, residential stability, HOME score, child care arrangements, and household composition), in addition

Incorporating the Social-Ecological Perspective

401

to the level of concurrent lead exposure, were independently associated with degree of recovery from 24-month cognitive deficits. The lead and PCB studies also provide us with information about the contribution of confounding factors to cognitive performance during the preschool period. These data are consistent with previous reports suggesting that children from less advantaged circumstances express deficits at lower blood lead levels and the deficits are more persistent than among children in higher social strata (115). However, meta-analysis of low lead level exposures and child I.Q. suggests that the actual decrement in I.Q. associated with lead exposure was smaller among disadvantaged populations as compared to non-disadvantaged populations (116). This may be because it is more difficult to discern the effects of a relatively minor factor on child cognitive functioning in populations with high but variable exposures to many more important factors. That is, it may be much easier to isolate the effects of lead itself in a “clean” population. On the other hand, even a small decrement in I.Q. score in a population with a low mean I.Q. may be sufficient to drop an individual child’s score below the cut-point for special education placement or the threshold for developmental delay.

Future Directions in the Modeling of Social and Chemical Risk Geographic information system (GIS) technology is a useful tool in the field of environmental science to understand the complex spatio-temporal relationships between environmental pollution, social conditions, and disease. GIS is a powerful computer mapping and analysis tool that allows large quantities of information to be viewed and analyzed within a geographic context. Data from multiple sources, geographic (spatial) as well as non-geographic (demographics, median income, racial distribution, location of toxic release inventory sites), can be integrated and modeled using several functions like automated address matching (residential sites of subjects), distance function (proximity to roadways, toxic release inventory sites, and bus depots), and buffer analysis. These methods have been used to estimate relationships between environmental contaminants and adverse health effects (117,118). Guthe and colleagues (119) assigned lead exposures in the Newark, NJ, area to predict populations of children at high risk of exposure. Glass and colleagues (120) used GIS to investigate residential environmental risk factors for lyme disease in Baltimore County, MD. Recently, Kohli and colleagues (121) used GIS methods to identify individuals living in areas with high background concentrations of radon in Sweden. The Centers for Disease Control has developed a mapping program, Epi-Map, describing the spatial distribution of disease occurrence (122). These reports indicate that the GIS approach can be an effective tool for quantifying exposures of individuals to environmental agents. These can simply include counts of the housing hazard or condition within a specified search radius of a woman’s house or, more appropriately, the concentration of the hazard measured as a density. While simple counts can be calculated using vector GIS, which uses points, lines, and polygons to represent map features, densities rely on raster GIS, which use grids made up of regular shaped cells to represent the change in values over space. In addition to overcoming issues with aggregation/zoning and scale effects, this approach to measuring neighborhood-level variables with raster densities at multiple scales also promises to increase the variance across the study sample. By using raster densities and testing the effects of different scales, this approach will help to distinguish the many different social and physical contexts from which the cohort families are drawn.

402

Rauh et al.

INTERACTIONS BETWEEN SOCIAL AND CHEMICAL RISKS Until very recently, progress in methods for the assessment of multiple cumulative risks has evolved somewhat separately from the literature addressing social contextual effects on human environmental health. Cumulative environmental risk models rarely include social factors (6), and studies that do address the role of social factors in environmental health typically treat such conditions as confounders rather than exploring their potentially interactive effects with one or more chemical exposures (123). In the first case, estimates of neurotoxic effects may be biased because of uncontrolled or residual confounding by unmeasured social factors (124). In the second case, failure to test effect modification may result in under- or over-estimation of neurotoxic effects in different populations so that findings lack ecological validity. Improving the validity of risk assessment in studies of human developmental neurotoxicity will require an integrated approach in which social conditions are incorporated into models of environmental risk assessment. The meaning of interactions involving social and physical environmental factors depends upon the level of analysis. At the individual biological level, interaction is perhaps more approporiately called synergism, and refers to a biological response produced by simultaneous exposure to two or more agents that exceeds the combined action of the agents when working independently (125). Synergistic agents work together mechanistically at the cellular level. Although this type of interaction is not the primary focus of this chapter, it is important to keep in mind that co-exposure to multiple chemical risks, regardless of social context, may involve synergistic mechanisms. Statistical interaction, observed as effect modification, does not usually refer to mechanisms of action at the biological level, but may be observed in populations. This use of the term refers to whether an effect measure varies in value over categories, based on the level of some other factor, either cross-sectionally or over time. The mechanism may or may not be biological, but can have important public health implications regardless of mechanism. It is this second type of interaction that is addressed here, specifically the modulation of chemical risk by social factors or vice-versa. The theoretical and empirical bases for the adverse health effects of possible interactions between social adversity and environmental exposures have been recently reviewed elsewhere (2). Susceptibility as Effect Modification Of particular relevance to the social-ecological perspective in environmental science is the notion of biological susceptibility, whereby one or more exposures renders the organism vulnerable to other, often subsequent, exposures to the same or different toxicants, but may be relatively harmless in the absence of the additional exposure (126). Sometimes, this phenomenon can be seen in the form of delayed, latent, or lagged effects in which the neurobehavioral damage is only apparent when the individual matures and is faced with new developmental challenges or stressors. In other cases, the individual is more vulnerable to the harmful effects of an exposure during some critical developmental period, in which case the vulnerability results from an interaction of the exposure with age. While identifying critical periods during which various exposures are likely to be most toxic is important for risk assessment, the often chronic nature of low-level exposures to many chemical agents and social adversities makes interaction effects with age (or time) difficult to model. Longitudinal studies of low-level lead exposure and PCB exposure on children’s cognitive function in the preschool years provide a reasonable example of susceptibility as a function of the timing of multiple exposures to a toxic agent. In a cohort of 170 middle

Incorporating the Social-Ecological Perspective

403

and upper-middle class children, Bellinger et al. (115) reported that the effects of prenatal low-level lead exposure on cognitive function decrease between 24 and 57 months, as measured by the McCarthy Scales of Children’s Abilities, except among children with higher concurrent lead exposure. The impact of early postnatal exposures on cognitive performance was however detected at 57 months, with main effect sizes of 6.4 I.Q. points, and an adjusted decrement of 2.95 points in 57-month general cognitive index score for each natural log unit increase in 24-month lead level. This suggests that the extent to which children with high levels of prenatal exposure “recovered” from deficits manifested at 24 months varied with level of concurrent lead exposure. Furthermore, the cognitive deficits associated with exposure to environmental lead in early childhood appear to be only partially reversed by a subsequent decline in blood lead level (127). Adverse neurobehavioral effects of low-level exposures may appear at various ages as the child matures, develops capacities, or faces challenges, with or without mediating effects on birth weight, length or head circumference (128,129). Wilhelm et al. (130) observed a 10–20% increase in the risk of term LBW and preterm birth in infants born to women living close to heavily-trafficked roadways and therefore likely exposed to higher levels of motor vehicle exhaust. Stronger effects were observed for women whose third trimesters fall during fall/winter months, who lived in high background air pollution areas, and/or who lived in more impoverished areas. Data from the Columbia Center for Children’s Environmental Health show an interactive effect between exposure to polycyclic aromatic hydrocarbons and environmental tobacco smoke on reduced birth weight (131). There is also some limited biological evidence for racial differences in susceptibility or response to specific toxic agents. For example, higher levels of cotinine have been reported in black smokers than in white smokers, after controlling for self-reported amount of smoking (47,132). In one study, African American children had two-fold higher cotinine levels than Caucasian children as a result of exposure to one cigarette per day (133). Similarly, after adjusting for cigarette dose, cotinine levels in pregnant women were higher in African Americans than in Caucasians, while the rate of decrease in infant birth weight per nanogram of cotinine was similar in the two groups (134). The findings of a higher internal dose of tobacco smoke constituent metabolites per unit exposure are consistent with the higher risk of smoking-related lung cancer in African Americans. They also point to the possibility that cigarette smoking among African Americans may have a more deleterious effect on fetal development (134). With respect to toxicant effect modification by aspects of the social environment, Sadler et al. have reported that prenatal ETS exposure was associated with small size-forgestational-age only among low-income women, with no ETS effect for the higher income group (135). Weiss compares the effects of lead toxicity in advantaged and disadvantaged communities, and demonstrates how, while harmful lead effects may appear to be stronger among more socio-economically advantaged children, the actual number of children added to the developmentally delayed category as a result of lead exposure is far greater (and more costly) in less advantaged populations (136). Jacobson and Jacobson (137) found that breast-fed children were less vulnerable to the adverse cognitive effects of prenatal consumption of PCBs, possibly because of higher quality parenting among breast-fed infants. They also found that the adverse neurobehavioral effects of prenatal PCB exposure were stronger among children with less verbal mothers. Other studies have reported modulation of the teratogenic effects of antenatal cocaine exposure on a range of neurobehavioral outcomes by social factors, such as public assistance, multiparity, and caregiving or early intervention (138,139).

404

Rauh et al.

Such interactions may reflect a kind of cumulative risk whereby children who are exposed prenatally to single or multiple neurotoxicants are rendered more susceptible to the developmental challenges posed by deprivation and/or stressful living conditions in the early years of life (140). Subsequent hardships may not be well-tolerated by children who have been prenatally exposed to chemical toxicants. Mayes provides a relevant behavioral teratogenic model in which prenatal cocaine exposure affects arousal regulatory mechanisms, resulting in the young child’s vulnerability to environments characterized by impaired parenting and other social stressors (141). According to this model, adaptation to challenging situations depends on an individual’s threshold for activation of the catecholamine and norepinephrin arousal system, and this threshold may be impaired by the prenatal toxicant exposure. Animal models have also shown that prenatal and early postnatal exposure to nicotine can modulate catecholamine gene expression or neuroendocrine regulation (142,143), and recent evidence from animal and human studies has shown that maternal stress may mediate associations between socioeconomic adversity and early childhood cognition (144,145). This model suggests that it may be important to identify early caretaking conditions that are developmentally protective, such that some children, despite prenatal toxicant exposure, are able to compensate and enjoy a more optimal developmental trajectory.

Biological Sources of Susceptibility to Neurotoxicants Social Stress From a biological perspective, the term “stress” is used to describe any physical or psychological challenge that threatens the stability of the internal milieu of the organism (homeostasis). Exposure to any kind of stressor may pose a challenge requiring adaptive physiological response. In practice, life poses a series of challenges from early development through the aging process, each requiring adaptive functioning and readjustment. However, individuals differ in the amount and adversity of exposures as well as their capacity to adapt. Various coping mechanisms and supports may be used to enhance adaptation along the way, and it may be that chronic stressors require very different kinds of coping than more acute situations. It is also clear that not all coping strategies are positive or effective in restoring balance. Distress, and possibly illness, occurs when stress levels exceed the capacity of the individual to adapt. Distress may be conceptualized as a state of homeostatic imbalance or dysregulation. The presence of distress, reflecting the inability of the individual to self-regulate in the face of some challenge, thus suggests a kind of susceptibility or characteristic hypothalamic-pituitary axis (HPA) response that can be long-lasting (146). In both children and adults, distress may be accompanied by a sense of vulnerability and powerlessness, and may actually increase the likelihood of engaging in other high risk health-related behaviors. The biology of the stress response (and some experimental evidence from animal studies) suggests that early hazardous exposures may affect the capacity of the individual to adapt or self-regulate with long-lasting harmful effects on health and well-being. For example, experimental models have shown that prenatal and early postnatal exposure to nicotine can alter catecholamine gene expression or neuroendocrine regulation involved in response to stress (143). Other evidence from animal models as well as clinical studies with humans suggests that gestational events may permanently influence the development of the HPA-axis and, in particular, the central corticotrophin-releasing factor systems, which then affect behavioral, autonomic, and endocrine functioning in later life (146).

Incorporating the Social-Ecological Perspective

405

Although we are limited to observational studies or natural experiments in humans, the notion that early hazardous exposure may impair a child’s capacity to adapt or selfregulate when faced with subsequent challenges is consistent with the observation that the negative consequences of early lead exposure become manifest with increasing developmental challenges. The toxicity of the exposure may be latent, or “silent,” and not apparent until additional demands supervene (147). Again, the resulting failure to adapt is experienced as a kind of dysregulation or distress. This biological model would strongly point to gestation and early childhood as periods of tremendous vulnerability. A clinical illustration of this biological model involves the observation that infants exposed to tobacco smoke (active maternal smoking) in utero are at higher risk for SIDS than unexposed infants (148). These infants show reduced physiological capacity to compensate for challenges (physiological stressors) that lead to decreases in blood pressure (149,150). This is manifested by altered patterns of heart rate and heart rate variability in response to blood pressure challenge in the first few days of life (151). The hypothesized mechanism is that prenatal toxic exposures lead to a defect in brainstem mechanisms that insure proper integration of cardiac, vascular, and respiratory activity. Fetal and early infant patterns of autonomic dysregulation appear to be relatively stable over time, and are associated with response to novelty at 4 months (152), infant difficultness, inability to adapt, reactivity, and attention problems at 12 months of age (153,154). This work has important implications for subsequent problems in processing frequently encountered stimuli, such as might be seen in attentional problems, impulsivity, or other learning problems. To date, it is not clear if early patterns of autonomic dysregulation are reflected in subsequent behavior, attention, or learning problems, especially under highly stressful early life conditions. This mechanistic pathway, linking fetal toxic exposures with later childhood behaviors and/or learning problems, has so far received little attention by environmental toxicologists. Recent evidence from animal and human studies has shown that prenatal and early postnatal stressful exposures may also modulate toxicant effects on the developing fetus (155–157). Nutrition Another source of individual susceptibility to a range of toxicants may include nutritional deficits (i.e., lower intake of essential micronutrients, such as antioxidants, essential fatty acids). Antioxidant vitamins (A, C, or E) remove free radicals and oxidant intermediates, thereby inhibiting chemical-DNA binding which has been associated with decreased weight, length, and head circumference at birth (158), with possible implications for subsequent neurodevelopment. They have also been shown to reduce PAH-DNA damage in smokers, alone or in combination with the GSTM1 null genotype (159). In addition, essential fatty acid (EFA), such as arachidonic acid (AA) and docosahexaenoic acid (DHA), are required for proper fetal and infant brain development, especially during the last trimester of pregnancy (160). Reduced levels of AA have been associated with deficits in fetal growth (161). In full-term infants, plasma levels of DHA at 16 weeks have been correlated with the psychomotor and gross motor scores of the Gesell Developmental Index at 48 to 60 weeks (162), and Bayley Mental Development Index (163). Poor nutrition may also amplify the toxicity of some specific environmental exposures such as lead, whose uptake is competitively inhibited by adequate calcium consumption. At the population level, there is some evidence that disadvantaged minorities are more susceptible to the toxic effects of certain exposures, at least partly because of nutritional or other deficiencies (114). Although some nutritional deficits are associated with poverty, there is considerable inter-individual variation in nutritional status within

406

Rauh et al.

socioeconomically disadvantaged populations. Taken together, these reports lend support to the idea that at least some adverse developmental effects of toxicants may be exacerbated by dietary deficits. Genetic Factors Genetic susceptibility may take the form of common polymorphisms that can affect metabolism, hence neurotoxicity as well as carcinogenicity in the individual. The P450 (CYP1A1) gene plays a role in activation of PAH to DNA-binding intermediates (164,165). The glutathione-S-transferase genes are involved in detoxification of PAH and other environmental carcinogens (166). The prevalence of certain of these polymorphisms ranges from 10–50%. A recent study reported a 1285 g reduction in birth weight and 5.2 weeks reduction in gestational age among smoking mothers with the CYP1A1 Aa/aa and GSTT1 absent genotypes, revealing a significant interaction between maternal active smoking and metabolic genes in a moderately large urban sample (167). This interaction has not yet been tested for ETS or PAH. As noted above, there is also some evidence for race/ethnic differences in biological response to specific toxicants, and it is likely that this reflects some type of gene– environment interaction. Race/ethnicity may also be a marker for some other condition or exposure that moderates the impact of the toxicant, such as chronic social adversity. As noted earlier in the chapter, higher levels of cotinine have been reported in African American smokers than in white smokers, after controlling for self-reported amount of smoking (132,168,169), specifically among African American women (134) and children (133). The finding of higher internal dose of tobacco smoke constituent metabolites per unit exposure in African Americans versus whites is consistent with the higher risk of smoking-related lung cancer in African Americans and points to the possibility that cigarette smoking among African Americans may have a more deleterious effect on fetal development. Effect Modification in Multi-Level Models This type of effect modification is perhaps most difficult to conceptualize and test because it implies a kind of cross-level effect whereby individual level effects are moderated by community-level conditions. This would mean, for example, that the effect of a toxic exposure on an individual would depend upon some attribute of the community in which that individual resides. This effect modification would not be explained by interactions with other individual-level psychological or social exposures. For example, Gee and Takeuchi reported that ecologically measured vehicular burden interacted with individually perceived stress such that people living in areas with greater vehicular burden and reporting the most traffic stress also had the lowest health status and greatest depressive symptoms (170). In a study of community characteristics and child maltreatment, The Project on Human Development in Chicago Neighborhoods found that neighborhood social networks interacted with individual Hispanic ethnicity to affect the amount of physical abuse used by individual families (171). The authors interpreted this finding to make the point that neighborhood-level interventions may be the most effective way to reduce rates of child abuse in certain populations. Such findings, from both the fields of environmental science and sociology, suggest that studies which examine factors at only one level (either individual level only or ecological level only) may underestimate the effect of the social environment and potentially miss an opportunity for intervention to reduce risk.

Incorporating the Social-Ecological Perspective

407

Other researchers would say that there is simply no evidence for this type of effect (113). Here it is worthwhile to rethink the many possible forms of community-level effects (such as those conferred by social norms, high crime rates, concentrated poverty, deteriorated environment, lack of local services, etc.), and to determine whether any apparent cross-level effects are actually explained by previously unmeasured individual factors. For example, if an interaction between neighborhood food supply and lead toxicity were observed, one would have to rule out the effects of individual diet. As suggested by Oakes (113) and Kaufman and Cooper (172), carefully conceptualized experimental studies involving social interventions may be the most useful design to test such hypotheses, in large part because of the enormous difficulty in ruling out the role of an almost infinite number of unobserved individual-level factors as alternative explanations. Such caution in no way undermines the thinking that strong social forces and structural conditions have significant health and developmental effect, but these effects (whether main, moderating, or mediating) are difficult to prove in observational epidemiological studies. The ecological perspective in studies of neurotoxicity can enrich our understanding of biological processes. Inclusion of this perspective may be the only way to effectively reduce some types of risk, but it remains a challenging area.

REFERENCES 1. Gee GC, Payne-Sturges DC. Environmental health disparities: a framework integrating psychosocial and environmental concepts. Envir Health Perspect 2004; 112:1645–1653. 2. O’Neill MS, Jetterr M, Kawachi I, et al. With input from participants of the Workshop on air pollution and socioeconomic conditions. Health, wealth and air pollution: advancing theory and methods. Envir Health Perspect 2003; 111:1861–1870. 3. Jerrett M, Burnett RT, Goldberg MS, et al. Spatial analysis for environmental health research: concepts, methods and examples. J Toxicol Envir Health Part A 2003; 66:1783–1810. 4. Halfon N, Hochstein M. Life course health development: an integrated framework for developing health, policy, and research [Rev]. Milbank Q 2002; 80:433–479. 5. Racial and Ethnic Approaches to Community Health (REACH) 2010: Addressing Disparities in Health, 2005. US Department of Health and Human Services, Centers for Disease Control and Prevention, Coordinating Center for Health Promotion, 2005. 6. Framework for Cumulative Risk Assessment. US Environmental Protection Agency. Washington DC, May 2003. 7. Shonkoff JP, Phillips D, eds. From neurons to neighborhoods: the science of early childhood development. In: Committee on Integrating the Science of Early Childhood Development. Washington, DC: National Academy Press, 2000. 8. Institute of Medicine of the National Academies. The role of environmental hazards in premature birth: workshop summary. RoundRoundtable on Environmental Health Sciences, Research and Medicine (Board on Health Sciences Policy). Washington, D.C.: National Academies Press, 2003. 9. Young Children Develop in an Environment of Relationships. National Scientific Council on the Developing Child. Working Paper No. 1. A Project of the Heller School for Social Policy and Management at Brandeis University, Waltham MA, 2004. 10. Ensuring Risk Reduction in Communities with Multiple Stressors: Environmental Justice and Cumulative Risks/Impact. National Environmental Justice Advisory Council Cumulative Risks/Impacts Work Group (Federal Advisory Committee to US EPA), Dec. 2004. 11. Fox MA. Evaluating cumulative risk assessment for environmental justice: a community case study. Envir Health Perspect 2002; 110:203–209.

408

Rauh et al.

12. Castorina R, Bradman A, McKone TE, Barr DB, Harnly ME, Eskenazi B. Cumulative organophosphate pesticide exposure and risk assessment among pregnant women living in an agricultural community: a case study from the CHAMACOS cohort. Envir Health Perspect 2003; 111:1640–1648. 13. Corburn J. Bringing Local Knowledge into Environmental Decision Making: Improving Urban Planning for Communities at Risk. J Plan Educ Res 2003; 22:420–433. 14. Chen JJ, Chen YJ, Rice G, et al. Using dose addition to estimate cumulative risks from exposures to multiple chemicals. Regul Toxicol Parmacol 2001; 34:35–41. 15. Fox MA, Tran NL, Groopman JD, Burke TA. Toxicological resources for cumulative risk: an example with hazardous air pollutants. Regul Toxicol Pharmacol 2004; 40:305–311. 16. Weiss B. Vulnerability of children and the developing brain to neurotoxic hazards. Environ Health Perspect 2000; 108:375–381. 17. Bellinger DC, Stiles KM. Epidemiologic approaches to assessing the developmental toxicity of lead. Neurotoxicology 1993; 14:151–160. 18. Needleman HL, Riess JA, Tobin MJ, Biesecker GE, Greenhouse JB. Blood lead levels and delinquent behavior. JAMA 1996; 275:363–369. 19. Adler NE, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic status and health in industrial nations: social, psychological and biological pathways. In: Annals of the New York Academy of Sciences, Vol. 896. New York: New York Academy of Sciences, 1999. 20. DiPietro JA, Costigan KA, Shupe AK, Pressman EK, Johnson TRB. Fetal neurobehavioral development: associations with socioeconomic class and fetal sex. Dev Psychobiol 1998; 33:79–91. 21. Lamb ME. Nonparental child care: context, quality, correlates, and consequences. In: Sigel I, Renninger KA, Damon, eds. In: Handbook of Child Psychology: Vol. 4 Child Psychology in Practice. New York: John Wiley, 1997:73–134. 22. Sameroff AJ, Seifer R, Bartko WT. Environmental perspectives on adaptation during childhood and adolescence. In: Luthar SS, Burack JA, Cicchetti D, Weisz JR, eds. Developmental Psychopathology: Perspectives on Adjustment, Risk and Disorder. Cambridge, UK: Cambridge University Press, 1997:507–526. 23. Duncan GJ, Brooks-Gunn J, Klebanov PK. Economic deprivation and early childhood development. Child Dev 1994; 65:296–318. 24. Duncan GJ, Brooks-Gunn J, eds. Consequences of Growing up Poor. New York: Russell Sage Foundation, 1997. 25. McLoyd VC. Socioeconomic disadvantage and child development. Am Psychol 1998; 53:185–204. 26. Coulton CJ, Pandey S. Geographic concentration of poverty and risk to children in urban neighborhoods. Am Beh Scient 1992; 35:238–257. 27. Huston AC, Duncan GJ, Granger R. Work-based antipoverty programs for parents can enhance the school performance and social behavior of children. Child Dev 2001; 72:318–336. 28. Slaughter-Defoe DT, Nakagawa K, Takanishi R, Johnson DJ. Toward cultural/ecological perspectives on schooling and achievement in African—and Asian-American children. Child Dev 1990; 61:363–383. 29. McLanahan S, Sandefur GD. Growing Up With a Single Parent: What Hurts, What Helps? Cambridge, MA: Harvard University Press, 1994. 30. Hoffman LW. The effects of maternal employment in the two-parent family. Am Psychol 1989; 44:283–292. 31. Evans GW, Palsane MN, Lepore SJ, Martin J. Residential density and psychological health: the mediating effects of social support. J Pers Soc Psychol 1989; 57:994–999. NICHD Early Child Care research Network. The relation of child care to cognitive and language development. 2000; 71:960–980. 32. Wasserman GA, Rauh VA, Brunelli S. Social support and living arrangements of adolescent and adult mothers. J Adolesc Res 1990; 5:55–66.

Incorporating the Social-Ecological Perspective

409

33. Taylor SE, Seeman T. Psychosocial resources and the SES-health relationship. In: Adler NE, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic Status and Health in Industrial Nations: Social, Psychological, and Biological Pathways. New York: The New York Academy of Sciences, 1999:210–225. 34. DiPietro JA. Baby and the brain: Advances in child development. Annu Rev Publ Health 2000; 21:455–471. 35. Lepore SJ, Evans GW, Schneider ML. Dynamic role of social support in the link between chronic stress and psychological distress. J Pers Soc Psychol 1991; 61:899–909. 36. Roberts RE, Vernon SW. Minority status and psychological distress reexamined: the case of Mexican-Americans. Res Comm Mental Health 1984; 4:131–164. 37. Dunn JR, Hayes MV. Identifying pathways for health inequalities. In: Adler NE, Marmot M, McEwen BS, Stewart J, eds. Socioeconomic Status and Health in Industrial Nations: Social, Psychological, and Biological Pathways. New York: The New York Academy of Sciences, 1999:399–402. 38. Marmot M, Smith GD, Stansfield S, et al. Health inequalities among British civil servants: the Whitehall II study. Lancet 1991; 337:1387–1393. 39. Garbarino J. An ecological perspective on the effects of violence on children. J Comm Psychol 2001; 29:361–378. 40. Wasserman GA, Liu X, Popovac D, et al. The Yugoslavia prospective lead study: contributions of prental and postnatal lead exposure to early intelligence. Neurotoxicol Teratol 2000; 22:811–818. 41. Guillette EA. Examining childhood development in contaminated urban settings. Environ Health Perspect 2000; 108:389–393. 42. Evans GW, Kantrowitz E. Socioeconomic status and health: the potential role of environmental risk exposure. Annu Rev Public Health 2002; 23:303–331. 43. Olden K, Poje J. Environmental justice and environmental health. Bull Society Occup Environ Health 1995; 4:3–4. 44. Lancaster T, Stead L, Shepperd S. Helping parents to stop smoking: which interventions are effective? Paediat Respir Rev 2001; 2:222–226. 45. Chen LH, Petitti DB. Case-control study of passive smoking and the risk of small-forgestational-age at term. Am J Epidemiol 1995; 142:158–165. 46. Pirkle JL, Flegal KM, Bernert JT, Brody DJ, Etzel RA, Maurer KR. Exposure of the US population to environmental tobacco smoke: the Third National Health and Nutrition Examination Survey, 1988 to 1991. JAMA 1996; 275:1233–1240. 47. Neidell MJ. Air pollution, health, and socio-economic status: the effect of outdoor air quality on childhood asthma. J Health Econ 2004; 23:1209–1236. 48. Wernette DR, Nieves LA. Breathing polluted air: minority disproportionately exposed. EPA J 1992; 18:16–17. 49. Weaver VM, Davoli CT, Murphy SE, et al. Environmental tobacco smoke exposure in innercity children. Cancer Epi Biom Prev 1996; 5:135–137. 50. Crawford FG, Mayer J, Santella RM, et al. Biomarkers of environmental tobacco smoke in preschool children and their mothers. J Natl Cancer Inst 1994; 86:1398–1402. 51. Massey DS, Denton NA. American Apartheid: Segregation and the Making of the Underclas. Cambridge, MA: Harvard University Press, 1993. 52. Maantay J. Zoning, equity, and public health. Am J Public Health 2001; 91:1033–1041. 53. Miller VD. Planning, power and politics: a case study of the land use and siting history of the North River Water Pollution Control Plant. Fordham Urban Law J 1994; 21:707–722. 54. Shepard PM. Issues of community empowerment. Fordham Urban Law J 1994; 21:739–755. 55. Caro RA. The Power Broker: Robert Moses and the Fall of New York. New York: Random House, 1974. 56. Metzer R, Delgado JL, Herrell R. Environmental health and hispanic children. Environ Health Perspect 1995; 103:25–32. 57. Polednak AP. Segregation, poverty and mortality in urban African Americans. New York: Oxford University Press, 1997.

410

Rauh et al.

58. Kahn AJ, Kamerman SB, eds. Beyond Child Poverty: The Social Exclusion of Children. New York: Institute for Child and Family Policy at Columbia University, 2002. 59. Evans GW, Hygge S, Bullinger M. Chronic noise and psychological stress. Psycholo Sci 1995; 6:333–338. 60. Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science 1997; 277:918–924. 61. Breysse P, Farr N, Galke W, Lanphear B, Morley R, Bergofsky L. The relationship between housing and health: children at risk. Envir Health Perspect 2004; 112:1583–1588. 62. Matte TD, Jacobs DE. Housing and health current issues and implications for research and programs. J Urban Health 2000; 77:7–25. 63. Citizen’s Committee for Children of New York. Keeping Track of New York City’s Children, 1997. 64. Workshop on Housing and Health: Interrelationship and Community Impact. New York: New York Academy of Medicine, 1989. 65. Rauh VA, Chew G, Garfinkel R. Deteriorated housing contributes to high cockroach allergen rates in inner-city households. ‘Using Community-based Participatory Research to Advance Environmental Justice’ (Monograph). Environ Health Perspect 2002; 110:323–327. 66. Whyatt RM, Barr DB, Camann DE, et al. Measurement of contemporary-use pesticides in personal air samples during pregnancy and blood samples at delivery among urban minority mothers and newborns. Environ Health Perspect 2003; 111:749–756. 67. Wilson WJ. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago: The University of Chicago Press, 1987. 68. Wallace D, Wallace R. Scales of geography, time, and population: the study of violence as a public health problem. AJPH 1998; 88:1853–1858. 69. Fullilove MT. Psychiatric implications of displacement: contributions from the psychology of place. Am J Psychiatry 1996; 153:1516–1523. 70. Wilkinson RG. Unhealthy societies: The Afflictions of Inequality. London: Routledge Press, 1996. 71. Kaplan GA, Pamuk ER, Lynch JW, Cohen RD, Balfour JL. Inequality in income and mortality in the United States: analysis of mortality and potential pathways. BMJ 1996; 312:999–1003. 72. Kennedy BP, Kawachi I, Glass R, Prothrow-Stith D. Income distribution, socioeconomic status, and self-rated health in the US: multi-level analysis. BMJ 1998; 317:917–921. 73. Krieger N, Rowley DL, Herman AA, Avery B, Phillips MT. Racism, sexism and social class: implications for studies of health, disease and well-being. Am J Prev Med 1993; 9:82–122. 74. Williams DR. Race and health: basic questions, emerging directions. Ann Epidemiol 1997; 7:322–333. 75. Geronimus A. To mitigate, resist or undo: Addressing structural influences on the health of urban populations. Am J Pub Health 2000; 90:867–872. 76. Krieger N. Embodying inequality: a review of concepts, measures, and methods for studying health consequences of discrimination. Int J Health Serv 1999; 29:295–352. 77. Wise PH. Confronting racial disparities in infant mortality: reconciling science and politics. Am J Preventative Med 1993; 31:7–16. 78. Kawachi I. In: Berkman LF, Kawachi I, eds. Income Inequality and Health. In Social Epidemiology. New York: Oxford University Press, 2000:76–94. 79. Geronimus AT, Bound J, Waidmann TA, Hillemeier MM, Burns PB. Excess mortality among blacks and whites in the U.S. N Engl J Med 1996; 335:1552–1558. 80. Diez Roux AV, Merkin SS, Hannan P, Jacobs DR, Kiefe CI. Area characteristics, individuallevel socioeconomic indicators, and smoking in young adults: the coronary artery disease risk development in young adults study. Am J Epidemiol 2003; 157:315–326. 81. DiPrete TA, Forristal JD. Multilevel models: methods and substance. Annu Rev Sociol 1994; 20:331–357. 82. Goldstein H. Multilevel statistical models. New York: Halstead Press, 1995.

Incorporating the Social-Ecological Perspective

411

83. Von Korff M, Koepsell T, Curry S, Diehr P. Multilevel analysis in epidemiologic research on health behaviors and outcomes. Am J Epidemiol 1992; 135:1077–1082. 84. Pickett KE, Pearl M. Multilevel analyses of neighborhood socioeconomic context and health outcomes: a critical review. J Epidemiol Community Health 2001; 55:111–122. 85. Diez-Roux A. Bringing context back into epidemiology: variables and fallacies in multilevel analysis. AJPH 1999; 88:216–222. 86. Bellinger D. Assessing environmental neurotoxicant exposures and child neurobehavior confounded by confounding? Epidemiology 2004; 15:383–384. 87. Link BG, Phelan JC. Social conditions as fundamental causes of disease. J Health Soc Behav 1995; Extra Issue:80–94. 88. Schwartz S, Carpenter KM. The right answer for the wrong questions: consequences of type three for public health research. AJPH 1999; 89:1175–1180. 89. Robert SA. Socioeconomic position and health: the independent contribution of community socioeconomic context. Ann Rev Sociol 1999; 25:489–516. 90. Anderson RT, Sorlie P, Backlund E, Johnson N, Kaplan GA. Mortality effects of community socioeconomic status. Epidemiology 1996; 8:42–47. 91. Aneshensel CS, Sucoff CA. The neighborhood context of adolescent mental health. J Health Soc Behav 1996; 37:293–310. 92. Cubbin C, LeClere FB, Smith GS. Socioeconomic status and injury mortality: individual and neighborhood determinants. J Epidemiol Community Health 2000; 54:517–524. 93. Culhane JF, Elo IT. Neighborhood context and reproductive health. Am J Obstet Gynecol 2005; 192:S22–S29. 94. Diez-Roux A, Nieto FJ, Muntaner C, et al. Neighborhood environments and coronary heart disease: a multilevel analysis. Am J Epidemiol 1997; 146:48–63. 95. Taylor SE, Repetti RL. Health psychology: what is an unhealthy environment and how does it get under the skin? Annu Rev Psychol 1997; 48:411–447. 96. Duncan C, Jones K, Moon G. Smoking and deprivation: are there neighborhood effects? Soc Sci Med 1999; 48:497–505. 97. LeClere FB, Rogers RG, Peters K. Neighborhood social context and racial differences in women’s heart disease mortality. J Health Soc Behav 1998; 39:91–107. 98. Jargowsky PA. Poverty and Place: Ghettos, Barrios, and the American City. New York: Russell Sage Foundation, 1997. 99. Jones K, Duncan C. Individuals and their ecologies: analyzing the geography of chronic illness within a multilevel modeling framework. Health Place 1995; 1:27–40. 100. Robert SA. Community-level socioeconomic status effects on adult health. J Health Soc Behav 1998; 39:18–37. 101. Sooman A, Macintyre S. Health and perceptions of the local environment in socially contrasting neighborhoods in Glasgow. Health Place 1995; 1:15–26. 102. Yen IH, Kaplan GA. Neighborhood social environment and risk of death: multilevel evidence from the alameda county study. Am J Epidemiol 1990; 149:898–907. 103. Massey DS, Condran GA, Denton NA. The effect of residential segregation of black social and economic well-being. Social Forces 1987; 66:29–56. 104. Collins JW, Jr., David RJ. Urban violence and African-American pregnancy outcome: an ecologic study. Ethn Dis 1997; 7:184–190. 105. Bird ST, Bauman KE. The Relationship between structural and health services variables and state-level infant mortality iu the United States. Am J Public Health 1995; 85:26–29. 106. LaVeist TA. The political empowerment and health status of African-americans: mapping a new territory. Am J Sociol 1992; 97:1080–1095. 107. O’Campo P, Xue X, Wang M-C, O’Brien Caughey M. Neighborhood risk factors for low birthweight in Baltimore: a mutilvariate analysis. Am J Public Health 1997; 87:1113–1118. 108. Elo IT, Rodgriguez G, Lee H. Racial and neighborhood disparities in birth weight in Philadelphia. Washington D.C.: Annual Meeting of the Populations Association of America, 2001.

412

Rauh et al.

109. Reijneveld SA, Verheij RA, de Bakkar DH. The impact of area deprivation on differences in health: does the choice of geographical classification matter? J Epidemiol Community Health 2000; 54:306–313. 110. Macintyre S, Maciver S, Sooman A. Area, class and health: should we be focusing on places or people? J Soc Policy 1993; 22:213–234. 111. Rauh VA, Andrews H, Garfinkel R. The contribution of maternal age to racial disparities in birthweight: a multi-level perspective. Am J Public Health 2001; 91:1815–1824. 112. Massey DS, Gross AB, Shibuya K. Migration, segregation, and the geographical concentration of poverty. Am Sociol Rev 1994; 59:425–445. 113. Oakes JM. The (mis)estimation of neighborhood effects: causal inference for a practicable social epidemiology. Soc Sci Med 2004; 58:1929–1952. 114. Bellinger D, Sloman J, Leviton A, Rabinowitz M, Needleman HL, Waternaux C. Low-level lead exposure and children’s cognitive function in the preschool years. Pediatrics 1991; 87:219–227. 115. Bellinger D, Leviton A, Waternaux C, Needleman H, Rabinowitz M. Longitudinal analyses of pre- and postnatal lead exposure and early cognitive development. New Engl J Med 1987; 316:1037–1044. 116. Schwartz J. Low-level lead and children’s I.Q.: A meta-analysis and search for a threshold. Envir Res 1994; 65:42–55. 117. Mather FJ, White LE, Langlois EC, et al. Statistical methods for linking health, exposure, and hazards. Environ Health Perspect 2004; 112:1440–1445. 118. Nuckols JR, Ward MH, Jarup L. Using geographic information systems for exposure assessment in environmental epidemiology studies. [Rev] Environ Health Perspectives 2004; 112:1007–1015. 119. Guthe WG, Tucker RK, Murphy EA, England R, Stevenson E, Luckhardt JC. Reassessment of lead exposure in New Jersey using GIS technology. Environ Res 1992; 59:318–325. 120. Glass GE, Schwartz BS, Morgan JM, III, Johnson DT, Noy PM, Israel E. Environmental risk factors for lyme disease identified with geographic information systems. Am J Public Health 1995; 85:944–994. 121. Kohli S, Sahlen K, Lofman O, et al. Individuals living in areas with high background radon: a GIS method to identify populations at risk. Comput Methods Programs Biomed 1997; 53:105–112. 122. Dean AG. Epi Infoe and Epi Map: Current status and plans for Epi Infoe 2000. J Pub Health Manage Prac 1999; 5:54–57. 123. Lynch JW, Kaplan GA, Shema SJ. Cumulative impact of sustained economic hardship on physical, cognitive, psychological and social functioning. NEJM 1997; 337:1889–1895. 124. Mink PJ, Goodman M, Barraj LM, Imrey H, Kelsh MA, Yager J. Evaluation of uncontrolled confounding in studies of environmental exposures and neurobehavioral testing in children. Epidemiology 2004; 15:385–393. 125. Kleinbaum DG, Kupper LL. Epidemiological Research. Belmont, CA: Wadsworth, Inc, 1982. 126. Bellinger DC. Effect modification in epidemiologic studies of low-level neurotoxicant exposures and health outcomes. Neurotoxicol Teratol 2000; 22:133–140. 127. Tong S, Baghurst PA, Sawyer MG, Burns J, McMichael AJ. Declining blood lead levels and changes in cognitive function during childhood: the Port Pirie Cohort Study. JAMA 1998; 280:1915–1919. 128. Wigg N, Wimpani G, McMichael A, Braghurst P, Robertson E, Roberts R. Port pirie cohort study: childhood blood lead and neuropsychological development at age two years. J Epidemiol Comm Health 1988; 42:213–219. 129. Needleman HL, Gatsonis CA. Low level lead exposure and the IQ of children: a meta-analysis of modern studies. JAMA 1990; 263:673–678. 130. Wilhelm M, Ritz B. Residential proximity to traffic and adverse birth outcomes in Los Angeles County, California, 1994–1996. Environ Health Perspectives 2002; 3:207–216.

Incorporating the Social-Ecological Perspective

413

131. Perera FP, Rauh V, Whyatt RM, et al. Molecular evidence of an interaction between prenatal environmental exposures and birth outcomes in a multiethnic population. Environ Health Perspectives 2004; 112:626–630. 132. Hecht SS, Carmella SG, Foiles PG, Murphy SE. Biomarkers for human uptake and metabolic activation of tobacco-specific nitrosamines. Cancer Res 1994; 54:1912s–1917s. 133. Knight JM, Eliopoulos C, Klein J, Greenwald M, Koren G. Passive smoking in children. Racial differences in systemic exposure to cotinine by hair and urine analysis. Chest 1996; 109:446–450. 134. English PB, Eskenazi B, Christianson RE. Black-white differences in serum cotinine levels among pregnant women and subsequent effects on infant birthweight. Am J Public Health 1994; 84:1439–1443. 135. Sadler L, Belanger K, Saftlas A, et al. Environmental tobacco smoke exposure and small-forgestational-age birth. Am J Epidemiol 1999; 150:695–705. 136. Weiss B. Vulnerability of children and the developing brain to neurotoxic hazards. Environ Health Perspect 2000; 108:375–381. 137. Jacobson JL, Jacobson SW. Breast-feeding and gender as moderators of teratogenic effects on cognitive development. Neurotoxicol Teratol 2002; 24:349–358. 138. Frank DA, Jacobs RR, Beeghly M, et al. Level of prenatal cocaine exposure and scores on the bayley scales of infant development: modifying effects of caregiver, early intervention, and birth weight. Pediatrics 2002; 110:1143–1152. 139. Eyler FD, Behnke M, Conlon NS, Woods K. Birth outcome from a prospective, matched study of prenatal crack/cocaine use: interactive and dose effects on health and growth. Pediatrics 1998; 101:229–237. 140. Tronick EZ, Beeghley M. Prenatal cocaine exposure, child development and the compromising effects of cumulative risk. Clin Perinatol 1999; 26:151–171. 141. Mayes LC. A behavioral teratogenic model of the impact of prenatal cocaine exposure on arousal regulatory systems. Neurotoxicol Teratol 2002; 24:385–395. 142. Gauda EB, Cooper R, Akins PK, Wu G. Prenatal nicotine affects catecholamine gene expression in newborn rat carotid body and petrosal ganglion. J Appl Physiol 2001; 91:2157–2165. 143. Serova L, Danailov E, Chamas F, Sabban EL. Nicotine infusion modulates immobilization stress-triggered induction of gene expression of rat catecholamine iosynthetic enzymes. J Pharmacol Exp Ther 1999; 291:884–892. 144. McEwen BS. Protective and damaging effects of stress mediators. N Engl J Med 1998; 338:171–179. 145. McEwen BS, Stellar E. Stress and the individual: Mechanisms leading to disease. Arch Intern Med 1992; 153:2093–2101. 146. Chrousos GP, Gold PW. The concepts of stress and stress systems disorders. JAMA 1992; 267:1244–1252. 147. Weiss B, Reuhl K. Delayed neurotoxicity: a silent toxicity. Principles of Neurotoxicology. New York: Dekker, 1994. 148. Kahn A, Sawaguchi T, Sawaguchi A, et al. Sudden infant deaths: from epidemiology to physiology. Forensic Sci Int 2002; 130:S8–S20. 149. Edner A, Katz-Salamon M, Lagercrantz H, Milerad J. Heart are response profiles during head upright tilt test in infants with apparent life threatening events. Arch Dis Child 1997; 76:27–30. 150. Fox GP, Matthews TG. Autonomic dysfunction at different ambient temperatures in infants at risk of sudden infant death syndrome. Lancet 1989; 2:1065–1067. 151. Fifer WP, Greene M, Hurtado A, Meyers MM. Cardiorespiratory responses to bidirectional tilt in infants. Early Hum Dev 1999; 55:265–279. 152. Kagan J, Reznick J, Sniderman N. The physiology and psychology of behavioral inhibition in children. Child Development 1987; 58:1459–1473. 153. DiPietro J, Hodgson D, Costigan K, Hilton S. Fetal neurobehavioral development. Child Development 1996; 67:2553–2567.

414

Rauh et al.

154. DiPietro J, Hodgson D, Costigan K, Johnson T. Development of fetal movement–fetal heart rate coupling from 20 weeks through term. Early Hum Dev 1996; 44:139–151. 155. Meaney MJ, Brake W, Gratton A. Environmental regulation of the development of mesolimbic dopamine systems: a neurobiological mechanism for vulnerability to drug abuse? Psychoneuroendocrinology 2002; 27:127–138. 156. McEwen BS. Protective and damaging effects of stress mediators. NEJM 1998; 338:171–179. 157. McEwen BS, Stellar E. Stress and the individual: Mechanisms leading to disease. Arch Intern Med 1992; 153:2093–2101. 158. Perera FP, Whyatt RM, Jedrychowski W, et al. Recent developments in molecular epidemiology: a study of the effects of environmental polycyclic aromatic hydrocarbons on birth outcomes in Poland. Am J Epidemiol 1998; 147:309–314. 159. Mooney LA, Santella RM, Covey L, et al. Decline on DNA damage and other biomarkers in peripheral blood following smoking cessation. Cancer Epidemiol Biomarkers Prev 1995; 4:627–634. 160. Wood KA, Youle RJ. The role of free radicals and p53 in neuron apoptosis in vivo. J Neurosci 1995; 15:5851–5857. 161. Crawford FG, Mayer J, Santella RM, et al. Biomarkers of environmental tobacco smoke in preschool children and their mothers. J Natl Cancer Inst 1994; 86:1398–1402. 162. Voigt RG, Jensen CL, Fraley JK, Rozelle JC, Brown FR, IIIrd, Heird WC. Relationship between omega3 long-chain polyunsaturated fatty acid status during early infancy and neurodevelopmental status at 1 yr of age. J Hum Nutr Diet 2002; 15:111–120. 163. Bjerve KS, Brubakk AM, Fougner KJ, Johnsen H, Midthjell K, Vik T. Omega-3 fatty acids: essential fatty acids with important biological effects, and serum phospholipid fatty acids as markers of dietary omega-3 fatty acid intake. Am J Clin Nutr 1993; 57:801S–805S. 164. Whyatt RM, Bell DA, Santella RM, et al. Polycyclic aromatic hydrocarbon-DNA adducts in human placenta and modulation by CYP1A1 induction and genotype. Carcinogenesis 1998; 19:1389–1392. 165. Perera FP. Molecular epidemiology: on the path to prevention? J Natl Cancer Inst 2000; 92:602–612. 166. Helzlsouer KJ, Selmin O, Huang HY, et al. Association between glutathione S-transferase M1, P1, and T1 genetic polymorphisms and development of breast cancer. J Natl Cancer Inst 1998; 90:512–518. 167. Wang X, Zuckerman B, Pearson C, et al. Maternal cigarette smoking, metabolic gene polymorphism, and infant birth weight. JAMA 2002; 287:195–202. 168. Wagenknecht LE, Manolio TA, Sidney S, Burke GL, Haley NJ. Environmental tobacco smoke exposure as determined by cotinine in black and white young adults: the CARDIA study. Environ Res 1993; 63:39–46. 169. Wagenknecht LE, Burke GL, Perkins LL, Haley NJ, Friedman GD. Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. Am J Public Health 1992; 82:33–36. 170. Gee GC, Takeuchi DT. Traffic stress, vehicular burden and well-being: a multilevel analysis. Soc Sci Med 2004; 59:405–414. 171. Molnar BE, Buka SL, Brennan RT, Holton JK, Earls F. A multilevel study of neighborhoods and parent-to-child physical aggression: results from the project on human development in Chicago neighborhoods. Child Maltreat 2003; 8:84–97. 172. Kaufman JS, Cooper RS. Seeking causal explanations in social epidemiology. Am J Epidem 1999; 150:113–120.

20 Assessing the Neurobehavioral Effects of Early Toxicant Exposure: A Perspective from Animal Research Barbara J. Strupp Division of Nutritional Sciences and Department of Psychology, Cornell University, Ithaca, New York, U.S.A.

Stephane A. Beaudin Division of Nutritional Sciences, Cornell University, Ithaca, New York, U.S.A.

INTRODUCTION Animal models play an essential role in determining whether exposure to a putative neurotoxicant impairs cognitive and/or emotional functioning. Such studies circumvent a thorny problem that commonly plagues studies with humans; namely, that exposure to the toxicant occurs within the context of sociodemographic factors that themselves place children at risk for impaired cognitive and emotional development (e.g., poverty, poor health care, low maternal education and IQ; maternal depression; low intellectual stimulation in the home). The demonstration of cognitive deficits in animal models, under conditions in which these confounded risk factors are absent, is often pivotal in convincing regulatory agencies that exposure to a particular substance does indeed cause adverse effects. Studies with animals can also elucidate the neural mechanisms that underlie altered neurobehavioral functioning and provide model systems for the development and testing of possible treatments (e.g., pharmacotherapy, chelating agents). Despite these potential contributions of animal models, some may have nagging concerns about the feasibility or appropriateness of this endeavor. Concerns may relate to whether laboratory animals, particularly rodents (the species of choice due to cost, sample size considerations, ethical concerns, etc.), are sufficiently advanced in terms of cognitive and emotional functioning to serve as a valid predictor of toxicant-induced effects in humans in these same domains. It is also reasonable to ask whether the brains of rats and mice are sufficiently similar to humans to support this endeavor. In light of these concerns, it is very encouraging that cross-species comparisons have found close correspondence regarding the neurobehavioral, neuropathological, and neurochemical effects of exposure to those toxicants where there are available data. For example, early developmental exposure to methylmercury has been found to produce similar neuropathological, sensory, and cognitive effects in rodents, monkeys, and humans (1,2). Similarly, compelling 415

416

Strupp and Beaudin

cross-species similarity has been found with regards to the specific neurobehavioral effects produced by early exposure to lead (3–5), alcohol (6), and PCBs (7). Further support for the validity of using rodent and non-human primate models to study the neurobehavioral effects of toxicant exposure is provided by studies demonstrating correspondence in the functional roles of various neural systems between rats, monkeys, and humans. Such correspondence is very clear for structures such as the amygdala (8,9), hippocampus (10–14) and basal ganglia (15). Although some concerns have been raised regarding the use of rodent models to study frontal cortical function (16), several studies have demonstrated strong parallels between rodents and primates regarding the functions of different areas of the frontal cortex (11,17–22). For example, experimental lesions of the orbitofrontal cortex (OFC) have been found to produce comparable cognitive deficits in rats and monkeys (22). Neurophysiological data also suggest common specializations of information integration between the OFC of rats and monkeys (23). Similarly, it has been argued that the PFC of rats is similar to the comparable region in monkeys as defined by connectional, neurochemical, and functional criteria (16,22).

CONSIDERATIONS IN TASK SELECTION AND DESIGN As in studies concerning other functional or morphological domains, the success of this endeavor depends on the selection of endpoints. In this chapter, we discuss two major considerations in the selection and design of cognitive and/or behavioral tasks and dependent measures: (1) sensitivity (being able to detect dysfunction if it exists); and (2) specificity (being able to say something about the nature of the dysfunction). Although many of the presented examples pertain to animal research, the issues that must be considered in designing a successful animal study in this area are similar, if not identical, to the considerations that drive the parallel studies with human subjects. Many of the examples are drawn from our own work, not necessarily because they are the gold standard for which to strive, but rather because these are approaches we have found useful and know them best. Sensitivity An overarching consideration in neurobehavioral toxicology studies—as in all areas of toxicology—is sensitivity: being confident that one will detect dysfunction if it exists. With respect to “neurobehavioral” function, sensitivity depends on tapping a broad range of functions. Different cognitive, affective, and sensory functions are biologically distinct; that is, they depend on different neural systems, and can therefore be independently altered. As a result, unless the investigator taps a broad range of functions, it is very likely that dysfunction will be missed. Although this general perspective tends to guide assessment of sensory function (i.e., different sensory systems are each assessed), appreciation of this issue is less apparent with respect to assessments of cognitive functioning, at least with respect to animal testing. a This problem is illustrated by the EPA

a

Guidelines for human neurotoxicology assessments reflect greater cognizance of these issues, as illustrated by the recommendations of a working group convened by the ATSDR in 1991, which lists many different cognitive, sensory, and affective domains that must all be included to constitute a thorough assessment (24).

A Perspective from Animal Research

417

Health Effects Test Guidelines for Developmental Neurotoxicity testing [(25); OPPTS 870.6300]. These guidelines recommend assessing “.developmental landmarks, motor activity, auditory startle test, a test of associative learning and memory, and neuropathology.” The implication is that some representative “learning and memory” test can serve as an index of the functional integrity of the many cognitive functions that contribute to normal intellectual functioning. This perspective ignores the fact that different aspects of cognitive functioning depend on distinct neural systems. As a result, a particular neurotoxicant can profoundly impair one aspect of cognitive, sensory, or affective functioning, but leave many others entirely intact. Consequently, performance will be altered on certain tasks but not others. This point is illustrated by the cognitive effects of damage to the hippocampus and surrounding cortical areas, collectively referred to as the medial temporal lobe. Damage limited to the medial temporal lobe results in a profound impairment in explicit or declarative memory function (i.e., global anterograde amnesia), which refers to the ability to remember events, facts, and people (14). However, individuals with medial temporal lobe damage may appear normal in casual conversation because so many other aspects of cognitive functioning are entirely intact, including speech and verbal comprehension, attention, perception, short-term memory, retrieval of remote memories (memories formed before the brain damage), and some forms of long-term memory, collectively referred to as implicit memory (ability to learn and remember cognitive and motor skills; priming, conditioning) (26). Indeed, such individuals perform normally on IQ tests, despite amnesia so profound that they remember little or nothing of the events that have occurred since their brain injury (27–29). Similarly, many discrimination tasks commonly used in animal studies do not reveal impairments in learning or long-term retention following complete bilateral removal of the medial temporal lobes; one example drawn from research with non-human primates is learning of a 2-choice pattern discrimination (30,31). Although mastery and retention of such tasks require memory, they appear to depend on implicit, rather than explicit, memory function; they are solved slowly and incrementally, like a habit. In contrast, tests specifically designed to assess explicit memory function, such as the delayed non-matching-to-sample task (DNMS), reveal dramatic impairment in animals and humans with damage to the medial temporal lobe (14,32–34). Implications of Picking a “Representative” Learning Task This recommendation to pick some representative “learning and memory” test as an index of “cognitive function” is of grave concern for toxicology studies, as illustrated by the aforementioned primate studies of medial temporal lobe damage as well as by studies of animal models of PKU or hypothyroidism, two conditions that produce profound mental retardation (MR) in humans. Of the roughly 28 behavioral tasks that have been used to assess cognitive functioning in these two MR animal models, more than half (approximately 17) failed to reveal impairment in the experimental group relative to controls (35,36). The failure to detect group differences with these latter tasks is unlikely to reflect inadequacies of the pharmacological manipulations used to reproduce these two disorders, as many other tasks did reveal significant group differences. A more likely explanation is that each of these early insults produced lasting deficits in some, but not all, cognitive and affective functions, and that these tasks differ in the extent to which they tap these areas of impairment. Although all of these tasks tap “learning and memory,” they differ in the extent to which they assess the myriad cognitive and affective functions that are important for optimal performance on cognitive tasks. Examples include associative

418

Strupp and Beaudin

ability, implicit and explicit memory function, selective attention, sustained attention, divided attention, attentional set shifting, inhibitory control, transfer of learning, allocentric and egocentric coding of spatial information, susceptibility to proactive interference, and emotion regulation. In addition, neither “learning” nor “memory” is a unitary construct. As discussed above, there are different types of memory that are biologically independent (e.g., explicit vs. implicit memory). Similarly, there are different types of learning, which depend on different neural circuitry. For example, the ability to form stimulus-reward associations is thought to depend on circuitry involving the basolateral amygdala and ventral striatum (8,37–39), whereas learning of stimulusresponse associations depends on a different circuit, centrally involving the basal ganglia (15,39–41). The sensitivity of any given test in revealing dysfunction in these two experimental MR models will depend on the extent to which that test taps the particular deficits produced by that early insult. These findings demonstrate that the risk of missing existing dysfunction is very great with the approach of selecting a “representative learning and memory test.” In the case of these animal models of MR syndromes, the majority of such tasks did not reveal impairment in the experimental group, in sharp contrast to the profound cognitive impairment that characterizes the analogous human condition. Had these been studies designed to assess the risk of some substance with unknown risk potential, the majority would have reached erroneous conclusions, with potentially devastating public health consequences. The solution, therefore, is to administer a battery of tasks that taps a broad range of cognitive and affective functions, dependent on distinct neural systems. Unfortunately, this labor-intensive approach cannot be circumvented by ranking behavioral tests in terms of sensitivity and then always relying on the one or two most sensitive tests. The test that will be most sensitive in any given case depends on the nature of the brain damage. A test that is exquisitely sensitive to one neurotoxicant may be totally insensitive to another and vice versa, if the two insults damage different neural systems. This phenomenon, referred to as a “double dissociation” in the field of neuropsychology, is illustrated by the results of a non-human primate study involving bilateral damage to the hippocampal formation (H) or the amgydaloid complex (42). Monkeys with the H lesion were dramatically impaired on tests of explicit memory function, such as the delayed-nonmatch-to sample test, but performed normally on a test of emotional behavior. In contrast, partial or complete damage to the amygdaloid complex impaired emotional behavior but not explicit memory. In some cases, available neuropathological and/or clinical data provide clues as to which areas of functioning are most likely to be altered, which can guide task selection. For example, if there is evidence that exposure to the toxicant alters hippocampal cytoarchitecture and/or transmitter systems, then behavioral studies with animals should surely include one or more tasks that tap hippocampal function, i.e., declarative memory; examples include the delayed non-match to sample (DNMTS) test, the radial arm maze, or the Morris water maze. In other cases, the available data may pertain to specific clinical syndromes that seem to be linked to exposure to the toxicant, based on epidemiological studies and/or clinical experience. For example, exposure to the suspected toxicant may be related to an increased risk of ADHD, suggesting that the follow-up animal studies focus on measures of impulse control and attentional dysfunction (43). Note, too, that clinical symptoms can guide task selection for animal studies even in cases where the symptoms are difficult to model in animals, although the approach must be more circuitous. Consider the situation where exposure to the toxicant is associated with low scores on the pattern construction subtest of the Differential Abilities Scales, with a pattern of deficits that implicates damage to parietal cortex. It is not at all clear how to devise a similar task for

A Perspective from Animal Research

419

rats or mice, and if one attempted to do so, it is very possible that the brain region(s) on which the task depended would be different in rodents and humans. Rather, as the pattern of deficits seen clinically implicated damage to parietal cortex, a better approach for the rodent studies would be to include tasks known to be sensitive to parietal damage in the particular species (rat or mouse); examples for the rat include the cheeseboard task (44) or the item memory task (45). The goal of tapping a broad range of cognitive and affective functions could theoretically be achieved through the use of either apical (or global) tests or tests designed to provide indices of specific cognitive functions. It is not possible to state a priori which type of test will be more sensitive; this will likely depend on the nature of the dysfunction. If the toxicant produces a small effect on many different cognitive functions, global measures such as IQ tests may be more sensitive because the many small effects summate in terms of the performance score. In contrast, if the resulting deficit is very circumscribed, it is likely that a test specifically designed to tap that function will be more sensitive than a global measure which only depends on this one affected function to a small degree. There are several other benefits of administering tests that provide indices of specific functions; these are discussed below. Cognitive Functions Neglected in Animal Assessments Although it may be recognized, from a theoretical perspective, that it is necessary to tap a broad range of functions, in practice, some functions are more frequently tapped than others in animal studies of toxicant exposure. It is most common to assess the animals’ ability to learn relatively simple associations involving very salient predictive cues; memory function is also frequently assessed. In contrast, several other functions are rarely evaluated, including some that appear to be especially vulnerable to early brain damage. Four such neglected functions are: (1) executive functions of the prefrontal cortices (including selective attention, planning, set maintenance, and inhibition), (2) transfer of learning, (3) regulation of affect or emotions, particularly as it interfaces with cognitive functioning; and (4) intrinsically-motivated learning (including latent or incidental learning). We suggest that the failure to evaluate these processes in many prior studies of early toxicant exposure may have resulted in an underestimate of the existing dysfunction and advocate the assessment of these functions in future studies. The first three are discussed below. Due to space constraints, intrinsically-motivated learning is not discussed here, but the interested reader can refer to several earlier publications in which we have discussed this aspect of cognition in relation to early brain damage (35,46–48). Executive Functions. There is suggestive evidence that some neural systems in the brain are more vulnerable than others to disruption by genetic and environmental factors. Pennington (49) describes five domains of brain function that account for nearly all learning disorders in children: phonological processing, executive functions, spatial cognition, social cognition, and long-term memory. He argues that these five domains (and their associated neural systems) are differentially vulnerable to developmental insult or variation. This postulate is based on the fact that observed developmental cognitive disorders are not divided equally between these five modular systems. Rather, the majority of such disorders appear to reflect dysfunction in just two domains: executive functions and phonological processing. Based on this pattern, he suggests that the neural systems on which these functions depend—the prefrontal cortices and the left hemisphere language system, respectively—are uniquely vulnerable to developmental insult (49). If this proposal is correct, assessment of these functions would be particularly important to

420

Strupp and Beaudin

include in neurobehavioral batteries devised to assess early toxicant exposure in either humans or experimental animals. It is therefore of concern that tests specifically designed to tap executive functioning are not commonly included in animal studies of toxicant exposure. Studies with human subjects, too, often rely on global cognitive measures such as IQ tests, which are relatively insensitive to executive function deficits (50,51). For example, patients suffering from injuries to the prefrontal lobes show little or no change in their IQ scores as a function of such lesions (52). The suggestion that executive functions subserved by the prefrontal cortices may be uniquely vulnerable to early insult is illustrated by the cognitive profile that characterizes individuals exposed to cocaine in utero. The impairment produced by prenatal cocaine exposure appears to be surprisingly selective, with deficits limited to two domains: (1) various aspects of attention (selective attention, sustained attention, ability to stay consistently “on task”); and (2) regulation of arousal or affect, including the reaction to committing an error (discussed below). Because interpretation of the human data in this area is complicated by the presence of numerous confounders (e.g., use of other licit and illicit drugs, maternal depression, violence in the home, low maternal education and IQ), we focus here on data from a rodent model, using an intravenous injection protocol that closely mimics the pharmacokinetic profile and physiological effects of recreational cocaine use (53). Adult rats exposed to cocaine in utero exhibited impaired sustained attention (54) as well as a deficit in staying consistently “on task” (55) in a visual attention task in which the time of onset, location, and duration of a brief visual cue varied randomly between trials. Selective attention was also impaired, based on the results of several types of tasks (56–58). For example, the performance of the cocaine-exposed animals was more disrupted than that of controls by the unpredictable presentation of olfactory distractors in a visual attention task (Fig. 1) (57). Notably, the groups did not differ on trials without an olfactory distractor,

9 control Premature responses (%)

8

0.5 mg/kg

7

** ** #

1.0 mg/kg

6

3.0 mg/kg

5 4 *

3 2

#

1 Non-distraction

0 or 2 sec after trial onset Onset of Olfactory Distractor

1 sec after trial onset

Figure 1 Mean (C/KSEM) percentage premature responses in a distraction task in which olfactory distractors were presented at various intervals before onset of the visual cue, on some trials. Prenatal cocaine exposure increased premature responses when distractors were presented, an effect that was most pronounced when the distractor was presented 1 sec after trial onset, the condition that placed the greatest demand on selective attention [Treatment Group!Distraction Condition, F(6, 1093)Z2.52, pZ0.02 # pZ0.06, -p!0.05, --p!0.01]. Source: Adapted from Ref. 57.

A Perspective from Animal Research

421

ruling out various alternative explanations for the impaired performance on distraction trials (e.g., impaired visual acuity, reduced motivation, motor dysfunction, impaired comprehension of task rules). Converging evidence for deficient selective attention was provided by the pattern of results in a series of extradimensional shift tasks: the cocaineexposed animals were impaired relative to controls when shifted from a task involving salient predictive cues (olfactory) and subtle irrelevant cues (spatial) to one in which the predictive cues were now the subtle cues and the salient cues were now irrelevant to the attainment of reward. In contrast, their learning rate did not differ from controls when shifted to a task in which the predictive cues (olfactory) were more salient than the irrelevant cues (spatial). This pattern of effects suggests that, for the cocaine-exposed animals, attention was “captured” by the most salient environmental cues, with the result that learning was impaired when these dominant cues were irrelevant to reward, but not when these cues were predictive of reward. This interpretation is supported by a more indepth analysis of the spatial-predictive task which revealed group differences in learning rate. The cocaine-exposed animals did not take longer than controls to move from chance performance to greater than chance performance; they were impaired only later in the task, when the distraction produced by the previously predictive olfactory cues made it more difficult for them to achieve the high performance level needed to reach criterion (Fig. 2). Note that the intact performance on the olfactory-predictive (spatial irrelevant) set-shifting task excludes various alternative explanations for the impaired performance on the spatial-predictive (olfactory irrelevant) task, such as impairments in attentional set shifting, motivation, associative ability, inhibitory control, and cognitive flexibility. In fact, in these same studies, the cocaine-exposed animals performed normally on several tasks tapping basic associative ability and memory function (55,56,59). Thus, the cognitive profile seen in individuals exposed to cocaine in utero exemplifies Pennington’s postulate that executive functions subserved by PFC may be uniquely vulnerable to early insults, and illustrates that, in some cases, many traditional tests of learning and memory may miss functionally important cognitive deficits. To assess the functional integrity of frontal cortex, we suggest including tests of attentional function rather than working memory, despite the evidence that both are subserved by the prefrontal cortex. This recommendation is based on evidence that

Cocaine

25

Saline Number of errors

20 15 10 5 0

Block 1

Block 2 Spatial EDS

Figure 2 Mean (C/KSEM) number of trials per block in an EDS task in which spatial cues were predictive and olfactory cues were irrelevant to reward. See text for details. -p!0.006. Source: Adapted from Ref. 56.

422

Strupp and Beaudin

commonly used memory tasks, such as delayed spatial alternation, may be less sensitive to subtle prefrontal cortical damage than other tests of prefrontal function, such as tests of attention and measures of species-typical behaviors. This inference is based on studies concerning early developmental lesions of the prefrontal cortex, as well as studies of in utero cocaine exposure which, as noted above, reveal a pattern indicative of prefrontal dysfunction, consistent with emerging neuroanatomical studies (60). When damage to the prefrontal cortex occurs during specific periods of early development, neural plasticity allows for some recovery of function, as demonstrated by the evidence that lesioning the prefrontal cortex during postnatal days 6–10 does not yield deficits in delayed spatial alternation tasks (DSAs) that show profound working memory deficits following similar lesions in adulthood (61–63). However, recovery of function following prefrontal cortical lesions at this time is not complete; such lesions result in permanent impairment on some tests of prefrontal function, such as those assessing the organization of species-typical behaviors (61,64,65). It appears that due to the partial recovery of function, only tests that are sensitive to subtle prefrontal dysfunction reveal impairment. This explanation is consistent with the pattern of effects described for prenatal cocaine exposure, if one posits that the attentional deficits are due to subtle prefrontal dysfunction. As described above, adult rats exposed to cocaine in utero exhibit lasting impairment in selective attention (56,57) and sustained or focused attention (54,55). In contrast, delayed spatial alternation performance did not reveal working memory dysfunction in these same animals (59). However, based on the evidence for subtle prefrontal cortical dysfunction following early lesions of this structure, it is very possible that working memory tasks that are more “executive” in nature, such as those requiring multiple operations to be performed in working memory, would also reveal dysfunction following early prefrontal cortical damage, produced by either surgical ablation or in utero cocaine exposure. Transfer of Learning (Cumulative Learning). Although low IQ scores can be produced by deficits in many different functional domains, one frequently cited hallmark of MR is an impaired ability to transfer learning from one situation to another (66,67) yet this process is rarely tested in animal models of early developmental toxicant exposure. In fact, it is most common to test experimentally naı¨ve animals on a single learning task. Because this testing situation does not assess the ability to benefit from prior relevant learning experiences, a process that is commonly impaired in MR syndromes (above citations), it is likely to result in an underestimate of the cognitive impairment produced by conditions that would manifest as MR in humans. This prediction proved to be true in rat models of both classic (68) and maternal (69) PKU, conditions known to produce MR in humans. In these studies, the animals were given a series of related olfactory discrimination tasks designed to permit positive transfer of learning between tasks. The study with classic PKU used a three-problem learning set task; the later study with maternal PKU used a nine-problem series of related tasks, comprised of three different types of tasks, with three exemplars of each. Learning transfer was defined as significantly faster learning of a given task by animals having mastered the earlier tasks in the series, relative to experimentally naı¨ve animals given only that one task. In both MR models, the PKU group benefited significantly less than controls from experience with similar discrimination problems, with the consequence that their impairment relative to controls increased across successive problems (Fig. 3). It should be emphasized that none of the tasks that comprised the sequence, when given as the only task to experimentally naı¨ve animals, revealed impairment in the PKU group, in either study. In contrast, each of these tasks revealed significant impairment in the PKU group when administered within the task series, after the animals had mastered similar tasks.

A Perspective from Animal Research 14

Trials to criterion

12

423

Control HP

10 8 6 4 2 0 First

Second

Third

Discrimination

Figure 3 Mean (C/KSEM) number of trials required to reach criterion for a learning set task administered to rats exposed to neonatal hyperphenylalaninemia (HP) (a model of classic PKU) and controls. The three 2-choice olfactory discrimination tasks were administered sequentially. Although the HP group did not differ from controls in the rate at which they learned the initial discrimination, they did not improve across the three tasks, contrary to controls. As a result, the HP group was significantly impaired relative to controls on both the second and third discriminations. -pZ0.03, --pZ0.02. Source: Adapted from Ref. 68.

Thus, a task series designed to allow for positive transfer of learning between tasks revealed impairments in these MR models that would have been missed if any of these tests had been administered singly. These findings support the importance of including assessments of learning transfer in studies of disorders known or hypothesized to cause MR or developmental delays, including early toxicant exposure. Regulation of Affect or Emotions. One aspect of brain function that has been conspicuously neglected in most studies of early toxicant exposure is emotion or affect. This omission is likely to have resulted in a significant underestimate of the functional effects of the toxicant under study; for example, impaired emotion regulation is integral to various types of psychopathology, including anxiety disorders, depression, conduct disorder, and delinquent behavior (70,71), which can be an important cause of impaired cognitive functioning and school performance (72–74). The suggestion that chemical exposure might cause social and behavioral problems in addition to cognitive deficits is not new. Over 60 years ago, in their landmark study on early childhood lead poisoning, Byers and Lord (75) raised the possibility that the poor school progress in children with a history of lead exposure might be related not only to their reduced cognitive abilities, but also to problems in regulating aggression. Several decades later, Weiss (76) suggested that exposure to environmental chemicals might contribute to social and behavioral pathologies, and recently Bellinger (77) argued that the assessment of emotional and affective functioning is an important and severely neglected area of inquiry in the field of neurotoxicology. Fortunately, the situation seems to be improving in recent years, with a growing number of studies of early toxicant exposure including measures of psychopathology. For example, several recent studies have reported links between indices of early lead exposure and the incidence of anti-social and delinquent behaviors in adolescence (77,78,80). Although these correlations are provocative, their interpretation is

424

Strupp and Beaudin

complicated by the frequent confounding of early toxicant exposure and disadvantaged social circumstances (e.g., poverty, low intellectual stimulation in the home, exposure to violence, maternal depression, etc.), factors which themselves place children at risk for delinquent and antisocial behavior (81,82). This is a common problem in toxicology research, as environmental health risks are borne disproportionately by members of the population who are poor (83). For this reason, animal models can be pivotal in gauging causality, assuming, however, that one has confidence that the selected behavioral measures are relevant for human psychopathology. Admittedly, such extrapolations are less straightforward than for indices of cognitive functioning such as learning rate or memory function. Assessing emotions relevant to human psychopathology. One approach is to assess the animals’ behavior in situations that induce emotions thought to be relevant to human psychopathology, such as anxiety, fear, or aggression. In light of the evidence that impaired regulation of these emotions is a risk factor for various types of psychopathology (70), the assessment of the animals’ reaction to these emotive situations seems an appropriate tack for modeling psychopathology risk in the animal. Moreover, there is evidence for cross-species correspondence in the neural systems that underlie these emotions (84-88). For some tests of emotions in animals, convincing evidence for the validity of cross-species extrapolations have been demonstrated. Such is the case for the elevated plus maze as a test of anxiety. In this task, a rat or mouse is placed on an elevated 4-arm maze in which two of the arms are covered, and two are open (i.e., exposed). Rodents prefer dark or enclosed spaces, consistent with evolutionary selection pressures to avoid predation. On the other hand, they have a propensity to explore novel environments, creating an approach-avoidance conflict with respect to exploring the open arms of the maze. The proportion of time that the animal spends in the open arms as well as the number of entries into the open arms, provide indices of anxiety (89,90). Anxiogenic and anxiolytic drugs alter these dependent measures in the predicted directions (91,92), supporting the validity of this tool as an index of anxiety. Another approach used to assess emotion or affect in animals examines behavioral responses to stressors such as electric shock, forced swimming, physical restraint, or the behavioral response to novelty (93). Lasting changes on such measures have been reported following early exposure to toxicants, such as lead (94–102) or cocaine (103–108). However, the way in which such findings can be used to predict effects in exposed children is less clear than for the plus maze; validation studies are lacking. Similarly, although altered behavior of animals in tests of social interactions is commonly cited as evidence for an increased risk of psychopathologies in which social behavior is impaired (e.g., autism), such inferences await validation. This is also the case for animal tests purported to provide indices of “depression,” such as the Porsolt swim test (109,110) and the learned helplessness model (111,112). Assessing reactivity to errors during testing. Another approach is to study the regulation of emotions within the context of cognitive testing; specifically, assessing the animals’ reaction to committing an error. This approach has the advantage that it may shed light on the basis of observed performance deficits, while simultaneously providing an assessment of emotion regulation. Although this approach to assessing emotion regulation also awaits validation in terms of extrapolation to humans, the underlying assumptions and interpretive leaps appear to be less than with some other measures in this area, as discussed above. We have studied the sequelae of committing an error in a wide variety of tasks, designed to tap a range of functions, including sustained attention, selective attention, attentional set-shifting, stimulus-reward learning,

A Perspective from Animal Research

425

and stimulus-response learning. These analyses were conducted on performance data collected after task rules had been mastered by the animals, such that errors were likely due to impulsivity or inattention, rather than due to learning dysfunction or uncertainty about task contingencies. It is very clear from these data that the outcome of the previous trial (i.e., correct or incorrect) is a powerful determinant of performance. Various performance measures demonstrate that both mice and rats (1) notice when they have committed an error (i.e., exhibit error-monitoring behavior); and (2) perform more poorly on trials that follow an error. For example, in attention tasks in which the location and onset time of a brief visual cue varied unpredictably between trials, several dependent measures varied significantly as a function of the outcome of the previous trial; specifically, on trials that followed an error (relative to trials following a correct response), the animals took longer to enter the testing alcove at trial onset, took longer to make a response, and were more likely to commit all types of errors: premature responses (responding before cue onset), inaccurate responses (responding after cue onset but to an incorrect port), and omission errors (missing the cue) (54,55,59,113). The nature of this effect might not have been predicted. A priori one might have hypothesized that an error on one trial might lead to a very low error rate on the next trial because the subject would slow down and respond less impulsively. This pattern, increased response latency and decreased error rate on post-error trials, has been reported in some human studies (114,115), and is thought to reflect an executive, errorcorrection system, localized to the anterior cingulate cortex (116,117). However, the pattern we observed in our rodent studies-increased response latency and increased error rate on post-error trials-has also been reported in some human studies (118), and likely reflects an emotional response to the error. This interpretation is supported by the finding that an electrophysiological marker of error detection, termed the error-related negativity (ERN), varies as a function of individual differences in negative affect and emotionality (119). Whether one observes increased or decreased accuracy on post-error trials in human subjects appears to depend on subtle differences in task parameters, such as the duration of the stimulus-response interval (RSI), whether or not a correction technique was employed, or both (114). In our rodent studies, the reaction to committing an error also appears to entail cognitive and effective components, based on data from an olfactory conditional discrimination task (120). In this task, only one of the two olfactory cues was presented on a given trial. One cue signified that a right nose-poke response would be rewarded, whereas the other olfactory cue signified that a nose-poke to the left port would be rewarded (i.e., conditional stimulus-response associations). An interesting statistical interaction was observed between the outcome of the previous trial and “stimulus repetition,” the latter term connoting whether the error trial and the post-error trial involved the same olfactory cue or a different cue (Fig. 4). Specifically, the disruptive effect of an error was significantly greater if the cue presented on error and post-error trials was the same, relative to the situation where the two trials involved different olfactory cues. The most likely explanation for this pattern of findings is that when the same cue was presented again on the post-error trial, the greater increase in response latency and error rate occurred because the animal was distracted by continued processing of the events of the previous trial. In contrast, when the alternate cue was presented on post-error trials, the “slate was wiped clean,” with the result that both response latency and accuracy were less affected. This observed interaction of stimulus repetition and previous trial outcome implies processing at the cognition-affect interface, and indeed reflects a level of cognitive processing that some may find surprising for the rat. A purely emotional response would

426

Strupp and Beaudin

100

Correct Response (%)

95

Alternate Repeated

90 85 80 75 70 0

Error Correct Outcome of the Previous Trial

Figure 4 Mean (C/KSEM) percent correct (accuracy) in a 2-choice olfactory conditional association task in which the odor presented on each trial signified whether a left or right response would be rewarded. Accuracy is presented as a function of (1) the outcome of the prior trial and (2) whether the same or alternate stimulus was presented on the current trial. Accuracy was significantly lower on trials following an error than on trials following a correct response. Moreover, this post-error reduction in accuracy was more pronounced when the same stimulus was presented as on the error trial, relative to when the alternate stimulus was presented. See text for additional details. -p!0.0005. Source: From Ref. 123.

likely have produced a main effect of previous trial outcome but not an interaction with the stimulus characteristics. In our animal studies, we have attempted to tap the cognition-affect interface by examining each performance measure (e.g., rate of various types of errors, response latency, latency at trial onset) as a function of prior trial outcome (correct or incorrect). This approach has illuminated the basis of group differences in performance in numerous studies and has proven to be a uniquely sensitive tool for detecting the lasting effects of various early insults. Measures of the degree of disruption produced by an error have revealed lasting effects of all insults that we have examined in this regard, including lead (113), cocaine (54,55,59), alcohol (121), and Down syndrome (122). One possible reason for the apparent sensitivity of these measures is that error monitoring and error reactivity—and emotion regulation in general—are thought to depend on a neural system that includes frontal cortex (70,71), consistent with Pennington’s theory discussed above. The insight provided by this type of analysis is illustrated by some recent findings from a study designed to test the efficacy of succimer (Chemete) chelation in ameliorating the lasting cognitive and affective changes produced by early lead exposure. Adult rats exposed to lead early in life exhibited a much greater disruption in performance following an error than controls, in both the olfactory conditional discrimination task described above (123), and in a sustained attention task (124,125). The approach of examining performance on each trial as a function of prior trial outcome revealed not only a functionally important effect of early lead exposure, but also an area of lead-induced dysfunction that was completely normalized by chelation with succimer (Fig. 5).

A Perspective from Animal Research

427

Correct responses (%)

100 90 80 70 Lead/ VEH Lead/ SUCC

60

Control / VEH 50 Error

Correct SRO Outcome of the Previous trial

Figure 5 Mean percent correct (C/KSEM) during the final block of trials in each daily session (trials 135–200) for a two-choice olfactory conditional associative task with surprising reward omission (SRO). For both the lead-exposed and control groups, performance was significantly lower on trials that followed an error than on trials that followed either a correct response or an SRO. However, this drop in performance following an error was significantly more pronounced for the lead-exposed animals than for controls (--p!0.01), indicating impaired emotion regulation. Importantly, chelation therapy with succimer normalized this area of impairment of the lead-exposed animals (Pb-veh vs. Pb-succ; -p!0.05). Source: From Ref. 123.

As discussed below, some areas of lead-induced dysfunction did not improve following chelation therapy. Elucidating the Nature of the Dysfunction Limits to Defining the “Behavioral Phenotype” in Toxicology Research The selection and design of neurobehavioral tasks is guided not only by concerns relating to sensitivity (i.e., detecting dysfunction), but also by the goal of gaining insight into the functional integrity of specific cognitive and/or affective processes. Before discussing the benefits of designing tasks in such a way that they provide this information, it is important to acknowledge the limits that exist to defining a behavioral phenotype produced by exposure to a particular neurotoxicant, the rationale often advanced for using this type of test battery. As Bellinger (77) concluded in his discussion of this issue, merely knowing that an individual has been exposed to a particular neurotoxicant at some point in his/her life “.provides little information that is useful in predicting the specific forms in which toxicity will be expressed.” This situation reflects the fact that many factors modify the nature of the neurotoxic effects that are produced, including aspects of the exposure (i.e., dose, timing, chronicity) and various sociodemographic characteristics of the population being studied (e.g., social class, sex). For example, it is well-known that early developmental exposure to a given toxicant or teratogen can produce qualitatively different behavioral and neural effects, depending on the particular developmental events that are taking place at the time of exposure (126,127). This point can be illustrated by the profiles of spared and impaired cognitive functions produced by exposure to lead during different periods of early development (113). Consequently, one may be limited to defining the impaired and

428

Strupp and Beaudin

spared functions for a given exposed individual (which may be useful clinically, as discussed below) or, for a given population of exposed individuals, identifying cognitive functions that, at a probabilistic level, are more likely to be affected than others.

Benefits of Using Tests which Provide Indices of Specific Functions Nonetheless, the use of tests that provide information about the integrity of specific functions can still offer significant benefits. First, as noted above, in cases where the dysfunction produced by the toxicant is selective, tests designed to specifically tap the affected function(s) are likely to produce a more sensitive index of the dysfunction than global or apical measures. Second, politicians are more likely to take notice of behavioral evidence indicative of a known behavioral/cognitive syndrome (i.e., impaired attention and impulse control, suggesting ADHD) than impaired performance on tests that, even if sensitive, do not provide insight into specific functions. Examples of tests in this latter category are the Rey-Osterrieth Complex Figure for children (128) and, in the case of animal testing, operant reinforcement schedules (129). In addition, unlike apical or global measures, the use of tasks (and dependent measures) that provide insight into the integrity of specific functions can be useful clinically to guide the choice of interventions. For example, if the task battery reveals that attention is impaired but that basic associative processes and memory are intact, then methylphenidate (Ritalin) would likely be an effective treatment; this would not be the case with the reciprocal pattern of spared and impaired functions. Delineating the spared and impaired functions can also shed light on the nature of the underlying brain damage. However, one caveat should be noted in this regard: this approach infers neural damage by cross-referencing information about specific functions that are altered by the toxicant of interest with knowledge concerning the neural systems that subserve those functions. However, for studies of toxicant exposure during early development, it is a concern that most of this knowledge is based on the effects of focal lesions sustained in adulthood; the effects of such lesions are often different from those produced by early developmental lesions of the same structure. For example, damage to the medial temporal lobe in adulthood leads to a rather pure amnesic syndrome in which explicit memory function is the only observed impairment (14,32,33), whereas the comparable lesion induced during early development produces autistic-like behavior in addition to impaired explicit memory function (130–132). Additional studies of lesions sustained during early development would significantly aid in this endeavor. A final benefit of selecting tasks that provide indices of specific functions pertains to studies concerning the efficacy of putative therapeutic interventions, such as administering chelating agents following heavy metal exposure. Some cognitive or affective dysfunctions produced by the toxicant may be ameliorated to a greater extent than others by the treatment, due to various factors, including, for example, the developmental stage of different neural systems at the time that the exposure occurred as well as the time that the intervention was implemented. Neural systems that have completed their development during the time of the toxicant exposure (and before the therapy were implemented) are unlikely to derive as much benefit as those that are still developing during the time that the treatment is given. A recent study of succimer chelation of lead-exposed rats illustrates this scenario (124,125,133). In such situations, erroneous conclusions would likely be reached unless a broad range of functions are independently assessed.

A Perspective from Animal Research

429

Tests Purporting to Assess Specific Functions Often Do Not Although tasks are frequently marketed or described as providing a rather specific index of some cognitive function, this is often not the case, as exemplified by the DSA task commonly cited as a measure of working memory function in both animals and humans. In this task, the subject is rewarded for alternating responses between two spatial locations on successive trials. For example, if the subject responded to the left location on trial n, then the right location is correct on trial nC1. Thus, the subject must remember the location of his/her previous response in order to respond correctly on the next trial. The delay between trials can be manipulated to vary the demand on memory function. However, poor performance in this task can result not only from memory dysfunction but also from deficits in any of the other processes also required for optimal performance in the task, including spatial perception, attention, behavioral modulation (i.e., impulsivity), susceptibility to proactive interference, inhibitory control (selecting the location that was rewarded on the previous trial), and motivation. Therefore, if the delay between consecutive trials is constant for the entire session, which is common, particularly in human studies (e.g., when administered within the CANTAB program), observed dysfunction is not very informative, as it could be due to alterations in any or all of these processes. Much greater insight can be gained about the nature of the dysfunction if the duration of the intertrial interval (ITI) is varied randomly across trials in the session. If the groups do not differ on trials with a 0 second ITI (i.e., no delay between consecutive responses), and the magnitude of the impairment increases with increasing ITI, then memory impairment would be implicated. However, if the performance of the two groups differs even on trials with no delay, then the deficit would not likely be due to memory specifically. Moreover, in this latter scenario, if the slopes across the varying delays are

85

Control Lead

Correct responses (%)

80 75 70 65 60 0 0

10 20 Intertrial Delay (Seconds)

40

Figure 6 Mean performance (percent correct C/KSEM) in a delayed spatial alternation task as a function of the intertrial delay. The inferior performance of the lead-exposed animals (p!0.01), which did not vary by delay, was most likely due to an impaired ability to regulate the frustration produced by having to wait for unpredictable and often long delays between trials, after having been trained on a task with all 0 second intertrial delays. This inference is based on (1) the unimpaired performance of the Pb-exposed rats in this training task, and (2) the results of an in-depth analysis of performance in the DSA task, which excluded several alternative explanations. Source: Adapted from Ref. 134.

430

Strupp and Beaudin

comparable for the two groups (parallel), one can exclude memory impairment as the basis of group differences despite the poor performance of the experimental group. This pattern is illustrated by the effects of post-weaning low-level lead exposure in the rat (Fig. 6) (134). Varying Task Parameters Systematically varying task parameters, such as described here for the DSA task is often an effective means of gaining insight into the integrity of specific functions because information is then provided concerning the particular conditions under which the subjects succeed and fail. This approach can often effectively exclude alternative explanations for poor performance, and thereby specify the nature of the impairment. For example, in visual attention tasks in which subjects must wait for and then respond to a brief visual cue, unpredictable in terms of both location and onset time, it is very informative to systematically vary several parameters of visual cue presentation, and then randomly present the different “levels” of each variable across trials; examples include varying (1) the delay between trial onset and presentation of the visual cue (which varies the demands placed on both inhibitory control and sustained attention); (2) the duration of cue illumination (which alters the difficulty of detecting the cue; hence, varying the demand on focused or sustained attention); and (3) whether or not a distractor is presented on a given trial (varying demands on selective attention). The effects of prenatal cocaine exposure in this task are illustrative. The exposed animals were impaired (relative to controls) on trials where the visual cue was presented very briefly (200 ms) but not on trials with longer cues (400 or 700 ms) (Fig. 7) (55). The unimpaired performance on trials with the longer cues allows one to exclude numerous potential interpretations (e.g., decreased motivation, impaired learning or memory of task rules, impaired motor function), rendering attentional dysfunction as the most likely basis of the observed impairment. 92

Saline

90

Cocaine

88 Accuracy (%)

86 84 82 80 78 76

*

74 0 200

400 Stimulus Duration (msec)

700

Figure 7 Percentage (C/KSEM) accuracy for a sustained attention task in which the presentation of a brief visual cue varied in location, onset time, and duration. A significant Treatment ! Duration interaction was observed (p!0.02). The rats exposed to cocaine in utero exhibited significantly reduced accuracy, relative to controls, specifically on trials with the briefest cue (pZ0.01). Source: Adapted from Ref. 55.

A Perspective from Animal Research

431

In these attention tasks, it is also informative to examine group differences as a function of “time on task,” e.g., early, middle, or end of session. Deficient sustained attention would manifest as impairment that increases across the testing session (54), whereas dysfunction that is specific to the beginning of the session would indicate difficulty in “settling in” to the task, possibly due to impaired regulation of the arousal produced by the recent handling, weighing, placement in the testing chamber, etc. Clues regarding the nature of the dysfunction is often provided by categorizing the types of errors committed (e.g., premature response, omission error, inaccurate response), and then evaluating each error type as a function of these various parameters (delay before cue onset, cue duration, trial block [portion of session]), as well as the outcome of the previous trial, as discussed above. For example, in this visual attention task, we found that adult male rats exposed to cocaine in utero committed more omission errors than controls only on trials in the final third of the testing session that occurred after an error (Fig. 8, the 3.0 mg/kg dose). These animals were not impaired in this final portion of the session on trials that followed a correct response, or earlier in the session regardless of prior trial outcome. This pattern implicates the additive effects of impairments in two areas: sustained attention and emotion regulation. Demarcating Phases of the Learning Process The interpretation of task results involving learning rate data may seem more straightforward; however, here, too, various types of dysfunction may underlie slower learning of the experimental group. For example, slower learning in serial reversal learning tasks could be due to an alteration in one or more of the following processes (in addition to impaired motivation or sensory acuity): (1) cognitive flexibility/ability to inhibit responses to previously rewarded stimuli, (2) attentional function (selectively attending to the predictive cues), (3) associative ability (associating the correct cue with reward), and (4) transfer of learning across tasks, or rule induction. One way to gain insight

12 control Omission errors (%)

10 8

0.5 mg/kg 1.0 mg/kg 3.0 mg/kg

6 4 2 0 Error Correct response Outcome of the Pervious Trial

Figure 8 Percentage omission errors in the final block of trials in each daily testing session (trials 135–200) in a sustained attention task, as a function of the outcome of the previous trial (correct or incorrect), for male rats exposed to cocaine prenatally and controls. The interaction of treatment group and prior trial outcome was significant [F(3,363)Z10.12, p!.0001]. The highest dose males committed significantly more omission errors than controls on trials following an error (pZ.002) but not on trials that followed a correct response (pZ.3). Source: Adapted from Ref. 54.

432

Strupp and Beaudin

into the specific cognitive functions that are altered as well as the neuroanatomical locus of the dysfunction is to analyze the length of discrete phases of the learning process (36,135,136). Mastery of reversal learning tasks can be demarcated into several phases: (1) a period of persistent responding to the previously correct cue immediately after the contingencies have been changed, called the perseverative phase; (2) a period during which performance is at chance levels, due to either random selection of the two stimuli or repeated responding to the same spatial location; and (3) a final phase during which performance is significantly greater than chance levels, characterized by increasingly accurate selection of the correct cue. A lengthened perseverative phase coupled with a normal post-persaverative phase would be indicative of deficient inhibitory control or inflexibility, as seen with damage to dorsolateral prefrontal cortex (137). A lengthened post-perseverative phase, coupled with a normal period of perseverative responding, would suggest associative dysfunction, as seen with damage to a circuit involving amygdala and OFC (138,139). Previous research involving a rodent model of early childhood lead exposure illustrates how this type of phase analysis can shed light on the integrity of specific cognitive functions. Chronic exposure to lead from conception produced both a doserelated shortening of the perseverative phase (a shortened period of responding to the previously correct cue), and a dose-related lengthening of the post-perseverative phase (Fig. 9) (136). These data illustrate that, although it is commonly assumed that slower reversal learning is due to inflexibility or deficient inhibitory control, this is not always the case. The slower learning of the lead-exposed animals was caused by impaired associative ability, not inflexibility; in fact, the exposed animals exhibited less perseverative responding to the previously correct cue than controls, despite their slower learning. In addition, these findings exemplify the greater sensitivity that can be provided by this type of learning phase analysis. Although the higher of the two exposure groups learned significantly more slowly than controls (in terms of trials to criterion), learning rate of the lower exposure group did not differ from that of controls despite the fact that they differed 400 350

# Trials

300

0 ppm 75 ppm 300 ppm

250 200 150 100 50 0 Perseverative

''Chance''

Final

Learning phase

Figure 9 Mean number of trials (C/KSEM) for each of the three phases of learning in a series of 2-choice olfactory reversal learning tasks. The perseverative phase, the period during which the rats persistently selected the previously correct cue, was significantly shorter for both lead-exposed groups relative to controls (p’sZ.05, and .06, respectively, for the 75 and 300 ppm groups). In contrast, the post-perseverative phase (comprising the chance and final learning phases) was significantly protracted for both lead-exposed groups relative to controls (pZ.08, and .002, respectively). Source: Adapted from Ref. 136.

A Perspective from Animal Research

433

significantly from controls in the durations of both the perseverative and postperseverative phases. Interestingly, these two effects cancelled each other out in terms of overall learning rate, with the result that impairment in this group was uncovered only in these analyses of the specific learning phases.

SOME CAUTIONARY NOTES Caveat About Automated Operant Tasks Automated operant tasks, such as those described in many of the cited examples, offer numerous benefits but also some limitations. One benefit is that they facilitate the systematic manipulation of task parameters, and thereby aid in defining the nature of the dysfunction. In addition, the fact that one collects hundreds of trials per animal on each task permits the delineation of different stages of learning, as well as increases the reliability of the results. The automated nature of the tasks also reduces experimenter effects and facilitates replication of findings. Additionally, the technology is ideal for assessing various aspects of attention (e.g., sustained attention, selective attention). However, this type of testing situation is not optimal for assessing all aspects of cognitive functioning, notably including hippocampal function. Automated operant tasks are generally learned quite incrementally and consequently do not depend on the hippocampus (30); i.e., they fall into the category of implicit, rather than explicit, learning. Tests likely to be sensitive to hippocampal dysfunction are those where the basic concept can be learned quickly (30), those that tap memory of discrete objects or events, such as the delayed-non-match-to-sample test, or maze tasks that tap spatial mapping strategies (140,141). Despite the fact that most of these automated operant tasks are mastered incrementally, like habits, they can be modified, after the basic task rules have been learned, to assess more complex, higher-order processing, including various aspects of attention. In addition, if multiple versions of the same task are presented successively, as in olfactory learning set tasks, the later tasks are learned very rapidly by both rats (69) and mice (142), thereby likely falling into the realm of explicit (i.e., hippocampal-dependent) learning and memory (30).

Using “Identical” Tasks for Different Species or Different Age Groups There is growing interest in developing tests for animals that are identical or “homologous” to those used with human subjects (143). It is hoped that this approach will facilitate translating the findings of the animal studies to the human condition being modeled. The reciprocal strategy is also being employed, namely, to develop tasks for humans that are identical or very similar to ones used in animal studies (144–146). Because animal studies have often provided information on the neural bases of specific patterns of dysfunction in these tasks (137,138), it is hoped that the use of these tasks in children will aid in identifying not only the areas of functional impairment but also the locus of the underlying neural damage. However, it is important to bear in mind that a given task (even if identical) may not be solved in the same way by animals and humans, or by children of different ages. Therefore, one cannot assume that merely because a toxicant-exposed animal performs normally on a given task that the same will hold true for children exposed to the same toxin, or vice versa. This point can be illustrated by considering the effects of medial temporal lobe damage on pattern discrimination learning. As discussed above, non-human primates with damage to the medial temporal lobe do not

434

Strupp and Beaudin

differ from controls in their rate of learning 2-choice pattern discriminations, whereas humans with similar brain damage are significantly impaired in this type of task (30). This difference reflects the fact that the monkeys learn this task incrementally, like a skill or habit (implicit learning), whereas humans learn this type of task very rapidly, moving almost instantaneously from chance-level responding to perfect performance. Because the task is solved differently by monkeys and humans, the neural systems on which the task depends differs between the two species. Similarly, because nearly all tasks depend on multiple cognitive processes, one cannot assume that poor performance on a given task by toxicant-exposed animals and humans necessarily reflects the same underlying dysfunction, at either a cognitive or neural level. To illustrate this point, consider the DNMTS performance of adult monkeys with medial temporal lobe damage, human toddlers, and infant monkeys. Each of these groups does poorly on the DNMTS task, but for different reasons. Adult monkeys with lesions of the medial temporal lobe do poorly on this task because of impaired explicit memory function, as indicated by the pattern of performance across the varying retention intervals: they do well on trials with short retention intervals (10–15 sec) and perform progressively worse as the retention interval increases (147). In contrast, human toddlers and infant monkeys fail even at the shortest retention interval (5 sec) (148,149). Their failure is almost certainly not due to poor memory because (1) toddlers of the same age can remember a variety of things for 5 sec, and (2) when they get older and finally succeed at the short delay, they also succeed at the longer delays (e.g., 30 sec). The impaired performance of the human toddlers and infant monkeys likely reflects immaturity of inhibitory control processes. This inference is based on the facts that (1) correct performance requires inhibition of the choice that was just rewarded (during the sample phase); and (2) the evidence, from several types of tasks, that the ability to inhibit prepotent responses is deficient in toddlers and infant monkeys (150–152). As this ability improves with age, the subjects do well at all the delays. For these reasons, before one can extrapolate the results of a given task across species or age groups, one must verify that the task is solved in the same way by these different groups of subjects. This is true both for predicting whether or not dysfunction will be observed in the other group of subjects, as well as for predicting the nature of the functional or neural damage. This latter caveat pertains to using animal data concerning the neural bases of dysfunction in the task to identify possible loci of neural damage in the target human population. Clues concerning whether different groups of subjects solve a given task in the same way may be provided by considering the specific trial conditions under which subjects succeed and fail, as well as by examining information, if available, on the effects that damage to particular brain regions has on task performance in these different groups of subjects (36). Interpreting the Results of Studies Evaluating Environmental Enrichment and Neurotoxicant Exposure Recently, there has been great interest in studies reporting that environmental enrichment (EE) can ameliorate the cognitive and neural damage produced by early exposure to neurotoxicants, such as lead (153,154). In these studies, EE, consisting of group housing in a large cage with toys, was found to reduce the performance difference between leadexposed rats and controls on a learning task, relative to that seen when the two groups were raised in impoverished conditions (i.e., social isolation and no environmental stimulation). This pattern has been reported for many types of brain damage, including surgical lesions of prefrontal cortex, sensorimotor cortex, reticular thalamus, or hippocampus, as well as

A Perspective from Animal Research

435

early exposure to drugs such as alcohol, MAM, alpha2 agonists, or beta blockers (155,156). Such findings have given rise to the hope that the dysfunction produced by exposure to neurotoxicants may be significantly ameliorated or even “reversed” by EE (154,157). Unfortunately, such hopes may rest on faulty logic. Several issues must be considered when drawing inferences from this type of EE study. One issue is whether the selected task (and task parameters) allowed for the detection of an enrichment effect in the controls (i.e., enrichment vs. isolation reared groups) that was at least as large at that seen in the brain-damaged animals. Many articles focus primarily on the extent to which the brain-damaged group benefited from the enrichment. This is important, but it is also crucial to consider whether the enrichment created group differences that were larger, smaller, or equivalent to that seen following rearing in the isolated conditions. In many EE studies, a floor or ceiling effect precluded the detection of an equivalent benefit of the EE for the controls. In such experiments, where the control group derives little benefit from the enrichment, the finding of amelioration of the deficits in the brain-damaged group (i.e., smaller group differences) should be viewed as an experimental artifact that would not apply to the real world. In fact, in studies where the task parameters permitted the detection of a sizeable enrichment effect in the controls, the impairment of the brain-damaged group (relative to controls) has often been larger in the enriched condition than the isolated condition (158–160), suggesting that the controls benefited from the enrichment more than the brain-damaged animals (Fig. 10). This type of finding is consistent with the prediction that the control animals would be superior to the brain-damaged animals in transferring knowledge and experiences from the enrichment situation to the subsequent learning tasks (35). In this latter scenario, enrichment improves the performance of the brain-damaged animals— which may be important clinically—but because it actually makes group differences larger, it seems inappropriate to call this curative. Parenthetically, whether enrichment enlarges or reduces group differences may depend not only on measurement issues 25

6-OHDA Saline

Error per problem

20

15

10

5

0 Isolated

Complex Environment

Figure 10 Average (C/KSEM) errors per problem in the Hebb–Williams maze for normal rats and those treated with 6-hydroxydopamine (6-OHDA), following rearing in a complex or an isolated environment. Group differences were seen only following rearing in the complex environment. Source: Adapted from Ref. 159.

436

Strupp and Beaudin

(as discussed above), but also on the nature of the benefit afforded by the enriched environment. If the derived benefit was cognitive in nature (i.e., transferring specific information from the EE to the task), the controls are likely to benefit to a greater extent than the brain-damaged animals, for the reasons noted above. In contrast, if the enrichment improved performance because it reduced emotionality (relative to the isolated condition), then the brain-damaged animals might benefit more than controls, particularly if the neural damage impaired arousal regulation (35). A second critical issue—even in cases where enrichment decreases group differences and ceiling effects do not seem to have artificially produced this type of interaction—is whether these results should be interpreted as evidence that EE was therapeutic or, alternatively, that the isolated condition exacerbated the dysfunction. Which interpretation one uses is not just a semantic issue; it has important implications for how this type of finding is translated into public policy. Pertinent evidence is provided by the few studies that have examined various brain parameters (e.g., cortical thickness) as a function of three levels of environmental stimulation: (1) the impoverished condition, consisting of either isolation (single housing and no enrichment) or standard laboratory housing (housing with one or two other animals, but no EE); (2) the standard enrichment condition, i.e., group housing in a large cage with changing toys; and (3) super-enriched lab conditions or wild-caught animals. In these studies, significant differences were seen between the enriched and impoverished conditions, but not between the standard enrichment condition and wild-caught animals (161–164). It follows that many of the behavioral and neural differences observed in the typical EE study may reflect the adverse effects of impoverished conditions rather than the beneficial effects of enrichment. By analogy, the same may be true for the studies showing an amelioration of some early insult by EE, namely, that the observed interaction between environment and early brain damage may reflect an exacerbation of the early insult by impoverished conditions, rather than an amelioration of the brain damage by enrichment. Which of these interpretations is correct for any given endpoint can be determined only by conducting this type of three-level enrichment study, involving impoverished, enriched, and super-enriched conditions. Some differences observed between animals reared in enriched and impoverished conditions, such as increased survival of newborn neuronal cells (165–167), provide a plausible mechanism by which EE could truly facilitate recovery of function following brain damage. However, unless it can be demonstrated that “super-enrichment” reduces group differences relative to an “enriched” condition, EE studies should not be cited as support for the notion that the effects of the early insult in children can be reversed by enrichment programs, as has been done; such programs would increase the child’s level of stimulation between levels of enrichment, not from impoverished to enriched conditions (when considered in terms of the animal studies).

SUMMARY In this chapter, we discussed the selection and design of cognitive tests for use in developmental neurotoxicology studies, raising issues that we believe are important for both animal and human studies. Two overarching considerations were discussed: (1) sensitivity (being able to detect dysfunction if it exists), and (2) specificity (being able to say something about the nature of the dysfunction). Creating a sensitive assessment battery begins with the acknowledgement that because different cognitive and affective functions can be independently altered, it is not possible to select some “representative” learning/memory task and assume that if “cognition” is altered, dysfunction will be detected

A Perspective from Animal Research

437

in this one task. Rather, it is necessary to assess a wide range of functions which, in turn, reflect the functional integrity of many distinct neural systems. A second implication of the biological distinctiveness of different cognitive and affective functions is that tests cannot be ranked in terms of sensitivity; the test that will be most sensitive in any given case depends on the nature of the brain damage. In practice, animal studies in this area have tended to rely on maze tests or other tests of associative learning and/or memory function; certain functional areas are rarely assessed. We identified four neglected areas of functioning that may be especially important to include in future studies, as they are commonly affected by early insults: (1) executive functions (including various components of attention); (2) transfer of learning across tasks; (3) emotion regulation (particularly within the context of cognitive testing, such as reactivity to committing an error); and (4) intrinsically-motivated learning (including latent learning and incidental learning). Assessing the ability to transfer learning from one situation to another, a hallmark dysfunction in human MR syndromes, seems particularly important for studies of early insults, including developmental toxicant exposure. In addition, there is evidence that functions subserved by frontal cortex, including attention and emotion regulation, may be uniquely sensitive to early developmental insult, and therefore particularly important to assess following early toxicant exposure. The failure to assess these functions in many prior studies is likely to have resulted in a significant underestimate of the functional consequences of the toxicant being studied. General statements about the behavioral phenotype or “signature” of a given toxicant are precluded by evidence that characteristics of the exposure (e.g., dose, duration) and the target population (e.g., age, SES) can alter the types of effects that are produced. Nonetheless, there are numerous benefits of designing cognitive tests in such a way that they provide indices of specific cognitive and affective functions. One approach for attaining this objective, illustrated with the test DSA task and automated visual attention tests, is to systematically vary task parameters (e.g., the retention interval, the delay prior to cue presentation, cue duration) so that one can characterize the specific conditions under which the exposed subjects differ from controls. For tasks in which learning rate is the primary dependent measure, insight into the nature of group differences can be gained by demarcating the learning process into distinct phases, and then comparing the rate at which the groups master each phase. The effects of early lead exposure on serial reversal learning illustrate the benefits gained by this approach. The final section of the chapter touched briefly on some “cautionary notes” concerning the use of animal models in this area. The first pertained to the use of automated operant testing methods. Although beneficial for many reasons, this technology may not be optimal for assessing all areas of functioning, notably including explicit memory, subserved by the medial temporal lobe. Nonetheless, it is possible to devise automated operant tasks that are sensitive to hippocampal dysfunction; some examples are given. A second cautionary note pertained to the use of identical tasks in different species or different age groups. Because a given task may be solved in different ways by these different groups of subjects, one cannot assume that the task results can be generalized. However, such extrapolations may be valid, if it can be demonstrated that the task is solved in the same way by the different groups of subjects, and that it depends on the same neural system(s). The final cautionary note pertained to the interpretation of studies designed to test the ability of EE to ameliorate the effects of early brain damage, including that produced by toxicant exposure. Although some studies have reported that EE reduces the magnitude of the behavioral or neural dysfunction, relative to that seen following rearing in impoverished environments, these studies do not allow the inference that

438

Strupp and Beaudin

enrichment programs for similarly-exposed children will reverse the functional or neural damage. This conclusion is based on the facts that: (1) in many such animal studies, ceiling effects precluded the detection of a sizeable enrichment effect in the controls; and (2) many of the differences seen between enriched and isolated animals appear to reflect the adverse effects of isolation, not the beneficial effects of enrichment. Accordingly, it is possible that the EE studies involving early brain damage reflect an exacerbation of the early insult by isolated rearing, not a therapeutic effect of enrichment.

REFERENCES 1. Burbacher TM, Rodier PM, Weiss B. Methylmercury developmental neurotoxicity: a comparison of effects in humans and animals. Neurotoxicol Teratol 1990; 12:191–202. 2. Rice DC. Sensory and cognitive effects of developmental methylmercury exposure in monkeys, and a comparison to effects in rodents. Neurotoxicology 1996; 17:139–154. 3. Winneke G, Lilienthal H, Kramer U. The neurobehavioural toxicology and teratology of lead. Arch Toxicol Suppl 1996; 18:57–70. 4. Banks EC, Ferretti LE, Shucard DW. Effects of low level lead exposure on cognitive function in children: a review of behavioral, neuropsychological and biological evidence. Neurotoxicology 1997; 18:237–281. 5. Davis JM, Otto DA, Weil DE, Grant LD. The comparative developmental neurotoxicity of lead in humans and animals. Neurotoxicol Teratol 1990; 12:215–229. 6. Driscoll CD, Streissguth AP, Riley EP. Prenatal alcohol exposure: comparability of effects in humans and animal models. Neurotoxicol Teratol 1990; 12:231–237. 7. Schantz SL. Developmental neurotoxicity of PCBs in humans: what do we know and where do we go from here? Neurotoxicol Teratol 1996; 18:217–227. 8. Cador M, Robbins TW, Everitt BJ. Involvement of the amygdala in stimulus-reward associations: interactions with the ventral striatum. Neuroscience 1989; 30:77–86. 9. Aggleton JP. The Amygdala: Neurobiological Aspects of Emotion, Memory and Mental Dysfunction. New York: Wiley-Liss, 1992. 10. Rothblat LA, Vnek N, Gleason TC, Kromer LF. Role of the parahippocampal region in spatial and non-spatial memory: effects of parahippocampal lesions on rewarded alternation and concurrent object discrimination learning in the rat. Behav Brain Res 1993; 5:93–100. 11. Kesner RP. Memory for frequency in rats: role of the hippocampus and prefrontal cortex. Behav Neural Biol 1990; 53:402–410. 12. Eichenbaum H, Fagan A, Cohen NJ. Normal olfactory discrimination learning set and facilitation of reversal learning after medial-temporal damage in rats: implications for an account of preserved learning abilities in amnesia. J Neurosci 1986; 6:1876–1884. 13. Eichenbaum H, Cohen NJ. From Conditioning To Conscious Recollection: Memory Systems of the Brain. New York: Oxford University Press, 2001. 14. Squire LR. Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans. Psychol Rev 1992; 99:195–231. 15. Packard MG, Knowlton BJ. Learning and memory functions of the basal ganglia. Ann Rev Neurosci 2002; 25:563–593. 16. Preuss TM. Do rats have a prefrontal cortex? The Rose-Woolsey-Akert program reconsidered J Cogn Neurosci 1995; 7:1–24. 17. Brown VJ, Bowman EM. Rodent models of prefrontal cortical function. Trends Neurosci 2002; 25:340–343. 18. Dunnett SB. Role of prefrontal cortex and striatal output systems in short-term memory deficits associated with ageing, basal forebrain lesions, and cholinergic-rich grafts. Can J Psychol 1990; 44:210–232. 19. Kolb B, Robbins T. The rodent prefrontal cortex. Behav Brain Res 2003; 146:1–2.

A Perspective from Animal Research

439

20. Murphy BL, Arnsten AF, Goldman-Rakic PS, Roth RH. Increased dopamine turnover in the prefrontal cortex impairs spatial working memory performance in rats and monkeys. Proc Natl Acad Sci USA 1996; 93:1325–1329. 21. Shaw L, Aggleton JP. The effects of fornix and medial prefrontal lesion on delayed nonmatching-to-sample by rats. Behav Brain Res 1993; 54:91–102. 22. Uylings HB, Groenewegen HJ, Kolb B. Do rats have a prefrontal cortex? Behav Brain Res 2003; 146:3–17. 23. Schoenbaum G, Setlow B, Ramus SJ. A systems approach to orbitofrontal cortex function: recordings in rat orbitofrontal cortex reveal interactions with different learning systems. Behav Brain Res 2003; 146:19–29. 24. White RF, Gerr F, Cohen RF, et al. Criteria for progressive modification of neurobehavioral batteries. Neurotox Teratol 1994; 16:511–524. 25. Environmental Protection Agency. Health effects test guidelines: Developmental neurotoxicity study (OPPTS 870.6300), 1998. 26. Cohen NJ, Squire LR. Preserved learning and retention of a pattern-analyzing skill in amnesia: dissociation of knowing how and knowing that. Science 1980; 210:207–210. 27. Milner B. The memory defect in bilateral hippocampal lesions. Psychiatr Res Rep Am Psychiatr Assoc 1959; 11:43–58. 28. Corkin S. What’s new with the amnesic patient H.M.? Nat Rev Neurosci 2002; 3:153–160. 29. Zola-Morgan S, Squire LR, Amaral DB. Human amnesia and the medial temporal region: enduring memory impairment following a bilateral lesion limited to field CA1 of the hippocampus. J Neurosci 1986; 6:2950–2967. 30. Zola-Morgan S, Squire LR. Preserved learning in monkeys with medial temporal lesions: sparing of motor and cognitive skills. J Neurosci 1984; 4:1072–1085. 31. Phillips RR, Malamut BL, Bachevalier J, Mishkin M. Dissociation of the effects of inferior temporal and limbic lesions on object discrimination learning with 24-h intertrial intervals. Behav Brain Res 1988; 27:99–107. 32. Mishkin M. Memory in monkeys severely impaired by combined but not by separate removal of amygdale and hippocampus. Nature 1978; 273:298–299. 33. Zola-Morgan S, Squire LR. Medial temporal lesions in monkeys impair memory on a variety of tasks sensitive to human amnesia. Behav Neurosci 1985; 99:22–34. 34. Zola-Morgan S, Squire LR, Ramus SJ. Severity of memory impairment in monkeys as a function of locus and extent of damage within the medial temporal lobe memory system. Hippocampus 1994; 4:483–495. 35. Strupp BJ, Levitsky DA. An animal model of retarded cognitive development. In: Collier CR, Lipsett L, eds. In: Advances in Infancy Research, Vol. 6. Norwood, N.J: ABLEX Publishing Corporation, 1990:149–186. 36. Strupp BJ, Diamond A. Assessing cognitive function in animal models of mental retardation. Ment Retard Dev Disabil Res 1996; 2:216–226. 37. Everitt BJ, Robbins TW. Amygdala-ventral striatal interactions and reward-related processes. In: Aggleton JP, ed. In: The Amygdala: Neurobiological aspects of emotion, memory, and mental dsyfunction. Chapter 15. New York: Wiley-Liss, 1992:401–429. 38. Gaffan D. Amygdala and the memory of reward. In: Aggleton JP, ed. In: The Amygdala: Neurobiological aspects of emotion, memory, and mental dsyfunction. Chapter 18. New York: Wiley-Liss, 1992:471–483. 39. McDonald RJ, White NM. A triple dissociation of memory systems: hippocampus. amygdala, and dorsal striatum. Behav Neurosci 1993; 107:3–22. 40. Kesner RP, Bolland BL, Dakis M. Memory for spatial locations, motor responses, and objects: triple dissociation among the hippocampus, caudate nucleus, and extrastriate visual cortex. Exp Brain Res 1993; 93:462–470. 41. Packard MG, McGaugh JL. Double dissociation of fornix and caudate nucleus lesions on acquisition of two water maze tasks: further evidence for multiple memory systems. Behav Neurosci 1992; 106:439–446.

440

Strupp and Beaudin

42. Zola-Morgan S, Squire LR, Alvarez-Royo P, Clower RP. Independence of memory functions and emotional behavior: separate contributions of the hippocampal formation and the amygdala. Hippocampus 1991; 1:207–220. 43. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association, 1994. 44. Kesner RP, Farnsworth G, DiMattia BV. Double dissociation of egocentric and allocentric space following medial prefrontal and parietal cortex lesions in the rat. Behav Neurosci 1989; 103:956–961. 45. Kesner RP, Gray ML. Dissociation of item and order memory following parietal cortex lesions in the rat. Behav Neurosci 1989; 103:907–910. 46. Strupp BJ, Levitsky DA. Early brain insult and cognition: a comparison of malnutrition and hypothyroidism. Dev Psychobiol 1983; 16:535–549. 47. Strupp BJ, Levitsky DA, Blumstein L. PKU, learning, and models of mental retardation. Dev Psychobiol 1984; 17:109–120. 48. Strupp BJ, Levitsky DA. Enduring cognitive effects of early malnutrition: a theoretical reappraisal. J Nutr 1995; 125:2221S–2232S. 49. Pennington BF. Genetics of learning disabilities. Semin Neurol 1991; 11:28–34. 50. Lezak MD. Neuropsychological Assessment. 2nd ed. New York: Oxford University Press, 1995. 51. Barkley RA. Additional evidence supporting the existence of the executive functions. In: Barkley RA, ed. ADHD and the Nature of Self-control. New York, USA: Guilford Press, 1997:108–153. 52. Stuss DT, Benson DF. The frontal lobes. New York: Raven, 1986. 53. Mactutus CF, Herman AS, Booze RM. Chronic intravenous model for studies of drug (Ab)use in the pregnant and/or group-housed rat: an initial study with cocaine. Neurotoxicol Teratol 1994; 16:183–191. 54. Gendle MH, Strawderman M, Mactutus CF, Levitsky DA, Booze RM, Strupp BJ. Alterations in reactivity to errors and sustained attention in an animal model of prenatal cocaine exposure. Dev Brain Res 2003; 147:85–96. 55. Morgan RE, Garavan H, Mactutus CF, Levitsky DA, Booze RM, Strupp BJ. Enduring effects of prenatal cocaine exposure on attentional control and reactivity to errors. Behav Neurosci 2002; 116:624–633. 56. Garavan H, Morgan RE, Mactutus CF, Levitsky DA, Booze RM, Strupp BJ. Prenatal cocaine exposure impairs selective attention: evidence from serial reversal and extradimensional shift tasks. Behav Neurosci 2000; 114:725–738. 57. Gendle MH, White TL, Strawderman M, et al. Enduring effects of prenatal cocaine exposure on selective attention and reactivity to errors: evidence from an animal model. Behav Neurosci 2004; 118:290–297. 58. Romano AG, Harvey JA. Prenatal exposure to cocaine disrupts discrimination learning in adult rabbits. Pharmacol Biochem Behav 1996; 53:617–621. 59. Gendle MH, Strawderman M, Mactutus CF, Levitsky DA, Booze RM, Strupp BJ. Prenatal cocaine exposure does not alter spatial working memory in adult rats. Neurotoxicol Teratol 2004; 26:319–329. 60. Levitt P, Reinoso B, Jones L. The critical impact of early cellular environment on neuronal development. Prev Med 1998; 27:180–183. 61. de Brabander JM, van Eden CG, de Bruin JP. Neuroanatomical correlates of sparing of function after neonatal medial prefrontal cortex lesions in rats. Brain Res 1991; 568:24–34. 62. Freeman JH, Jr., Stanton ME. Medial prefrontal cortex lesions and spatial delayed alternation in the developing rat: recovery or sparing. Behav Neurosci 1992; 106:924–932. 63. Carter CS, Freeman JH, Jr., Stanton ME. Neonatal medial prefrontal lesions and recovery of spatial delayed alternation in the rat: effects of delay interval. Dev Psychobiol 1995; 28:269–279.

A Perspective from Animal Research

441

64. Kolb B, Whishaw IQ. Neonatal frontal lesions in the rat: sparing of learned but not speciestypical behavior in the presence of reduced brain weight and cortical thickness. J Comp Physiol Psychol 1981; 95:863–879. 65. Kolb B. Sparing and Recovery of Function. In: Kolb B, Tees RC, eds. The Cerebral Cortex of the Rat. Cambridge, MA: MIT Press, 1990:537–561. 66. Campione JC, Brown AL. Learning ability and transfer propensity as sources of individual differences in intelligence. In: Brooks PH, Sperber RD, McCauley C, eds. Learning and Cognition in the Mentally Retarded. Baltimore, MD: University Park Press, 1984:265–294. 67. Campione JC, Brown AL, Ferrara RA, Jones RS, Steinberg E. Breakdowns in flexible use of information: intelligence-related differences in transfer following equivalent learning performance. Intelligence 1985; 9:297–315. 68. Strupp BJ, Himmelstein S, Bunsey M, Levitsky DA, Kesler M. Cognitive profile of rats exposed to lactational hyperphenylalaninemia: correspondence with human mental retardation. Dev Psychobiol 1990; 23:195–214. 69. Strupp BJ, Bunsey M, Levitsky DA, Hamberger K. Deficient cumulative learning: an animal model of retarded cognitive development. Neurotoxicol Teratol 1994; 16:71–79. 70. Davidson RJ, Putnam KM, Larson CL. Dysfunction in the neural circuitry of emotion regulation-a possible prelude to violence. Science 2000; 289:591–594. 71. Davidson RJ. Affective style and affective disorders: perspectives from affective neuroscience. Cognition Emotion 1998; 12:307–320. 72. Aronen ET, Vuontela V, Steenari MR, Salmi J, Carlson S. Working memory, psychiatric symptoms, and academic performance at school. Neurobiol Learn Mem 2005; 83:33–42. 73. Seipp B. Anxiety and academic performance: a meta-analysis of findings. Anxiety Research 1991; 4:27–41. 74. Burt DB, Zembar MJ, Niederehe G. Depression and memory impairment: a meta-analysis of the association, its pattern, and specificity. Psychol Bull 1995; 117:285–305. 75. Byers RK, Lord EE. Late effects of lead poisoning on mental development. Am J Dis Child 1943; 66:471–483. 76. Weiss B. Intersections of psychiatry and toxicology. Int J Mental Health 1985; 14:7–25. 77. Bellinger DC. Future directions for neurobehavioral studies of environmental neurotoxicants. Neurotoxicology 2001; 22:645–656. 78. Needleman HL, Riess JA, Tobin MJ, Biesecker GE, Greenhouse JB. Bone lead levels and delinquent behavior. JAMA 1996; 275:363–369. 79. Needleman HL, McFarland C, Ness RB, Fienberg SE, Tobin MJ. Bone lead levels in adjudicated delinquents. A case control study. Neurotoxicol Teratol 2002; 24:711–717. 80. Dietrich KN, Ris MD, Succop PA, Berger OG, Bornschein RL. Early exposure to lead and juvenile delinquency. Neurotoxicol Teratol 2001; 23:511–518. 81. Farrington DP, Loeber R. Epidemiology of juvenile violence. Child Adolesc Psychiatr Clin N Am 2000; 9:733–748. 82. Rutter M. Poverty and child mental health: natural experiments and social causation. JAMA 2003; 290:2063–2064. 83. Evans GW, Kantrowitz E. Socioeconomic status and health: the potential role of environmental risk exposure. Ann Rev Public Health 2002; 23:303–331. 84. Phelps EA, Delgado MR, Nearing KI, LeDoux JE. Extinction learning in humans: role of the amygdale and vmPFC. Neuron 2004; 43:897–905. 85. Blanchard DC, Blanchard RJ. What can animal aggression research tell us about human aggression. Horm Behav 2003; 44:171–177. 86. Blanchard DC, Hynd AL, Minke KA, Minemoto T, Blanchard RJ. Human defensive behaviors to threat scenarios show parallels to fear- and anxiety-related defense patterns of non-human mammals. Neurosci Biobehav Rev 2001; 25:761–770. 87. Graeff FG. On serotonin and experimental anxiety. Psychopharmacology (Berl) 2002; 163:467–476. 88. Kalin NH, Shelton SE, Davidson RJ. The role of the central nucleus of the amygdala in mediating fear and anxiety in the primate. J Neurosci 2004; 24:5506–5515.

442

Strupp and Beaudin

89. Hogg S. A review of the validity and variability of the elevated plus-maze as an animal model of anxiety. Pharmacol Biochem Behav 1996; 54:21–30. 90. Dawson GR, Tricklebank MD. Use of the elevated plus maze in the search for novel anxiolytic agents. Trends Pharmacol Sci 1995; 16:33–36. 91. Pellow S, Chopin P, File SE, Briley M. Validation of open: closed arm entries in an elevated plus-maze as a measure of anxiety in the rat. J Neurosci Meth 1985; 14:149–167. 92. Pellow S, File SE. Anxiolytic and anxiogenic drug effects on exploratory activity in an elevated plus-maze: a novel test of anxiety in the rat. Pharmacol Biachem Behav 1986; 24:525–529. 93. Willner P. Validity, reliability and utility of the chronic mild stress model of depression: a 10-year review and evaluation. Psychopharmacology 1997; 134:319–329. 94. Barrett J, Livesey PJ. Low level lead effects on activity under varying stress conditions in the developing rat. Pharmacol Biochem Behav 1985; 22:107–118. 95. Commissaris RL, Tavakoli-Nezhad M, Barron AM, Pitts DK. Effects of chronic low-level oral lead exposure on prepulse inhibition of acoustic startle in the rat. Neurotoxicol Teratol 2000; 22:55–60. 96. Yu SY, Mizinga KM, Nonavinakere VK, Soliman KRA. Decreased endurance to cold water swimming and delayed sexual maturity in the rat following neonatal lead exposure. Toxicol Lett 1996; 85:135–141. 97. Salinas JA, Huff NC. Lead and conditioned fear to contextual and discrete cues. Neurotoxicol Teratol 2002; 24:541–550. 98. Hastings L, Cooper GP, Bornschein RL, Michaelson IA. Behavioral effects of low level neonatal lead exposure. Pharmacol Biochem Behav 1977; 7:37–42. 99. Reiter LW, Anderson GE, Laskey JW, Cahill DF. Developmental and behavioral changes in the rat during chronic exposure to lead. Environ Health Persp 1975; 12:119–123. 100. Winneke G, Brockhaus A, Baltissen R. Neurobehavioral and systemic effects of longterm blood lead-elevation in rats. I. Discrimination learning and open field-behavior. Arch Toxicol 1977; 37:247–263. 101. Lasky RE, Laughlin NK. Exploring a partially enclosed space by lead-exposed female rhesus monkeys. Neurotoxicol Teratol 2001; 23:177–183. 102. Ferguson SA, Bowman RE. Effects of postnatal lead exposure on open field behavior in monkeys. Neurotoxicol Teratol 1990; 12:91–97. 103. Campbell JO, Bliven TD, Sliveri MM, Snyder KJ, Spear LP. Effects of prenatal cocaine on behavioral adaptation to chronic stress in adult rats. Neurotoxicol Teratol 2000; 22:845–850. 104. Elsworth JD, Morrow BA, Roth RH. Prenatal cocaine exposure increases mesoprefrontal dopamine neuron responsivity to mild stress. Synapse 2001; 42:80–83. 105. Huber J, Darling S, Park K, Soliman KF. Altered responsiveness to stress and NMDA following prenatal exposure to cocaine. Physiol Behav 2001; 72:181–188. 106. Molina VA, Wagner JM, Spear LP. The behavioral response to stress is altered in adult rats exposed prenatally to cocaine. Physiol Behav 1994; 55:941–945. 107. Overstreet DH, Moy SS, Lubin DA, Gause LR, Lieberman JA, Johns JM. Enduring effects of prenatal cocaine administration on emotional behavior in rats. Physiol Behav 2000; 70:149–156. 108. Spear LP, Campbell J, Snyder K, Silveri M, Katovic N. Animal behavior models. Increased sensitivity to stressors and other environmental experiences after prenatal cocaine exposure. Ann NY Acad Sci 1998; 846:76–88. 109. Porsolt RD, Bertin A, Jalfre M. Behavioral despair in mice: a primary screening test for antidepressants. Arch Int Pharmacodyn Ther 1977; 229:327–336. 110. Porsolt RD, Anton G, Blavet N, Jalfre M. Behavioural despair in rats: a new model sensitive to antidepressant treatment. Eur J Pharmacol 1978; 47:379–391. 111. Overmier JB, Seligman ME. Effects of inescapable shock upon subsequent escape and avoidance responding. J Camp Physiol Psychol 1967; 63:28–33.

A Perspective from Animal Research

443

112. Telner JI, Singhal RL. Psychiatric progress. The learned helplessness model of depression. J Psychiatr Res 1984; 18:207–215. 113. Morgan RE, Garavan H, Smith E, Driscoll LL, Levitsky DA, Strupp BJ. Enduring effects of early lead (Pb) exposure on sustained attention, response initiation, and reactivity to errors. Neurotoxicol Teratol 2001; 23:519–532. 114. Laming D. Choice reaction performance following an error. Acta Psychologia 1979; 43:199–224. 115. Robertson IH, Manly T, Andrade J, Baddeley BT, Yiend J. ‘Oops!’: performance correlates of everyday attentional failures in traumatic brain injured and normal subjects. Neuropsychologia 1997; 35:747–758. 116. Bush G, Luu P, Posner M. Cognitive and emotional influences in anterior cingulate cortex. Trends Cogn Sci 2000; 4:215–222. 117. Fernandez-Duque D, Baird JA, Posner MI. Executive attention and metacognitive regulation. Conscious Cogn 2000; 9:288–307. 118. Rabbitt PMA, Rogers B. What does a man do after he makes an error? An analysis of response programming. Q J Exp Psychol 1977; 29:727–743. 119. Luu P, Collins P, Tucker DM. Mood, personality, and self-monitoring: negative affect and emotionality in relation to frontal lobe mechanisms of error monitoring. J Exp Psychol Gen 2000; 129:43–60. 120. Beaudin SA, Stangle D, Strupp BJ. Lasting effects of early lead exposure on reactivity to errors, but not omission of reward. Neurotoxicol Teratol 2001; 23:291–294. 121. Strupp BJ, Korahais J, Levitsky DA, Ginsberg LS. Attentional impairment in rats exposed to alcohol prenatally: lack of hypothesized masking by food deprivation. Ann NY Acad Sci 1989; 562:380–382. 122. Driscoll LL, Carroll JC, Moon J, Crnic LS, Levitsky DA, Strupp BJ. Impaired sustained attention and error-induced stereotypy in the aged Ts65Dn mouse: a mouse model of Down syndrome and Alzheimer’s disease. Behav Neurosci 2004; 118:1196–1205. 123. Beaudin SA, Stangle DE, Strawderman MS, Smith D, Levitsky DA, Strupp BJ. Succimer chelation normalizes the enduring reactivity to errors in a rodent model of pediatric lead exposure: Evidence from an olfactory conditional associative task. Submitted for Publication. 124. Stangle DG, Strawderman M, Smith D, Strupp BJ. Succimer chelation significantly ameliorates the lasting cognitive and affective dysfunction produced by early lead exposure in a rodent model. Toxicol Sci 2003; 72:128. 125. Strupp BJ, Stangle DG, Strawderman M, Smith D, Levitsky. Get the lead out: Succimer chelation significantly ameliorates the lasting cognitive and affective dysfunction produced by early lead exposure in a rodent model. Neurotoxicol Teratol 2003; 25:394–395. 126. Rodier PM. Chronology of neuron development: animal studies and their clinical implications. Dev Med Child Neurol 1980; 22:525–545. 127. Rodier PM. Developing brain as a target of toxicity. Environ Health Persp 1995; 103:73–76. 128. Bellinger DC, Bernstein JH, Kirkwood MW, Rappaport LA, Newburger JW. Visual-spatial skills in children after open-heart surgery. Dev Behav Pediatr 2003; 24:169–179. 129. Cory-Slechta DA, Weiss B, Cox C. Performance and exposure indices of rats exposed to low concentrations of lead. Toxicol Appi Pharmacol 1985; 78:291–299. 130. Bachevalier J, Beauregard M, Alvarado MC. Long-term effects of neonatal damage to the hippocampal formation and amygdaloid complex on object discrimination and object recognition in rhesus monkeys (Macaca Mulatta). Behav Neurosci 1999; 113:1127–1151. 131. Bachevalier J, Malkova L, Mishkin M. Effects of selective neonatal temporal lobe lesions on socioemotional behavior in infant rhesus monkeys (Macaca mulatta). Behav Neurosci 2001; 115:545–559. 132. Alvarado MC, Wright AA, Bachevalier J. Object and spatial relational memory in adult rhesus monkeys is impaired by neonatal lesions of the hippocampal formation but not the amygdaloid complex. Hippocampus 2002; 12:421–433.

444

Strupp and Beaudin

133. Stangle D, Strawderman MS, Smith D, Levitsky DA, Beaudin SA, and Strupp BJ. Succimer chelation improves cognition and arousal regulation in lead-exposed rats but produces lasting cognitive impairment in the absence of lead exposure, Submitted for publication. 134. Alber SA, Strupp BJ. An in-depth analysis of lead effects in a delayed spatial alternation task: assessment of mnemonic effects, side bias and proactive interference. Neurotoxicol Teratol 1996; 18:3–15. 135. Garavan H, Morgan RE, Hermer-Vazquez L, Levitsky DA, Strupp BJ. Enduring effects of early lead exposure: evidence for a specific deficit in associative ability. Neurotoxicol Teratol 2000; 22:151–164. 136. Hilson J, Strupp BJ. Analyses of response patterns clarify lead effects in olfactory reversal and extra-dimensional shift tasks: assessment of inhibitory control, associative ability, and memory. Behav Neurosci 1996; 111:532–542. 137. Dias R, Robbins TW, Roberts AC. Dissociation in prefrontal cortex of affective and attentional shifts. Nature 1996; 380:69–72. 138. Jones B, Mishkin M. Limbic lesions and the problem of stimulus-reinforcement associations. Exp Neurol 1972; 36:362–377. 139. Higley MJ, Hermer-Vazquez L, Levitsky DA, Strupp BJ. Recovery of associative function following early amygdala lesions in rats. Behav Neurosci 2001; 115:154–164. 140. Morris RG, Garrud P, Rawlins JN, O’Keefe J. Place navigation impaired in rats with hippocampal lesions. Nature 1982; 297:681–683. 141. Olton DS, Samuelson RJ. Remembrance of places past: spatial memory in rats. J Exp Psycho Anim Behav Process 1976; 4:297–317. 142. Larson J, Sieprawska D. Automated study of simultaneous-cue olfactory discrimination learning in adult mice. Behav Neurosci 2002; 116:588–599. 143. Sharbaugh C, Viet SM, Fraser A, McMaster SB. Comparable measures of cognitive function in human infants and laboratory animals to identify environmental health risks to children. Environ Health Perspect 2003; 111:1630–1639. 144. Overman WH, Bachevalier J, Sewell F, Drew J. A comparison of children’s performance on two recognition memory tasks delayed nonmatch-to-sample versus visual paired-comparison. Dev Psychobiol 1993; 26:345–357. 145. Paule MG, Chelonis JJ, Buffalo EA, Blake DJ, Casey PH. Operant test battery performance in children: correlation with IQ. Neurotoxicol Teratol 1999; 21:223–230. 146. Squire LR, Zola-Morgan S, Chen KS. Human amnesia and animal models of amnesia: performance of amnesic patients on tests designed for the monkey. Behav Neurosci 1988; 102:210–221. 147. Zola-Morgan S, Squire LR, Amaral DG, Suzuki WA. Lesions of perirhinal and parahippocampal cortex that spare the amygdala and hippocampal formation produce severe memory impairment. J Neurosci 1989; 9:4355–4370. 148. Overman W, Bachevalier J, Turner M, Peuster A. Object recognition versus object discrimination: comparison between human infants and infant monkeys. Behav Neurosci 1992; 106:15–29. 149. Diamond A, Towle C, Boyer K. Young children’s performance on a task sensitive to the memory functions of the medial temporal lobe in adults—the delayed nonmatching-to-sample task—reveals problems that are due to non-memory-related task demands. Behav Neurosci 1994; 108:659–680. 150. Diamond A, Goldman-Rakic PS. Comparison of human infants and rhesus monkeys on Piaget’s AB task: evidence for dependence on dorsolateral prefrontal cortex. Exp Brain Res 1989; 74:24–40. 151. Diamond A, Doar B. The performance of human infants on a measure of frontal cortex function, the delayed response task. Dev Psychobiol 1989; 22:271–294. 152. Espy KA, Kaufmann PM, McDiarmid MD, Glisky ML. Executive functioning in preschool children: performance on A-not-B and other delayed response format tasks. Brain Cogn 1999; 41:178–199.

A Perspective from Animal Research

445

153. Schneider JS, Lee MH, Anderson DW, Zuck L, Lidsky TI. Enriched environment during development is protective against lead-induced neurotoxicity. Brain Res 2001; 896:48–55. 154. Guilarte TR, Toscano CD, McGlothan JL, Weaver SA. Environmental enrichment reverses cognitive and molecular deficits induced by developmental lead exposure. Ann Neurol 2003; 53:50–56. 155. Walsh RN, Greenough WT. Environments as Therapy for Brain Dysfunction. New York: Plenum, 1975. 156. Will B, Galani R, Kelche C, Rosenzweig M. Recovery from brain injury in animals: relative efficacy in environmental enrichment, physical exercise or formal training (1990–2002). Prog Neurobiol 2004; 72:167–182. 157. Koller K, Brown T, Spurgeon A, Levy L. Recent developments in low-level lead exposure and intellectual impairment in children. Environ Health Perspect 2004; 112:987–994. 158. Sjoden PO, Lindqvist M. Behavioral effects of neonatal thyroid hormones and differential postweaning rearing in rats. Dev Psychobiol 1978; 11:371–383. 159. Mohammed AK, Jonsson G, Archer T. Selective lesioning of forebrain noradrenaline neurons at birth abolishes the improved maze learning performance induced by rearing in complex environment. Brain Res 1986; 398:6–10. 160. O’Shea L, Saari M, Pappas BA, Ings R, Strange K. Neonatal 6-hydroxydopamine attenuates the neural and behavioral effects of enriched rearing in the rat. Eur J Pharmacol 1983; 92:43–47. 161. Rosenzweig MR, Bennett EL. How plastic is the nervous system? In: Almi, Finger S, eds. Comprehensive Handbook of Behavioral Medicine. Spectrum Publications, 1980:149–185. 162. Rosenzweig MR, Bennett EL, Alberti M, Morimoto H, Renner MJ. Effects of differential environments and hibernation on ground squirrel brain measures. Soc Neurosci Abstr 1982; 8:669. 163. Rosenzweig MR, Bennett EL, Renner MJ, Alberti M. Laboratory enrichment and the natural baseline: effects of differential environments on brain measures in two species of ground squirrels. Unpublished manuscript, 1987. 164. Renner MJ, Rosenzweig MR. Enriched and Impoverished Environments: Effects on Brain and Behavior. New York: Springer-Verlag, 1987. 165. Kempermann G, Gage FH. Experience dependent regulation of adult hippocampal neurogenesis: effects of long-term stimulation and stimulus withdrawal. Hippocampus 1999; 9:321–332. 166. Kempermann G, Brandon EP, Gage FH. Environmental stimulation of 129/SvJ mice causes increased cell proliferation and neurogenesis in the adult dentate gyrus. Curr Biol 1998; 8:939–942. 167. Nilsson M, Perfilieva E, Johansson U, Orwar O, Eriksson PS. Enriched environment increases neurogenesis in the adult rat dentate gyrus and improves spatial memory. J Neurobiol 1999; 39:569–578.

21 The Developmental Neurotoxicology of Chemicals Disrupting Thyroid Hormone Signaling R. Thomas Zoeller Biology Department, University of Massachusetts-Amherst, Amherst, Massachusetts, U.S.A.

Thyroid hormone (TH) is essential for normal brain development (1–4). Exposure to severe TH insufficiency during fetal and/or neonatal development in humans leads to a condition known as cretinism, characterized in part by severe mental retardation (5). Likewise, exposure to severe TH insufficiency during fetal and/or neonatal development in experimental animals leads to severe behavioral impairment indicative of neurodevelopmental deficits (6). Considering these observations, it is reasonable to predict that any environmental chemical affecting TH action during brain development may produce adverse effects on cognitive development (7). However, there are several reasons that testing this prediction is not currently a simple matter. Therefore, the goal of this chapter is to provide an analysis of the challenges to investigators engaged in basic research focused on providing information to be used in risk assessment. The point of departure for this chapter is the recommendation by the U.S. EPA Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) for assays to be used to screen for thyroid toxicants.

THE EDSTAC FINAL RECOMMENDATION FOR SCREENING ANTI-THYROID ACTIVITIES The authors of the EDSTAC final report accurately stated that, at the time, “all known antithyroid compounds so far reported in vertebrates affect circulating levels of thyroxine (T4).” A large number of thyroid toxicants have been identified (8), and in all cases, these chemicals have been identified by their ability to reduce circulating levels of T4. Moreover, the term “environmental goitrogenesis” (9) was developed to describe chemicals and mechanisms that lead to reduced circulating levels of TH, producing a 447

448

Zoeller

compensatory over-stimulation of the thyroid gland and goiter formation. Considering this, the recommendation by the EDSTAC for identifying thyroid toxicants was to measure circulating levels of T4 and TSH. Recognizing that changes in circulating levels of hormones should not be considered adverse per se, they recommended that thyroid histopathology be used as an endpoint that reflects the physiological consequences of altered thyroid function. Little has changed during the past 5 years with regard to thyroid toxicology. Measurements of serum THs and thyroid histopathology remain the end points for identifying thyroid toxicants and for gaining insight into risk assessment (10). However, the authors of the EDSTAC final report also wisely cautioned that, “Despite the volume of literature reviewed, the rapid pace of research into TH action makes it predictable that the present screens will become obsolete, both because more effective assays will likely be developed and because new information about TH action may reveal mechanisms of thyroid disruption not identified by the recommended T1S battery.” Part of this prediction has been realized—new information about the mechanisms of thyroid disruption have been identified. However, more effective screens have not been developed. Thus, identification and risk characterization of thyroid toxicants is compromised by the lack of valid neurodevelopmental endpoints of thyroid toxicity that may be employed in newly designed screens for thyroid toxicants. Strawson et al. (10) recently reviewed the logic and evidence employed to characterize risk to human health following exposure to the environmental contaminant perchlorate, which is known to block iodide uptake into the thyroid gland potentially reducing circulating levels of TH (11). As they describe, an important weakness in the experimental data is that the relationship between perchlorate exposure, circulating levels of TH, and neurodevelopment is not characterized. Thus, two critical issues in thyroid toxicology are: (1) To what degree must circulating levels of TH be reduced before adverse effects on brain development are observed? This question is as relevant to humans as it is for experimental animals. (2) What are the most sensitive endpoints of brain development that would clearly reflect adverse effects of thyroid toxicants? These questions are discussed below in sequence. (1) To what degree must circulating levels of TH must be reduced before adverse effects on brain development are observed? A key decision for any risk assessment of thyroid toxicants is to distinguish adaptive from adverse effects on the physiology of humans or of experiment animals. This question is derived from the observation that the hypothalamic–pituitary–thyroid (HPT) axis has two predominant mechanisms that can potentially compensate for fluctuating levels of TH. These mechanisms include the negative feedback action of TH on the hypothalamus and pituitary, and the responses of tissue deiodinases to changing circulating levels of TH. These mechanisms are believed to allow “compensation” to occur; that is, reductions in circulating TH are compensated for by adaptive mechanisms within the HPT axis and within tissues such that adverse consequences are prevented. This concept requires that these compensatory mechanisms are more sensitive to reduced circulating TH levels than any other tissues so that it “senses” changes in TH levels and “adapts” to protect tissues including the developing brain. Currently, this hypothesis is whimsical, for there are no experimental data to support it and, in fact, recent reports do not support it. These adaptive mechanisms are described below.

Thyroid Hormone Signaling

449

THE HPT NEGATIVE FEEDBACK SYSTEM The HPT axis is a classic neuroendocrine axis; the hypothalamus controls the pituitary gland, which in turn controls the thyroid, and feedback mechanisms between thyroid secretions and the hypothalamus and pituitary maintain the activity of this axis within narrow limits (12). Hormones secreted from the thyroid gland, thyroxine (T4) and triiodothyronine (T3), are synthesized by coupling two iodinated tyrosyl residues that make up the larger hormone precursor, thyroglobulin (Tg). Tg, a large glycoprotein of nearly 3000 amino acids (13), is stored in the fluid filling the central core of the thyroid follicle (the colloid), and at the time of hormone release, iodinated Tg is taken up into the cell from the colloid and digested by lysosomal enzymes, liberating T3 and T4 into the blood (14). Tg is also secreted into the serum by transcytosis (15), with measurable and changing levels in serum (16), and may even have direct hormonal actions (17). Perhaps more importantly, serum Tg is a measure of thyroid function in humans (16). The pituitary glycoprotein hormone, thyrotropin (TSH) (18), regulates the synthesis and secretion of THs by activating adeylate cyclase in thyroid follicular cells (19). Although T4 is the predominant form of TH in the serum, T3 is the active hormone at the receptor. TH exerts a negative feedback effect on pituitary secretion of TSH (20,21), and on the hypothalamic secretion of the releasing factor, thyrotropinreleasing hormone (TRH) (20,22,23). The negative feedback effect of TH is mediated by the beta receptor subtype of the TH receptor (TRb). The TRb subtype is predominant in the hypothalamic paraventricular nucleus (23) and in the pituitary (24,25). Thus, the compensatory response to fluctuations in circulating TH mediated by this negative feedback mechanism is dependent upon the TRb isoform. This is especially clear in people with thyroid resistance syndrome caused by a defect in the TRb ligand binding domain (26), resulting in elevated T4 and elevated TSH. The causal relationship between the TRb defect and thyroid resistance was confirmed in animals by targeted insertion of the human TRb mutation into a mouse line (27). In addition to the negative feedback mechanisms, small changes in circulating levels of TH affect the expression of the deiodinases (28), enzymes that convert T4 to the hormonally active T3 (type I, D1; or type II, D2), or T4 (or T3) to the hormonally inactive reverse T3 or T2 (type III, D3) (29). Changes in deiodinase expression and activity may represent an important tissue-autonomous mechanism controlling the sensitivity of target tissues to TH fluctuations (30), including the developing brain (31). Taken together, these two mechanisms—negative feedback and tissue deiodinases—may interact to protect tissues from small fluctuations in serum TH. Thus, when a thyroid toxicant causes a reduction in circulating levels of TH, the question is, to what extent must circulating TH decline before adverse effects are produced? This question can only be answered by empirical measures of TH action, and the answer to the question will likely depend on the endpoint and upon the developmental timing of the TH insufficiency. (2) What are the endpoints of brain development that would provide valid insight into the adverse effects of thyroid toxicants? Identifying specific endpoints of TH action in the developing brain requires both a consideration of the molecular and cellular mechanisms of TH action as well as the developmental events influenced by TH. Therefore, a brief review of these topics is provided for background. MECHANISM OF THYROID HORMONE ACTION Most, but perhaps not all, of the important effects of TH on brain development are mediated by their receptors (TRs) (32–35). TRs are members of the steroid/thyroid

450

Zoeller

superfamily of ligand-dependent transcription factors (36,37). They are encoded by two genes, designated a and b c-erbA (38,39), which generate at least four functional TRs: TRa1, TRb1, TRb2, and TRb3 (34,35,40). The TRa gene has 10 exons; TRa1 is composed of exons 1–9, whereas TRa2 is generated by the addition of a long c-terminal domain (exon 10) that disrupts the ligand-binding domain of the TR [see review by Flamant and Samarut (41)]. Thus, TRa2 isoform does not bind to TH and is generally not considered to be a bona fide TR. There is an internal promoter that drives the transcription of two additional short forms of the TRa gene (42). These short forms, designated TRDa1 and TRDa2, are encoded by exons 8–9 and 8–10, respectively. These proteins are able to bind to TH (T4 and reverse T3), but do not bind DNA. However, they may mediate important intracellular and non-genomic actions of THs and may be an important target of thyroid toxicants. In contrast, there are three promoters that drive the expression of the three functional TRs from the TRb gene (TRb1, TRb2 and TRb3) (43). In addition, the TRb3 transcript is differentially spliced to produce a TRDb3 isoform; again, this small product of the TRb gene binds to TH, but not to DNA. TH exerts effects on different aspects of brain development via different TRs. The a and b TRs exhibit distinct temporal and spatial patterns of expression in the developing rat CNS (44). For example in the neocortex, neuroblasts migrating from the germinal ventricular zone to the cortical plate on gestational day 13 express intense TRa1 mRNA, which continues through day 19. During this time, TRb1 mRNA is confined to the site of neuroblast proliferation—the ventricular layer. Therefore, it appears that neuroblasts switch TRs as they leave their birth site to migrate to their final destination. Similarly in the hippocampus, elevated TRa1 expression is coincident with the outward migration of postmitotic neurons from the hippocampal ventricular germinal zone, whereas TRb1 expression is weak. Leonard et al. (45) have suggested that TRa2 is expressed exclusively in glial cells, while TRa1 and TRb1 are expressed more predominantly in neurons. Thus, TH may influence different developmental processes by different receptor-mediated events, all well before the fetal HPT system begins to function on GD 17–20 (46). This implies that different levels and combinations of TR isoform expression may account in part for the pleiotropic effects of TH (37,47). Recent studies demonstrate empirically that TH actions on brain development are mediated separately by different TR isoforms. TRb knockout mice (TRb-/-) have resistance to TH (48–50), meaning that they have elevated levels of both T4 and TSH. In contrast, mice with deletion of the TRa1 and TRa2 isoforms (TR0/0) are hypersensitive to TH in several of the tissues examined (51) or less prone to the effects of TH deprivation (52). Moreover, mice completely deficient in both TRa and TRb (TR null) exhibit more severe resistance to TH than those lacking TRb only (53). Taken together, these data suggest that both isoforms play selective and overlapping roles in controlling serum TSH, but that the TRb form plays the dominant role. It is important also to recognize that TR knock out mice do not exhibit alterations in brain structure that resemble hypothyroid animals (35). Not only do TR knock-out mice not show affects of brain damage associated with hypothyroidism, but targeted deletion of specific TR isoforms can protect the brain from hypothyroidism in these strains (52). These observations led to the hypothesis that the unliganded TR mediates the adverse consequences of TH insufficiency on brain development (and on the function of other tissues). That is, it is not the absence of TH itself that causes the damage, but rather the presence of the unliganded TR. To test this hypothesis, Hashimoto et al. (54) constructed a mouse carrying a TRb gene with a targeted mutation in the ligand binding domain (TRbD337). This mutated TRb is unable to bind to TH, but remains capable of binding to DNA and to the co-repressor N-CoR. These investigators found that the TRbD337 mouse

Thyroid Hormone Signaling

451

exhibits the same severe defects in brain development as observed in wild type animals made hypothyroid. Therefore, the unliganded TR exerts the adverse effect on brain development in the absence of TH under normal conditions. The Bernal group in Madrid has begun to use these mouse lines to identify the TR isoforms that mediate effects of TH on brain development (55–57). Cerebellar granule cells are largely affected by TH acting on the TRa1 isoform. This includes their migration from the external germinal layer (EGL) across the mitral layer to populate the internal granule layer (IGL). However, it is not clear whether this also includes their proliferation in the EGL as well as apoptosis in the IGL, two important elements of granule cell development under the influence of TH (58,59). In contrast, Purkinje cells are affected by TH acting predominantly on the TRb isoform. Thus, it is likely that specific TH actions in the developing brain can ultimately be attributed to mediation by the TRa or TRb subtypes. These types of studies would map the developmental time and place of TH action during brain development and would characterize the TR subtype mediating the effect. In turn, this will be valuable information because thyroid toxicants such as bisphenol A, which binds directly to the TR (60,61), may conceivably exert TR subtypespecific effects on TH signaling. If so, the pattern of effects on brain development will be a mosaic of effects in the brain that would not be recognized as mediated by interfering with TH signaling.

THYROID HORMONE EXERTS TIME- AND DOSE-DEPENDENT EFFECTS ON BRAIN DEVELOPMENT We have known for a very long time that the thyroid gland is important for brain development. Paracelsus is credited with publishing the first proposal that endemic goiter is causally linked to cretinism in 1527 (62), although it is likely that this relationship was widely known before his report (63). We now know that endemic goiter is indicative of hypothyroidism; thus, the relationship between endemic cretinism and hypothyroidism described in modern times forms an important basis for the conclusion that TH is essential for brain development (5,64). Modern studies of the relationship between TH and brain development in humans reveals an important fundamental conclusion: that the effect and severity of the mental defect caused by TH insufficiency is determined by the developmental timing and severity of TH insufficiency (65). This conclusion is critical for the field of thyroid toxicology because it requires that investigations incorporate this principle; that is, the effect and severity of thyroid toxicants on brain development will be dependent upon the timing and dose of toxicant exposure. The clinical literature provides the strongest support for this principle. The original concept of the “critical period” of TH action on brain development was proposed to identify the postnatal period during which TH supplement must be provided to a child with congenital hypothyroidism (CH) to prevent mental retardation (66). However, as neuropsychological tools have become more sensitive, it has become apparent that even mild TH insufficiency in humans can produce measurable deficits in very specific neuropsychological functions (67–73), and that the specific consequences of TH deficiency depends on the precise developmental timing of the deficiency (65). Clinical studies of children born to women with low TH during pregnancy and studies of children with CH provide these insights. If the TH deficiency occurs early in pregnancy, the offspring display problems in visual attention, visual processing (viz., acuity and strabismus), and gross motor skills. If it occurs later in pregnancy, children are at additional risk of subnormal visual (viz., contrast sensitivity), and visuospatial skills as

452

Zoeller

well as slower response speeds and fine motor deficits. Finally, if TH insufficiency occurs after birth, language and memory skills are most predominantly affected (74–76). Although the experimental literature lags behind clinical studies in providing a mechanistic explanation for each of these observations, recent studies confirm that the specific action of TH on brain development depends upon developmental timing, and studies informing us about molecular mechanisms of TH action are generating hypotheses concerning possible mechanisms to account for these pleiotropic actions. For example, the progeny of hypothyroid dams (rats) show significant deficits in maze learning, are less cautious in emotionality testing, and are more active in open-field exploration (77), whereas animals rendered hypothyroid postnatally exhibit learning deficits and hypoactivity (78–80). There are a very large number of papers describing experiments designed to identify effects of TH on anatomical measures of brain development. Perinatal hypothyroidism alters the packing density and size of neuronal perikarya within specific brain areas, as well as fiber density and orientation within adult cortical layers (81–83). Quantitative and qualitative abnormalities have been seen in projections of the corpus callosum (CC) from both pre-and postnatal hypothyroidism (84–86). The number of axons in CC was reduced by 76%, especially in the posterior sector compared to the remainder (95% vs. 63%) (86–89). A few recent studies have begun to address more clearly the consequences of mild TH insufficiency in the rat dam on fetal brain development. For example, Lavado-Autric et al. (90) took advantage of the fact that cortical neurons occupying different lamina are born at different times during development. Using timed exposure to bromodeoxyuridine (BrdU), which is incorporated by dividing cells into their newly-synthesized DNA and marking them for life, the authors were able to show that a significant proportion of BrdUC cells in the cortex of pups derived from TH insufficient dams did not migrate far enough. In a followup study, Auso et al. found that even 3 days of treatment of pregnant rats with the goitrogen methimazole (MMI) was enough to cause a delay in cortical neuronal migration in the fetal brain, despite the fact that the subsequent TH insufficiency was not robust enough to cause serum TSH to become elevated in the dams (91). Thus, in this case, the HPT axis of the dam did not compensate for the TH insufficiency in the fetus. These studies, especially in humans, indicate that TH exerts actions in the developing brain that depend on the development time under consideration. This view is perhaps best articulated by Howdeshell (92). The implication for neurotoxicology is that the endpoint of thyroid toxicity must be designed to capture these effects occurring at different developmental times. These issues are developed more fully using the example of polychlorinated biphenyls.

POLYCHLORINATED BIPHENYLS (PCBS) PCBs are industrial chemicals consisting of paired phenyl rings with various degrees of chlorination (93). Although the production of PCBs was banned in the mid-1970s and body burden of PCBs are declining, these contaminants are routinely detected in the environment at significant concentrations (94,95). Epidemiological studies have indicated that children developmentally exposed to PCBs suffer from neuropsychological deficits such as a lower full-scale IQ, reduced visual recognition memory, attention deficits, and motor deficits (96–101) that overlap with those affected by maternal TH insufficiency (67,68,102). Therefore, several investigators have speculated that PCBs may impact brain development by interfering with TH signaling (103–105). PCBs, or specific PCB congeners, in maternal and cord blood are sometimes found to be associated with lower

Thyroid Hormone Signaling

453

TH levels in both the mother and infant (106,107). Although several epidemiological studies have failed to identify an association between TH and PCB body burden (73,108–111), experimental studies consistently find that PCB exposure decreases circulating levels of thyroxine (T4) in rats (112–114). Therefore, it is possible that PCB body burden is negatively associated with serum TH levels in humans, but that the variability inherent in human populations makes this association difficult to reveal. Developmental exposure to PCBs in experimental animals induces a hearing loss (115–118), a reduction in choline acetyltransferase in the cerebral cortex (119), an increase in testicular growth (120), and effects on ovarian follicular maturation (121), all consistent to some degree with effects produced by low circulating TH. Moreover, T4 replacement can at least partially ameliorate these effects (119,122), indicating that PCBs can influence development in part by causing a reduction in serum TH. However, hypothyroidism induced by exposure to the goitrogen propylthiouracil (PTU) causes a significant increase in serum TSH levels (123), reduced body and brain weight and brain size of rat pups (124), and a delay in eye opening and tooth eruption (125). In contrast, PCB exposure at doses that significantly reduce serum TH do not always produce these effects (116,126–128). Therefore, there is a significant discrepancy between the ability of PCBs to reduce circulating levels of TH and their ability to produce symptoms of hypothyroidism. Some authors have proposed that this discrepancy may be attributable to PCBs acting as imperfect agonists/antagonists on TH receptors (TRs) (129). If so, then PCBs should affect endpoints of TH action in the developing brain in a manner that is not predicted by effects on serum TH. To test this hypothesis, we evaluated the effect the PCB exposure on the expression of a variety of TH-responsive genes in the developing fetus (130) and neonatal rat (128). Interestingly, we found that PCB exposure, using the commercial mixture Aroclor 1254, produces a TH-like effect on the expression of a variety of genes in vivo, including RC3/Neurogranin and Oct-1 in the fetal cortex, and RC3/Neurogranin and Myelin Basic Protein (MBP) in the postnatal brain. In addition, we have found that PCB exposure increases the expression of Malic Enzyme [a well characterized TH-responsive gene in the liver (131)] in the maternal liver. Thus, despite causing a reduction in circulating TH levels, PCBs cause an increase in the expression of genes normally up-regulated by TH. These findings are consistent with a direct action of PCBs on the TR. However, only one report had addressed this proposal directly (132), finding that two hydroxylated PCB congeners (4 0 -OH-PCB 14 and 4 0 -OH-PCB106) exhibit a relatively low affinity for human TRb1 (KiZ32mM). We tested this hypothesis using a very large battery of parent PCB congeners and their metabolites (130), finding that none of these chemicals significantly displaced T3 from either the TRa1 or TRb1, or affecting their affinity for T3. Moreover, we have narrowed the TH-like effects of Aroclor 1254 to a combination of 6 congeners in vivo: PCBs 77, 126, 108, 118, 153, and 132. We do not currently propose that these effects are mediated by dioxin-like congeners (CB 77 and 126) because if we increase the proportion of these congeners in the mixture of 6, the TH-like effect is blocked. This single example illustrates several important points relevant to the issue of deciding whether effects of antithyroid agents are adverse or adaptive. First, without formal endpoints of TH action, not only can one not decide whether the effects of a toxicant on circulating levels of TH are adaptive, but it is also not possible to identify whether the effect of toxicant on circulating levels of TH mediate effects on TH signaling in the developing brain. Considering that endpoints of TH signaling in the developing brain are not routinely incorporated into studies of thyroid toxicants, it is not possible to

454

Zoeller

make evidence-based conclusions about whether the effects of toxicants on serum TH levels are adaptive or adverse (10). Second, this example illustrates that total serum T4 is not a good estimate of TH action in tissues. PCBs appear to selectively reduce circulating levels of T4 both by displacing T4 from the serum binding protein transthyretin (133,134) and by inducing the T4-directed UDP-glucuronocyltransferse (135) in liver. Many PCB studies report that serum TSH levels are not elevated despite low T4 (116,136). Some speculate that this is due to normal serum T3 in PCB-treated animals. However, serum T4 appears to be the important determinant of serum TSH considering that D2-knockout mice exhibit elevated levels of both serum T4 and TSH (137), indicating that intra-pituitary conversion of T4 to T3 is an essential step in the negative feedback action of T4 on the pituitary gland. Thus, it may be that serum displacement of T4 from TTR increases or maintains levels of serum free T4 such that serum TSH remains normal. This possibility remains poorly understood because important serum measures of thyroid function are not routinely assayed, including free hormone (T4 and T3), serum thyrotropin (TSH), estimates of serum binding proteins, and serum Tg. This issue may be particularly important if there exist serum measures of thyroid function (hormone and binding protein level) that provide a more reliable predictor of TH action in tissues. Recent studies indicate that a number of environmental chemicals can directly affect the TR at low concentrations, including PCBs. Iwasaki et al. (138) found that very low concentrations (10K10M) of a hydroxylated PCB congener could suppress T3induced gene expression, using an in vitro expression system. Moreover, this group found that it appears to do so by causing the TR (TRb1) to dissociate from DNA (139). Thus, it may be that PCBs cause a “TH-like” effect on the expression of specific genes in our experiments because it prevents the unliganded TR from suppressing gene expression without cause activation per se. Another important observation are those reported concerning bisphenol A (BPA) (60,61). Parent BPA binds to the TRb1 with relatively low affinity and acts as an antagonist in vitro (60), and we have found that BPA exposure causes serum T4 levels to increase, consistent with an antagonistic effect on the pituitary TRb (unpublished data). Yamada-Okabe et al. (140) reported that 2,3,7,8 tetrachloro-p-dioxin can enhance the T3-mediated gene expression in vitro. This report also characterized the positive and negative effects of PCB exposure on T3regulated gene expression.

CONCLUSION TH is essential for normal brain development, and recent studies demonstrate that TH insufficiency produces adverse consequences on the developing brain depending on the timing and severity of the insufficiency. Despite this importance, toxicological endpoints of TH action in the developing brain have not been validated. Thus, we do not even know the degree to which TH levels must decline before brain development is compromised. Moreover, considering the data indicating that TH effects are exerted on various developmental events by specific TR isoforms, the development of end points of TH action in the developing brain must take this into consideration. Recent studies indicate that various toxicants may exert direct—even novel—actions on the TR and this may be site-specific. It is essential that we correct these weaknesses so that toxicological research can credibly evaluate risk to human (and wildlife) populations to thyroid disrupters.

Thyroid Hormone Signaling

455

REFERENCES 1. Kilby MD. Thyroid hormones and fetal brain development. Clin Endocrinol (Oxf) 2003; 59:280–281. 2. Bernal J, Guadano-Ferraz A, Morte B. Perspectives in the study of thyroid hormone action on brain development and function. Thyroid 2003; 13:1005–1012. 3. Smith JW, Evans AT, Costall B, Smythe JW. Thyroid hormones, brain function and cognition: a brief review. Neurosci Biobehav Rev 2002; 26:45–60. 4. Bernal J. Action of thyroid hormone in brain. J Endocrinol Invest 2002; 25:268–288. 5. Delange FM. Endemic cretinism. In: Braverman LE, Utiger RD, eds. The Thyroid: A Fundamental and Clinical Text. 8th ed. Philadelphia: Lippincott Williams and Wilkins, 2000:743–754. 6. Darbra S, Garau A, Balada F, Sala J, Marti-Carbonell MA. Perinatal hypothyroidism effects on neuromotor competence, novelty-directed exploratory and anxiety-related behaviour and learning in rats. Behav Brain Res 2003; 143:209–215. 7. Zoeller RT, Dowling ALS, Herzig CTA, Iannacone EA, Gauger KJ, Bansal R. Thyroid hormone, brain development, and the environment. Environ Health Perspect 2002; 110:355–361. 8. Brucker-Davis F. Effects of environmental synthetic chemicals on thyroid function. Thyroid 1998; 8:827–856. 9. Gaitan E, ed. Environmental Goitrogenesis. Boca Raton, FL: CRC Press, 1989. 10. Strawson J, Zhao Q, Dourson M. Reference dose for perchlorate based on thyroid hormone change in pregnant women as the critical effect. Regul Toxicol Pharmacol 2004; 39:44–65. 11. Wolff J. Perchlorate and the thyroid gland. Phamacol Rev 1998; 50:89–105. 12. Andersen S, Pedersen KM, Bruun NH, Laurberg P. Narrow individual variations in serum T(4) and T(3) in normal subjects: a clue to the understanding of subclinical thyroid disease. J Clin Endocrinol Metab 2002; 87:1068–1072. 13. Dunn JT, Dunn AD. Thyroglobulin: chemistry, biosynthesis, and proteolysis. In: Braverman LE, Utiger RD, eds. The Thyroid: A Fundamental and Clinical Text. 8th ed. Philadelphia: Lippincott, Williams and Wilkins, 2000:91–104. 14. Taurog A. Hormone synthesis: thyroid iodine metabolism. In: Braverman LE, Utiger RD, eds. The Thyroid: A Fundamental and Clinical Text. 8th ed. Philadelphia: Lippincott Williams and Wilkins, 2000:61–85. 15. Herzog V. Transcytosis in thyroid follicle cells. J Cell Biol 1983; 97:607. 16. Spencer CA. Thyroglobulin. In: Braverman LE, Utiger RD, eds. The Thyroid: A Fundamental and Clinical Text. 8th ed. Philadelphia: Lippincott Williams and Wilkins, 2000:402–413. 17. Sellitti DF, Suzuki K, Doi SQ, et al. Thyroglobulin increases cell proliferation and suppresses Pax-8 in mesangial cells. Biochem Biophys Res Commun 2001; 285:795–799. 18. Wondisford FE, Magner JA, Weintraub BD. Thyrotropin. In: Braverman LE, Utiger RD, eds. The Thyroid: A Fundamental and Clinical Text. 7th ed. Philadelphia: Lippincott-Raven, 1996:190–206. 19. Rapoport B, Spaulding SW. Mechanism of action of thyrotropin and other thyroid growth factors. In: Braverman LE, Utiger RD, eds. Werner and Ingbar’s The Thyroid: A Fundamental and Clinical Text. Philadelphia: Lippincott-Raven, 1986:207–219. 20. Koller KJ, Wolff RS, Warden MK, Zoeller RT. Thyroid hormones regulate levels of thyrotropin-releasing hormone mRNA in the paraventricular nucleus. Proc Natl Acad Sci USA 1987; 84:7329–7333. 21. Segersen TP, Kauer J, Wolfe HC, et al. Thyroid hormone regulates TRH biosynthesis in the paraventricular nucleus of the rat hypothalamus. Science 1987; 84:78–80. 22. Rondeel JMM, deGreef WJ, van der Schoot P, Karels B, Klootwijk W, Visser TJ. Effect of thyroid status and paraventricular area lesions on the release of thyrotropin-releasing hormone and catecholamines into hypophysial portal blood. Endocrinology 1988; 123:523–527.

456

Zoeller

23. Lechan RM, Qi Y, Jackson IMD, Mahdavi V. Identification of thyroid hormone receptor isoforms in thyrotropin-releasing hormone neurons of the hypothalamic paraventricular nucleus. Endocrinology 1994; 135:92–100. 24. Hodin RA, Lazar MA, Wintman BI, Darling DS, Chin WW. Identification of a thyroid hormone receptor that is pituitary-specific. Science 1989; 244:76–79. 25. Hodin RA, Lazar MA, Chin WW. Differential and tissue-specific regulation of the multiple rat c-erbA messenger RNA species by thyroid hormone. J Clin Invest 1990; 85:101–105. 26. Refetoff S, Weiss RE, Usala SJ. The syndromes of resistance to thyroid hormone. Endocrine Rev 1993; 14:348–399. 27. Safer JD, O’Connor MG, Colan SD, Srinivasan S, Tollin SR, Wondisford FE. The thyroid hormone receptor-beta gene mutation R383H is associated with isolated central resistance to thyroid hormone. J Clin Endocrinol Metab 1999; 84:3099–3109. 28. Burmeister LA, Pachucki J, St Germain DL. Thyroid hormones inhibit type 2 iodothyronine deiodinase in the rat cerebral cortex by both pre- and posttranslational mechanisms. Endocrinology 1997; 138:5231–5237. 29. St Germain DL, Galton VA. The deiodinase family of selenoproteins. Thyroid 1997; 7:655–668. 30. Kohrle J. Local activation and inactivation of thyroid hormones: the deiodinase family. Mol Cell Endocrinol 1999; 151:103–119. 31. Kester MHA, Martinez de Mena R, Obregon MJ, et al. Iodothyronine levels in the human developing brain: major regulatory roles of iodothyronine deiodinases in different areas. J Clin Endocrinol Metab 2004; 89:3117–3218. 32. Chatterjee VK, Tata JR. Thyroid hormone receptors and their role in development. Cancer Surv 1992; 14:147–167. 33. Whitfield GK, Jurutka PW, Haussler CA, Haussler MR. Steroid hormone receptors: evolution, ligands, and molecular basis of biologic function. J Cell Biochem 1999; 33:110–122. 34. O’Shea PJ, Williams GR. Insight into the physiological actions of thyroid hormone receptors from genetically modified mice. J Endocrinol 2002; 175:553–570. 35. Wondisford FE. Thyroid hormone action: insight from transgenic mouse models. J Investig Med 2003; 51:215–220. 36. Evans RM. The steroid and thyroid hormone receptor superfamily. Science 1988; 240:889–895. 37. Lazar MA. Thyroid hormone receptors: multiple forms, multiple possibilities. Endocr Rev 1993; 14:184–193. 38. Sap J, Munoz A, Damm K, et al. The c-erbA protein is a high affinity receptor for thyroid hormone. Nature 1986; 324:635–640. 39. Weinberger C, Thompson CC, Ong ES, Lebo R, Gruol DJ, Evans RM. The c-erbA gene encodes a thyroid hormone receptor. Nature 1986; 324:641–646. 40. Forrest D, Reh TA, Rusch A. Neurodevelopmental control by thyroid hormone receptors. Curr Opin Neurobiol 2002; 12:49–56. 41. Flamant F, Samarut J. Thyroid hormone receptors: lessons from knockout and knock-in mutant mice. Trends Endocrinol Metab 2003; 14:85–90. 42. Chassande O, Fraichard A, Gauthier K, et al. Identification of transcripts initiated from an internal promoter in the c-erbAa locus that encode inhibitors of retinoic acid receptor-a and triiodothyronine receptor activities. Mol Endocrinol 1997; 11:1278–1290. 43. Williams GR. Cloning and characterization of two novel thyroid hormone receptor b isoforms. Mol Cell Biol 2000; 20:8329–8342. 44. Bradley DJ, Towle HC, Young WS. Spatial and temporal expression of alpha- and betathyroid hormone receptor mRNAs, including the beta-2 subtype, in the developing mammalian nervous system. J Neurosci 1992; 12:2288–2302. 45. Leonard JL, Farwell AP, Yen PM, Chin WW, Stula M. Differential expression of thyroid hormone receptor isoforms in neurons and astroglial cells. Endocrinol 1994; 135:548–555. 46. Fisher D, Dussault J, Sack J, Chopra I. Ontogenesis of hypothalamic-pituitary-thyroid function and metabolism in man, sheep, rat. Rec Progr Horm Res 1977; 33:59–107.

Thyroid Hormone Signaling

457

47. Lazar MA. Thyroid hormone receptors: Update 1994. Endocr Rev Monogr 1994; 3:280–283. 48. Forrest D, Erway LC, Ng L, Altschuler R, Curran T. Thyroid hormone receptor beta is essential for development of auditory function. Nature Genet 1996; 13:354–357. 49. Forrest D, Hanebuth E, Smeyne RJ, et al. Recessive resistance to thyroid hormone in mice lacking thyroid hormone receptor ß: evidence for tissue-specific modulation of receptor function. EMBO J 1996; 15:3006–3015. 50. Gauthier K, Chassande O, Plateroti M, et al. Different functions for the thyroid hormone receptors TRalpha and TRbeta in the control of thyroid hormone production and post-natal development. EMBO J 1999; 18:623–631. 51. McKenna NJ, O’Malley BW. Combinatorial control of gene expression by nuclear receptors and coregulators. Cell 2002; 108:465–474. 52. Morte B, Manzano J, Scanlan T, Vennstrom B, Bernal J. Deletion of the thyroid hormone receptor alpha 1 prevents the structural alterations of the cerebellum induced by hypothyroidism. Proc Natl Acad Sci USA 2002; 99:3985–3989. 53. Gothe S, Wang Z, Ng L, et al. Mice devoid of all known thyroid hormone receptors are viable but exhibit disorders of the pituitary–thyroid axis, growth, and bone maturation. Genes Dev 1999; 13:1329–1341. 54. Hashimoto K, Curty FH, Borges PP, et al. An unliganded thyroid hormone receptor causes severe neurological dysfunction. Proc Natl Acad Sci USA 2001; 98:3998–4003. 55. Guadano-Ferraz A, Benavides-Piccione R, Venero C, et al. Lack of thyroid hormone receptor alpha1 is associated with selective alterations in behavior and hippocampal circuits. Mol Psychiatry 2003; 8:30–38. 56. Manzano J, Morte B, Scanlan TS, Bernal J. Differential effects of triiodothyronine and the thyroid hormone receptor beta-specific agonist GC-1 on thyroid hormone target genes in the brain. Endocrinology 2003; 144:5480–5487. 57. Morte B, Manzano J, Scanlan TS, Vennstrom B, Bernal J. Aberrant maturation of astrocytes in thyroid hormone receptor {alpha}1 knock out mice reveals an interplay between thyroid hormone receptor isoforms. Endocrinology 2004; 145:1386–1391. 58. Nicholson JL, Altman J. The effects of early hypo- and hyperthyroidism on the development of rat cerebellar cortex. I. Cell proliferation and differentiation. Brain Res 1972; 44:13–23. 59. Xiao Q, Nikodem VM. Apoptosis in the developing cerebellum of the thyroid hormone deficient rat. Front Biosci 1998; 3:A52–A57. 60. Moriyama K, Tagami T, Akamizu T, et al. Thyroid hormone action is disrupted by bisphenol A as an antagonist. J Clin Endocrinol Metab 2002; 87:5185–5190. 61. Kitamura S, Jinno N, Ohta S, Kuroki H, Fujimoto N. Thyroid hormonal activity of the flame retardants tetrabromobisphenol A and tetrachlorobisphenol A. Biochem Biophys Res Commun 2002; 293:554–559. 62. Cranefield P, Federn W. Paracelsus on Goiter and Cretinism: A Translation and Discussion of “De Struma Vulgo Der Kropf. Bull Hist Med 1963; 37:463–471. 63. Merke F. The history of endemic goitre and cretinism in the thirteenth to fifteenth centuries. Proc R Soc Med 1960; 53:995–1002. 64. DeLong GR, Ma T, Cao XY, et al. The neuromotor deficit in endemic cretinism. In: Stanbury JB, ed. The Damaged Brain of Iodine Deficiency. New York: Cognizant Communications, 1994:9–17. 65. Chan S, Rovet J. Thyroid hormones in the fetal central nervous system development. Fetal Mater Med Rev 2003; 14:177–208. 66. Brown AW, Bronstein IP, Kraines R. Hypothyroidism and cretinism in childhood. VI. Influence of thyroid therapy on mental growth. Amer J Dis Child 1939; 57:517–523. 67. Haddow JE, Palomaki GE, Allan WC, et al. Maternal thyroid deficiency during pregnancy and subsequent neuropsychological development of the child. N Engl J Med 1999; 341:549–555. 68. Pop VJ, Kuijpens JL, van Baar AL, et al. Low maternal free thyroxine concentrations during early pregnancy are associated with impaired psychomotor development in infancy. Clin Endocrinol (Oxf) 1999; 50:149–155.

458

Zoeller

69. Smit BJ, Kok JH, Vulsma T, Briet JM, Boer K, Wiersinga WM. Neurologic development of the newborn and young child in relation to maternal thyroid function. Acta Paediatr 2000; 89:291–295. 70. Glinoer D. Potential consequences of maternal hypothyroidism on the offspring: evidence and implications. Horm Res 2001; 55:109–114. 71. Calvo RM, Jauniaux E, Gulbis B, et al. Fetal tissues are exposed to biologically relevant free thyroxine concentrations during early phases of development. J Clin Endocrinol Metab 2002; 87:1768–1777. 72. Pacaud D, Huot C, Gattereau A, et al. Outcome in three siblings with antibody-mediated transient congenital hypothyroidism. J Pediatr 1995; 127:275–277. 73. Matsuura N, Konishi J. Transient hypothyroidism in infants born to mothers with chronic thyroiditis—a nationwide study of twenty-three cases. The transient hypothyroidism study group. Endocrinol Jpn 1990; 37:369–379. 74. Rovet J. Congenital hypothyroidism: long-term outcome. Thyroid 1999; 9:741–748. 75. Heyerdahl S, Oerbeck B. Congenital hypothyroidism: developmental outcome in relation to levothyroxine treatment variables. Thyroid 2003; 13:1029–1038. 76. Oerbeck B, Sundet K, Kase BF, Heyerdahl S. Congenital hypothyroidism: influence of disease severity and L-thyroxine treatment on intellectual, motor, and school-associated outcomes in young adults. Pediatrics 2003; 112:923–930. 77. Friedhoff AJ, Miller JC, Armour M, Schweitzer JW, Mohan S. Role of maternal biochemistry in fetal brain development: effect of maternal thyroidectomy on behaviour and biogenic amine metabolism in rat progeny. Int J NeuropsychoPharmacol 2000; 3:89–97. 78. Sinha AK, Pickard MR, Kim KD, et al. Perturbation of thyroid hormone homeostasis in the adult and brain function. Acta Med Austriaca 1994; 21:35–43. 79. Sinha AK, Pickard MR, Hubank MJ, et al. Maternal hypothyroxinemia and brain development: II. Biochemical, metabolic and behavioural correlates. Acta Med Austriaca 1992; 19:49–54. 80. Hendrich CE, Jackson WJ, Porterfield SP. Behavioral testing of progenies of Tx (hypothyroid) and growth hormone- treated Tx rats: an animal model for mental retardation. Neuroendocrinology 1984; 38:429–437. 81. Eayrs JE, Taylor SH. The effect of thyroid deficiency induced by methyl thiouracil on the maturation of the central nervous system. J Anat 1951; 85:350–358. 82. Eayrs JT, Horne G. The development of the cerebral cortex in hypothyroid and starved rats. Anat Rec 1955; 121:53–61. 83. Eayrs JT. The cerebral cortex of normal and hypothyroid rats. Acta Anat 1955; 25:160–183. 84. Gravel C, Hawkes R. Maturation of the corpus callosum of the rat: I. Influence of thyroid hormones on the topography of callosal projections. J Comp Neurol 1990; 291:128–146. 85. Gravel C, Sasseville R, Hawkes R. Maturation of the corpus callosum of the rat: II. Influence of thyroid hormones on the number and maturation of axons. J Comp Neurol 1990; 291:147–161. 86. Berbel P, Guadano-Ferraz A, Angulo A, Ramon Cerezo J. Role of thyroid hormones in the maturation of interhemispheric connections in rats. Behav Brain Res 1994; 64:9–14. 87. Berbel P, Guadano-Ferraz A, Martinez M, Quiles JA, Balboa R, Innocenti GM. Organization of auditory callosal connections in hypothyroid adult rats. Eur J Neurosci. 1993; 5:1465–1478. 88. Berbel P, Marco P, Cerezo JR, DeFelipe J. Distribution of parvalbumin immunoreactivity in the neocortex of hypothyroid adult rats. Neurosci Lett 1996; 204:65–68. 89. Berbel P, Auso E, Garcia-Velasco JV, Molina ML, Camacho M. Role of thyroid hormones in the maturation and organisation of rat barrel cortex. Neuroscience 2001; 107:383–394. 90. Lavado-Autric R, Auso E, Garcia-Velasco JV, et al. Early maternal hypothyroxinemia alters histogenesis and cerebral cortex cytoarchitecture of the progeny. J Clin Invest 2003; 111:1073–1082. 91. Auso E, Lavado-Autric R, Cuevas E, Escobar Del Rey F, Morreale De Escobar G, Berbel P. A moderate and transient deficiency of maternal thyroid function at the beginning of fetal neocorticogenesis alters neuronal migration. Endocrinology 2004; 145:4037–4047.

Thyroid Hormone Signaling

459

92. Howdeshell KL. A model of the development of the brain as a construct of the thyroid system. Environ Health Perspect 2002; 110:337–348. 93. Chana A, Concejero MA, de Frutos M, Gonzalez MJ, Herradon B. Computational studies on biphenyl derivatives. Analysis of the conformational mobility, molecular electrostatic potential, and dipole moment of chlorinated biphenyl: searching for the rationalization of the selective toxicity of polychlorinated biphenyls (PCBs). Chem Res Toxicol 2002; 15:1514–1526. 94. Breivik K, Sweetman A, Pacyna JM, Jones KC. Towards a global historical emission inventory for selected PCB congeners—a mass balance approach. 2. Emissions. Sci Total Environ 2002; 290:199–224. 95. Sjodin A, Jones RS, Focant JF, et al. Retrospective time-trend study of polybrominated diphenyl ether and polybrominated and polychlorinated biphenyl levels in human serum from the United States. Environ Health Perspect 2004; 112:654–658. 96. Huisman M, Koopman-Esseboom C, Lanting CI, et al. Neurological condition in 18-monthold children perinatally exposed to polychlorinated biphenyls and dioxins. Early Hum Dev 1995; 43:165–176. 97. Jackson TA, Richer JK, Bain DL, Takimoto GS, Tung L, Horwitz KB. The partial agonist activity of antagonist-occupied steroid receptors is controlled by a novel hinge domainbinding coactivator L7/SPA and the corepressors N-CoR or SMRT. Mol Endocrinol 1997; 11:693–705. 98. Osius N, Karmaus W, Kruse H, Witten J. Exposure to polychlorinated biphenyls and levels of thyroid hormones in children. Environ Health Perspect 1999; 107:843–849. 99. Korrick SA, Altshul L. High breast milk levels of polychlorinated biphenyls (PCBs) among four women living adjacent to a PCB-contaminated waste site. Environ Health Perspect 1998; 106:513–518. 100. Ayotte P, Muckle G, Jacobson JL, Jacobson SW, Dewailly E. Assessment of pre- and postnatal exposure to polychlorinated biphenyls: lessons from the Inuit cohort study. Environ Health Perspect 2003; 111:1253–1258. 101. Walkowiak J, Wiener JA, Fastabend A, et al. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on psychodevelopment in early childhood. Lancet 2001; 358:1602–1607. 102. Morreale de Escobar G, Obregon MJ, Escobar del Rey F. Is neuropsychological development related to maternal hypothyroidism or to maternal hypothyroxinemia. J Clin Endocrinol Metab 2000; 85:3975–3987. 103. Porterfield SP, Hendry LB. Impact of PCBs on thyroid hormone directed brain development. Toxicol Ind Health 1998; 14:103–120. 104. Porterfield SP. Thyroidal dysfunction and environmental chemicals—potential impact on brain development. Environ Health Perspect 2000; 108:433–438. 105. McKinney JD, Waller CL. Molecular determinants of hormone mimicry: halogenated aromatic hydrocarbon environmental agents. J Toxicol Env Hlth—Part B 1998; 1:27–58. 106. Koopman-Esseboom C, Morse DC, Weisglas-Kuperus N, et al. Effects of dioxins and polychlorinated biphenyls on thyroid hormone status of pregnant women and their infants. Pediatr Res 1994; 36:468–473. 107. Schantz SL, Widholm JJ, Rice DC. Effects of PCB exposure on neuropsychological function in children. Environ Health Perspect 2003; 111:357–376. 108. Sala M, Sunyer J, Herrero C, To-Figueras J, Grimalt J. Association between serum concentrations of hexachlorobenzene and polychlorobiphenyls with thyroid hormone and liver enzymes in a sample of the general population. Occup Environ Med 2001; 58:172–177. 109. Longnecker MP, Gladen BC, Patterson DG, Jr, Rogan WJ. Polychlorinated biphenyl (PCB) exposure in relation to thyroid hormone levels in neonates. Epidemiology 2000; 11:249–254. 110. Hagmar L, Rylander L, Dyremark E, Klasson-Wehler E, Erfurth EM. Plasma concentrations of persistent organochlorines in relation to thyrotropin and thyroid hormone levels in women. Int Arch Occup Environ Health 2001; 74:184–188.

460

Zoeller

111. Steuerwald U, Weihe P, Jorgensen PJ, et al. Maternal seafood diet, methylmercury exposure, and neonatal neurologic function. J Pediatr 2000; 136:599–605. 112. Bastomsky CH, Murthy PVN, Banovac K. Alterations in thyroxine metabolism produced by cutaneous application of microscope immersion oil: effects due to polychlorinated biphenyls. Endocrinol 1976; 98:1309–1314. 113. Bastomsky CH. Effects of a polychlorinated biphenyl mixture (Aroclor 1254) and DDT on biliary thyroxine excretion in rats. Endocrinology 1974; 95:1150–1155. 114. Brouwer A, Morse DC, Lans MC, et al. Interactions of persistent environmental organohalides with the thyroid hormone system: mechanisms and possible consequences for animal and human health. Toxicol Ind Health 1998; 14:59–84. 115. Goldey ES, Kehn LS, Rehnberg GL, Crofton KM. Effects of developmental hypothyroidism on auditory and motor function in the rat. Toxicol Appl Pharmacol 1995; 135:67–76. 116. Goldey ES, Kehn LS, Lau C, Rehnberg GL, Crofton KM. Developmental exposure to polychlorinated biphenyls (Aroclor 1254) reduces circulating thyroid hormone concentrations and causes hearing deficits in rats. Toxicol Appl Pharmacol 1995; 135:77–88. 117. Crofton KM, Ding D, Padich R, Taylor M, Henderson D. Hearing loss following exposure during development to polychlorinated biphenyls: a cochlear site of action. Hear Res 2000; 144:196–204. 118. Crofton KM, Kodavanti PR, Derr-Yellin EC, Casey AC, Kehn LS. PCBs, thyroid hormones, and ototoxicity in rats: cross-fostering experiments demonstrate the impact of postnatal lactation exposure. Toxicol Sci 2000; 57:131–140. 119. Juarez de Ku LM, Sharma-Stokkermans M, Meserve LA. Thyroxine normalizes polychlorinated biphenyl (PCB) dose-related depression of choline acetyltransferase (ChAT) activity in hippocampus and basal forebrain of 15-day-old rats. Toxicol 1994; 94:19–30. 120. Cooke PS, Zhao Y-D, Hansen LG. Neonatal polychlorinated biphenyl treatment increases adult testis size and sperm production in the rat. Toxicol Appl Pharmacol 1996; 136:112–117. 121. Baldridge MG, Stahl RL, Gerstenberger SL, Tripoli V, Hutz RJ. Modulation of ovarian follicle maturation in Long-Evans rats exposed to polychlorinated biphenyls (PCBs) in-utero and lactationally. Reprod Toxicol 2003; 17:567–573. 122. Goldey ES, Crofton KM. Thyroxine replacement attenuates hypothyroxinemia, hearing loss, and motor deficits following developmental exposure to Aroclor 1254 in rats. Toxicol Sci 1998; 45:94–105. 123. Connors JM, Hedge GA. Feedback regulation of thyrotropin by thyroxine under physiological conditions. Am J Physiol 1981; 240:E308–E313. 124. Schwartz HL. Effect of thyroid hormone on growth and development. In: Oppenheimer JH, Samuels HH, eds. Molecular Basis of Thyroid Hormone Action. New York: Academic Press, 1983:413–444. 125. Varma SK, Murray R, Stanbury JB. Effect of maternal hypothyroidism and triiodothyronine on the fetus and newborn in rats. Endocrinology 1978; 102:24–30. 126. Kolaja KL, Klaassen CD. Dose-response examination of UDP-glucuronosyltransferase inducers and their ability to increase both TGF-beta expression and thyroid follicular cell apoptosis. Toxicol Sci 1998; 46:31–37. 127. Hood A, Klaassen CD. Effects of microsomal enzyme inducers on outer-ring deiodinase activity toward thyroid hormones in various rat tissues. Toxicol Appl Pharmacol 2000; 163:240–248. 128. Zoeller RT, Dowling AL, Vas AA. Developmental exposure to polychlorinated biphenyls exerts thyroid hormone-like effects on the expression of RC3/neurogranin and myelin basic protein messenger ribonucleic acids in the developing rat brain. Endocrinol 2000; 141:181–189. 129. McKinney JD, Waller CL. Polychlorinated biphenyls as hormonally active structural analogues. Environ Health Perspect 1994; 102:290–297. 130. Gauger KJ, Kato Y, Haraguchi K, et al. Polychlorinated biphenyls (PCBs) exert thyroid hormone-like effects in the fetal rat brain but do not bind to thyroid hormone receptors. Environ Health Perspect 2004; 112:516–523.

Thyroid Hormone Signaling

461

131. Song M-KH, Dozin B, Grieco D, Rall JE, Nikodem VM. Transcriptional activation and stabilization of malic enzyme mRNA precursor by thyroid hormone. J Biol Chem 1986; 263:17970–17974. 132. Cheek AO, Kow K, Chen J, McLachlan JA. Potential mechanisms of thyroid disruption in humans: interaction of organochlorine compounds with thyroid receptor, transthyretin, and thyroid-binding globulin. Environ Health Perspect 1999; 107:273–278. 133. Brouwer A, Morse DC, Lans MC, et al. Interactions of persistent environmental organohalogens with the thyroid hormone system: mechanisms and possible consequences for animal and human health. Toxicol Ind Health 1998; 14:59–84. 134. Chauhan KR, Kodavanti PR, McKinney JD. Assessing the role of ortho-substitution on polychlorinated biphenyl binding to transthyretin, a thyroxine transport protein. Toxicol Appl Pharmacol 2000; 162:10–21. 135. Hood A, Klaassen CD. Differential effects of microsomal enzyme inducers on in vitro thyroxine (T(4)) and triiodothyronine (T(3)) glucuronidation. Toxicol Sci 2000; 55:78–84. 136. Liu J, Liu Y, Barter RA, Klaassen CD. Alteration of thyroid homeostasis by UDP-glucuronosyltransferase inducers in rats: a dose-response study. J Pharmacol Exp Ther 1995; 273:977–985. 137. Schneider MJ, Fiering SN, Pallud SE, Parlow AF. Targeted disruption of the type 2 selenodeiodinase gene (DIO2) results in a phenotype of pituitary resistance to T4. Mol Endocrinol 2001; 15:2137–2148. 138. Iwasaki T, Miyazaki W, Takeshita A, Kuroda Y, Koibuchi N. Polychlorinated biphenyls suppress thyroid hormone-induced transactivation. Biochem Biophys Res Commun 2002; 299:384–388. 139. Miyazaki W, Iwasaki T, Takeshita A, Kuroda Y, Koibuchi N. Polychlorinated biphenyls suppress thyroid hormone receptor-mediated transcription through a novel Mechanism. J Biol Chem 2004; 279:18195–18202. 140. Yamada-Okabe T, Aono T, Sakai H, Kashima Y, Yamada-Okabe H. 2,3,7,8-tetrachlorodibenzo-p-dioxin augments the modulation of gene expression mediated by the thyroid hormone receptor. Toxicol Appl Pharmacol 2004; 194:201–210.

SECTION FIVE: DEVELOPMENTAL NEUROTOXICITY: CLINICAL PRACTICE, RISK ASSESSMENT ETHICS

22 Public Health Perspectives on Developmental Neurotoxicology Jennifer S. Mammen Environmental Health Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland, U.S.A.

Lynn R. Goldman Environmental Health Sciences, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, U.S.A.

PUBLIC HEALTH BURDEN OF DEVELOPMENTAL DISABILITIES/ LEARNING DISABILITIES IN CHILDREN Developmental disabilities are among the most important child health problems in the US. The U.S. Centers for Disease Control and Prevention (CDC) has estimated that 17% of children have a developmental disability and 2% have a “severe” disability requiring special education. According to the American Psychiatric Association, 3–7% of children have attention deficit hyperactivity disorder (ADHD) with higher rates reported in some communities. According to the CDC, as many as 6 in 1000 children have some form of autism spectrum disorder. The identification of preventable causes of such developmental and learning disabilities is crucial to their prevention. For example, lower levels of folate have been linked to neural tube defects and rates of this birth defect have sharply declined since the introduction of folic acid supplements into the U.S. diet. Broadly speaking, there are diverse influences on neurologic development, including genetics, lifestyle, infections in utero and in early childhood, and head trauma. Known causes of developmental defects include genetic (e.g., Trisomy 21), lifestyle (fetal alcohol, smoking), environmental (hyperthermia, UV radiation, X-rays, certain chemicals) and nutritional (deficiencies or excesses) factors (2). In most cases, we do not have information for parents about why their children have disabilities. Increasingly it is appreciated that while there is strong evidence for a genetic component to most developmental environmental influences that trigger these genetic traits. For malformations, although 15– 25% are estimated to be purely genetic in origin, and 10% to broad environmental factors (mostly maternal conditions and infectious agents), the vast majority, 65–75%, are of unknown origin and are thought to be due to complex interactions among multiple genes, multiple teratogens, gene and environmental factors, or spontaneous mutations (1). 463

464

Mammen and Goldman

Others have concluded that “3% of all developmental defects are attributable to exposure to toxic chemicals and physical agents, including environmental factors, and 25% of developmental defects may be due to a combination of genetic and environmental factors” (2). Thus, environmental exposures to chemicals and other substances must be considered within the broader context of all agents and factors that may be affecting neurodevelopment. A further consideration for policy makers is the extent to which risk factors may be controllable or amenable to change. For example, a number of persistent pollutants—lead, polychlorinated biphenyls and dioxins, and methyl mercury—have been linked with developmental neurotoxicity in children. While we do not know the role of these, and other environmental exposures, in developmental disabilities, the environmental persistence changes available policy tools and shifts the focus form clean up to prevention of contamination in order to prevent disease. There are numerous policy efforts that have been undertaken to address such concerns over the last ten years. These efforts have involved: assessment of risk of potential developmental neurotoxic agents; expanding funding for tracking developmental disabilities and researching their etiologies; and controlling persistent pollutants, many of which are developmental neurotoxicants. This paper reviews our progress in these three areas over the last ten years, and looks at how this information can be applied to policy decisions in order to optimise our children’s development.

ASSESSMENT OF RISK OF POTENTIAL DEVELOPMENTAL NEUROTOXIC AGENTS Environmental health policy responds to two questions: which of the xenobiotics currently in the environment pose sizable and immediate health risks; and which of the many new compounds being synthesized and added to waste stream every year will pose health threats in the future? That is to say, policy makers are charged with establishing priorities for current problems and screening for future ones. Establishing priorities among ongoing environmental health threats requires being able to perform some kind of assessment of the potential for harm. at a minimum to estimate the relative threat of each problem. The framework that is used in the United States for such assessment is risk assessment. Formally, this is done in four steps: hazard identification, dose response assessment, exposure assessment, and risk characterization (3). Hazard identification, traditionally, the Environmental Protection Agency (EPA) has relied on population data on rates of intake (e.g., drinking!gallons of water a week) together with environmental sampling (e.g., water containing y ppm of a contaminant of concern) to estimate exposures. These assessments make a number of broad assumptions that may not apply to all populations. Children often have greater rates of exposure to contaminants in food, water, air, and household and play environments. Because of differences in metabolism, may be more or less exposed to hazardous substances. To the extent that the industrial chemicals or pesticides being reviewed are current environmental contaminants, improved epidemiology that provides direct measures of both exposure and outcome and the dose-response relationships in children vastly improves the quality of the risk assessment being performed, and is therefore of high value to public health policy makers. Hazard identification and dose-response assessments involve the determination of health outcomes that may be associated with particular exposures. Most risk assessments rely on hazard data from animal studies, occasionally on the effects seen in occupational settings, where adults are exposed to higher levels of toxicant, and much less frequently on epidemiological data from studies of children. The validity of data from animal

Public Health Perspectives on Developmental Neurotoxicology

465

experimentation and adult studies for determining the health effects of chronic low-level exposures to human children is a very contentious issue in the regulatory community of activists and industry. Not only may the hazards differ for children but also children may have different dose response curves, that is, they may be more or less susceptible. Developmental neurotoxicity is an area that, by definition, involves increased susceptibility of children compared with adults, and may also imply an increased vulnerability of humans compared to animals given the more complex functions performed by the human brain. Epidemiologic studies of children may be more direct but also present challenges for determining how to measure exposure at low levels and assess developmental endpoints appropriately, as discussed below. Being able to interpret the implications of available data for public health is a major challenge for non-scientist policy makers. The risk characterization, in which all the foregoing information is summarized by the agency to evaluate the overall impacts, needs to take into account the uncertainties in the data that result from extrapolations such as those between animals and humans. Under the Food Quality Protection Act (FQPA) (the pesticide safety law), the EPA must employ a tenfold safety factor to assure that increased susceptibility and exposure to children has been accounted for in the risk analysis. A “different” factor may be employed by the EPA if so warranted by the data (4). Various groups have complained that this process is either an under- or over-estimate of risk (which is claimed is the result largely of a priori policy goals) with no greater likelihood of being accurate about the risk than the original correction factor itself. There are no such specific provisions in the law under which industrial chemicals are regulated, the Toxic Substances Control Act (TSCA). Rather, under that provision, the EPA is required to show that a chemical poses an “unreasonable risk” in order to regulate it (which has been interpreted to mean balancing with the costs to industry). How to bridge the gap between the common sense idea of “unreasonable risk” and the reality of imperfect scientific understanding is a major public policy challenge. Hazard Identification and Dose Response Assessment To identify potential hazards requires a basis of scientific information. The criteria proposed by Brent for identification of causes of congenital malformation (1) are equally applicable to putative developmental neurotoxicants and include: controlled epidemiology studies that consistently show a relationship between particular effects and exposure to humans; secular trend data that show a relationship between changing levels of exposure to an agent and the specific effect over time; animal developmental toxicity studies (which can indicate whether there is a general hazard to the development of the fetus but may not give valid information about the specific adverse effect); and evidence for a dose-response and biological plausibility, meaning information on issues such as mechanisms of action, receptor interactions that shed light on the association (1). According to Rodier, there are a number of modes of action of agents that may be associated with maldevelopment of the central nervous system, including agents that interfere with neuronal cell proliferation early in pregnancy; agents that interfere with appropriate migration of neurons; agents that may modify neuronal connections, transmitter levels and receptor numbers; agents that may alter myelin deposition; and agents that may affect postnatal development (5). Brent (1) also identified a number of “teratogenic principles” which also are useful for identification of potential teratogens. These include: (1) teratogens exhibit a dose response and a threshold below which effects do not occur; (2) the embryonic stage of exposure is critical to whether the teratogenic effect will be produced; (3) there is specificity between the teratogen and the effect to be produced; and (4) presence of malformations other than those

466

Mammen and Goldman

produced by a specific teratogen suggests a different mechanism. Dose response assessment relies on the same sort of information. Particularly useful (but often lacking) is information regarding toxicity in animals at doses in the same range as those for humans (1). Laboratory Models In 1993, the National Research Council published its report Pesticides in the Diets of Infants and Children. It concluded that the EPA was doing an inadequate job in assessing the risks of pesticides to children and, among many recommendations, called for developmental neurotoxicity testing of pesticides. Although the report was on pesticides, the same was true for other substances—industrial chemicals, pharmaceutical agents, food additives, consumer products and workplace exposures. While policy makers historically had been focused on prevention of cancer as a target, they had not addressed developmental toxicity in most of the efforts to assure safety. Ideally, the preventive goal of public health would require the detection of neurotoxic compounds before there is widespread environmental exposure of populations on which to do epidemiology. This means that developmental outcomes in children need to be correlated to laboratory models that can be used to test potential toxins. Current animal models of neurodevelopmental toxicity include various rodent paradigms for learning and memory—such as the ability to do mazes—and behavioral assessments. Primates are sometimes also used in such research. One obvious problem is that some human cognitive domains, for example abstract reasoning, do not have clear equivalents in the rat or mouse. It is a plausible hypothesis that if a rat has trouble remembering where a platform is located in a pool after lead exposure then a child will also have difficulty with learning, but it is less certain that any compound that impacts human intellectual ability will have similarly observable effects in rodent or other toxicological models. This latter inference needs to be present for a screening tool to be of value in public health. Since 1993, efforts have been under way to refine and apply a standard developmental neurotoxicity test battery for pesticides. What progress has been made in this area since 1993? The EPA in 1991 issued its first test guideline for the assessment of developmental neurotoxicity; however, it was triggered by other test findings and was not employed very frequently (6). More than ten years of effort had gone into its development. Then, in 1996, Congress passed the FQPA, which required that children’s risks be assessed for pesticides. In 1998, the EPA’s developmental neurotoxicity guideline was reissued as OPPTS 870.6300. The guideline requirements have been summarized by EPA scientists as follows: “In this study, pregnant rats are administered the chemical by an oral route beginning on gestation day 6 and continuing through postnatal day 10. The offspring are therefore exposed to the chemical, via the maternal circulation and/or milk, during in utero and early postnatal development. Examination of the dams in this study includes minimal clinical and functional observations and weekly body weight measurements. Endpoints evaluated in the offspring between birth and approximately day 60 of age include measures of physical development (including growth and survival, and the age of sexual maturation), functional observations, motor activity, auditory startle reflex function, and learning and memory. At postnatal day 11 and at study termination, the offspring are subjected to extensive neuropathological examination including simple morphometric analysis (consisting at a minimum of a reliable estimate of the thickness of major layers at representative locations within the neocortex, hippocampus, and cerebellum) (6).”

In 1999, the EPA issued a data call-in for 38 organophosphate pesticides. According to EPA scientists, this order “included a request for specific modifications to the

Public Health Perspectives on Developmental Neurotoxicology

467

guideline study, in order to address a number of issues and concerns that were identified, including increasing the duration of dosing to postnatal day 21, increasing the number of animals examined for neuropathology (from 6/sex/dose to 10/sex/dose), demonstrating the adequacy of postnatal dosing for substances that are not present to any significant extent in the milk of the dams (including an assessment for the need for direct exposure of pups to the test substance), and conducting a comparative evaluation of cholinesterase inhibition (or other biomarkers) and behavior in adults and young organisms” (6). We have yet to see the full results of that effort. EPA scientists have published a report detailing quality problems with 34 developmental neurotoxicity tests submitted by 16 laboratories between 1991 and 2002. There were major quality control problems, most notably a lack of positive control data for all endpoints for the majority of test laboratories and the fact that, among those submitting positive control data, the “positive control” was ineffective 13% of the time, indicating either a problem with the control or with the assay. The EPA reviewers identified multiple deficiencies in data and analyses indicating a lack of proficiency among most of the laboratories (7). No data call-ins for developmental neurotoxicity have been issued for most other pesticide classes or other pesticide ingredients. For other substances there has been little progress in testing for developmental neurotoxicity at the EPA. The EPA, the chemical industry, and Environmental Defense are producing much new information about chemicals under the voluntary High Production Volume (HPV) Chemical Testing Program. The program, modeled after the SIDS (Screening Inventory Data Set) that was agreed to by Organization for Economic Cooperation and Development (OECD) nations in the 1980s, does not include a specific assessment of developmental neurotoxicity and neither the EPA nor the OECD have proposed criteria that would trigger such evaluation, post screening. The EPA and the American Chemistry Council are also carrying out a voluntary test program called the Voluntary Children’s Chemical Evaluation Program (VCCEP); however, this pilot effort has yet to yield any new information on neurodevelopmental effects in children over the three years of its existence. A third effort, required under both FIFRA and the Safe Drinking Water Act, is the Endocrine Disruptor Screening and Testing Program (EDSTP). This program has been under development for the last seven years but is not specifically targeted to neurological outcomes. In September 2005, the EPA announced that it will begin to identify and screen the first 50–100 chemicals. However, identification of potential endocrine disruptors is of particular interest since, at least on a theoretical basis, these are chemicals that are most likely to modify. Meanwhile, the FDA is engaged in a process to assess pediatric effectiveness and hazards associated with certain drugs that are already on the market. This activity is occurring under the Best Pharmaceuticals for Children Act (PL-No. 107–109) that was enacted in 2002. This is an effort that will have enormous benefit for children. While the FDA has not yet adopted a protocol for developmental neurotoxicity testing of drugs or other agents it has considered this end point in several recent decisions. Developmental neurotoxicity testing is controversial. Some question whether the current EPA developmental neurotoxicity protocol is adequate for detection of disruption of all components of neurodevelopment, and over the full developmental time span (6). (For example, dosing begins later and ends prior to full development of the nervous system.) Others have been concerned about the numbers of animals required by the test and whether the test could be done more humanely in conjunction with other studies. Other controversies have to do with the circumstances under which the test is warranted, with some (mostly from industry) asserting that the test should be triggered only with clear indications, from earlier testing, that a substance is likely to be neurotoxic. A committee of the National Research Council in 2000 recommended that novel approaches

468

Mammen and Goldman

be considered to the broader problem of assessing developmental toxicity of chemicals, utilizing so-called “model systems” that might “include in-vitro and cellular assays, nonmammalian (e.g., fruit fly, roundworm, and zebrafish) developmental assays, mammalian (e.g., the mouse) developmental assays, and in-depth mammalian tests of mechanism and susceptibility” (2) For developmental neurotoxicity such systems could provide much needed information about mechanisms of action that are critical in identifying developmental neurotoxicants, dose-response relationships, and critical windows of exposure. It has been suggested that particularly for developmental neurotoxicity such models are needed since the conventional test battery has missed important modes of action such as inhibition of DNA expression by the pesticide chlorpyrifos at levels below those associated with cholinesterase inhibition (8). The OECD in 1996 began a process to develop an internationally harmonized guideline for developmental neurotoxicity evaluations, and in September 2003 the OECD reviewed a final draft guideline. The new OECD guideline has some improvements over the EPA’s guideline. These may include extension of the period of postnatal dosing, assessment of preweaning reflex development, a change in the timing of the first neuropathological evaluation (from postnatal day 11 to postnatal day 21 or 22), tiered morphometric evaluation of offspring brains, and increased flexibility in the methods used to assign animals to testing (6). This guideline is scheduled for completion in 2006. Although these efforts to develop and improve upon developmental neurotoxicity testing guidelines are important, it is not yet clear how often and when the developmental neurotoxicity test will be used. Overall, it is fair to conclude that there has been progress on development of developmental neurotoxicity testing processes, but very little output in terms of new information at this point in time. Clearly, research efforts to develop better predictive models, as well as a better understanding of the fundamental mechanisms for neurotoxicity, are needed. Epidemiological Investigations As noted above, it is not only prevention of developmental disabilities, but also promotion of optimal brain functioning (and capacity to live a happy and productive life) that are goals of policy makers. Translating these goals to science is not easy. The functions of the brain are quite diverse, ranging from control over gross motor behavior to subtle abstract reasoning. Because environmental exposures can in theory disrupt the development of any or all of these functions, it is the goal of an epidemiological investigation of environmental effects to discover which such functions are affected by individual toxicants. From this one would like ultimately to infer what the impact is likely to be on individuals and the population as a whole, in terms of neurological functioning and disabilities. As Bellinger recently argued, it is important for many reasons to assess subtle changes in function (9). From a societal perspective, small shifts in average functioning can cause large burdens across the population if there are resulting increases in the extent of widespread vulnerability, for example of students to school failure and drop-outs, or of teenagers to difficulty with emotional control and violence. Furthermore, such small average shifts that may be observed in the middle of a distribution of functioning and behavior are often linked to similar shifts at the tails of the distribution (assuming no change in variance). Thus, a decline in population average implies a decrease in the number of exceptional performers and an increase in the number of poor performers. On the practical side, while small changes in population norms may be more difficult to assess technically than clinical disease states, it can be difficult for a population-based study to obtain sufficient power to observe small increases in rare diseases.

Public Health Perspectives on Developmental Neurotoxicology

469

Ideally, to assess risks of subtle neurological changes, we need assessment tools that make fine measurements of normal function. Since individual toxins might affect specific areas of function, termed “domains,” while leaving others untouched, the ideal research tool would assess each domain’s function independently. There are many theoretical schemas that divide brain functions into discrete domains. A workshop sponsored by the Agency for Toxic Substances and Disease Registry (ATSDR) (10) published a neurobehavioral test battery for use in environmental field studies; ATSDR proposed a list of eight cognitive domains, including affect (mood), memory, attention, social skills/ adjustment, learning, communication, executive control, activity regulation, and problem solving. Research in intelligence has also led to proposals that divide intellectual function into domains. This research has used factor analysis to identify distinct functional areas that are independent of other domains that are assessed in IQ batteries. This analysis has been used to highlight the strengths and weaknesses of current test batteries for IQ, none of which assess all of the posited domains. This work suggests that new tasks could be designed to provide a battery that would encompass sufficient numbers of tasks for each domain to produce valid and reliable scores across the range of brain function. Unfortunately, such a tool does not yet exist, and while clinicians can use individual tasks drawn from multiple batteries to probe an individual’s strengths and weaknesses for the purposes of designing an educational plan, such a practice is not possible in quantitative research, where the normative data are battery-specific. In practice, researchers have assessed brain function in children by using videotapes and scoring of behavior, parent and teacher interviews of behavior, neurological evaluations, records of school achievement, and standardized batteries known broadly as IQ tests. In each of these areas there are a number of tools that are available for different age ranges; each has advantages and disadvantages in field studies. Neurological assessments are available starting from birth. These have generally been designed for the purpose of identifying children with clinical disease and are relatively insensitive to small perturbations in function. Videotaping protocols have been used especially to look at changes in behavior, for example gender-based play, which has been hypothesized to change as a result of exposure to hormonally active compounds. Parent and teacher interviews are the main stay of the ATSDR protocol, and provide fairly low cost information, but are more open to bias. IQ-test test batteries are widely used in children as young as one-year-old to test both motor and cognitive function. A reliance of many tests on language to test cognitive function can impede good cross-cultural testing and also of course make it more difficult to assess individual domains. New test batteries continue to proliferate as computer-based assessments are developed, and tests are developed to target specific populations of disability. The public’s health is served best when genuine health risks are detected, and our relative intolerance of risking harm to children suggests that increased sensitivity would take precedence over greater specificity in the choice of a research tool for detecting a neurological outcome. In practice, this requires balancing the sophistication of the research tool against the size of the population to be evaluated. The more comprehensive the data collected, the more resource intensive the protocol. For example, asking parents to rate their children’s behavior on a questionnaire is more easily accomplished than videotaping and scoring children’s behavior in a structured situation. While videotaping protocols will create more quantitative and objective data, a structured interview such as is recommended by the ATSDR is experimentally more feasible in many instances, allowing for a larger population to be examined. It is possible that the larger size of the resultant study would increase the power to detect small differences between those with different levels of exposure to the toxicant of concern, assuming the tool is sensitive to real

470

Mammen and Goldman

differences in behavior. On the other hand the greater sensitivity gained from using more quantifiable and objective data as from videotaping also can increase the power of a study. Obviously there are tradeoffs between various approaches that might be taken, even just in the context of increasing the sensitivity of studies. IQ tests, the most widely used measures of mental development, are most stable and reliable when global scores are reported. Using IQ tests, one strategy that frequently has been used to increase the sensitivity to neurotoxin-mediated effects has been to analyze sub-scales independently, which may give a picture of domain-specific effects. This is a less-than-perfect solution since individual tasks, such as copying block design or similes have relatively low re-test reliability on their own, while the psychomotor and verbal subscales arrived at by normalizing the scores on multiple related tasks have relatively greater stability. For both lead and PCBs, associations between exposure and performance on either full-scale or sub-scale scores have been reported, but in neither case does the literature as a whole reveal a consistent pattern of deficits that would correlate with a specific domain. For example, in preschoolers, using the McCarthy Scales, two Michigan cohort studies reported associations with PCB exposure (11,12) while a North Carolina study did not (13). In the Jacobson study, negative associations between cord blood PCB level and the verbal and memory sub-scores were observed at age 4, but effects were not significant for the General Cognitive Index (GCI). The Oswego Cohort study reported a negative association between cord blood levels of highly chlorinated PCBs and GCI as well as perceptual performance and quantitative subscales at 3 but not 4.5 years of age. With these inconsistent results, it is hard to know whether sub-scale analysis is legitimately providing evidence about performance on individual cognitive domains or, more likely, about general cognitive issues. Moreover, such analysis could in theory increase the statistical probability of a positive result by increasing the number of comparisons being made. In this situation, it is important to focus on the larger context in terms of patterns of responses on a number of subtests as opposed to picking out individual test results, out of the broader context (14). Many other factors will affect the sensitivity of a test to detect important but small differences in the performance of subsets of children. Including children from different cultural backgrounds in single study requires tests that are either culturally neutral or adaptable. The disadvantages children with first languages other than English face on many standard IQ tests is well-known, but other factors can also vary; for example, behavioral norms may be very different between ethnic groups, an important consideration in the scoring of videotaped data. Other sources of variation are: varied test conditions across geographic areas; age-appropriateness of tests; and inherent variability in performance when children are tested at certain ages. All of these are likely to lead to a loss of power to detect differences in the means between groups within a reasonable sized population. A second significant difficulty in good assessment of the impact of environmental exposures on development is the length of time needed in follow-up for these studies. Because retrospective exposure assessments are generally less reliable, prospective studies beginning perhaps in utero and continuing through school age would be ideal for assessing the range of development. One North Carolina cohort with early PCB exposure measures has been followed from birth and through adolescence (15,16). Such a study faces the problem of attrition over so many years so that there is less power to detect later effects and more potential for participation bids. Assessment tools that could be used at a young age to measure aspects of brain function that predict later development and success would address this problem by reducing the total time frame. Neurological evaluations have been used to assess children immediately after birth. However, these evaluations have utilized tools that were

Public Health Perspectives on Developmental Neurotoxicology

471

designed to identify major clinical outcomes and do not necessarily identify infants with suboptimum development due to chemical or pesticide exposures. The two most commonly used neonatal protocols are the Brazelton Neonatal Behavioral Assessment Scale (NBAS) and the Prechtl Neonatal Optimality Score (NOS). Brazelton designed the NBAS to be a “clinical, interactive, process-oriented instrument” which was designed to “reflect the neonate’s capacity to recover from. stress” over the several days after delivery (17). There are 27 items scored to assess the infant’s interactive, motor, and organizational (state and physiological control) capacities, and the examiner’s goal is to elicit the infant’s best performance, which Brazelton recognized can be influenced by the environment and the timing of the exam relative to things such as last feeding time and presence of the parents. He had many reservations about the use of this instrument as a one-time assessment of neonates, saying, “a single score, as an indication of goodness or badness is not what is obtained with the scale”, stressing instead the importance of the “recovery curve” (17). In addition, his experience indicated that rigorous centralized training was necessary to obtain reliable inter-examiner scoring, and that a stretch of cross-cultural or at-risk examinations could lead to drift in the scoring norms being used by any individual examiner. Nonetheless, this assessment has been found to be useful in studies such as the North Carolina Breast Feeding study, which found increased hypotonicity and hyporeflexia with higher PCB levels, (18) and lower psychomotor scores at 2 years of age (19). These specific differences did not persist at ages 3–5 (13). From the standpoint of policy implications it is not clear what significance early hypotonicity and hyporeflexia has in terms of later developmental disability and reduced intellectual performance. The NOS is more of a clinical neurological exam, with the most comprehensive version encompassing 250 items. As with the BNAS, inter-rater reliability is problematic, (20) with one study finding a correlation coefficient of only 0.31 (21). This scale defines clinically apparent neurological abnormalities. About 5% of a term population has been identified as abnormal, most having significant perinatal risk factors, which suggests that a very large study population would be necessary to detect environmental effects. While there is some evidence that children with abnormal neonatal scores are more likely to have later sequelae than children with normal scores, (22) the scale is much better as a negative predictor of later developmental disability than as a positive one, since many “abnormal” children at birth are normal by 6 months of age—one study finding 70% “recovery” (23). Another early measurement tool, the Fagan Test of Infant Intelligence (FTII), evaluates habituation and recognition of visual stimuli by infants, endpoints that are thought to be indicative of early intellectual development. A meta-analysis of the ability of these scores to predict IQ found that the weighted (for study size) average correlation coefficient was 0.36 between FTII and IQ scores between one and eight years later, (24) with a strong inverse relationship between study size and correlation coefficient. While this tool has been used more rarely, three cohort studies on PCBs have utilized the FTII. One reported an association between increased PCB levels and decreased novelty preference at 7 months (25) and a second found an association at 12 but not 6 months that accounted for just 2% of the variance) (26). Neither study group reports the predictive value of the FTII in later evaluations of their cohorts. A third group attempted to use the FTII but found that the interrater and re-test reliability were close to zero (27). Thus while it appears that the FTII is useful, it is not at all clear what its predictive value is for later IQ and neurological outcomes in children, within the ranges where effects are demonstrated in these studies. In conclusion, small shifts in the distribution of cognitive function within a population can have large public health burdens. Protecting the public health therefore

472

Mammen and Goldman

dictates relatively sensitive research tools. The ideal assessment tool for large populationbased studies of low-level exposures to neurotoxins would be easily administered, culturally neutral, with domain-specific testing capacities. In practice, researchers face a difficult trade off between more detailed assessments and larger study populations, as well as a host of other population- and testing condition-specific considerations in picking the instrument. A lack of a priori information about which domains are likely to be affected by a particular neurotoxin also impedes optimal study design. Public health will be best served when neurotoxicity can be anticipated by laboratory-based research before compounds are allowed to contaminate the environment. This will require significant research on the correlations between human and animal cognitive functions in order to design model systems on which compounds can be screened. Assessment of Exposure to Developmental Neurotoxic Agents Theoretically, exposure to a xenobiotic can be measured at many points along the path to causing harm. The term exposure usually refers to measures of the levels of environmental contamination. A dose is the amount of xenobiotic that enters the body through various routes, such as food consumed, air breathed or dust touched. Internal dose is used to describe the amount of compound absorbed from each route, whether in the gut or the lungs or the skin. This is usually measured as an aggregate, either in circulation in the blood or as excreted in urine or stool. Effective dose is the concentration of the xenobiotic that arrives at the target tissue. Many pharmacokinetic differences between individuals will contribute to variation in the internal and effective doses for the same original exposure, because absorption, metabolism, volume of distribution, and rates of elimination can vary quite widely between individuals. Measuring the effective dose directly integrates such individual variation into the exposure assessment. Since the effective dose is responsible for causing the injury to the target tissue, it should be the best predictor of outcome (28). Thus, because individual variability modifies and complicates other exposure measures, and effective dose is most closely related to outcome, the ideal exposure assessment would measure the effective dose. This is especially important when designing a study to demonstrate causal effects of a xenobiotic on health because population-based studies contain many variables and confounders, and so the more accurately dose can be measured, the more power in the data to observe any real effect. Unfortunately, as will be discussed below, effective dose is almost impossible to measure directly in humans, and even internal dose, which may often be an adequate biomarker of exposure, can be difficult to obtain, in particular during the in utero period so critical for brain development. Research is needed to push our technical abilities to measure internal and effective dose for many compounds. In addition, research to understand the pharmacokinetic variables that moderate effective dose is needed to improve the measurement and modeling of this variability. The second critical feature of exposure assessment is the timing of exposure relative both to the environmental contamination event(s) and to the resultant injury. The concept of critical developmental windows derives from the observation that organs are most vulnerable to teratogens while they are actively being formed but often are relatively (or completely) resistant to the same toxicants either earlier or later during development (29). An assessment of exposure that is designed to correlate a toxicant with a developmental outcome would be most predictive of the effect if the exposure during the critical window were assessed directly. At a second best, we would want to know that the exposure assessment done was well correlated with the exposure during the time of concern. As will be discussed below, how well we can do this now depends on both the pharmacokinetics of the individual compounds and the extent to which the critical window is well-defined

Public Health Perspectives on Developmental Neurotoxicology

473

biologically. For neurological development, the question of timing is complicated because brain development begins in utero and continues throughout childhood and adolescence, with different regions undergoing maximal growth at different times. Therefore, it is probable that independent neurological and cognitive functions have unique critical windows, but most of these are not defined at present. Timing the exposure assessments to the contamination events is relevant particularly for episodic exposures of compounds that are readily excreted and therefore do not result in steady-state levels. The non-persistent pesticides such as chlorpyrifos, discussed below, are an important example. In practice it is difficult if not impossible to achieve the ideal of accounting for individual variability in the exposure assessment and correctly timing the assessment to the window of maximal effect in human studies. Without specific knowledge about when the critical windows occur, the two possible approaches are (1) to measure exposures over the extended period of development or (2) to reconstruct past exposures at various points during child development. How well either approach can be achieved for an individual compound is most significantly a result of pharmacokinetic properties because these determine the availability and stability of biomarkers to measure exposure. In particular, biological half-life, which results from a combination of metabolism and excretion rates, is a critical determinant of how robustly an ongoing or episodic exposure can be determined for an extended critical period such as experienced during brain development. For a compound with a long half-life, levels in blood, urine or other tissues will likely be stable over weeks or months. Both lead and polychlorinated biphenyls are in this category. Except in the face of an acute, high-level poisoning, storage tissues like bone (lead) and fat (PCBs) act as buffers against short-term changes in blood levels. Thus a single point measurement of levels provides a relatively reliable picture of the past weeks or months. Steady-state environmental exposures to compounds with short half-lives will also result in blood or excretion levels that are fairly constant over time; however, intermittent exposures can result in much more variable levels in tissues. Therefore, the total dose from intermittent exposures to compounds that are rapidly cleared will be much more difficult to measure accurately without a significant investment of time and resources. In addition, depending on the length of the “critical” developmental window, one might either miss the relevant measurement if it is short, or fail to fully characterize the exposure if it is long. Longer-term changes in both body burden and blood levels of persistent compounds occur as a result of the net effects of continued exposures, elimination and growth (increased volume of distribution). Rarely, rapid changes occur when conditions cause acute turnover of the storage compartment, for example with increased mobilization of bone calcium and lead into the blood stream during pregnancy. In addition, the solubility in the blood will vary with protein, cell and fat content, relevant variables that are easily measured and corrected for in practice. In one-recent study, blood draws separated by two years in children aged 7 and then 9 had correlations between rZ0.61 for b-HCH and rZ0.86 for DDE (30). Following a defined one-year exposure to polybrominated biphenyls through contaminated meat in Michigan in 1973, consecutive serum levels drawn in 1977, 1978 and 1979 for a cohort of 3683 individuals of diverse ages found very small decreases in levels year to year (31). These data again reinforce the validity of blood levels as relatively stable biomarkers for xenobiotics with long biological half-lives. The storage tissues in which such compounds accumulate are also theoretically available to be sampled as markers of cumulative exposure. Research on PCBs has demonstrated excellent correspondences between fat and blood levels, such that fat sampling is rarely done any more due to its increased invasiveness (32). For heavy metals, radiographic measurements of bone lead, or the accumulated levels in shed tissues such as hair and baby teeth are occasionally used for cumulative exposure assessment. These are

474

Mammen and Goldman

not of clinical value but for the purpose of epidemiological study have been found to provide a useful metric for chronic, cumulative exposure to lead. In contrast, it is significantly more difficult to determine long-term exposures to more labile compounds. Compounds with short biological half-lives are often rapidly metabolized and excreted in the urine, which has been the traditional source for measuring internal dose to such xenobiotics as organophosphate pesticides (33). There is a range of technical drawbacks with urinary biomarkers. For example, some of the urinary excretion biomarkers for organophosphates have been metabolites that are common to the class as a whole, since hydrolysis releases an organic moiety and combinations of six dialkyl phosphate (DAP) metabolites depending on the methyl- and thio-substitution of the parent compound. However, DAP analysis is not an accurate reflection of cumulative organophosphate toxicity, since the proportional exposure to parent compounds with different potencies contributing to the pool of metabolites cannot be analyzed. In addition, the metabolite measured is not always proportionally generated with the actual toxicant (either parent or other metabolite) because metabolism involves multiple enzymes that can vary in function between individuals. This is the case for chlorpyrifos, where the quantity of toxic oxon metabolite is determined by a polymorphism in paraoxonase-1 (34,35) that is not reflected in the excretion of the measured metabolite 3,5,6-trichloro-2-pyridinol (TCPy). Finally, spot urine collection needs to be corrected for urine concentration, which is traditionally done using creatinine levels (36). Fortunately, there are more specific assays that are now available for some of the organophosphate pesticides. Recent development of sensitive high-resolution mass spectrometry detection methods has allowed for serum and plasma levels of some organophosphate pesticides to be determined (37). While such a method is a better assessment of internal dose than the urinary makers, the short half-life of organophosphates still contributes to a lack of stability in the measurement over time. The Minnesota Children’s Pesticide Exposure Study found poor correlation between repeated urinary measurements of organophosphate pesticides (38). In fact, the variability between measurements was greatest for those children with the highest average exposure levels— that is, those in whom a single high exposure was captured also had very low levels on other occasions. The short half-life of these compounds also meant that even known acute exposures were not well correlated with biomarkers. For 16 children who reported being present for pesticide applications during the week of monitoring, average urinary metabolite levels were not different from those who did not report acute exposure (38). The lack of stability over time means that accurately describing longitudinal exposure to these compounds over the years involved in development is a daunting practical problem. One approach has been to back off of biomarkers to questionnaires and environmental sampling in the hopes that these might provide more complete assessments of ongoing exposures. However, by stepping back in the exposure pathway, more variability is introduced that could misclassify individuals as to effective dose. In addition, one must know the major exposure pathways in order to be sure of assessing the largest components of dose. The MNCPES was not able to correlate any external exposure measurement (e.g., dust, duplicate diets, hand-wash samples) or self-reported historical use with urinary metabolite measures, (39) which suggests that we do not yet have a good handle on the exposure pathways that most impact internal dose. Measurement stability is less of a problem when studying more highly and consistently exposed sub-populations, for example, those with agricultural exposures or living in distressed urban environments, where large scale applications lead to sustained airborne exposures. For example, in New York City it was possible to observe a seasonal spike in the urinary pyrethroid metabolite PBA, but not other OP pesticides, during the

Public Health Perspectives on Developmental Neurotoxicology

475

summer of 2000, when pyrethroids were used widely by the city to control West Nile Virus, which was not present the previous summer when city-wide spraying was not performed (40). Similarly, in a largely agricultural community in Washington State, distance of residence from organophosphate-treated apple orchards predicted house dust concentrations and was marginally significantly related to children’s urine DMTP metabolite levels, the DAP detected with the greatest frequency (41). Unique to studies on development, the effects of in utero exposure can be profound. Alcohol consumption and fetal alcohol syndrome is a classic example of neurotoxicity following an exclusively in utero exposure; mothers do not develop the same neurologic effects as those that are caused in the developing child. Animal studies with known neurotoxins suggest that many compounds hamper development beginning in utero. The PCB literature of human exposure suggests a larger effect for in utero exposures than postnatal ones (13,19,25,42–45). In utero exposure assessment provides a distinct challenge in the development of biomarkers. Amniotic fluid contains all soluble metabolites excreted in the fetal urine after week 20 and other compounds filtered through the placenta. It is available from some pregnancies but general collection in the context a population-based study is not possible, and thus it is an impractical source in such studies, although it may prove able to validate other biomarkers of exposure. Similarly while it is possible to sample umbilical cord blood in utero, the risks to the fetus are unacceptable for research purposes. Maternal serum levels constitute the external environment of the fetus and as such may be a source of exposure information, especially for those compounds that may be in equilibrium between mother and fetus. Lipophilic compounds are readily soluble in fat, and thus biological membranes are freely permeable to such chemicals. If inter-individual variation in the partitioning of such compounds between tissues is small relative to the differences in blood concentrations between individuals, blood levels in the mother would be a good surrogate for effective dose in the fetus, since placental and brain partitioning would be largely driven by the maternal serum level. Studies that have looked at levels of PCBs in both maternal and cord blood with the most sensitive modern techniques find that levels are highly correlated (46). Therefore, maternal serum is probably a reasonable surrogate measure for the in utero brain exposures to such lipophilic compounds. More study is needed to investigate the extent to which fetal exposure to other potential neurotoxicants is correlated with circulating maternal levels. Cord blood, placental levels, and meconium are source of retrospective analysis available at birth. Cord blood is quite lean and so detection limits have made the use of this source difficult historically for lipophilic compounds; for example, PCB detection rates have averaged 30% in the past (47,48). Newer techniques have allowed for almost 100% detection of PCB in cord blood at the present time, making this a reasonable means of sampling this fetal biomarker (27,49). For compounds with short biological half-lives, the problems of extrapolating a nine-months exposure from a single measurement persist with the use of cord blood, which is only reasonably sampled in a population-study at the time of birth. One study on the distribution of methyl mercury in hamsters reported accumulation of MeHg in the fetus compared to the mother, but relatively similar levels between fetal brain and placenta, suggesting that placenta could act as a surrogate target tissue for in utero brain exposure (50). Meconium provides a cumulative record of compounds excreted in the stool starting in the second trimester. Pilot studies have demonstrated the feasibility of detecting many compounds in meconium, both persistent and non-persistent, (51,52) and this is a rich source of in utero exposure data that needs further research and development.

476

Mammen and Goldman

In summary, accurate exposure assessment is fundamental to risk assessment which in turn informs policy makers about potential environmental hazards. Persistent compounds with long half-lives in the body can be measured using biomarkers that provide relatively robust estimates of long-term exposures, including estimates of in utero exposures in maternal serum. The stability of these measurements allows for epidemiological investigation of the effects of these compounds on neurodevelopment. In contrast, investigating the neurotoxicity of those compounds which are rapid metabolized and excreted is much more challenging because it is difficult to reliably assess exposure history, particularly to environmental contaminants. Some subpopulations such as inner-city residents may experience sufficiently consistent exposure levels to toxicants such as organophosphate pesticides that the stability of their biomarkers of exposure will enable such epidemiology to be accomplished. However even in these populations exposure assessments are likely to miss peak exposures during critical windows of development, resulting in decreased statistical power of studies (due to random misclassification of exposed subjects as unexposed). Meconium is a new potential source of cumulative in utero exposure data to these compounds and deserves more research attention. Finally, a major challenge to policy makers in assessing the risks is isolating the effects from environmental toxins from each other and all of the other factors that influence brain growth and development, such as the social and educational environment, since disadvantages in situation are often linked with increased environmental pollution.

TRACKING DEVELOPMENTAL DISABILITIES In 1999 one of us (LG) authored a report for the Pew Environmental Health Commission that called for the establishment of a nationwide health tracking system that would track the rates of developmental disabilities and birth defects in children, environmental exposures, and support research to determine the causes of these diseases (53). Since that time, a fair amount of progress has been made. The U.S. CDC has established a number of Centers of Excellence for tracking rates of autism and other developmental disabilities. These are new efforts but it can be expected that over the next few years we will begin to see better data regarding rates and trends of these diseases. Also, the State of California funded a comprehensive effort to assess rates and trends of autism and is funding research on autism. A nongovernmental organization, CAN, is also funding research. On the exposure side, the CDC also is funding state environmental public health surveillance projects and Centers of Excellence that are moving toward the development of a National Environmental Health Tracking system. Progress also has been made in the area of funding of fundamental research and particularly the funding of several NIEHS/EPA Children’s Centers that are specifically assessing developmental disabilities. At this time, the U.S. has reliable statistics for birth defects in many but not all states, but little or no information about rates of developmental disabilities in most jurisdictions.

POLICY ISSUES Principles There are several principles of environmental policy making that are particularly operative when it comes to issues of developmental neurotoxicity of chemicals. First is the precautionary principle. Given the existence of uncertainty, how should policy makers

Public Health Perspectives on Developmental Neurotoxicology

477

proceed? According to the precautionary principle, as articulated by United Nations Conference on Environment and Development (UNCED) in 1992: “In order to protect the environment, the precautionary approach shall be widely applied by States according to their capabilities. Where there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation” (54). Although this principle is not embodied in environmental law in the US the US has asserted that it follows a “precautionary approach”. The UNCED also adopted a principle related to intergenerational equity: “The right to development must be fulfilled so as to equitably meet developmental and environmental needs of present and future generations” (54). This principle is particularly relevant to the issue of developmental neurotoxicity since it is the wellbeing of future generations that would receive negatively impact if we inadvertently causes increases in developmental disabilities and chronic neurologic diseases. A purely economic analysis may miss the significance of such impacts because discounting can make the costs of control today appear to be too great compared to benefits that are decades into the future (55). Another principle involved in policy making in the U.S., is monetization of benefits and costs of environmental policies. Certainly the costs of developmental neurotoxicants in the environment are likely to be quite large. According to Landrigan et al. the total cost of environmental related diseases in U.S. children is in the range of $48.8 to $64.8 billion per year. Most of this is attributable to neurotoxicity: $43.7 billion for pediatric lead exposure and between $4.6 and $18.4 billion for major developmental disabilities not related to lead (depending on the attributable risk fraction from environmental exposures) (56). Some laws, such as the TSCA, as well as successive presidential Executive Orders since the time of President Carter, have required that regulatory agencies do formal economic analyses for certain actions. In the case of the Executive Orders, such analyses have been triggered by major rules (as defined by total impact of at least $100 million or the judgment of the US Office of Management and Budget). Economic analyses can play a number of roles including attempting to weight costs and benefits of an action (so-called cost-benefit analysis) weighing the relative cost effectiveness of alternative solutions to a problem and identification of economic inequities in impact that can inform decision making. Monetization creates challenges for regulation of developmental neurotoxic agents for a number of reasons including the difficulty of projecting impacts across a diverse population of humans from animal toxicity tests or more limited epidemiology studies and the difficulty in quantifying benefits of regulation such as prevention of IQ points lost. Another issue is, as noted above, that the standard practice of discounting that is used in economic analysis can result in devaluing the worth of future lives and the lives of our children (55). Laws and Regulations The basic structure of the domestic laws of the United States with respect to chemicals was established in 1972 for pesticides (Federal Insecticide, Fungicide and Rodenticide Act or FIFRA) and 1976 for industrial chemicals TSCA. In the face of widespread concern about the proliferation of chemicals and pesticides in commerce, and the unknown risks, Congress had given the U.S. EPA authority over testing of chemicals and pesticides, review of new introductions, and assessment and management of risks of existing chemicals. In 1988, Congress had amended and strengthened FIFRA. In 1986, it enacted the Emergency Preparedness and Community Right to Know Act (EPCRA), thereby establishing the Toxic Release Inventory (TRI) for tracking the releases and transfers of

478

Mammen and Goldman

chemicals from industry. In 1990, it adopted the Pollution Prevention Act (PPA). In 1996, Congress enacted the FQPA which amended FIFRA and the Federal Food, Drug and Cosmetics Act (FFDCA) to strengthen public health protections for pesticides (as described above). Certain other substances (pharmaceuticals, food additives, medical devices, and cosmetics) were exempted from these statutes by Congress because they are regulated by the Food and Drug Administration under FFDCA. Further, the Consumer Products Safety Commission is responsible for regulation of chemicals as they occur in consumer products, under the Consumer Products Safety Act, and the Occupational Safety and Health Administration for regulation of workplace hazards under the Occupational Safety and Health Act. Together, these statutes form the legal framework for regulation of chemicals and pesticides in the US. There are also laws that establish non-regulatory responsibility for chemicals. For example, the National Institute of Occupational Safety and Health (NIOSH), which is a component of the CDC, has been given specific responsibilities for assessment and study of hazards related to OSHA’s mission. The Agency for Toxic Substance and Disease Registries (ATSDR) of the CDC has statutory responsibilities related to Superfund and other health assessments relevant to the work of the EPA. The National Institute of Environmental Health Sciences (NIEHS) has several statutory roles in the process. It houses the National Toxicology Program (which coordinates chemical testing within the government as well as the Report on Carcinogens (57). Recently the National Toxicology Program has begun to assess chemicals for developmental and reproductive hazards, for example, its recent assessments of phthalates (58) and methanol (59). International Efforts The “dirty dozen” POPs chemicals are twelve chemicals and pesticides that have been selected for control under the international Stockholm Convention on Persistent Organic Pollutants. Countries got together to create a convention because of the way that such pollutants move around the globe and because of the difficulty in controlling exposure; once they have entered the environment, you can’t just turn off the tap. The chemicals included are: the pesticides aldrin, chlordane, DDT, dieldrin, endrin, heptachlor, mirex, toxaphene, and hexachlorobenzene; the industrial chemicals PCBs; and the inadvertent contaminants dioxins and furans. Most of these are scheduled for a complete phaseout with a couple of notable exceptions. DDT has a public health exemption to allow for its use in malaria control. Dioxins and furans are not phased out but rather “phased down” in recognition of the fact that they are produced inadvertently and therefore are impossible to ban completely. The U.S. government has signed, but not ratified, this agreement, which came into force in 2004. Lead is one of the oldest and most familiar of persistent and toxic substances. Regulation of lead in gasoline, water and food sources, paint, and consumer products has resulted in dramatic decrease in lead levels in the U.S. over time. However, there are still high rates of lead exposure in many parts of the world, much of it from use of leaded gasoline. In 2002, UN Environmental Program governments agreed to phase out leaded gasoline globally. This action will result in dramatic reductions of blood lead levels in children worldwide. Even in the U.S., there are still pockets of lead exposure, mostly due to lead-contaminated housing. Additionally we still need to focus on other sources, for example, the recent news of high lead levels in drinking water in the District of Columbia. At last count, the CDC estimated that some 300,000 children under six have blood lead levels greater than 10 micrograms/deciliter (60) and evidence supports a lower target of 5 (61).

Public Health Perspectives on Developmental Neurotoxicology

479

Mercury is another important persistent substance. It is methylmercury that is of concern as a developmental neurotoxicant. For mercury, some progress has been made in the last ten years. The EPA has tightened the regulations on air emissions from municipal and medical waste incinerators (which used to be the most important sources), and industry has voluntarily reduced emissions from choralkalai production plants. In certain areas, progress has been delayed. The regulation of older power plants that were grandfathered by the 1972 Clean Air Act is gridlocked, with the EPA’s legislative efforts having been rejected by the courts and the EPA’s “Clear Skies” proposal not having moved in Congress in 2003. In 2002, the United Nations Environment Program completed a global assessment on mercury, which concluded that there is sufficient evidence of significant global adverse impacts from mercury to warrant further international action to reduce the risks to humans and wildlife from the release of mercury to the environment. In 2003, UNEP began a mercury program, which provides technical assistance to support national efforts to reduce human activities that cause mercury releases to the environment. Currently the European Union is undertaking a fundamental re-examination of its legislation for control of chemicals and considering the adoption of a new REACH (Registration, Evaluation and Authorization of Chemicals) approach that would result in a much larger base of information being made available about the toxicity of chemicals in their society. Such an approach would dramatically increase the ability to do hazard and risk assessments, if enacted. At the time of this writing it is not clear the extent to which it will increase the fund of knowledge about, and control over exposures to, developmental neurotoxic agents in the environment.

CONCLUSIONS Adverse neurological outcomes are very prevalent among our children, and there are enormous gaps in our knowledge about what causes such outcomes. Yet we are making progress in a number of areas, involving tracking efforts, research, testing, and risk management. What we do know to date about known factors has largely been derived from epidemiology studies of neurological outcomes among children exposed in utero and in the first few years of life. Clearly, such investigations provide critical information regarding developmental neurotoxicity. What is needed are larger longitudinal data bases to inform us of the associations between early childhood exposures—to persistent chemicals but also to pathogens, to nutrients, to pharmacologic agents, to injury, and to stress—and subsequent neurodevelopmental outcomes and disabilities. In addition, we need new toxicology models that will utilize in vitro and in vivo models to identify chemicals that are most likely to have modes of action that may be associated with developmental neurotoxicity. Finally, such information would be most useful within a framework that would prudently protect the fetus and developing child from harm, so as to safeguard the health and welfare of future generations.

REFERENCES 1. Brent RL. Environmental causes of human congenital malformations: the pediatrician’s role in dealing with these complex clinical problems caused by a multiplicity of environmental and genetic factors. Pediatrics 2004; 113:957–968.

480

Mammen and Goldman

2. National Research Council. Committee on Developmental Toxicology. Scientific Frontiers in Developmental Toxicology and Risk Assessment. Washington, DC: National Academies Press, 2000. 3. National Research Council. Risk Assessment in the Federal Government: Managing the Process. Washington DC: National Academy Press, 1983. 4. Goldman LR. EPA seeks public health views on new pesticide law. Public Health Rep 1996; 111:512–514. 5. Rodier PM. Environmental causes of central nervous system maldevelopment. Pediatrics 2004; 113:1076–1083. 6. Kimmel CA, Makris SL. Recent developments in regulatory requirements for developmental toxicology. Toxicol Lett 2001; 120:73–82. 7. Crofton KM, Makris SL, Sette WF, Mendez E, Raffaele KC. A qualitative retrospective analysis of positive control data in developmental neurotoxicity studies. Neurotoxicol Teratol 2004; 26:345–352. 8. Slotkin TA. Guidelines for developmental neurotoxicity and their impact on organophosphate pesticides: a personal view from an academic perspective. Neurotoxicol 2004; 25:631–640. 9. Bellinger DC. What is an adverse effect? A possible resolution of clinical and epidemiological perspectives on neurobehavioral toxicity Environ Res 2004; 95:394–405. 10. ATSDR. Pediatric Environmental Neurobehavioral Test Battery. Atlanta, Georgia: US Department of Health and Human Services, 1996. 11. Stewart PW, Reihman J, Lonky EI, Darvill TJ, Pagano J. Cognitive development in preschool children prenatally exposed to PCBs and MeHg. Neurotoxicol Teratol 2003; 25:11–22. 12. Jacobson JL, Jacobson SW, Humphrey HE. Effects of in utero exposure to polychlorinated biphenyls and related contaminants on cognitive functioning in young children. J Pediatr 1990; 116:38–45. 13. Gladen BC, Rogan WJ. Effects of perinatal polychlorinated biphenyls and dichlorodiphenyl dichloroethene on later development. J Pediatr 1991; 119:58–63. 14. Rothman KJ. No adjustments are needed for multiple comparisons. Epidemiology 1990; 1:43–46. 15. Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated biphenyls (PCBs) and dichlorodiphenyl dichloroethene (DDE) in human milk: effects of maternal factors and previous lactation. Am J Public Health 1986; 76:172–177. 16. Gladen BC, Ragan NB, Rogan WJ. Pubertal growth and development and prenatal and lactational exposure to polychlorinated biphenyls and dichlorodiphenyl dichloroethene. J Pediatr 2000; 136:490–496. 17. Brazelton TB. The Brazelton neonatal behavior assessment scale: introduction. Monogr Soc Res Child Dev 1978; 43:1–13. 18. Rogan WJ, Gladen BC, McKinney JD, et al. Neonatal effects of transplacental exposure to PCBs and DDE. J Pediatr 1986; 109:335–341. 19. Gladen BC, Rogan WJ, Hardy P, Thullen J, Tingelstad J, Tully M. Development after exposure to polychlorinated biphenyls and dichlorodiphenyl dichloroethene transplacentally and through human milk. J Pediatr 1988; 113:991–995. 20. Kakebeeke TH, Jongmans MJ, Dubowitz LM, Schoemaker MM, Henderson SE. Some aspects of the reliability of Touwen’s examination of the child with minor neurological dysfunction. Dev Med Child Neurol 1993; 35:1097–1105. 21. Moninex WM, Smolders-de Hass H, Bonsel GJ, Zondervan HA. Interobserver variability of the neurological optimality score. Eur J Obstet Gynecol 1999; 85:167–171. 22. Touwen BCL, Lok-Meijer TY, Huisjes HJ, Olinga AA. The recovery rate of neurologically deviant newborns. Early Hum Dev 1982; 7:131–148. 23. Stave U, Ruvalo C. Neurological development in very-low-birthweight infants. Application of a standardized examination and Prechtl’s optimality concept in routine evaluations. Early Hum Dev 1980; 4:229–241. 24. McCall RB, Carriger MS. A meta-analysis of infant habituation and recognition memory performance as predictors of later IQ. Child Dev 1993; 64:57–79.

Public Health Perspectives on Developmental Neurotoxicology

481

25. Jacobson JL, Jacobson SW. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. N Engl J Med 1996; 335:783–789 see comments. 26. Darvill T, Lonky E, Reihman J, Stewart P, Pagano J. Prenatal exposure to PCBs and infant performance on the fagan test of infant intelligence. Neurotoxicology 2000; 21:1029–1038. 27. Winneke G, Bucholski A, Heinzow B, et al. Developmental neurotoxicity of polychlorinated biphenyls (PCBS): cognitive and psychomotor functions in 7-months old children. Toxicol Lett 1998; 102–103:423–428. 28. Sexton K, Callahan MA, Bryan EF. Estimating exposure and dose to characterize health risks: the role of human tissue monitoring in exposure assessment. Environ Health Perspect 1995; 103:13–29. 29. Rice D, Barone S, Jr. Critical periods of vulnerability for the developing nervous system: evidence from humans and animal models. Environ Health Perspect 2000; 108:511–533. 30. Karmaus W, DeKoning EP, Kruse H, Witten J, Osius N. Early childhood determinants of organochlorine concentrations in school- aged children. Pediatr Res 2001; 50:331–336. 31. Kreiss K, Roberts C, Humphrey HE. Serial PBB levels, PCB levels, and clinical chemistries in Michigan’s PBB cohort. Arch Environ Health 1982; 37:141–147. 32. Patterson DGJ, Todd GD, Turner WE, Maggio V, Alexander LR, Needham LL. Levels of nonortho-substituted (coplanar), mono- and di-ortho-substituted polychlorinated biphenyls, dibenzo-p-dioxins, and dibenzofurans in human serum and adipose tissue. Environ Health Perspect 1994; 102:195–204. 33. Barr DB, Barr JR, Maggio VL, et al. A multi-analyte method for the quantification of contemporary pesticides in human serum and plasma using high-resolution mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2002; 778:99–111. 34. Costa LG, Li WF, Richter RJ, Shih DM, Lusis A, Furlong CE. The role of paraoxonase (PON1) in the detoxication of organophosphates and its human polymorphism. Chem Biol Interact 1999; 119–120:429–438. 35. Furlong CE, Cole TB, Jarvik GP, Costa LG. Pharmacogenomic considerations of the paraoxonase polymorphisms. Pharmacogenomics 2002; 3:341–348. 36. IRCP. Report of the Task Group on Reference Man. Amsterdam, NY: Elsevier Science Inc, 1994. 37. Barr DB, Needham LL. Analytical methods for biological monitoring of exposure to pesticides: a review. J Chromatogr B Analyt Technol Biomed Life Sci 2002; 778:5–29. 38. Adgate JL, Barr DB, Clayton CA, et al. Measurement of children’s exposure to pesticides: analysis of urinary metabolite levels in a probability-based sample. Environ Health Perspect 2001; 109:583–590. 39. Sexton K, Adgate JL, Eberly LE, et al. Predicting children’s short-term exposure to pesticides: results of a questionnaire screening approach. Environ Health Perspect 2003; 111:123–128. 40. Berkowitz GS, Obel J, Deych E, et al. Exposure to indoor pesticides during pregnancy in a multiethnic, urban cohort. Environ Health Perspect 2003; 111:79–84. 41. Lu C, Fenske RA, Simcox NJ, Kalman D. Pesticide exposure of children in an agricultural community: evidence of household proximity to farmland and take home exposure pathways. Environ Res 2000; 84:290–302. 42. Jacobson SW, Fein GG, Jacobson JL, Schwartz PM, Dowler JK. The effect of intrauterine PCB exposure on visual recognition memory. Child Dev 1985; 56:853–860. 43. Jacobson JL, Humphrey HE, Jacobson SW, Schantz SL, Mullin MD, Welch R. Determinants of polychlorinated biphenyls (PCBs), polybrominated biphenyls (PBBs), and dichlorodiphenyl trichloroethane (DDT) levels in the sera of young children. Am J Public Health 1989; 79:1401–1404. 44. Lanting CI, Patandin S, Fidler V, et al. Neurological condition in 42-months-old children in relation to pre- and postnatal exposure to polychlorinated biphenyls and dioxins. Early Hum Dev 1998; 50:283–292. 45. Patandin S, Lanting CI, Mulder PG, Boersma ER, Sauer PJ, Weisglas-Kuperus N. Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in Dutch children at 42 months of age. J Pediatr 1999; 134:33–41 see comments.

482

Mammen and Goldman

46. Steuerwald U, Weihe P, Jorgensen PJ, et al. Maternal seafood diet, methylmercury exposure, and neonatal neurologic function. J Pediatr 2000; 136:599–605. 47. Fein GG, Jacobson JL, Jacobson SW, Schwartz PM, Dowler JK. Prenatal exposure to polychlorinated biphenyls: effects on birth size and gestational age. J Pediatr 1984; 105:315–320. 48. Stewart P, Reihman J, Lonky E, Darvill T, Pagano J. Prenatal PCB exposure and neonatal behavioral assessment scale (NBAS) performance. Neurotoxicol Teratol 2001; 22:21–29. 49. Grandjean P, Weihe P, Burse VW, et al. Neurobehavioral deficits associated with PCB in 7-year-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001; 23:305–313. 50. Dock L, Rissanen RL, Vahter M. Demethylation and placental transfer of methyl mercury in the pregnant hamster. Toxicology 1994; 94:131–142. 51. Whyatt RM, Barr DB. Measurement of organophosphate metabolites in postpartum meconium as a potential biomarker of prenatal exposure: a validation study. Environ Health Perspect 2001; 109:417–420. 52. Hong Z, Gunter M, Randow FF. Meconium: a matrix reflecting potential fetal exposure to organochlorine pesticides and its metabolites. Ecotoxicol Environ Safety 2002; 51:60–64. 53. Goldman LR, Apelberg B, Koduru S, Sorian R. Healthy From the Start: Why America Needs a Better System to Track and Understand Birth Defects and the Environment. Baltimore, MD: Pew Environmental Health Commission, 1999. 54. United Nations Conference on Environment and Development. Rio Declaration on Environment and Development. Rio de Janeiro, Brazil: United Nations, 1992. 55. Heinzerling L. Discounting life. Yale Law J 1912 (1998–1999); 108:1911–1916. 56. Landrigan PJ, Schechter CB, Lipton JM, Fahs MC, Schwartz J. Environmental pollutants and disease in American children: estimates of morbidity, mortality, and costs for lead poisoning, asthma, cancer, and developmental disabilities. Environ Health Perspect 2002; 110:721–728. 57. National Toxicology Program. Report on Carcinogens, Tenth Edition: U.S. Department of Health and Human Services, Public Health Service, National Toxicology Program, 2002. 58. Kavlock R, Boekelheide K, Chapin R, et al. NTP center for the evaluation of risks to human reproduction: phthalates expert panel report on the reproductive and developmental toxicity of di-n-hexyl phthalate. Reprod Toxicol 2002; 16:709–719. 59. Shelby M, Portier C, Goldman L, et al. NTP-CERHR expert panel report on the reproductive and developmental toxicity of methanol. Reprod Toxicol 2004; 18:303–390. 60. U.S. Center for Disease Control and Prevention. Blood lead levels in young children—United States and selected states, 1996–1999. Morb Mortal Wkly Rep 2000; 49:1133–1137. 61. Lanphear BP, Dietrich K, Auinger P, Cox C. Cognitive deficits associated with blood lead concentrations !10 microg/dL in US children and adolescents. Public Health Rep 2000; 115:521–529.

23 Research to Clinical Practice: A Pediatric Environmental Health Perspective April A. Harper Division of General Pediatrics, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A.

Michael W. Shannon Division of Emergency Medicine, Children’s Hospital Boston, Harvard Medical School, Boston, Massachusetts, U.S.A.

INTRODUCTION Children are particularly vulnerable to hazardous substances within their environment. This heightened susceptibility results from the fact that children are in a progressive state of growth and development, in utero through adolescence. The obvious implication of this is that children can receive substantially larger exposures pound for pound than adults to toxins present in their water, food, or air. Unfortunately, hazardous substances such as lead, solvents, pesticides, and polychlorinated biphenyls (PCBs) have now found their way into the homes, schools, and playgrounds of all children. Within the discipline of Pediatrics the availability of specialized expertise to evaluate the relationships between child health and well-being and the impact of the physical, chemical, and biological contaminants in the environment has been limited. This deficit promoted the recent development of a new subspecialty, Pediatric Environmental Health (PEH), whose focus is to improve and protect children’s health through prevention, education, diagnosis, and treatment of environmentally-related diseases in children. PEH is both a clinical and public health area. Health practitioners, governmental and. advisory agencies, and parents are recognizing its importance. From this increased awareness, the Pediatric Environmental Health Specialty Units (PEHSUs) were established. In 1998, in partnership with the Agency of Toxic Substances and Disease Registry (ATSDR), the Association of Occupational and Environmental Clinics created the PEHSUs to encourage environmental medical specialists to work collaboratively with pediatricians to further develop proficiency in the recognition, diagnosis, and treatment of children affected by environmental hazards. The program has now grown to include a national network of 11 multidisciplinary operating units and clinics (Table 1) (1,2). Children today inhabit a world that is fundamentally different from that of past generations, facing hazards in the environment that did not exist or whose clinical effects 483

484 Table 1

Harper and Shannon Pediatric Environmental Health Subspecialty Unit Locations

United States Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7 Region 8 Region 9 Region 10 Locations Outside the United States

Boston, MA New York, NY Washington, D.C. Atlanta, GA Chicago, IL Tyler, TX Iowa City, IA Denver, CO Irvine and San Francisco, CA Seattle, WA Alberta, Canada Morelos, Mexico

Source: Adapted from Ref. 46.

were neither known nor suspected only a few decades ago. For PEH professionals to accurately determine if a child is at risk as a result of an exposure, future research that recognizes patterns of environmental disease in children; assesses the impact of children’s unique vulnerability to environmental toxicants; identifies critical developmental periods of exposure; and quantifies dose-response relationships in children are needed. This information will contribute to a better understanding of the relationship between environmental exposures and health outcomes in children and provide guidance to clinical management and interventions (44,45).

THE ROLE OF SCIENTIFIC RESEARCH IN CLINICAL PRACTICE Both clinicians and the general public have become increasingly aware that neurodevelopmental disorders may result from exposures to a variety of chemical hazards that are now ubiquitous within the environment. They include the metals lead, mercury, and manganese; pesticides such as organophosphates and others that are widely used in homes and schools; dioxin and PCBs in the food chain; and solvents, including ethanol and others used in paints, glues, and cleaning solutions (30,54). These agents may be directly toxic to cells or interfere with hormones as endocrine disruptors, neurotransmitters, or other growth factors (3,19,21,25,62,66,72,74,75). At least 75,000 new synthetic chemicals have been developed and dispersed into the environment; fewer than half of these chemicals have ever been tested for their potential toxicity to humans, and fewer still have been assessed for their toxicity to children (3,67,82). During the last few decades, much has been learned about the neurobehavioral consequences of chemical exposure to adults in the workplace. By comparison, much less is known about the neurotoxicity of a chemical exposure, other than lead and mercury, in children. What we have learned is that neurotoxicity can occur any time during the life cycle, from conception to senescence, and its manifestations can change with age. Different responses to the same neurotoxicant can occur, depending upon the dose and timing of exposure. In addition, different agents can produce similar patterns of effects (6). Expression of neurotoxicity can encompass multiple levels of organization and complexity, including structural, biochemical, and physiological and behavioral, and

A Pediatric Environmental Health Perspective

485

there is a risk that the resulting neurobehavioral dysfunction can be permanent and irreversible (6,7,12). Even with the improvements made in neurotoxicity research and risk assessment, there is still a worldwide concern about the potential neurotoxic effects of environmental contaminants in children (36). Currently, there are relatively large gaps in our knowledge regarding the effects of chemical exposures on the development of the nervous system and the resulting psychological and behavioral correlates. Consequently, our tools to measure neurobehavioral dysfunction in children are often crude, insensitive, and poorly validated.

SPECIFIC RESEARCH NEEDS Practitioners of PEH need to improve their overall understanding of the relationship between environmental exposures and health outcomes in children. Future research that identifies patterns of environmental diseases in children, determines critical developmental periods, and addresses the unique vulnerability of children will strengthen the ability of a clinician to predict a potential of neurotoxic injury and prognosis following an exposure. Specific research advances regarding risk assessment will require a more childoriented approach that quantifies dose-response relationships in children, develops and validates the use of new biomarkers, and explores the influence of gene-environment interactions on toxicity. Finally, the establishment of a more comprehensive and sensitive battery of neurobehavioral tests and measures is required. Such advances will contribute to the improvement of clinical oriented preventions and interventions within the developing field of PEH. Patterns of Environmental Diseases in Children Disease patterns in children have changed drastically in the industrially developed nations of the world over the last century. Despite a lower infant mortally rate and increased life expectancy, children today are confronted with serious diseases that are often chronic and debilitating in nature (82). Examples include asthma, the incidence of which has almost doubled in the last 10–15 years; childhood cancer, for which the incidence of certain types has substantially increased; and the increases in the prevalences of developmental, learning, and behavioral disabilities (57–59,82). The causes of the common chronic disease and developmental disabilities among children in the United States today are not clearly understood. The last few decades have seen increasing recognition that children’s exposures to environmental toxicants play a role in the development of certain chronic disease. The U.S. National Academy of Sciences (NAS) estimated in 2000 that 3% of neurobehavioral disorders in American children are caused directly by an environmental exposure and that another 25% are caused by the interaction between a child’s genetic susceptibility and environmental factors (60,82). Lead is now recognized as a significant cause of neurobehavioral dysfunction, producing measurable loss of IQ at very low blood levels, as well as alterations in behavior (21,24–25,31,32). In utero exposures to chemicals such as methyl mercury, resulting from regular maternal fish consumption, have been implicated in language, attention, and memory impairments, and children exposed to PCBs during fetal life show IQ deficits, hyperactivity, and attention deficits when tested years later (18,33,35,52,61–66,83). Many diseases that are triggered by environmental toxicants require decades to develop. Examples include mesothelioma caused by exposure to asbestos and leukemia

486

Harper and Shannon

caused by benzene (39,47). Because children have more future years of life than do most adults, they have more time to develop chronic diseases that may be triggered by early environmental exposures (82). Although the enhanced susceptibility of infants and children to environmental toxicants has been shown in multiple studies, the extent to which environmentally-induced pediatric illness contributes to adult disease has not been well-characterized. Environmental epidemiology research plays a critical role in identifying associations between disease and morbidity, distinguishing populations at high risk, and discovering casual factors and routes of exposure to environmental contaminants (11,14,17). Additionally, epidemiologic research can contribute to the integration of population data, providing clinical and laboratory insights into the understanding of mechanisms of disease causation (14,16). Future epidemiologic studies designed to establish the prevalence and incidence of neurologic dysfunction in children resulting from an exposure would be extremely beneficial in describing correlation of reported neurotoxic effects with exposure data. There is a tremendous need for well designed and executed studies that address the role of environmental exposures in childhood chronic illness (14). Research that characterizes more precisely the link between an exposure and resultant disease should be performed to accurately define the pathway from exposure to disease. This will enable the clinician to effectively implement prevention and intervention measures. Prospective longitudinal studies consisting of children with known or even suspected exposures, particularly in utero exposures, are also needed to elucidate environmental exposures in early life and their relationship to adult disease (14,53,67). Identifying Critical Periods of Exposure Certain periods during development of the nervous system are more sensitive to environmental insults because of their dependence on the sequential progression of developmental processes (i.e., synaptogenesis and myelination). These developmental processes can extend from the embryonic period through adolescence (7–9,12). Exposure to an environmental toxicant, particularly in the fetal period, can disrupt these processes and result in developmental neurotoxicity and injury. The actual mechanisms by which these processes are perturbed by an agent, and to what extent, remain unknown for many chemicals (51). As a result, the clinical correlation of errors created by neurotoxic agents on developmental processes is not always clearly defined. The human fetus can be exposed to neurotoxicants through placental transport and higher quantities of these compounds can be transferred to the infant during breast-feeding (65,68,70–74). Rarely have early gestational versus late gestational versus lactational exposures been examined. Studies have shown that lactational exposures can contribute significantly to a child’s burden (68–70). For example, Schecter et al. reported that lactational transfer represents an important source of PCBs and dioxins to the developing infant. While the level of exposure decreases after weaning, the young child is still more highly exposed on a daily basis than the adult (70). Understanding the critical timing of exposures will aid practitioners in determining where the most effective and feasible exposure prevention may occur along the developmental spectrum. Multigenerational testing in animal models exploring prenatal and postnatal developmental toxicity has long been utilized to gather valuable information pertaining to critical developmental periods of vulnerability (27,51). Unfortunately, the current multigenerational tests performed in animals are conducted on relatively few environmental contaminants. In addition, the application of animal data to human toxicity

A Pediatric Environmental Health Perspective

487

is still somewhat unclear and controversial. For the clinician, three key gaps in information need to be explored: (1) the relevance of animal models for predicting outcomes in humans, (2) the lack of comparative developmental profiles in animal models, and (3) application of testing protocols, both of animals and humans, in evaluating functional outcomes at relevant life stages (51,56). Once these issues are clarified by appropriate experimentation, the practitioner will be able to efficiently utilize animal data to reduce the range of uncertainties in determining toxicity in a child (29).

Impact of the Unique Vulnerability of Children The vulnerability of children to environmental containments is related to both qualitative and quantitative differences between children and adults (6,50,78,83). For example, the dietary habits of infants are clearly different from those of an adult. In addition to ingesting food and drink, children also have a greater opportunity of exploratory ingestion of unusual materials such as soil and dust (50). Children eat more food per unit of body weight and typically drink more water than do adults (Table 1) (6,12,37). Other exposure differences can be explained by dissimilarity in behaviors. Children spend significant time crawling and utilizing their hands for increased mobility. This behavior causes them to come into contact with many residential hazards, such as pesticide residues on carpets and lawns, that are not routinely present in an adult’s exposure pathway (15,19,37). Heightened vulnerability to exposure is compounded by age-related differences in susceptibility. Pharmacokinetic factors that contribute to differences in absorption, distribution, metabolism, and excretion can affect the amount and form (e.g., an active metabolite) of a chemical reaching a target site and could influence the toxic response (78,83). For example, pesticides such as carbamate and organophosphate thioether compounds must be metabolized by the cytochrome P-450 system (phase I and/or phase II enzymes) to become biologically active and produce a response (13,45,75,76). Many of the enzymes involved in the cytochrome P-450 system remain at low levels during prenatal and neonatal development (78,80,83). They do, however, attain adult levels fairly rapidly postnatally, at 6–12 months of age (50,78). In adults, cytochrome P-450 catalytic activity can be induced but tends to remain relatively stable (80). Therefore, the rate of formation of toxic metabolites differs in adults and children. The NAS report Pesticides in the Diets of Infants and Children, concluded that there can be large differences between children and adults in metabolite production; children may experience quantitatively and qualitatively different exposures to chemicals than do adults; that children may have differential sensitivity to chemically-induced toxicity compared to adults; and that standard approaches to risk assessment and regulation may not always account explicitly for potential age-related differences in exposure and toxicity (76,77). The challenge in evaluating potential adverse effects of chemicals on children is to design studies that will identify and quantify the influence of differential sensitivities on toxicological outcome. Tests involving intentional exposure to children will never be conducted (20,27). Consequently, it is important to identify animal models and design protocols (route, intensity, and duration of exposure) that are appropriate for extrapolating to human infants and children. Incorporating animals at all stages of growth and development into the currently used testing protocols would be useful for establishing the extent to which differential sensitivities between children and adults modify overall toxicity (43).

488

Harper and Shannon

Development and Use of Biomarkers Direct measurements of biomarkers in the form of biochemical, molecular, and cytogenetic probes are needed to conduct an accurate exposure assessment in children. The use of biomarkers is particularly valuable in a clinical setting when (a) exposures occur through multiple routes and pathways; (b) there is significant potential for dermal absorption or ingestion (i.e., from hand-to-mouth activity); or (c) there is a high ratio of sampling burden to subject ability (cognitive and physical) (10,28,32,41,42). Three types of biomarkers are useful in making the diagnosis of neurotoxic illness: biomarkers of exposure, biomarkers of effect, and biomarkers of susceptibility (8,13,41). Biomarkers of exposure are measures of internal dose, the biologically active dose, or the product of an interaction between a chemical and target molecule or receptor (13). The most common biomarkers of exposure are obtained through the determination of internal dose levels of a toxicant or its metabolite in a biological media (5,15,28). Biomarkers of effect are measurable biochemical or physiological perturbations within the nervous system that can be associated with an adverse health effect (5,15). In a clinical setting, this refers to qualitative or quantitative change noted from baseline or expected levels of performance that result from a given chemical exposure; for example, results from neuropsychological testing or neuroimaging studies (13,15,22). Biomarkers of susceptibility are indicators of increased or decreased risk and reflect an individual’s level of sensitivity to a chemical. An intrinsic genetic characteristic, such as a polymorphism, that results in an increase in the internal dose, the biologically effective dose, or the target tissue response is an example of a marker of increased susceptibility (4,9,41). Biomarkers that are specific to a particular environmental exposure can aid in the identification of the early stages of neurological impairment and in the understanding of basic mechanisms of exposure and response. Despite the clinical relevance of biomarkers, they should be interpreted with caution. Very few specific neurotoxicity biomarkers have been developed because of the complexity of the nervous system. Nonetheless, the use of biomarkers in clinical practice is essential (10,28,41,42). The development of adequate laboratory procedures for the application of suitable tests to measure biomarkers is fundamental to assure accurate and objective findings (8,9,15). Each new biomarker requires a comprehensive validation process. Biomarkers of exposure or effect must be validated in terms of their ability to assess the true exposure or disease (sensitivity) and their ability to assess the lack of exposure or disease (specificity). Establishment of normal baseline values and characterization of the distribution of biomarkers in laboratory animals, and determination of how they can be extrapolated to humans of all ages, will help to identify exposed persons so that risk can be predicted and disease prevented. An example of a biomarker that has been extensively validated for both exposure and effect is blood lead concentration. The ability to measure whole blood lead levels has had an enormous impact on the prevention, management, and treatment of lead poisoning in children (21,24). This example demonstrates the importance of internal dose biomarkers for clinicians and the many advantages they can provide for effective treatment and regulation. It is necessary to consider individual variability when using biomarker data, and as a result it is unlikely that general neurotoxicity biomarkers will be developed (5). Although many studies have used biomarkers in the adult clinical setting, limited research has been conducted using biomarkers to identify children affected by an environmental exposure. PEH research needs to continue to investigate and establish new biomarkers of exposure, effect, and susceptibility that can be integrated into future studies investigating

A Pediatric Environmental Health Perspective

489

environmental factors and childhood disease etiology. To do so, extensive and continued research on the basic mechanisms by which chemicals and their metabolites interact with developing tissues and organs of children and other organisms should be performed. The data will further our understanding of the biochemical interactions involved in the development of environmentally-induced illness. Information such as this would facilitate the clinician’s ability to detect a toxic exposure, monitor a disease treatment, perform preventive screening in high-risk populations, such as pregnant women, and to include environmental toxic exposures in differential diagnosis (42). Determining the Influence of the Gene-Environmental Interaction Most diseases affect the individual in ways that depend upon the age, susceptibility, environmental exposures, and genotype. The effect of the genotype on disease may occur through mechanisms of cellular regulation and growth, DNA replication and repair, and metabolism and biotransformation of endogenous or exogenous agents (4,80). Many genetic polymorphisms are known to contribute to enhanced or decreased sensitivity to toxicants and environmentally-related disease (4,80). For example, pesticides, such as parathion, are metabolized by the cytochrome p-450 system to paraoxon, which is a potent cholinesterase inhibitor. The enzyme paraoxonase (PON) catalyzes the hydrolysis of toxic metabolites and protects against pesticide toxicity, such as parathion. Polymorphisms in the genotype of this enzyme have been shown to alter serum PON activity, to affect the distribution and persistence of parathion metabolites in farm workers (45), and thus to affect susceptibility to chronic pesticide poisoning. The application of molecular biology assay techniques to aid in clinical diagnosis has dramatically increased, and improved methods of DNA diagnosis are now available in the clinical setting (3,4,42). Researchers should continue to explore gene–environment interactions and to expand our understanding regarding the contribution of genotype differences to an individual’s disease susceptibility. Future PEH research needs to identify, develop, and validate genetic markers of susceptibility in children. Children and Risk Assessment Relying heavily on existing toxicity information developed on a specific environmental contaminant, practitioners make decisions regarding level of toxicity and weigh available evidence about the contaminant’s potential to cause adverse health effects. This process is referred to as a risk assessment. Risk assessment is an attempt to evaluate the hazardous properties of a chemical and to determine the risks that result from exposure to it (81). Risk assessment is based on clinical and epidemiologic studies in which the effects of a toxic chemical are evaluated directly in humans, but more often it is based on toxicological studies of a chemical in animal models (81–82). To determine the potential risk to an infant or child resulting from a particular exposure may require many assumptions and extrapolations (27,40,56). Consequently, the results of a risk assessment performed for an infant or child are often controversial. There are four phases in a risk assessment: hazard identification, dose-response assessment, exposure assessment, and risk characterization. When applying the process of risk assessment to children, obvious data gaps within each phase should be highlighted: 1. The hazard identification phase is to determine whether exposure to an agent can increase the incidence of an adverse health effect (9). In PEH, the first step in hazard identification is clinical observation. Unfortunately, clinical recognition

490

Harper and Shannon

can occur only after disease symptomotology becomes evident. Diseases caused by environmental toxicants are frequently indistinguishable from illnesses caused by other factors. Outside of an acute poisoning, relying on clinical recognition to identify a link between an environmental toxicant or exposure and disease often proves challenging. Additionally, many years can elapse between exposure to a toxicant and the appearance of disease, making the assessment of past exposures extraordinarily difficult (56,82). Testing the possible toxicity of all new chemical compounds in animal models before they are ever used commercially or residentially can facilitate the identification of potential hazards much more efficiently and systematically. To do so, research needs to improve the linkage of toxicity testing studies to realworld exposures by developing an expanded set of research tools; developing and improving test methods to characterize the potential for chemicals to cause toxicity as a result of exposure during development; and evaluating alternative screening and testing methods that offer the potential to more rapidly, efficiently, and effectively assess potential health impacts of chemical exposures (67,76,77,81–82). This information will assist in the identification of chemical hazards before human exposure, disease, and death occurr. 2. The dose-response assessment phase evaluates toxicity information and characterizes the relationship between received dose of the contaminant and occurrence of adverse health effects within an exposed population (9). The doseresponse relationship is of particular importance for children. Unfortunately, as discussed previously, there is a distinct lack of information about the effects of most chemicals on the developing system. Toxicity testing of chemicals generally fails to consider the special vulnerability of infants and children; therefore, it provides little information about the hazards of toxic chemicals in this age group (81–83). For example, the overwhelming majority of pesticides have never been tested in animal newborns (76–77). Testing characteristically begins at age six to eight weeks, which corresponds roughly to five years of age in humans (82). Rarely have studies been conducted in which experimental animals were exposed to pesticides in utero or early in life and then followed over a lifetime to assess the late effects of early exposures. This is the situation that typically occurs in real life when infants are exposed to significant quantities of pesticides during development (76,77,81–82). Consequently, little is known of the delayed effects of early exposures to pesticides and other environmental toxicants. Future research is needed to better characterize the potential toxicity of the environmental chemicals to which children are frequently exposed. For example, studies evaluating both acute and delayed neurotoxic consequences of exposure to agents such as mercury, PCBs, solvents, pesticides and endocrine disruptors are needed. Basic science studies should include a comprehensive assessment of the toxickinetics relative to exposure dose during all stages of development. This will assist in determining the proper margin of safety and “effective” internal dose in children. Mechanistic toxicity studies in animals are needed for the evaluation of both single chemicals and mixtures. Many exposure scenarios involve more than one agent. Furthermore, development of methods and principles to determine the relationship between low-dose exposure and health risks to children, including concerns of unanticipated toxicity interactions and vulnerabilities of susceptible subpopulations, is an important research need. A better understanding of dose-response relationship would enhance the

A Pediatric Environmental Health Perspective

491

accuracy of risk assessments performed on infants and children in a clinical setting. 3. The exposure assessment phase evaluates exposure to the toxin in terms of exposure source, magnitude, duration, frequency and pathways of exposure (9,13,15). Exposure assessment must be expanded to include consideration of the unique dietary factors and behaviors of children. For example, the proximity of children to the ground increases their exposure, compared to an adult’s exposure, to chemical contaminants that may be present in dust and soil (81–83). Exposure assessment techniques often involve the use of biomarkers, which currently have restricted application in children. This limitation should be addressed. For example, because biologic monitoring data are often insufficient in children, the use of parent-child study pairs who are similar genetically and may have similar exposures can be utilized to permit an assessment of the importance of age in determining susceptibility (10,28). Exposure monitoring study equipment can be modified and designed so it is appropriately sized for children. Children differ from adults in their absorbed dose for a given external exposure level, resulting in responses that differ. Therefore, biomarker measurements are particularly important clinically for an accurate exposure assessment in children and should be incorporated into future study designs (10,28,42,56). 4. The risk characterization phase is the integration of the three phases previously described, hazard identification, dose-response assessment, and exposure assessment, into an estimation of the adverse effects likely to occur in a given population, including attendant uncertainties (9,13,76,81–82). When the risk to a child is different from that to an adult, the risk characterization should include those factors that differentiate the two. However, because of data gaps outlined in the previous phases, usually very little information pertaining to the risk to children is included in the final analysis. Thus, risk characterization often tends ignore children (3,10,41,82). This is clearly evident as classical risk assessment is designed to protect healthy male adults with a single chemical exposure. Children are exposed to numerous environmental hazards, often concurrently, in varying doses at different stages of their development (82). Future approaches for assessing risk need to expand to incorporate consideration of concurrent, multiple, and cumulative exposures that children frequently encounter (82). Desirable research advancements in risk assessment include: improved ability to extrapolate laboratory animal data to human risks; determination of the impact of the vulnerabilities of children on exposure dose (e.g., route, intensity, duration, and frequency) and risk outcomes; and integrated use of health and exposure data in methods that estimate the distribution of risk. Collectively, this information will provide insight into a particular compound’s toxicity and help characterize the pattern of illness induced by an agent in exposed children.

Measures of Neurobehavioral Dysfunction and Prognostic Indicators In children, neurotoxicity can manifest itself in a multitude of ways. A child can demonstrate impaired speech and language, attention problems, difficulties in learning, delayed acquisition of developmental milestones, and anti-social temperament (6,8,9,14,15). Chemical neurotoxicity may affect different neurobehavioral functions

492

Harper and Shannon

during the developmental phases from the neonatal period through adolescence. Neurodevelopmental toxicity may not always result in immediate impairments but may have a delayed appearance in manifestation of a particular function or skill (7,8,12,15). It may in fact take months or even years to detect the neurotoxic effect or outcome in a child exposed to a toxic agent in utero (22,23). The endpoints frequently used to test for developmental neurotoxicity in exposed children can be divided into two categories: apical and narrow-band tests (15,22,23). Examples of some apical tests are IQ and the Bayley Scales (Table 2). In apical testing, the patient must show successful performance on a variety of skills (22). Narrow-band tests tend to focus more on specific changes in neural development. Examples of these would be measurements of attention or memory. The selection of the appropriate outcome measure or test is critical in the context of developmental neurotoxicology studies. Because of age-related increases in the complexity of the developing nervous system and in functional capabilities, only a few tests that are appropriate for infants can be adminstered to older children as well. Consequently, very few infant tests are predictive of future development. Another problem in the clinical application of developmental neurotoxicology studies is the difficultly in accurately measuring all key potential confounders known to influence human cognition and development. These would include variables such as quality of the home environment, socioeconomic status, nutritional status, and parental IQ. The importance of these variables has been demonstrated in longitudinal prospective studies examining the association between environmental lead exposure and IQ (24,25). Many behavioral and cognitive assessment instruments are crude and inadequately validated in exposed children. The ability of a PEH clinician to use these tools to predict outcome and long term neurotoxic injury is limited. Most developmental evaluations utilize general intelligence as measured by IQ and/or cognitive ability assessments such as the Mental Developmental Index (MDI) of the Bayley Scales of Infant Development and Denver Developmental Screening Test (DDST) (26). These measures are crude when predicting if an insult has occurred specifically as a result of a previous exposure. There is also a poor correlation between the results of these types of evaluations, conducted in children less than 4 yr of age, and the results of evaluations conducted at older ages. An important research need is, therefore, the development of more sensitive, multidimensional neurobehavioral assessment tools that can account for age at the time of testing and which test specific neurobehavioral domains. Nevertheless, there has been significant progress in the last decade in developing validated methods for detecting neurotoxicity and expanding the depth of scientific knowledge. We need to expand the outcome measures used in assessing neurotoxicity endpoints to routinely include standardized neuropsychological tests, validated

Table 2

Selected Developmental and Psychological Tests for Children

Test Denver developmental screening test-revised Early intervention developmental profile Bayley scale of infant development Stanford–Binet intelligence test Wechsler preschool and primary scale of intelligence Wechsler intelligence scale for children–Revised

Age %5 2 mo to 3 yr !42 mo 2 yr to adulthood 3 yr 10 mo to 6 yr 7 mo 6 yr to 16 yr 11 mo

A Pediatric Environmental Health Perspective Table 3

493

Differences Between Children and Adults

Surface area: body mass ratio (m2/kg)a Respiratory ventilation rates Respiratory volume (mL/kg/breath)b Alveolar surface area (m2)c Respiration rate (breaths/min)d Respiratory minute ventilation ratee Drinking water (tap) Mean intake (mL/kg/day)f Fruit consumption (g/kg/day)g Citrus fruits Other fruits (including apples)h

Infants

Children

Teens

Adults

0.067 Infant 10 3 40 133 !1 yr 43.5 !1 yr 1.9 12.9

0.047 – – – – – 1–10 yr 35.5 3–5 yr 2.6 5.8

0.033 – – – – – 11–19 yr 18.2 12–19 yr 1.1 1.1

0.025 Adult 2 10 75 2 20–64 yr 19.9 40–69 yr 0.9 1.3

a

Square meters per kilogram. Milliliters per kilogram per breath. c Per square meter. d Breaths per minute. e Milliliters per kilogram body weight per square meter lung surface area per minute. f Milliliters per kilogram per day. g Grams per kilogram per day. h Milligrams per day. Source: From Refs. 47–50. b

computer-assisted test batteries, neurophysiological and biochemical tests, and imaging (9). These methods can provide data that are useful, within the clinical setting, in identifying neurotoxicity in exposed children (Table 3).

CONCLUSION Learning, behavior, and developmental disabilities in children are clearly the result of complex interactions among chemical, genetic, and social-environmental factors that influence children during vulnerable periods of development. The amount of new information and the rate of new discoveries involving children’s health and the environment in the last decade have been staggering. The susceptibility of the developing child to environmental exposures has become a major concern of the public health community and the environmentally aware public. This increased awareness can act as a powerful link to drive research and policy change and development. Better coordination of local and global data collection on environmental exposures in children is needed in order to meet the research needs in children’s environmental health. The identification of patterns of environmental diseases in children; improvement in the assessment of children’s exposures to environmental toxicants; a clear determination of developmental periods of vulnerability; and a better quantification of dose-response relationships in children are all required in order to improve our understanding, as clinicians, of the relationship between environmental exposures, risk, and health outcomes in children. With adequate information, we can make informed decisions about chemical exposures during all stages of a child’s development and minimize the risk of environmentallyinduced illness in children (2,11,17,29–36,38,40,43,44,46,53,54,55,74,79).

494

Harper and Shannon

REFERENCES 1. American Academy of Pediatrics Committee on Environmental Health. Pediatric Environmental Health. 2nd ed. American Academy of Pediatrics: Elk Groove Village, IL, 2003. 2. Shannon MW, Woolf AD, Goldman R. Children’s environmental health: one year in a pediatric environmental health specialty unit. Amb Pediatr 2003; 3:53–56. 3. Suk WA, Murray K, Avaikian MD. Environmental hazards to children’s health in the modern world. Mutat Res 2003; 544:235–242. 4. Suk WA, Collman GW. Genes and the environment: their impact on children’s health. Environ Health Perspect 1998; 106:817–820. 5. Costa LG. Research in neurotoxicology: the role of mechanistic studies to bridge the gap between the laboratory and epidemiological investigation. Environ Health Perspect 1996; 104:55–67. 6. Graeter LJ, Mortensen ME. Kids are different: developmental variability in toxicology. Toxicol 1996; 111:15–20. 7. Schwetz BA, Harris MW. Developmental toxicology: status of the field and contribution of the National Toxicology Program. Environ Health Perspect 1993; 100:269–282. 8. National Research Council. Evaluating Chemical and other Agent Exposures for Reproductive and Developmental Toxicity. Washington: National Academy Press, 2001. 9. IPCS, Environmental Health Criteria 223. Neurotoxicity Risk Assessment for Human Health: Principles and Approaches. Geneva: World Health Organization, International Programme on Chemical Safety, 2001. 10. Weaver VM, Buckey TJ, Groopman JD. Approaches to environmental exposure assessment in children. Environ Health Perspect 1998; 106:827–832. 11. Tilson HA, Moser VC. Comparison of screening approaches. Neurotoxicology 1992; 13:1–14. 12. Henck JW. Developmental neurotoxicology. In: Massaro EJ, Schardein JL, eds. In: Handbook of Neurotoxicology, Vol. 2. Totowa, NJ: Humana Press, 2001. 13. Feldman RG. Occupational and Environmental Neurotoxicology. Philadelphia, PA: Lippincott-Raven Publishers, 1999. 14. Landrigan PJ, Carlson JP, Bearer CF, et al. Children’s health and the environment: a new agenda for prevention research. Environ Health Perspect 1998; 106:787–794. 15. Altmann L, Sveinsson K, Kramer U, Winneke G, Wiegand H. Assessment of neurophysiologic and neurobehavioral effects of environmental pollutants in 5 and 6year old children. Environ Res 1997; 73:125–131. 16. Flegal KM, Keyl PM, Nieto FJ. Differential misclassification arising from nondifferential errors in exposure measurement. Am J Epidemiol 1991; 134:1233–1242. 17. Rothman KJ, Greenland S. 2nd ed Modern Epidemiology. New York: Lippincott Williams & Wilkins, 1998. 18. Jacobson JL, Jacobson SW. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. N Engl J Med 1996; 335:783–789. 19. Tilson HA. Developmental neurotoxicology of endocrine disruptors and pesticides: identification of information gaps and research needs. Environ Health Perspect 1998; 106:807–811. 20. Bellinger DC, Dietrich KN. Ethical challenges in conducting pediatric environmental health research. Neurotoxicol Teratol 2002; 24:443. 21. Needleman HL, Schell A, Bellinger D, Leviton A, Allred EN. The long-term effects of exposure to low doses of lead in childhood: An 11-year follow up report. N Engl J Med 1990; 322:695–702. 22. Jacobson JL, Jacobson SW. Assessing neurotoxicity in children. In: Tilson HA, Harry GJ, eds. Neurotoxicology. New York: Taylor and Francis, 1999. 23. Winneke G. Endpoints of developmental neurotoxicity in environmentally exposed children. Toxicol Lett 1995; 77:127–136. 24. Bellinger DC. Interpreting the literature on lead and child development: The neglected role of the “experimental system”. Neurotoxicol Teratol 1995; 17:201–212.

A Pediatric Environmental Health Perspective

495

25. Canfield RL, Henderson CR, Jr., Cory-Slechta DA, Cox C, Jusko TA, Lanphear BP. Intellectual impairment in children with blood lead concentrations below 10 microg per deciliter. N Engl J Med 2003; 348:1517–1526. 26. Behrman RE. 15th ed Nelson Textbook of Pediatrics. Philadelphia: W.B. Saunders, 1996. 27. Russell RW. Essential roles for animal models in understanding human toxicities. Biobehav Rev 1991; 15:7–11. 28. Bearer CF. Biomarkers in pediatric environmental health: a cross-cutting issue. Environ Health Perspect 1998; 106:813–816. 29. Miller M, Solomon G. Environmental risk communication for the clinician. Pediatr 2003; 112:211–217. 30. Agency for Toxic Substances and Disease Registry. Case Studies in Environmental Medicine; Lead Toxicity, 2002. Atlanta, GA. 31. Shannon MW, Graef J. Lead intoxication in children with pervasive developmental disorders. J Toxicol Clin Toxicol 1996; 24:177–181. 32. Shannon MW, Woolf AD, Binns H. Chelation therapy in children exposed to lead. N Engl J Med 2001; 345:1212–1213. 33. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Mercury 2001. Atlanta, GA. 34. Budtz-Jorgensen E, Grandjean P, Weihe P, Keiding N. Association between mercury concentrations in blood and hair in methylmercury-exposed subjects at different ages. Environ Res 2004; 95:385–393. 35. Frumkin H, Manning CC, Williams PL, et al. Diagnostic chelation challenge with DMSA: a biomarker of long-term mercury exposure? Environ Health Perspect 2001; 109:167–171. 36. Hviid A, Stellfeld M, Wohlfahrt J, Melbye M. Association between thimerosal-containing vaccine and autism. JAMA 2003; 290:1763–1766. 37. Bearer CF. How are children different from adults? Environ Health Perspect 1995; 103:7–12. 38. Ogborn CJ, Duggan AK, DeAngelis C. Urinary cotinine as a measure of passive smoke exposure in asthmatic children. Clin Pediatr 1994; 33:220–226. 39. Shu XO, Perentesis JP, Wen W, et al. Parental exposure to medications and hydrocarbons and ras mutations in children with acute lymphoblastic leukemia: a report from the Children’s Oncology Group. Cancer Epidemiol Biomarkers Prev 2004; 13:1230–1235. 40. Wright JM, Schwartz J, Dockery DW. The effect of disinfection by-products and mutagenic activity on birth weight and gestational duration. Environ Health Perspect 2004; 112:920–925. 41. Chern C-M, Proctor SP, Feldman RG. Exposure assessment in clinical neurotoxicology: environmental monitoring and biological markers. In: Chang L, Slikker W, Jr., eds. Neurotoxicology: Approaches and Methods. San Diego, CA: Academic Press, 1995. 42. Lubin B, Lewis R. Biomarkers and pediatric environmental health. Environ Health Perspect 1995; 103:99–104. 43. Ryan D, Farr N. Confronting the ethical challenges of environmental health research. Neurotoxicol Teratol 2002; 24:471–473. 44. Goldman LR. Linking research and policy to ensure children’s environmental health. Environ Health Perspect 1998; 106:857–862. 45. Lee BW, London L, Paulauskis J, Myers J, Christiani DC. Association between human paraoxonase gene polymorphism and chronic symptoms in pesticide-exposed workers. J Occup Environ Med 2003; 45:118–122. 46. Agency for Toxic Substances and Disease Registry. http://www.atsdr.cdc.gov/HEC/natorg/ pehsu.html, 2000. 47. Agency for Toxic Substances and Disease Registry. Case Studies in Environmental Medicine; Pediatric Environmental Health. Atlanta, GA, 2000. 48. US Environmental Protection Agency. Exposure Factors Handbook. Washington (DC): US Environmental Protection Agency. pp. 3–7. Vol. 1: General Factors. Report No.: EPA/600/ P-95/002Fa, 1997.

496

Harper and Shannon

49. Silvaggio T, Mattison DR. Comparative approach to toxicokinetics. In: Paul M, ed. Occupational and Environmental Reproductive Hazards: A Guide for Clinicians. Baltimore: Williams and Wilkins, 1993. 50. Snodgrass WR. Physiological and biochemical differences between children and adults as determinants of toxic response to environmental pollutants. In: Guzelian PS, Henry CJ, Olin SS, eds. Similarities and differences between children and adults: implications for risk assessment. Washington, D.C.: ILSI Press, 1992. 51. Selevan S, Kimmel CA, Mendola P. Identifying critical windows of exposure for children’s health. Environ Health Perspect 2000; 108:451–455. 52. World Health Organization. Environmental Health Criteria 118: Inorganic Mercury. Geneva: World Health Organization, 1991. 53. Dargan PI, Jones AL, Bogle RG, House IM. The effects of DMSA on urinary lead and mercury excretion in healthy volunteers. J Toxicol Clin Toxicol 2001; 39:521. 54. Agency for Toxic Substances and Disease Registry. Toxicological Profile for Trichloroethylene. Atlanta, GA, 1997. 55. Centers for Disease Control and Prevention. Managing Elevated Blood Lead Levels Among Young Children: Recommendations from the Advisory Committee on Childhood Lead Poisoning Prevention. Atlanta: CDC, 2002. 56. Morford LL, Henck JW, Breslin WJ, DeSesso JM. Hazard identification and predictability of children’s health risk from animal data. Environ Health Perspect 2004; 112:266–271. 57. Centers for Disease Control. Surveillance for Asthma–United States, 1960–1995. Morb Mortal Wkly Rep 1998; 47:1–27. 58. DeVesa SS, Blot WJ, Stone BJ, Miller BA, Tarove RE, Fraumeni JF, Jr. Recent cancer trends in the United States. J Natl Cancer Inst 1995; 87:175–182. 59. Buxbaum L, Boyle C, Yeargin-Allsopp M, Murphy CC, Roberts HE. Etiology of mental retardation among children Ages 3–10: The Metropolitan Atlanta Developmental Disabilities Surveillance Program. Atlanta, GA: Centers for Disease Control and Prevention, 2000. 60. National Academy of Sciences Committee on Developmental Toxicology. Scientific Frontiers in Developmental Toxicology and Risk Assessment. Washington, D.C.: National Academy Press, 2000. 61. Swain WR. Effects of organochlorine chemicals on the reproductive outcome of humans who consumed contaminated Great Lakes fish: an epidemiologic consideration. J Toxicol Environ Health 1991; 33:587–636. 62. Rogan WJ, Gladen BC, Hung KL, et al. Congenital poisoning by polychlorinated biphenyls and their contaminants in Taiwan. Science 1988; 241:334–336. 63. Jacobson SW, Jacobson JL, Schwartz PM, Fein GG. Intrauterine exposure of human newborns to PCBs: measures of exposure. In: D’Itri FM, Kamrin M, eds. PCBs: Human and Environmental Hazards. Boston: Butterworth, 1983. 64. Marsh DO, Myers GJ, Clarkson TW, Amin-Zaki L, Tikriti S, Majeed MA. Dose-response relationship for human fetal exposure to methylmercury. Clin Toxicol 1981; 18:1311–1318. 65. Amin-Zaki L, Elhassani SB, Majeed MA, Clarkson TW, Doherty RA, Greenwood MR. Mercury poisoning in mothers and their suckling infants. In: Holmstedt B, Lauwerys R, Mercier M, Roberfroid M, eds. Mechanisms of Toxicity and Hazard Evaluation. Amsterdam: Elsevier, 1980. 66. Marsh DO. Dose-response relationships in humans: methylmercury epidemics in Japan and Iraq. In: Eccles CU, Annau Z, eds. The Toxicity of Methyl Mercury. Baltimore, MD: Johns Hopkins University Press, 1987. 67. Kreutzer, RA, et al. First National Research Conference on Children’s Environmental Health: Research, Practice, Prevention, Policy. Conference Report. 1997. Available online: http:// www.cehn.org/cehn/Resconfreport.html. 68. Yakushiji T. Contamination, clearance and transfer of PCB from human milk. Rev Environ Contam Toxicol 1988; 101:139–164.

A Pediatric Environmental Health Perspective

497

69. Li X, Weber LW, Rozman KK. Toxicokinetics of 2,3,7,8-tetrachlorodibenzo-p-dioxin in female Sprague-Dawley rats including placental and lactational transfer to fetuses and neonates. Fundam Appl Toxicol 1995; 27:70–76. 70. Schecter A, Papke O, Lis A, et al. Decrease in milk and blood dioxin levels over two years in a mother nursing twins: estimates of decreased maternal and increased infant dioxin body burden from nursing. Chemosphere 1996; 32:543–549. 71. Jacobson JL, Fein GG, Jacobson SW, Schwartz PM, Dowler JK. The transfer of polychlorinated biphenyls (PCBs) and polybrominated biphenyls (PBBs) across the human placenta and into maternal milk. Am J Public Health 1984; 74:378–379. 72. Schecter A, Pa¨pke O, Ball M. Evidence for transplacental transfer of dioxins from mother to fetus: chlorinated dioxin and dibenzofuran levels in livers of stillborn infants. Chemosphere 1990; 21:1017–1022. 73. Yakushiji T, Watanabe I, Kuwabara K, et al. Postnatal transfer of PCBs from exposed mothers to their babies: influence of breast- feeding. Arch Environ Health 1984; 39:368–375. 74. Renwick AG, Dorne JL, Walton K. An analysis of the need for an additional uncertainty factor for infants and children. Reg Toxicol Pharmacol 2000; 31:286–296. 75. Scheuplein R. Pesticides and infant risk. Is there a need for an additional safety margin. Regul Toxicol Pharmacol 2000; 31:267–279. 76. Charnley G, Putzrath RM. Children’s health, susceptibility, and regulatory approaches to reducing risks from chemical carcinogens. Environ Health Perspect 2001; 109:187–192. 77. National Research Council. Pesticides in the Diets of Infants and Children. Washington, D.C.: National Academy Press, 1993. 78. Crom WR. Pharmacokinetics in the child. Environ Health Perspect 1994; 102:111–118. 79. Daly AK, Fairbrother KS, Smart J. Recent advances in understanding the molecular basis of polymorphisms in genes encoding cytochrome P450 enzymes. Toxicol Lett 1998; 28:143–147. 80. Nakajima T, Aoyama T. Polymorphisms of drug-metabolizing enzymes in relation to individual susceptibility to industrial chemicals. Industrial Health 2000; 38:143–152. 81. National Research Council. Science and Judgment in Risk Assessment. Washington, D.C.: National Academy Press, 1994. 82. Behrman R. The Future of Children: Critical Issues for Children and Youths. Los Altos, CA: The David and Lucile Packard Foundation, 1995. 83. Done AK, Cohen SN, Strebel L. Pediatric clinical pharmacology and the “therapeutic orphan”. Ann Rev Pharmacol Toxicol 1977; 17:561–573.

24 Evaluating Neurotoxic Effects: Epidemiological, Epistemic, and Economic Issues Herbert L. Needleman Department of Psychiatry, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania, U.S.A.

INTRODUCTION When the effects of a toxicant are less obvious than death, assessment of its toxicity is guaranteed to be a contentious enterprise. The notion that toxins could produce important health outcomes short of fatality was slow to develop and only recently accepted. The discipline of toxicology, after all, began as an offspring of the widespread clandestine practice of poisoning one’s rivals. One of its more prominent exponents was Nero’s mother, Agrippina. It is only recently that toxicologists abandoned as the main metric of potency the LD50, the dose that killed half of one’s experimental animals. At about the same time, case reports of lethal exposures in humans became rare, and were replaced with epidemiological investigations of toxicity at doses below the fatal. Dioscerides, Nero’s physician and author of the Materia Medica, established himself as perhaps the first neurotoxicologist with his warning that “Lead makes the mind give way.” Neurotoxicology emerged in the 1960s as a legitimate field. This paralleled in time the rapid growth of environmental awareness and increasing skepticism about industrial chemicals. The thalidomide disaster of the late 1950s furnished an ironic background to Dupont’s grandiose promise, “Better Living through Chemistry.” In 1970 Richard Nixon founded the Environmental Protection Agency (EPA) and inaugurated Earth Day. Neurotoxicology is a relatively young branch of the mother toxicology tree. Recognizing that the central nervous system was a vulnerable target for xenobiotics, and that the developing nervous system was a particularly susceptible target, came late into focus. The first meeting devoted to neurobehavioral toxicology was organized by Bernard Weiss and Victor Laties in 1975 (1). At the same time, the realization grew that thousands of untested chemicals were being inserted into our environment. Animal studies of methylmercury (2) and human studies of alcohol (3) were published during the same period that Earth Day was celebrated and the EPA was founded. Toxicologists rapidly learned the tools of experimental psychology, and sensitive measures of behavior were put to work studying toxicants. 499

500

Needleman

Lead, because it is the best-studied toxicant, offers useful examples of methodological and other problems that could be encountered in the study of other agents. I offer some observations gleaned from lead toxicology for examination.

THE GROWTH OF UNDERSTANDING OF CHILDHOOD LEAD TOXICITY One hundred years ago, the first published reports of lead poisoning in children in Brisbane Australia were greeted with disdain and frank skepticism. The orthodox view at that time was that lead poisoning afflicted only workers or imbibers of adulterated wine. The lethal threat of lead to children was generally accepted at the turn of the 19th century. After this, it was generally believed that childhood lead poisoning had only two outcomes: death or complete recovery without residua. Randolph Byers challenged this dogma in 1943 with the first demonstration of learning and behavior problems in children who survived acute illness (4). The long-term consequences of acute lead toxicity were accepted, but the prevailing belief was that sequelae were found only in those children who had displayed signs of acute encephalopathy. In the late 1970s and 1980s the deleterious impact of lead in children in the absence of overt toxic symptoms was convincingly demonstrated (5). We may expect other toxicants to follow similar paths of discovery, and the reality of low-dose asymptomatic effects to encounter similar incredulity. Causal associations between toxicants and neurobehavioral dysfunction will predictably be challenged by skeptics.

THE CONUNDRUM OF CAUSALITY The central task of science is to develop a trustworthy picture of nature in order to enable dependable predictions, and in the case of the health sciences, effective remedies. To accomplish this practical mission requires the separation of causal from random or spurious associations. This is difficult work, and in the area of environmental toxicants, the task is often vexed by external influences, only some of which are scientific. Some are long held social myths, some are methodological predispositions, and some are cynical arguments spawned by vested interests. A large segment of the scientific community believes that the pursuit of their craft isolates them from external influences, and that their work is neutral. This is naive. Social and economic factors, often operating out of awareness, thoroughly influence the conduct of research, the nature of problems selected, the tools employed, and the conclusions drawn. In evaluating associations between toxicants and toxic effects, two types of errors are encountered: errors of the first kind (a or false positive errors), and of the second kind (b or false negative errors). Science has traditionally endowed Type I error avoidance with prestige, considering it as evidence of scientific rigor. Avoiding Type II errors is granted considerably less credit. This dissociation can be readily seen in the table below which shows the ratio of b (Type II) to a (Type I) errors at frequently encountered sample sizes (Table 1). It is indisputable that spurious causal associations should be carefully guarded against. In the area of public health, however, where exposures may affect large numbers, the costs of missed subtle associations are particularly serious and important to avoid. In practice, at subject and effect sizes frequently employed, the demonstrated ratio of b to

Evaluating Neurotoxic Effects Table 1 Number of subjects 50 80 100 200

501

Ratio of b to a Errors at Commonly Employed Sample and Effect Sizes Effect size, selected a risk RZ0.10, aZ0.05 18 17 17 16

RZ0.10, aZ0.01 97 96 94 88

RZ0.3, aZ0.05 9 4 3 0.2

RZ0.3, aZ0.01 67 44 61 31

a error is extraordinarily high; science has a large tolerance for b errors. Only at relatively large effect and sample sizes are the two error types in the same range. The Illusion of Causal Proof The demand for proof of a causal association, in this case, between lead and behavioral deficit, has been a potent impediment, often cynically employed, to informed standard setting. Defining causality is itself no simple matter. The English mathematician, philosopher, and gymnast William Kingdom Clifford wrote, in 1872, that “The word represented by ‘Cause’ has sixty-four meanings in Plato and forty-eight in Aristotle.” Science has to a large extent moved past the search for single causes, and acknowledges that in our work we attempt to tease out the constellation of necessary and sufficient factors to reliably predict the effect under study. David Hume’s assertion that causality is a relationship between mental as opposed to physical events established the impossibility of achieving causal proof. Ernst Cassirer went further: We must think of causality as a guide-line which leads us from cognition to cognition and thus only indirectly from event to event, a proposition which allows us to reduce individual statements to general and universal ones and to represent the former by the latter. Understood in this sense, every genuine causal proposition, every natural law, contains not so much a prediction of future events as a promise of future cognitions.

Despite the caution that cognitions or mental events are not available to empirical proof, industrial interests regularly attempt to exploit this limitation by demanding proof of causality before setting regulatory standards. Among the industrial arguments used to challenge epidemiological studies of lead and cognitive function are these:

† Causality operates in the other direction, that is, elevated body burdens of lead are due to preexistent cognitive deficit. This is a disingenuous extrapolation from the finding that mentally retarded children frequently mouth nonfood substances. Proponents of this solecism ignore the fact that most published studies examined children who are not retarded, that forward studies of children whose blood lead levels were measured at birth showed a dose-dependent decrease in cognition, and that experimental studies of primates given lead at controlled doses showed impaired cognition † Because control of confounders is incomplete, the causal assertions reported are spurious. Confounder control will always be incomplete. Multivariate space is infinite, but subject numbers are limited. It is always possible to nominate an additional variate. In order to qualify as a confounder, a covariate is required to be correlated with lead and to affect development. Behavioral scientists know a great deal about the factors that influence child development, and many studies,

502

Needleman

CUMULATIVE FREQUENCY DISTRIBUTION (%)

using different covariate models, find similar associations between lead and outcome. It is unlikely that any important covariate has been overlooked. † Treating the criterion of P!0.05 as a sacrament. RA Fisher is credited with establishing this criterion. In 1925, Fisher wrote, indicating the cutoff of 0.05, (i) “It is convenient to take this point as limit in judging whether a deviation is to be considered significant or not. Deviations exceeding twice the standard deviation are thus formally regarded as significant (6)”. (ii) P!0.05 is a convenient convention, not a universal truth. Jerome Cornfeld, attempting to explain this arbitrary number, wrote that Fisher’s publisher allowed him to reprint only two lines from his table of P-values: 0.05 and 0.01. Despite the slow slide of P-values into statistical desuetude, this issue continues to be raised to disparage reports of toxic associations, and dismiss studies with P-values greater than 0.05 as evidence for no effect. † The effects demonstrated are small, and therefore of no biological consequence. This is disingenuous. Critics of the lead/cognition hypothesis have criticized the decrement of 4–6 points in exposed subjects as trivial. Figure 1, taken from actual data, shows that a shift in median IQ of four points is associated with a 400% increase in the rate of children with severe deficit (IQ!80), and that shifting the curve to the right prevents 5% of the population from achieving superior function (IQO125). † Errors in measurement result in erroneous causal conclusions. This is a commonly applied argument. All phenomena are measured with some error, but all measurement error does not yield false positive judgments. For a measurement error to produce a Type I error, it must be nonrandom and nonsystematic. Random and nonsystematic errors are null biasing; they increase the Type II risk. Most errors in epidemiological studies, because they are random and nonsystematic, tend to increase Type II risk. They diminish the reported effect size. 100 90

High Lead Low Lead

80 70 60 50 40 30 20 10 50

Figure 1

60

70

80 90 100 110 120 130 140 VERBAL IQ

IQ scores in children with high and low lead levels: the cumulative frequency.

Evaluating Neurotoxic Effects

503

† Arguments that not all studies have found a statistically significant effect; therefore, no conclusion can be drawn. Most, but not all of the modern studies have shown a statistically significant association between lead and IQ. Counting votes or balloting as an approach to pooling studies is inferior thinking; it severely degrades the data. If 10 studies of lead and IQ were done, and four reported an effect at P!0.05, while six reported non-significant findings, many would interpret this constellation as arguing to no effect. The actual probability of such an outcome, under the null hypothesis is PZ0.006. All meta-analyses have found a robust effect of lead on IQ (7–9). The arguments summarized above, employed repeatedly by the lead industry in one form or another, explain to a considerable extent why, after tetraethyl lead (TEL) was introduced and workers began to die, it required 60 years to ban it from gasoline (10). Similar arguments are employed in opposition to the regulation of pesticides, and in discussions of global warming. They can be expected to find employment, with some modifications, in the adjudication of other environmental questions. A sophisticated and sensible response to the question of causal judgment can be found in the well-known contribution of Bradford Hill. AB Hill and Causal Inference In 1966, recognizing the difficulties in seeking causal proof, Sir Austin Bradford Hill recommended 9 factors to be taken into consideration in evaluating the causal status of observed associations. Hill’s own estimate of the role of his list is quite modest: “What aspects of that association should we especially consider before deciding that the most likely interpretation of it is causation?” Despite this modest advice, his guidelines have achieved canonical status, and over time have been widely used and uncritically swallowed whole. Hill’s Criteria

† Strength of association. The stronger the association, the more attractive is

† † † † † † † † †

a causal interpretation. Hill then qualifies this: “We must not be too ready to dismiss a cause and effect hypothesis on the grounds that the observed association appears to be slight.” Consistency. Has the association been observed by different persons, in different circumstances, and times? Specificity. “We must not, however, over-emphasize the importance of the characteristic.diseases may have more than one cause.” Time precedence. The cause must precede the effect. Biological gradient. The dose-response effect Plausibility. “It will be helpful if the causation we suspect is biologically plausible. But this is a feature I am convinced we cannot demand.” Coherence. “ [T]he cause and effect interpretation should not seriously conflict with the generally known facts of the natural history and biology of the disease.” Experiment. “For example: . some preventive action is taken. Does it in fact prevent?” Analogy. This is the weakest category: arguing from similar agents producing similar effects Surprisingly, Hill’s paper does not deal with the question of confounders. The

504

Needleman

criterion of non-spuriousness is central. Have the important confounders been identified and controlled by design or statistical analysis? What is the impact of adjustment on effect size? Hill is at pains to deprecate the utility of significance tests: “No formal tests of significance can answer these questions. Such tests can, and should, remind us of the effects that the play of chance can create, and they will instruct us in the likely magnitude of those effects. Beyond that they contribute nothing to the “proof” of our hypothesis.”

SCIENCE, UNIVERSITIES, AND THE MARKET PLACE There are other influences that inhibit the untrammeled conduct of science. Robert Merton defined the fundamental values of science as universalism, communalism, disinterestedness, and organized skepticism. Universalism means that the assertions of science are separated from national or cultural influences. Communalism refers to the free and unhindered sharing of data and understanding. Disinterestedness means that the enterprise of science goes forward without consideration of personal or external values. Merton’s portrait of science is an idyllic vision of a community of scholars, speaking freely, sharing ideas, sheltered from external pressures and interests. It is this pastoral vision that has attracted many to the academic life. Realizing this vision has been severely tampered with until it now exists only in nostalgia (11). A university has multiple personalities (12). In its classical mode, the university’s function is the pursuit of knowledge for its intrinsic value. The Baconian ideal seeks knowledge in the service of society: the development of new products and efficiencies in production. The military mode focuses on the development of knowledge to serve the national defense. The fourth model portrays the university in the service of public needs. Each of these models has its own priorities, its own ethos, and often is in contest with the others for prestige, space, and budget. The increased emphasis on the use of the university to meet national needs of production was the casis belli of the student uprising at Berkeley in the 1960s. The transformation of American research universities from sheltered places for seekers of knowledge for its own sake to exemplars of the Baconian ideal has taken place in lock-step with the decline of federal support of research. As federal dollars became scarce, commercial interests were quick to fill the gap. In 1980, the government, concerned about productivity and competition from other countries, passed the Bayh-Dole Act, permitting universities to patent the fruits of federally funded research. The interlocking of universities and corporations tightened, and many faculty members became entrepreneurs. The communal nature of science withered as investigators were barred from free publication of their work, from sharing biological specimens, and from discussing their results with others. The market place had embraced the university. And vice versa. This single act has permanently altered the structure and function of universities, and warped the goals and ethos of many academic scientists. It is now possible for a scientist in his laboratory to visualize becoming wealthy. There are many doleful examples of the altered ethos in which biomedical research is conducted; it is enough to remember that after he successfully produced and tested the killed polio vaccine in 1954, Jonas Salk refused to patent his discovery. An instructive exercise would be to imagine this occurring a half century later.

Evaluating Neurotoxic Effects

505

Sheldon Krimsky, a scholar of the issue of the impacts of corporate influence on science and scientists, surveyed 800 biotechnology faculty in 1992 and found that 47% consulted with industry, 25% received industry–supported grants or contracts, and 8% held equity in a company whose products were related to their research. There is no place for Merton’s criterion of disinterestedness in this environment. Does a relationship with industry distort scientific judgment? Does it result in biased conclusions? A recent meta-analysis surveyed 37 publications dealing with support and outcome and found a statistically significant relationship between industrial sponsorship and pro-industry conclusions (13). Stelfox and colleagues examined 70 articles dealing with the safety of calcium channel blockers. Ninety-six percentage of the authors who supported the safety of the agents had financial relationships with manufacturers of the drugs. Only 37% of those authors who expressed critical beliefs had an industrial association (14). Accepting commercial support affects not only the conclusions drawn; it colors the quality of reporting. A recent survey of clinical trial agreements between medical schools and industry sponsors found that industry–supported institutions routinely fail to meet the guidelines issued by the International Committee of Medical Journals for study design, access to data, and control over publication (15). The history of regulating lead in gasoline further illuminates these findings. In 1925 when TEL was first manufactured, workers at all three plants became psychotic, and many died terribly in acute mania. A moratorium was called on its manufacture. The Surgeon General convened a meeting of academic and industry scientists to discuss the best action. Medical scientists from Harvard, Yale, and Columbia warned of the severe consequences of producing and distributing this toxin. They seemed to be prevailing until a corporate officer from Standard Oil declared that this compound was “a gift from God,” needed to nourish our industrial competitiveness. Prominent on behalf of the lead industry was Robert Kehoe, a pathologist from Cincinnati, whom General Motors hired to study the stricken workers at the Dayton, Ohio, TEL plant. Kehoe claimed that TEL was purely an industrial hazard; it could be safely controlled by the proper hygiene, and presented no threat to the general public. Ethyl Corporation and General Motors conveyed its gratitude by underwriting an institute at the University of Cincinnati, making Kehoe director, and naming it after him. Kehoe claimed that he knew more about the subject than anyone else. From the 1920s until the 1970s he was practically the only investigator receiving financial support to study lead. In the 1970s the National Institute of Environmental Health Sciences began to fund independent investigators of the question. The evidence of the true toxicity of lead grew and became generally understood by the scientific community. When the study of lead turned to effects at low, clinically silent doses, the industry, through its trade organization, the International Lead Zinc Organization, recognized this as a severe threat to the production of lead in gasoline. The EPA moved to set an air lead standard as a prelude to regulating lead in gasoline. After strenuous debate, during which three separate drafts of the Air Lead Criteria Document were written, the EPA accepted the reality of toxicity at low dose. The final document resulted in lowering the acceptable air lead concentration to 1.5 mg/m3. When the second Criteria Document was drafted in 1982, a small number of industry–affiliated investigators challenged the toxicity of lead at low dose (16). A review of the reports of this question disclosed that practically all of the publications that found no harmful effect at these low concentrations were written by the beneficiaries of industry support. Interestingly, a close scrutiny of these papers reveals that their own data supported the toxicity of lead at these low doses (17,18). The spokesmen for the International Lead Zinc Organization and Ethyl Corporation loudly and repeatedly pronounced that no one had ever been sickened by lead in the air,

506

Needleman

and that removing it from gasoline would have no impact on the body lead burden of adults and children. After a 20-yr struggle, in which most of the issues touched on in this paper were confronted, lead was removed from gasoline (19). The mean national blood lead level, which was 15 mg/dL in 1975, is now !2 mg/dL. The monetized benefit for this lowering of blood lead levels for a one-year cohort of American children has been estimated to be between $119 billion and $ 318 billion (20). SUMMARY Mammals have populated the planet for 250 million years. Over this time those who were able to develop adaptive mechanisms to protect them from “natural” toxins in their environment were favored by natural selection. In contrast, most neurotoxicants are manmade. Their history extends back less than 100 yr, and we possess few adaptive mechanisms to deal with them. If a man-made agent is biologically active, it is bound to be toxic at some dose. The toxic metals are exceptions: they are as old as the planet, but their natural concentrations are at the trace level. Human activity has concentrated them in the biosphere. Patterson has shown that modern man has 1000 times the lead in his body as our pretechnological ancestors (21). Measurement error in studies is unavoidable. It is an ingredient in experimental studies, but is especially potent in epidemiological studies. Because in the great majority of studies the errors are random (neither systematically greater or less) and nonsystematic (not increased systematically in exposed or unexposed subjects), they tend to bias results towards the null and reduce the measured effect size. The application of more sensitive and rigorous measures of exposure and performance, the use of larger samples, and better multivariate techniques have enabled investigators to discern effects at lower body burdens of lead. Equally important, with the removal of lead from gasoline, blood lead levels declined drastically. It became possible to compare exposed subjects to referent groups with blood lead levels !2 mg/dL. Reports of lead effects on IQ associated with blood lead concentrations of 5 mg/dL or below were published (22,23,24). In these studies, the slope of the IQ/lead regression was steeper !7 mg/dL than above it. If a threshold for lead toxicity exists, it has not been demonstrated. Given the omnipresence of these new toxicants, and the categorical impossibility of achieving causal proof of their toxicity, what should be done? If society awaits causal proof of the toxicity of an agent, millions of people will have paid the price of exposure. In 1976 the mean blood lead level was 15.6, 5 mg/dL above the CDC definition of toxicity. Rational prevention resides in the tension between Type I and Type II risks. Type I risks have been called producer risks. If a safe agent is defined as toxic, production will be limited or ended, jobs will be lost, and profits reduced. Type II risks are consumer risks. If a dangerous agent is declared safe, the agent will be distributed, consumers will suffer, and people will be sickened. The nature of the regulatory decisions our society chooses under the governing uncertainties it faces will depend on our values and will ultimately define us as a civilization. REFERENCES 1. Weiss B, Laties V. Behavioral pharmacology: the current status. FASEB Monographs, Bethesda MD 1975. 2. Spyker J, Sparber S, Goldberg A. Subtle consequences of methylmercury exposure: behavioral deviations in offspring of treated mothers. Science 1972; 177:621–623.

Evaluating Neurotoxic Effects

507

3. Jones K, Smith D, Ulleland C, Streissguth A. Patterns of malformation in offspring of chronic alcoholic mothers. Lancet 1973; 1:1267–1272. 4. Byers RK, Lord EE. Late effects of lead poisoning on mental development. Am J Dis Child 1943; 66:471–483. 5. National Research Council. Measuring Lead Exposure in Infants, Children, and Other Sensitive Populations Washington D.C. Washington: National Academy of Sciences Press, 1993. 6. Fisher RA. Statistical Methods for Research Workers. Edinburgh: Oliver & Boyd, 1925. 7. Schwartz J. Low-level lead exposure and children’s IQ: a meta-analysis and search for a threshold. Environ Res 1994; 65:42–55. 8. Needleman HL, Gatsonis C. Low level lead exposure and the IQ of children. JAMA 1990; 263:673–678. 9. Pocock SJ, Smith M, Baghurst P. Environmental lead and children’s intelligence: a systematic review of the epidemiological evidence. BMJ 1994; 309:11197–11898. 10. Needleman HL. Clamped in a straitjacket: the insertion of lead into gasoline. Environ Res 1997; 74:95–103. 11. Washburn J. The kept university. The Atlantic Mon 2000; March. 12. Krimsky S. Science and the Private Interest. Lanham, MD: Rowman-Littlefield Publishing Co, 2003. 13. Bekelman J, Li Y, Gross C. Scope and impact of financial conflicts of interest in biomedical research. JAMA 2003; 289:454–465. 14. Stelfox H, Chua G, O’Rourke G, Detsky A. Conflict of interest in the debate over calcium antagonists. N Engl J Med 1998; 338:101–106. 15. Shulman K, Seils D, Timbie JW, et al. A national survey of provisions in clinical-trial agreements between medical schools and industry sponsors. N Engl J Med 2002; 347:1335–1341. 16. United States Environmental Protection Agency. Air Quality Criteria for Lead. Environmental Criteria and Assessment Office, EPA 600/8/-83-028A. 17. Ernhart CB, Landa B, Schell NB. Subclinical levels of lead and developmental deficit A multivariate followup reassessment. Pediatrics 1981; 67:911–919. 18. Smith M, Delves T, Lansdown R, Clayton B, Graham P. The effects of lead exposure on urban children: The Institute of Child Health: Southampton Study. Dev Med Child Neurol 1983; 25:1–54. 19. Needleman HL. The removal of lead from gasoline: historical and personal reflections. Environ Res 2000; 84:20–35. 20. Grosse SD, Matte TD, Schwartz J, Jackson RJ. Economic gains resulting from the reduction in children’s exposure to lead in the United States. Environ Health Perspect 2002; 110:563–569. 21. Patterson CC. Contaminated and natural environments of man. Arch Environ Health 1965; 11:344–360. 22. Canfield RL, Henderson CR, Cory-Slechta DA, Cox C, Jusko TA, Lanphear BP. Intellectual impairment in children with blood lead concentrations below 10 micrograms per deciliter. N Engl J Med 2003; 348:1517–1526. 23. Lanphear BP, Dietrich KN, Auinger P, Cox C. Cognitive deficits associated with blood lead concentrations !10 (g/dL in U.S. children and adolescents. Publ Health Rep 2000; 111:521–529. 24. Bellinger DC, Needleman HL. Intellectual impairment and blood lead levels. N Engl J Med 2003;349–500.

25 Knowledge, Norms, and the Politics of Risk: Ethical Issues in Policy-Relevant Science Virginia Ashby Sharpe Center for Clinical Bioethics, Georgetown University Medical Center, Washington, D.C., U.S.A.

I know of no safe depository of the ultimate powers of the society but the people themselves; and if we think them not enlightened enough to exercise their control with a wholesome discretion, the remedy is not to take it from them, but to inform their discretion. —Thomas Jefferson, 1720

INTRODUCTION In the 1970s, the American medical profession was confronted with the fact that the internal norms that had guided its practice for centuries were ill-suited to the demands of participatory democracy. As is well-documented in texts on the emergence of the contemporary field of bioethics (1,2), the medical professions’ paternalistic attitude regarding what was best for patients in terms of both the information and the interventions they received began to be supplanted in the 1960s and 1970s by the call for the medical profession to be accountable to democratic norms of self-determination, transparency, and justice. This demand reflected the broader calls for democratic reform that defined the era: civil rights for African–Americans, women, and others who were marginalized in American society; more open and democratic practices by government agencies charged with protecting public health and the environment; and international recognition of the need for regulations governing the use of humans in medical experimentation. In 1948, in response to the war crimes tribunals against Nazi doctors, the Nuremberg Code articulated as its first principle that “the voluntary consent of the human subject is absolutely essential” (3). It took decades before the U.S. codified this principle into law—decades during which egregious violations of the principle were exposed in the Tuskegee experiments and a host of other research on vulnerable groups such as the terminally ill, retarded children, the poor, the institutionalized elderly, and military recruits. By the late 1970s a national commission on the protection of human research subjects had articulated 509

510

Sharpe

the ethical principles that would be formalized in regulations governing the growing enterprise of federally-funded research. These principles were manifested in the requirements of informed consent, the justifiability of research risks, and fairness in the selection of subjects. The demand for physicians—whether they were engaged in clinical practice or in research—to be accountable to public norms established outside the canons of professional practice represented a paradigm shift in both the ethics and the politics of medicine. There is a comparable shift occurring today as scientists—particularly in fields such as ecology, epidemiology, and toxicology—serve as a resource for policy making on such high-stakes issues as global climate change, endocrine disruption, and agricultural biotechnology. In this chapter, I offer an analysis of this shift with particular attention to toxicology and epidemiology. I draw on the distinction between normal and post-normal science to examine the ethical challenges that confront scientists individually and collectively when their work is relevant to the policy or regulatory setting. I argue that the uncertain and contested character of knowledge, as well as the growing significance of risk characterization, together emphasize the need for norms of transparency and access to participation in science-based decision making. I discuss the problem of conflicts of interest in science—particularly as regulated industries play an increasing and documented role in manufacturing scientific uncertainty and shaping science to serve specific corporate interests. Finally, I discuss “community-based research” and the particular ethical complexities that confront this model of inquiry. My perspective on these issues is shaped by fifteen years of work on ethical issues in biomedical and environmental science and two years as director of the Center for Science in the Public Interest’s project on Integrity in Science, which examines the role of industry in scientific research and sciencebased policy.

THE NORMATIVE CONTEXT OF POLICY-RELEVANT SCIENCE Normal Science and the Norms of Governing Science In The Structure of Scientific Revolutions, Thomas Kuhn (4) coined the phrase “normal science” to refer to the activities of scientists engaged in puzzle-solving to ensure the steady accumulation of relatively stable and certain facts within the context of established methods. The “normal science” view of responsible scientific conduct was probably best captured in Robert Merton’s 1942 essay “The Normative Structure of Science” (5) where he described a set of four principles governing scientific practice: (1) universalism, (2) commun[al]ism, (3) disinterestedness, and (4) organized skepticism. Universalism ensures that scientific claims are evaluated on the basis of established and impersonal criteria without regard to the attributes of the scientist. Commun[al]ism is the norm of cooperation, collaboration, and openness in the conduct of research and the nonproprietary nature of scientific findings. Disinterestedness is the pursuit of science subject to scrutiny and policing by one’s expert peers. Organized skepticism is an intellectual commitment to critical examination of all claims and the suspension of judgment until knowledge has been certified. This conception of science as objective, disinterested, and value-free is the basis of scientific positivism and has had remarkable staying power, providing the justification—often explicit—for deference to scientists and reliance on scientific expertise as an independent source of information about objective reality. This view has been reinforced by the many scientific triumphs that have provided the basis for advances, for example, in medicine, transportation, sanitation, and agriculture.

Knowledge, Norms, and the Politics of Risk

511

Over the last 75 years there have been three strong and related challenges to the view of science as objective and disinterested: (1) the view of science as a social construction; (2) the view that science is insufficient as a basis for complex decision making; and (3) the view that science is a servant yoked to particular social projects. These challenges to the ideal of scientific isolationism give rise to new ethical demands as science is used to further the aims of democratic society. Science as a Social Construction The most significant challenge to the ideal of scientific objectivity has been the analysis by sociologists and philosophers of science of the production of scientific knowledge. Kuhn’s own analysis of scientific revolutions, for example, showed that theories are retained well beyond their explanatory capacities simply because they conform to the prevailing paradigm. Others, including Mary Douglas (6,7), Sandra Harding (8), Ulrich Beck (9), and Helen Longino (10), have illustrated that assumptions about what problems are worth addressing, which experimental endpoints will be selected for study, what populations will be sampled, what effects will be considered significant, and what “confidence level” will legitimate a statistical test all reflect values on which methodologies, research programs, scientific disciplines, and grant-making, in short, the scientific enterprise, is literally constructed. As Sandra Harding observed in her book Whose Science, Whose Knowledge?, political and social interests are not “add-ons” to an otherwise transcendent, value-neutral science: “Scientific beliefs, practices, institutions, histories and problems are constituted in and through.political and social projects and always have been” (8). The view of science as a social construction has been controversial in part because it opens all scientific claims to the charge of relativism. If science is “simply” a social construction, guided by various political interests and cultural biases, the thinking goes, then scientific knowledge loses its credibility and integrity as an authoritative basis for decision making. Science is Insufficient as a Basis for Complex Decision Making Less controversial than the strong claim that science is socially constructed is the view that even the best science is inadequate to the task of complex decision making—either in policy setting or for individuals. Scientific information by itself imposes no imperatives. It is not an end in itself, nor does it dictate the goals we should pursue. Rather, it is the means to ends that are decided with but not by science. In other words, science is a resource for technical information only, and the uses of that information are determined on the basis of a host of social, economic, and cultural values. For example, even though it has been well established that lead is a neurotoxin in children, human exposures to lead are allowed to persist in municipal water supplies, housing stocks, and until 1996 in lead-based gasoline. It is, in part, economic values, political inequalities, and conventional wisdom about exposures that have produced this set of circumstances even in the face of scientific consensus regarding lead’s adverse effects. For example, lead-based paint is prevalent in many older inner city apartments where poor people live. They have little economic and political power. Regulation and clean-up enforcement requires that elected government officials with limited financial resources confront land and apartment owners, who may or may not have the economic means or incentives to take corrective action. Likewise, conventional wisdom is that lead concentrations in municipal water supplies are typically very low, with most of the lead entering water as a result of plumbing configurations in individual houses. However, as the recent revelations about lead in the District of Columbia water supply indicate, that municipality decided, as a cost-saving measure, to

512

Sharpe

introduce chloramine into the water supply even though that chemical was known to leach lead out of pipes. Scientific evidence was in these cases an inadequate mandate for comprehensive change. The pressure to acknowledge the limits of science as a basis for decision making has been exerted perhaps most forcefully in the fields of medicine and clinical research, where the impact of science on individual human beings is so direct and apparent. As mentioned above, when the harms associated with clinical research and increasingly invasive technological treatments began to show up after World War II, the public demanded the imposition of new norms; these norms shifted authority for risk decisions from exclusive control by clinicians and scientists to greater control by those individuals bearing the risk. Science as Servant The ideal of normal science as independent inquiry motivated by disinterestedness and methodological skepticism has long provided the justification for science and scientists to be self-regulating. As the application of science in medical decision making and clinical research revealed, self-regulation failed to accommodate democratic norms of respect for the decision-making authority of patients and research subjects who bear the risk of scientific interventions. Many physicians and other scientists who adhere to the paradigm of professional self-regulation have resisted the imposition of other forms of accountability on their work. The motivations for resistance are often complex involving economic self-interest, the protection of the health and welfare of individuals and the public, and the preservation of independent science from political manipulation. Perhaps the most acute concern is that scientific professionals and thus science itself will become captive to interests that are hostile to the mission of science to further the bounds of certified knowledge. The problem, of course, is that the mission of science is itself contested and fundamentally shaped by the political and social projects which science is called to serve (11). In 1980, for example, when the passage of the Bayh–Dole Act allowed federally-funded researchers to obtain commercial patents and licenses for publiclyfunded research, the explicit rationale for the Act was “to promote free competition and enterprise” by facilitating the commercialization of science (12). Echoing this shift, a recent report on Integrity in Scientific Research by the National Academy of Sciences omitted the traditional norms of disinterestedness and collaboration in its criteria for scientific integrity, stating that researchers must “balance collaboration and collegiality with competition and secrecy” (13). Growing criticism of the commercialization of science (14–17) is evidence of an ongoing struggle precisely about the mission(s) of science and the norms that should govern it. Among other things, this struggle lends support to the strong view that the practice, the norms, and the claims of science are constructed on the basis of cultural, social, and ideological commitments. Norms and Post Normal Science When we consider ethical norms that should govern the conduct of science, we can’t assume that science functions in a vacuum and is therefore sufficient to oversee itself. Nor can we assume that the norms that have traditionally governed science are satisfactory to confront the complex social issues that confront us in the 21st century. Indeed, we have to assume that those norms have in some measure brought us to the crisis where we are now. As sociologist Ulrich Beck has observed, we now live in a “risk society” which has resulted in large part from the impact of science and technology on our world bringing

Knowledge, Norms, and the Politics of Risk

513

Decision Stakes

with it a host of toxic threats. Because risk assessments always involve normative judgments about the scope, nature, magnitude, scale, probability, and acceptability of harms, science has become both more necessary and less sufficient as a basis for social decision making (9). In contrast to the problems of normal science—problems and methods believed to be internal to scientific disciplines—the problems we face today have been described as problems of “post-normal science.” Funtowicz and Ravetz (Fig. 1) offer a functional account of science relative to the decision stakes and the degree of complexity involved. In normal science, where scientific expertise is (rightly or wrongly) regarded as fully sufficient to address fairly insular puzzle-solving, the decision stakes are low and the objects under consideration are deliberately simplified. In consultant science such as clinical research or practice, decision making has moved outside the stable, controlled conditions of a laboratory experiment, the stakes are higher, and decisions involve greater complexity and uncertainty. In post-normal science, Funtowicz and Ravetz explain, “facts are uncertain, values in dispute, stakes high, and decisions urgent” (18,20). In environmental policy making, for example, say Funtowicz and Ravetz, where the mass of details requires separate analysis and management relative to broad strategic goals such as “sustainability,” “nothing can be managed in a convenient isolation; issues are mutually implicated; problems extend across many scale levels of space and time; and uncertainties and value-loadings of all sorts and all degrees of severity affect data and theories alike.” Under these circumstances, they argue, “the traditional guiding principle of research science, the goal of achievement of truth or at least of factual knowledge, must be substantially modified. In post-normal conditions, such products may be a luxury, indeed an irrelevance. Here, the guiding principle is a more robust one, that of quality” (21). For Funtowicz and Ravetz, quality-control of policy-relevant science is embodied in what they call “extended peer communities,” an idea that moves beyond professional peer review to a more public and democratic process of problem-solving. As Ravetz describes it, the model of extended peer communities “extends the franchise” (20) of policy-relevant knowledge. This call for increased public participation in science-based policy and regulation echoes an earlier and forceful argument made by David Bazelon, Chief Judge of Court of Appeals for the D.C. Circuit. In a 1979 Science article on “Risk and Responsibility,” (22) Judge Bazelon, writing during the period of landmark legislation regarding openness in government (the Freedom of Information Act, the Federal Advisory Committee Act, the Open Meetings Law) observed that we are not only reluctant to delegate risk assessment to “so-called disinterested scientists.[but] the very concept of

Post-Normal Science Professional Consultancy Normal Science System Uncertainties

Figure 1

Three types of problem-solving strategies. Source: From Refs. 18, 19.

514

Sharpe

objectivity embodied in the word ‘disinterested’ is now discredited. . Since we have no shaman we must have confidence in the decision-making process” (22). Similarly, in describing the ethical implications of “risk society” Ulrich Beck observes “there is no expert on risk. there are always competing and conflicting claims, interests and viewpoints of the various agents of modernity and affected groups” (9). As a result, risk determinations can “no longer be isolated from one another through specialization, and developed and set down according to their own standards of rationality. They require cooperation across the trenches of disciplines, citizen’s groups, factories, administration, and politics” (9). Transparency and increased mechanisms of participation are norms central to this process and are reflected, for example, in the emergence of citizen’s groups to address issues of toxic pollution and environmental justice and in a variety of federal and state mandates on public participation in agency decision making (23–27). Feminist philosopher of science, Sandra Harding, offers “strong objectivity” as a feature of research that is more inclusive and therefore more responsive to social justice. Strong objectivity, she says, extends “the notion of scientific research to include systematic examination of.powerful background beliefs” that for example, downplay the ways in which minorities are disproportionately exposed to the risks of environmental pollution (8). Research that meets the conditions of “strong objectivity” will begin from the perspective of marginalized and systematically oppressed groups, she says, because these groups provide evidence that is routinely silenced by more powerful parties hoping, thereby, to secure their own interests. The lead industry provides an example of research that fails to meet this norm: when the lead industry’s own research indicated lead’s neurotoxic effects on children, for example, it shifted its research focus exclusively to adults and continued to conceal its findings regarding children (28). By contrast, “community-based research,” such as that conducted by epidemiologist Steve Wing on the health effects of radiation release and concentrated hog production, begins with the health concerns of the affected community (29–32). Strong objectivity, says Harding, is at the methodological level a form of “intellectual participatory democracy” that will improve our understandings and explanations by counting as evidence the claims of local knowledge that are traditionally excluded from the elite notion of scientific rationality. The upshot of these insights is that neither scientific evidence alone nor the institution of science itself gives authority to research and decision making. In a democracy, that authority must come from the quality of the research and the decisionmaking processes themselves.

NEW MANDATES IN POLICY-RELEVANT SCIENCE The Environmental Endocrine Hypothesis as a New Scientific Paradigm The environmental endocrine hypothesis—that synthetic estrogens found in plastics, pesticides, and other chemicals disrupt the body’s ability to deliver hormones to the bloodstream—provides an illustration of the foregoing ideas with respect to the field of toxicology. The paradigm of the monotonic dose-response curve reflecting exposure to a single chemical—the ancient principle that the dose makes the poison—has resulted in a methodological focus on acute toxicity to the exclusion of long-term low dose exposures and synergies among multiple chemicals. Likewise, widespread fears about cancer risk have, over the last 50 years, elevated cancer as the most significant endpoint in the study of

Knowledge, Norms, and the Politics of Risk

515

toxic chemicals. Until recently, this has caused scientists, institutions, funders, and policy makers to largely ignore the reproductive, neurophysiological, and behavioral effects of chemicals. In the early to mid-1990s, however, the growing evidence of endocrine disruption was deemed sufficient to prompt the passage of two landmark pieces of legislation: the Food Quality Protection Act (FQPA) (33) and the Safe Drinking Water Estrogenic Substances Screening Program Act to amend the Safe Drinking Water Act (34). Both statutes require, among other things, that the Environmental Protection Agency (EPA) screen and test chemicals for endocrine-disrupting properties and to screen for estrogenic chemicals in food and source drinking water. One of the most significant features of the FQPA and a reason for its support by industry groups was its abandonment of the “zero-tolerance standard” for food additives in favor of a “negligible-risk standard.” As Sheldon Krimsky describes in his book, Hormonal Chaos, on the science and policy implications of the endocrine disrupter hypothesis, since the 1958 passage of the Food and Drug Amendments, the pesticide industry had been bound by the Amendment’s “Delaney clause” which “required zerotolerance for food additives that had been shown to be human or animal carcinogens at any dose” (35). In 1992 the Natural Resources Defense Council sued the EPA on the grounds that the Delaney Act should be applied also to pesticide residues since they are a type of food additive. The success of NRDC’s case would have meant the removal of many agricultural pesticides from the marketplace. The shift from the precautionary “zero-risk” to the less stringent “negligible-risk” strengthens the link between toxicology and policy making by introducing a wider margin of toxic effects into risk assessment. This book on Clinical and Developmental Neurotoxicology is a reflection of the need to confront the challenges posed by the nexus of toxicology (and other sciences) and policy not only around the environmental endocrine hypothesis but also around uncertainty and risk more generally. This chapter is an attempt to explain both the normative context of policy-relevant science and the norms that should govern it.

Ethical Issues in Post-Normal Science There are ample resources on the ethical conduct of scientific research with regard to humans and animals, and much of this is codified in federal regulations. Such issues concern the scientific justification for research (whether the projected benefits of research outweigh the potential harms), informed consent, maximizing benefit and minimizing harm to research subjects, the fair distribution of the benefits, and harms of research. In addition, codes of ethics have been issued by scientific societies to address questions internal to scientific practice such as the sharing of research materials, assignment of authorship credit, fabrication of data, etc. Such standards and guidelines are also disseminated in mandatory research ethics courses for federally-funded researchers. My focus in what follows will not be on these issues, but rather on a set of problems that become prominent in “post-normal science,” that is, regulatory and policy-relevant science where “facts are uncertain, values in dispute, stakes high, and decisions urgent.” I do this with specific reference to the role that corporate interests can play in the collection, characterization, and dissemination of data on risk, the conflicts of interest faced by scientists engaged with interest groups, and the role of community participation in research and policy making.

516

Sharpe

CORPORATE CONSTRUCTION OF RISK AND CONFLICTS OF INTEREST The assessment, management, and communication of risk is a prominent feature of regulatory or policy-relevant science. Given what we know about the construction of scientific “facts,” it should come as no surprise that the interests of stakeholders play a significant role in shaping their construction of risk. For example, anti-regulatory industry groups tend to emphasize the resilience of risk-bearers, whereas public health groups emphasize primary prevention, that is, efforts to prevent risk exposure. As a result, in statistical calculations of risk, industry stakeholders are more tolerant of Type I errors, false negatives, that underestimate risk, whereas public health stakeholders are more tolerant of Type II errors, false positives that overestimate risk (35). The question of whether risk should be regulated often depends on public opinion and may at times be referred to the courts. Thus stakeholder groups attempt to construct risk to shape public opinion, policymaking, or the outcome of litigation (36). Most of the evidence on the deliberate manipulation of risk information comes from formerly confidential corporate documents revealed during litigation. The tobacco litigation is the most prominent example. Although there is no reason, in principle, to believe that other stakeholders, such as public health or environmental groups, do not use tactics to influence risk perception, such groups rarely have the financial resources or the institutional structures to do so at the level available to industry. As Bero has pointed out, “sponsoring, publishing and criticizing research are costly endeavors [and] corporate interest groups are more likely to have the resources to launch expensive, coordinated efforts. In contrast, public health groups, which tend to act independently, are less likely to gather these resources” (36). So when we think about the “extended peer communities” suggested by Funtowicz and Ravetz, or “stakeholder theory” in general as a framework for deliberative decision making, it is important to keep in mind that stakeholders have vastly different political and economic power. Documents revealed through legal discovery in cases against the manufacturers of tobacco, asbestos, and other hazardous substances reveal the variety of ways in which companies have borrowed the authority of science to create uncertainty about risk as a means of advancing their marketing, regulatory, and eventually defensive legal aims. The Council for Tobacco Research (CTR), for example, was established by the tobacco industry in 1954 after a scientist at Memorial Sloan Kettering Cancer Center reported evidence indicating the carcinogenicity of tobacco tars. The publicly stated aim of the CTR, which was managed through the public relations firm Hill and Knowlton, was to fund independent research that would examine “all phases of tobacco use and health,” stating that “we accept an interest in people’s health as a basic responsibility, paramount to every other consideration in our business” (38). Internal tobacco documents reveal a different story. After establishing the CTR, the tobacco industry soon created a Special Projects Division within its legal department. The purpose and effect of this strategy was to shield the scientific data produced by the Division behind attorney-client privilege, thus concealing any unwelcome data and hiding researchers ties to industry (38). The CTR also used its funding to secure compliance from scientists in the collection and interpretation of data. In 1968 the CTR signed an agreement with Mason Research in Worcester, MA, to evaluate “smoking machines” for animal inhalation studies. During the study, company lawyers came to the lab to ensure that only the most cancer-resistant rodents were used, and dictated other aspects of the study design. Although Mason researcher Miasnig Hagopian concedes that the industry was “directing the course of the

Knowledge, Norms, and the Politics of Risk

517

study” to the point where its objectivity was compromised, he admits that he and other researchers “looked the other way” in the hopes of a renewed contract (38). Another contract with researcher Freddy Homberger at Bio-Research Institute in Cambridge, MA, produced a high cancer incidence in Syrian rats. Under the conditions of the research contract, Homberger was required to submit his manuscript to the company for review prior to publication. “They were quite open about [the fact that they changed our funding from a grant to a contract] so they could control publication,” said Homberger. Under pressure from CTR lawyers, who did not want anything in published studies to be called “cancer,” Dr. Homberger changed the wording in the final proofs from “cancer” to “microinvasive tumor.” He later admitted that the lawyer had told him that “he would never get a penny more” if the paper was published without the change. Homberger’s interpretation of the Syrian hamster data was later used by the industry in the Rose Cipollone lawsuit—the first products liability lawsuit against the tobacco industry. Using Homberger’s paper, an industry expert witness stated that smoking had produced nothing but “microinvasive” tumors in the hamsters (38). A recent and similar example from the pharmaceutical industry concerns drug maker Pfizer’s corporate damage control after a major study published in the Journal of the American Medical Association indicated that expensive calcium channel blockers and angiotensin-converting enzyme (ACE) inhibitors, such as Pfizer’s Cardura, to treat high blood pressure were no better, and in some cases, more risky than inexpensive diuretics. To blunt the results of the study and prevent information about the findings from reaching physicians, Pfizer got the American College of Cardiology to soften the wording of an “alert” urging doctors to “discontinue” use of Cardura. In the reissued alert the wording was changed from “discontinue” to “reassess.” Pfizer has contributed more than $500,000 a year to the College of Cardiology in recent years (39). There are many examples of this type of activity from tobacco, lead, asbestos, chemical, food, and pharmaceutical industry-sponsored science (28,40–50). In addition, the passage of the Bayh–Dole Act in 1980 and the economic potential of biotechnology have resulted in a host of new incentives for scientists to join forces with industry in entrepreneurial activities. As a result, attention to financial conflicts of interest has dramatically increased over the last five years with scientific and medical bodies issuing new statements and standards.

CONFLICT-OF-INTEREST DISCLOSURE AND THE RIGHT-TO-KNOW In a now classic analysis, Dennis Thompson has defined a conflict of interest as a condition under which professional judgment about one’s primary interest (e.g., the integrity of research) may be unduly influenced by a secondary interest (e.g., financial gain) (51). Although financial interests are not the only factors that can inappropriately influence judgment, and financial ties to industry are not the only monetary ties that can be biasing, such conflicts are recognized as a troubling feature of the commercialization of science. As such, researchers’ and research institution’s financial ties to industry have become the focus of institutional and regulatory efforts to bolster the trustworthiness of science. Prohibition and management are two means of regulating financial conflicts of interest. Proposals to adopt standards prohibiting financial conflicts of interest have been the most controversial and the most difficult to establish. For example, only 1 out of the 10 major medical research universities in the U.S. has a policy banning researchers from holding stock, stock options, or company positions related to their research (52). When the New England Journal of Medicine established a policy prohibiting editorial authors from

518

Sharpe

having a financial stake in the matter on which they are commenting, there was strong criticism that the journal was engaging in censorship (53). Under the editorship of Jeffrey Drazen, who has himself received considerable research support from the pharmaceutical industry (54), the New England Journal of Medicine abandoned that restrictive policy in June, 2002, citing the difficulty of finding authors without industry ties. Because organizations, many of which have their own institutional conflicts of interest, have been reluctant to ban certain financial ties, management of conflicts has become the prevailing institutional response with disclosure of financial ties the principal conflict-management strategy. It should be obvious that disclosure alone cannot resolve a potentially biasing conflict-of-interest, but it is an essential first step in making transparent the financial circumstances under which research and its oversight takes place. That said, disclosure—as a necessary element of an open and transparent process—must be understood as disclosure to the public (in journals, on CVs, in public presentations, to advisory bodies, and the media) as opposed to disclosure or reporting to a closed, institutional oversight body. The American Association of Medical Colleges’ efforts to grapple with institutional conflicts of interest (55) and the General Accounting Office’s recommendations regarding the risks of such conflicts in human research (56) are an acknowledgement that oversight bodies themselves are doing an inadequate job of managing the potentially biasing financial affiliations of individual investigators because they also have significant conflicts of interest that compromise their ability to be watchdogs. As we have seen in the examples from tobacco industry-funded science, hope for continued funding can influence researchers in their characterization of findings, even if companies themselves do not explicitly require such control. Some of the pressure for such funding comes from academic institutions themselves, which may have their own financial stake in the outcome of faculty research. Because financial ties can shape how a scientist collects, characterizes, and disseminates data on risk, and how an interested institution oversees the conduct of research, disclosure of financial conflicts of interest has become a vital aspect of the public’s “right to know” and a hallmark of technically and ethically “good” science in the post-normal age (57). Beyond disclosure as a means of managing conflicts of interest, many professional bodies have also made it clear that in order to preserve the independence of research, investigators should not enter into agreements that cede control of research or its publication to sponsors. This is especially important in the academic setting, where there is still an expectation of free inquiry—an expectation that has been strategically used by industry to enhance the credibility of research findings. As more academic research institutions enter into research agreements with industry and facilitate entrepreneurial arrangements for their faculty, it becomes increasingly important to assure that contract and grant agreements are free from binding clauses that cede control of research. Likewise, in the service of transparency, university-sponsored research programs should make public all of the sources and amounts of funding that support research and researchers at their institutions. Conflict of interest in research and its oversight is only one among a host of issues made prominent by the high stakes of policy-relevant science. Sheldon Krimsky (35) has outlined an additional set of ethical issues that arise in this area. They concern (1) the release of uninterpreted data to individuals and communities; (2) the volatility of research hypotheses that threaten to reinforce ethnic, gender, etc. stereotypes; (3) the prepublication release of research; and (4) the evidentiary requirements for risk characterization. There is not enough space here to elaborate on each of these pressing issues but they represent concerns at the frontiers of research ethics in an era characterized by enormous

Knowledge, Norms, and the Politics of Risk

519

uncertainties about risk and tremendous financial and human stakes in how they are to be handled.

COMMUNITY-BASED RESEARCH One of the most pioneering and democratically-informed developments in recent scientific research is the emergence of “community-based research.” I will briefly describe the normative basis of this new approach and highlight some of the particular ethical challenges that it poses. My purpose is not to offer a comprehensive analysis of such investigation or its ethical complexities, but simply to offer this as an approach to scientific inquiry that takes seriously the possibility of scientific engagement to promote fundamental human rights. Unlike traditional academic research, which is driven by the intellectual interests of scientists, or industry-sponsored research, which seeks to develop or secure the marketability of products, community-driven or participatory research is oriented to the self-defined needs of a particular group. As one advocacy organization puts it, communitybased research is “coupled relatively tightly with community groups that are eager to know the research results and to use them in practical efforts to achieve constructive social change” (58). Community-based research might focus on environmental or public health initiatives, land-use planning, or epidemiologic surveys to ascertain information on disease incidence. Steve Wing, an epidemiologist at the University of North Carolina (UNC), has undertaken community-driven studies of industrialized hog production to determine (1) whether there are risk disparities in the siting of concentrated animal feed operations (CAFO) and (2) what the health effects of exposure to CAFOs might be (30). The siting issue was the basis of an environmental justice study whose questions originated with the exposed communities. That study was funded by the National Institutes of Environmental Health Sciences (NIEHS) to a partnership formed by the Concerned Citizens of Tillery, the Halifax North Carolina County Health Department, and the UNC School of Public Health. In the health effects study, Wing and his team collaborated with community-based organizations in three rural areas to develop and implement a questionnaire on respiratory and other symptoms potentially associated with proximity to hog CAFO. The rural health survey was funded by the North Carolina State Health Department. The findings from the environmental justice study indicated that CAFOs are disproportionately located in proximity to poor communities and communities of color. Hog production facilities were also concentrated in areas where most people depend on wells for drinking water. These finding were significant for the design and conduct of the subsequent health effects study because they reinforced the minority communities’ sense of disenfranchisement and suspicion toward institutions including health departments, universities, and researchers. An essential aspect of the health effects study—which depended on the participation of these community members—was, therefore, the establishment of trust and trustworthiness through a research process that guaranteed confidentiality to the participants and accurately reported results in a timely way. The finding from the health effects study indicated that “residents near the hog CAFO reported increased numbers of headaches, runny noses, sore throats, excessive coughing, diarrhea, and burning eyes” (30). Wing’s experience conducting high-stakes, state-funded university research to examine the possible negative effects on communities of one of the state’s largest and fastest growing industries represents a cautionary tale about conflicts of interest,

520

Sharpe

confidentiality, and the scope of institutional review boards—three issues central to the quality of a research process. Hog production in North Carolina has grown approximately 300% in the last 15 years. The industry has economic ties and governance roles at the state’s research universities and ties at all levels of government in North Carolina and the U.S. Congress (59). These affiliations create pressure on researchers whose work potentially threatens the industry. Pressure comes from the industry itself in the form of threatened lawsuits, from academic administrators and superiors concerned about funding, and from researchers themselves who self-censor to avoid trouble. For example, the pork industry response to Wing’s health effects study findings was swift. An industry representative, who was also a member of the UNC Board of Governors, called Wing to explain that good business requires CAFO siting on the cheapest land. The UNC Vice-Chancellor set up a meeting with Wing indicating concern about an upcoming presentation of study data before the North Carolina General Assembly—which votes on UNC appropriations. As one researcher told Wing, the legal threats from the hog industry caused this untenured faculty member to drop controversial hog research. Likewise, North Carolina State researcher JoAnn Burkholder, whose pioneering work has implicated run-off from hog farming in widespread aquatic pollution, has generated industry backlash such that Burkholder advises untenured colleagues to stay away from politically unpalatable research (60). In addition to discouraging researchers, such pressure can also threaten the confidentiality of informants, thereby producing a chilling effect on individuals’ and communities’ willingness to participate in research. In response to Wing’s findings, attorneys from the North Carolina Pork Council wrote to the researchers requesting, under the North Carolina Public Records Law, documents from the health study including “the identities.of persons interviewed” and maps of the study locations. As Wing points out, the “North Carolina Public Records Statute does not protect documents collected in the course of research involving human subjects” (30). As such, it is up to researchers, and their universities—which may have their own institutional conflicts of interest—to safeguard participant confidentiality in publicly-funded research. Another structural obstacle in safeguarding confidentiality in such research is the traditional scope of institutional review boards. Such boards, which evolved in response to biomedical research on individual human subjects, tend to focus on potential research harms to individuals directly participating in research. In the context of environmental research, however, entire communities can be placed at risk by research results that are adverse to powerful industries that operate in those communities. As IRBs take responsibility for reviewing an increasing number of public health research protocols involving epidemiological and toxicological methods, the question of risk to communities will become increasingly important.

CONCLUSION Policy-relevant science takes place within a context of extremely high stakes: high human and environmental stakes—when proposed or existing activities of industry or government pose potential threat or benefit—often disproportionately—to particular populations; high political stakes—when elected representatives must make controversial policy decisions; and high economic stakes—when industries and others stand to gain or lose by virtue of science-based regulation. The ethical stakes in policy-relevant science are also extremely high. In the face of often significant uncertainty about risk and opposing values and

Knowledge, Norms, and the Politics of Risk

521

interests of stakeholders, the process by which science contributes to the policy process becomes all-important. In this paper I have argued that an ethical process—one that is accountable to public norms of transparency, respect, and fairness—entails candor about the context in which science is produced. This includes disclosure about the financial ties and potentially biasing conflicts of interest behind scientific research as well as explicit efforts to democratize the generation of scientific knowledge by initiating research with and from the point of view of traditionally disenfranchised and marginalized groups. These two ingredients of an ethically informed practice of science are especially important as science and the norms governing it face increasing pressure to conform to the norms of the marketplace: secrecy, entrepreneurial research directed at marketable products, and risk characterizations that are favorable to industry interests.

ACKNOWLEDGMENTS I would like to thank Jill Schneiderman, Darryl Farber, and David Bellinger for their very helpful comments on earlier drafts of this paper.

REFERENCES 1. Rothman DM. Strangers at the Bedside: A History of How Law and Bioethics Transformed Medical Decision-Making. New York: Basic Books, 1991. 2. Sharpe VA, Faden AI. Medical Harm: Historical, Conceptual and Ethical Dimensions of Iatrogenic Illness. New York: Cambridge University Press, 1998. 3. Trials of War Criminals before the Nuremberg Military Tribunals under Control Council Law No. 10. Washington, D.C.: U.S. Government Printing Office, 1949; 2 pp. 181–182. (Accessed February 15, 2004, at http://ohsr.od.nih.gov/nuremberg.php3). 4. Kuhn T. The Structure of Scientific Revolutions. Chicago: University of Chicago Press, 1962. 5. Merton RK. The normative structure of science. In: Merton R, Storer NW, eds. The Sociology of Science. Chicago: University of Chicago Press, 1973:267–278. 6. Douglas M, Wildavsky A. Risk and Culture. Berkeley: University of California Press, 1982. 7. Douglas M. Risk and Blame: Essays in Cultural Theory. London: Routledge, 1992. 8. Harding S. Whose Science? Whose Knowledge? Thinking from Women’s Lives. Ithaca: Cornell Univ. Press, 1991. 9. Beck U. Risk Society: Towards a New Modernity. London and Newbury Park, CA.: Sage Publications, 1992. Translated by Mark Ritter. 10. Longino H. Science as Social Knowledge. Princeton, NJ: Princeton University Press, 1990. 11. Merton RK. Science and the social order. Philosophy of Science 1938; 5:321–337. 12. Patent and Trademark Laws Amendments of 1980 (the Bayh-Dole Act) (35 USC 200–212). (Accessed on May 20, 2004, at http://www4.law.cornell.edu/uscode/35/200.html). 13. Institute of Medicine. Integrity in Scientific Research. Washington, DC: National Academies Press, 2002. 14. Center for Science in the Public Interest Project on Integrity in Science. See the website for the Integrity in Science Project at www.integrityinscience.org. 15. Rosenstock L, Jackson LL. Attacks on science: The risks to evidence-based policy. Am J Public Health 2002; 92:14–18. 16. Bok D. Universities in the Marketplace: The Commercialization of Higher Education. Princeton, NJ: Princeton University Press, 2003. 17. Krimsky S. Science in the Private Interest: Has the Lure of Profits Corrupted Biomedical Research? Lanham: Rowman & Littlefield Publishers, 2003.

522

Sharpe

18. Funtowicz SO, Ravetz JR. Three types of risk assessment and the emergence of post normal science. In: Krimsky S, Golding D, eds. Social Theories of Risk. Westport, CT: Praeger, 1992:254. 19. Dempster B. Toward a Post-Normal Science: New (?) Approaches to Research, 1998. (Accessed May 21, 2004, at http://www.fes.uwaterloo.ca/u/mbldemps/meth/pnsresearch/ index.html#Science). 20. Ravetz JR. What is post-normal science? Futures 1999; 31:647–653. 21. Funtowicz S, Ravetz J. Post-normal science–environmental policy under conditions of complexity, 2000. (Accessed February 16, 2004, at http://www.nusap.net/sections. php?opZviewarticle&artidZ13). 22. Bazelon DL. Risk and responsibility. Science 1979; 205:277–280. 23. Federal Actions to Address Environmental Justice in Minority Populations and Low-Income Populations: Executive Order 12898, February 11, 1994, at http://www.fs.fed.us/land/envjust. html. 24. Pellerano MB. Public participation—Part 1 #782 (January 29, 2004), Public participation-Part 2, #783 (January 22, 2004) Rachel’s Environment & Health News. (Accessed February 26, 2004, at http://www.rachel.org/bulletin/index.cfm?StZ3). 25. Lee C, ed. Proceedings of the First National People of Color Environmental Leadership Summit. New York: United Church of Christ Commission for Racial Justice, 1992: xiii. The Principles of Environmental Justice are at http://www.rachel.org/library/getfile.cfm?IDZ153. 26. U.S. Environmental Protection Agency. Public Involvement Policy of the U.S. Environmental Protection Agency [document number EPA 233-B-03-002]. Washington, D.C.: U.S. Environmental Protection Agency, May 2003, at http://www.rachel.org/library/getfile. cfm?IDZ319. 27. Rutledge P, Barnes JA, Benavides T, et al. Addressing Community Concerns: How Environmental Justice Relates to Land Use Planning and Zoning. Washington, D.C.: National Academy of Public Administration, 2003. (Accessed May 20, 2004, at http://www.napawash. org/Pubs/EJ.pdf and http://www.rachel.org/library/getfile.cfm?IDZ320). 28. Markowitz G, Rosner D. In: Deceit and Denial: The Deadly Politics of Industrial Pollution. Berkeley, CA: University of California Press, 2002:46. 29. Wing S, Wolf S. Intensive livestock operations, health, and quality of life among eastern North Carolina residents. Environ Health Perspect 2000; 108:233–238. 30. Wing S. Social responsibility and research ethics in community driven studies of industrialized hog production. Environ Health Perspect 2002; 110:437–444. 31. Wing S. Objectivity and Ethics in Environmental Health Science. Environ Health Perspect 2003; 111:1809–1818. 32. Sharpe VA. What are we doing when we’re doing human subjects research? Neurotoxicol Teratol 2002; 24:451–454. 33. U.S. Environmental Protection Agency. Food Quality Protection Act, Public Law 104–170, at http://www.epa.gov/oppfead1/fqpa/gpogate.pdf and background information http://www.epa. gov/oppfead1/fqpa/backgrnd.htm. 34. U.S. Environmental protection Agency. Safe Drinking Water Act Amendments of 1996, Public Law 104–182. (Accessed May 20, 2004, at http://www.epa.gov/safewater/sdwa/text.html and background information http://www.epa.gov/safewater/sdwa/summ.html). 35. Krimsky S. In: Hormonal Chaos: The Scientific and Social Origins of the Environmental Endocrine Hypothesis. Baltimore, MD: Johns Hopkins University Press, 2000:73. 36. Bero LA. The role of special interest groups in influencing data on risk. Delivered at the World Health Organization and United Kingdom Department of Health Consultation on Risks to Health, London, 23–24 October, 2001. 37. Montini T, Bero LA. Policy makers’ perspectives on public health advocates’ roles in regulation development. Tobacco Control 2001; 10:218–224. 38. Freedman AM, Cohen L. Smoke and mirrors: how cigarette makers keep health question ‘open’ year after year. Wall Street Journal 1993;A1. February 11. 39. Lenzer J. Spin doctors soft pedal data on antihypertensives. BMJ 2003; 326:170.

Knowledge, Norms, and the Politics of Risk

523

40. Barnes DE, Bero LA. Industry funded research and conflict of interest: an analysis of research sponsored by the tobacco industry through the Center for Indoor Air Research. J Health Policy Law 1996; 21:515–542. 41. Bero LA, Rennie D. Influences on the quality of published drug studies. Int J Technol Assess Health Care 1996; 12:209–237. 42. Blumenthal D, Campbell EG, Anderson MS, Causino N, Louis KS. Withholding research results in academic life science: Evidence from a national survey of faculty. JAMA 1997; 277:1224–1228. 43. Brennan TA. Buying editorials. N Engl J Med. 1994; 331:673–675. 44. Choudhry NK, Stelfox HT, Detsky AS. Relationships between authors of clinical practice guidelines and the pharmaceutical industry. JAMA 2002; 287:612–617. 45. Horton R. Lotronex and the FDA: a fatal erosion of integrity. Lancet 2001; 357:1544–1545. 46. Mundy A. Dispensing With the Truth: The Victims, the Companies and the Dramatic Battle over Fen-Phen. New York: St. Martin’s Press, 2001. 47. Nestle M. Food Politics: How the Food Industry Influences Nutrition and Health.: University of California Press, 2002. 48. Petersen M. Madison Ave. has growing role in the business of drug research. New York Times, November 22, 2002. 49. Petersen M. Making drugs, shaping the rules. New York Times, February 1, 2004. 50. Rennie D. Thyroid storm. JAMA 1997; 277:1238–1243. 51. Thompson D. Understanding financial conflicts of interest. N Engl J Med. 1993; 329:573–576. 52. Lo B, Wolf LE, Berkeley A. Conflict-of-interest policies for investigators in clinical trials. N Engl J Med. 2000; 343:1616–1620. 53. Marshall E. Journals joust over conflict-of-interest rules. Science 1997; 276:523–525. 54. Cauchon D. A very real interest in new drugs Researchers’ financial ties to industry are big concern in academia. USA Today 2000; 3A. Johannes L. New England Journal of Medicine appoints Drazen as Editor in Chief. Wall Street Journal, May 12, 2000 p 1. 55. Task Force on Financial Conflicts of Interest in Clinical Research. Report on Institution’s Financial Interests in Human Subjects Research. Washington, DC: American Association of Medical Colleges, October, 2002. (Accessed May 20, 2004, at http://www.aamc.org/members/ coitf/2002coireport.pdf). 56. U.S. General Accounting Office. Biomedical Research: HHS Direction Needed to Address Financial Conflicts of Interest. Report No. GAO-02-89, November 2001, (Accessed May 20, 2004, at http://www.aau.edu/research/GAO.pdf). 57. Sharpe VA. Oversight, disclosure, and integrity in science. In: AAS Science and Technology Policy Yearbook, 2003. Washington, DC: AAAS. (Accessed March 15, 2004, at http://www. aaas.org/spp/yearbook/2003/yrbk03.htm). 58. Sclove RE, Scammell ML, Holland B. Community-based research in the United States: an introductory reconnaissance, including twelve organizational case studies and comparison with the Dutch science shops and the mainstream research system. Amherst MA: The Loka Institute, 1998. (Accessed March 8, 2004, at www.loka.org/crn/pubs/execsumm.htm). 59. Stith P, Warrick J, Sill M. Boss hog: North Carolina’s pork revolution. Raleigh News and Observer, February 19,21,22,24, 26, 1995. (Accessed March 15, 2004, at http://www.pulitzer. org/year/1996/public-service/works/about.html). 60. Gannett News Service. Scientist stirred up pork industry wrath. December 9, 2002. (Accessed May 20, 2004, at http://www.ncsu.edu/news/dailyclips/1202/121202.htm#DJ5).

Index

Abortion. See spontaneous abortion Absorption, neurotoxic metals, 232–233 Academic performance, 115–116, 156–158, 182, 344 Acetaldehyde, 4–5 Acetyl cholinesterase (AChE), 254–256, 276 Acrodynia, 12 Aculturation, neurobehavioral tests, 310 Acute lymphoblastic leukemia (ALL), 131–148 Adaptive mechanisms, 183–184, 508 Adjusted regression coefficients, 386 Adolescence, 17, 27, 32, 155–157, 327 Adopted children, 211 Adulteration of samples, 291–293 Adults, 4, 142, 309, 341–359, 495 AEDs. See antiepileptic drugs Affect, 72–73, 425–429 Age groups, 435–436, 439 Agency for Toxic Substances and Disease Registry (ATSDR), 471, 485 Age-related effects, 138, 307–308, 343 Aggregate-level studies, 401–402 Aggression, 425–429 Aging, 14–15, 17–18 Alcohol, 37, 169–191, 325. See also fetal alcohol spectrum disorders Alkaloids, 133 Alkylating agents, 133 ALL. See acute lymphoblastic leukemia Alzheimer’s disease, 341–342, 344–345, 351–352 American medical profession, 511–512 ‘Ancient’ toxin tolerance, 508 Angiotensin-converting enzyme (ACE), 519 Animal cognitive studies Alzheimer’s disease/early life lead exposure, 345 antiepileptic drug teratology, 106, 107

[Animal cognitive studies] cocaine teratology, 201–208 critical period of exposure, 343 cross-species comparisons, 417–418 cumulative learning, 424–425 early toxicant exposure, 417–447 environmental enrichments, 436–438 executive functions, 421–424 fetal organic solvent exposure, 89 multigenerational testing, 488–489 neglected function, 421 Parkinson’s disease risks, 346 ‘representative’ learning tasks, 419–421 sensitivity, 418–429 specific cognitive functions, 430–435 task selection, 418–435 tobacco, 153–154 Animal feed, 521–522 Anthracyclines, 133 Antiepileptic drug teratology, 103–130 Antimetabolites, 133–134 Anti-social behavior, 425–429 Anti-thyroid activity, 449–463 Anxiety, 426 Apical tests, 494 Arousal regulation, 200, 209–210, 217 Arsenic, 231, 232 Asparaginase, 132–134, 138 ATSDR. See Agency for Toxic Substances and Disease Registry Attention cocaine teratology, 200, 207–210 complex constructs to explain, 72 deficit disorders, 153, 155–156, 176–178 tasks, 432–433 Auditory processing, 156–157 Automated address matching, 403 Automated operant tasks, 435, 439 525

526 BAEPs. See brainstem auditory evoked potentials Banked biological samples, 348–349 Barbiturates, 107–116 Batteries of cognitive tasks, 420–421 Bayh–Dole Act, 506, 514, 519 Bayley Scales of Infant Development (BSID), 35, 305–307, 494 Behavioral Assessment and Research System (BARS), neurobehavioral tests, 312 Behavioral phenotypes, 429–430, 439 ‘Behavioral signature’, lead neurotoxicity in children, 71 Behavior aspects children’s unique patterns, 489 cocaine, 197, 206–208, 211–216 delayed methylmercury poisoning, 13–14 fetal alcohol spectrum disorders, 183–185 neurobehavioral test batteries, 303–320 neuroimaging role, 321–322, 327–332 organic solvent toxicity, 89 parental smoking effects, 149, 154–158 PCB exposure, 44–49 teratology study concepts, 194–197 Bench Mark Dose (BMD), 8, 58, 389–392 Bergland, Fredrik, 3 Best Pharmaceuticals for Children Act, 469 Bias, 172, 349, 374, 507 Binge drinking behavior, 173, 186 Bioconcentration, 3 Biological monitoring, 86–87 Biomarkers. See exposure biomarkers Biomonitoring, 234–248, 258, 264, 272–275 Blinding, 308 Block groups, 402 Blood-brain barrier, neurotoxic metals, 232 Blood levels insecticide exposure assessment, 264, 265, 273–275 lead in children, 67, 73 mercury, 8, 240–241, 245–248 methylmercury, 382–387, 389–390 neurotoxic metals, 235, 237, 240–241, 245–248 BMD. See Bench Mark Dose Body measurements, 104 Bone lead, 70 Bonferroni method, 382 Boston Naming Test (BNT), 35, 328–329, 331–332, 383, 387–389 Brain activation patterns, 323, 327–332 damage, 321, 325–327

Index [Brain] development, 342–344, 451–456 metabolism, 325–326 Brainstem auditory evoked potentials (BAEPs), 27, 36 Brazelton Neonatal Behavioral Assessment Scale, 307 Breast milk, 51, 53–57, 258, 264 BSID. See Bayley Scales of Infant Development Built environment, 397–402 Byers, Randolph, 502 Cadmium, 231 CAFO. See concentrated animal feed operations Calcium channel blockers, 507, 519 California Verbal Learning Test (CVLT), 383, 387–389 Calomel, 12 Cambridge Neuropsychological Automated Battery (CANTAB), 72 Cancer terminology, 518–519 Cannabinoid testing, 292 Cannabis, 287–302 CANTAB. See Cambridge Neuropsychological Automated Battery Carbamates, 253–257, 274 Carbamazepine (CBZ), 106, 107, 109, 115, 118–119, 122 Cardiac effects, 107–109 Caregiving interactions, 212–213 Carson, Rachel, 3 Catecholamines, 200–201 Causality, 362–363, 502–506 CBCL. See Child Behavior Check List CBZ. See carbamazepine Census-block groups, 401–402 Central nervous system (CNS), 87–88, 107–108, 122, 131–132, 169 CHAMACOS birth cohort study, 259 ‘Change-in-estimate’ approach, 370, 372, 374 Chemical disasters, 3–6 Chemical risk modeling, 403 Chemical toxicant exposure, 395–416 Chemistry of organic solvents, 83–84 Chemotherapy, 131–148 Child Behavior Check List (CBCL), 48, 157–158 Children clinical research perspective, 485–499 differences from adults, 495 insecticide exposure, 254, 261–263

Index [Children] lead neurotoxicity, 67–82 lead toxicity, 502 leukemia chemotherapy, 131–148 Parkinsonism, 345–346 PCBs effects on brain development, 453–456 prenatal methylmercury exposure, 381–394 public health burden of learning disabilities, 465–466 Chisso factory, Japan, 4 Choline, fish, 9 Chronic disease, 487–488 Clarkson, Thomas W., 6 Classes of organic solvents, 84–86 Cleft palate, 107 Clinical exposure assessment, 233 Clinical practice, 485–499 CNS. See central nervous system Cocaine animal cognitive function studies, 422–423, 426, 433 extent of use, 197–199 maternal biological measures, 291–294 maternal reporting, 288–291 mother–child relationship effects, 211–216 neonatal biological measures, 294–296 neuroimaging, 325 prenatal exposure, 193–229 prevalence, 288 teratogenic mechanisms, 199–211 in utero exposure assessment, 287–302 Co-exposure, 402–403 Cognition-related executive skills, 181 Cognitive aspects animal tests, 417–447 antiepileptic drug teratology, 107, 110–124 cocaine teratology, 210–211 early life exposure effects on adults, 341–342, 346–347, 353 fetal alcohol spectrum disorders, 173–174, 178–179 leukemia therapy outcomes, 133, 135–138 parental smoking, 150, 156–158 ‘Cognitive reserve’ hypothesis, 344–345 Cohort studies, 25–29, 245–246 Color discrimination, 95–96 Commercialization of science, 514 ‘Commonly used’ covariates, 364–367 Communalism (scientific), 512 Community-based research, 521–522 Complex decision making, 513–514 Computer-based neurobehavioral tests for children, 309, 312–313

527 Computerized tomography (CT), 322, 326 Concentrated animal feed operations (CAFO), 521–522 Concentration–response/concentration–effect relationship, 73–74 Conduct disorders, 155–156 Confidence levels, 362–363 Conflicts of interest, 518–521 Confounding variables alcohol effect studies, 171, 172 biological susceptibility, 404–406 early life exposure effects on adults, 349–350 human neurotoxicological studies, 361–378 lead studies, 75–77, 503–504 methylmercury, 9–11, 36–38 Monte-Carlo studies, 370–371 neurobehavioral tests for children, 308–309, 315 PCB exposure studies, 51–54, 59 social/chemical risk interaction, 404 social-ecological perspective, 402–403 Congeners, 455, 456 Congener-specific analysis, 49–51, 59 Congenital malformations risk, 91–93 Contamination of samples, 292, 293 Context, social-ecological, 395–416 Continuous variables, 367–368 Continuum of effects, 105 Control, 361–378 groups, 140–141, 304, 306 variables, 367–368 See also confounder control Cord blood exposure assessment, 477 insecticides, 258 mercury levels, 8 methylmercury levels, 382–387, 389–390 neurotoxic metals, 236, 237, 241, 245–248 Cortical atrophy, 134 Cortical morphology, 201–202 Corticosteroids, 132–134, 137–138 Council for Tobacco Research (CTR), 518–519 Covariates, 363, 368–375, 402–403. See also confounding variables Crack cocaine, 198 Cranial irradiation therapy (CRT), 131–133, 135, 139, 143 Cranio-facial abnormalities, 104, 107–109, 116–118, 121, 169, 174–175 Creatine, 70, 274 Criteria for entry of covariates, 369–375

528 ‘Critical Developmental Windows for Exposure’ concept, 341 Critical exposure periods, 342–344, 488–489 Critical outcomes, 71–73 Cross-cultural issues, 309–310, 314–315 Cross-sectional studies, 29–32, 36–38, 304 Cross-species comparisons, 417–418 CRT. See cranial irradiation therapy CT. See computerized tomography CTR. See Council for Tobacco Research Cultural issues, 309–310, 314–315 Cumulative learning, 424–425 Cumulative risk assessment, 396 Cytochrome P-450 system, 489, 491 DA. See dopamine ‘Dancing disease’, 4 DDE. See p,p’-dichlorodiphenyldichloroethane DDST. See Denver Developmental Screening Test DDT. See p,p’-dichlorodiphenyltrichloroethane Deciduous teeth, 235, 348–349 Delayed non-matching-to-sample task (DNMS), 419–420, 436 Delayed toxicity, 5, 13–16 Delinquency, 185 Democratic norms, 511–512 Demographics, 198–199, 308, 365–366 Dental amalgams, 1, 11–12 Denver Developmental Screening Test (DDST), 35, 494 Depakote (valproic acid), 122 Derived variables, 402 Detoxification enzymes, 257 Dexamethasone, 134 Diaries, 239, 272 p,p’-Dichlorodiphenyldichloroethane (DDE), 53–54, 258–259 p,p’-Dichlorodiphenyltrichloroethane (DDT), 54, 254, 257, 258 Dichotomous variables, 367–368 Diet, 2–13, 18, 25–42, 233–234, 239–240, 365, 407–408 Difference versus deficit, 142 Digital hypoplasia, 104, 107–109, 116, 120, 121 Digit Cancellation task, 47 Dilantin, 107 Dioxin-like PCB congeners, 49–50 Dioxins, 48, 53 ‘Dirty dozen’ chemicals, 480–481 Disasters, 3–6, 43

Index Disclosure in conflicts of interest, 519–521 Disease patterns in children, 487–488 Disinterestedness (scientific), 512, 514 Distress, 406 Distributions, 75 DNMS. See delayed non-matching-to-sample task Domain-specific testing, 306–308 Dopamine (DA), 200–203, 206–207, 209, 210 Dopamine receptors (D1/D2), 201, 203, 206–207, 210 Dopaminergic pathways, 151 Dose, 389–392, 474 Dose–response relationships cocaine effects, 195, 211 fetal alcohol spectrum disorders, 172–173 neurotoxic metals, 237–239, 247–248 pediatric risk assessment, 492–493 potential neurotoxins, 467–474 pregnancy effects of insecticides, 260 structural equation models, 381–394 thyroid hormone, 453–454 ‘Double dissociation’, 71 Drinking water exposure, 233–234 Drug abuse, 150, 325–326. See also cannabis; cocaine Drug–drug interactions, 135 Drug testing, 291–296 Duplicate diet studies, 240 Dysmorphic fetal alcohol spectrum disorders, 169–170 Early toxicant exposure, 417–447 Ecological level, 395–416 Economics, 501–509 EDSTAC. See Endocrine Disruptor Screening and Testing Advisory Committee Educational performance, 115–116, 156–158, 182, 344 EE. See environmental enrichment Effect, 387–389, 490. See also confounding variables; exposure–effect relationship Effective internal dose, 343 Elderly people, 341–359 Emotion, 72–73, 181, 425–429 Empirical approach, 369 Encephalopathy, 67 Endocrine disruption, 48, 449–463, 516–517 Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC), 449–450 Endocrine Disruptor Screening and Testing Program (EDSTP), 469

Index Endpoints, 450–451, 494 Entry of covariates, 369–375 Environmental endocrine hypothesis, 516–517 Environmental enrichment (EE), 436–438 Environmental exposures, 326–332 Environmental Protection Agency (EPA), 468–469, 501 Epidemiological studies control aspects, 361–378 goals, 470–474 infertility risk from organic solvent exposure, 90 lead, 503–505 neurotoxicity evaluation issues, 501–509 neurotoxic metals, 233 PCB effects on neurophysiological function, 43–66 pediatric disease patterns, 488 structural equation models, 381–394 Epilepsy, 103–130 Epipodophyllotoxins, 133 Epistemic issues, 501–509 Error reactions, 426–428. See also measurement error Ethanol, 169–191, 325 Ethical issues, 511–525 Evaluation of neurotoxicity, 501–509 Examiners, 308, 311–312 Executive functions, 180–181, 210, 421–424 Experimental/non-experimental studies comparison, 362–363 Exposure assessment, 474–478 biomarkers, 68–71, 75–76, 348–349, 352, 490–491 insecticides, 263–275 methylmercury neurotoxicity studies, 32–33 neurotoxic metals, 231–251 phase in pediatric risk assessment, 493 Exposure–effect relationship, 57–58 Exposure error, 391–392 Exposure estimation measurement error, 385–386 Exposure misclassification, 236, 241 Exposure to ethanol (ETOH), 325 Extensions, 384–385 Facial anomalies, 104, 107–109, 116–118, 121 FAE. See fetal alcohol effect Fagan Test of Infant Intelligence (FTII), 47, 307, 473 False negative (Type II) errors, 171, 502–505, 508, 518

529 False positive (Type I) errors, 502–505, 508, 518 Family structure, 37, 397 Faroe Islands study, 6–8, 10, 17, 245–247 DDE exposure, 53 Imaging Study, 327 neurotoxicity studies, 26–27, 29, 33, 35, 39 PCB exposure, 48, 51–52 prenatal exposure, 381–394 whaling record relationship, 348 FAS. See fetal alcohol syndrome FASD. See fetal alcohol spectrum disorders Feces analysis, 235 Fecundability ratio (FR), 90 Female-dominated occupations, 84, 94 Fetal alcohol effect (FAE), 94, 169, 186 Fetal alcohol spectrum disorders (FASD), 169–191, 325 Fetal alcohol syndrome (FAS), 169, 172, 186 critical period, 343 methylmercury comparison, 2 neuroimaging, 325 organic solvent effects comparison, 94 Fetal anticonvulsant syndrome, 116 Fetus antiepileptic medications, 103–130 cannabis/cocaine, 287–302, 325 hydantoin syndrome, 116 insecticides, 254, 256, 258–261, 264 maternal smoking, 151–153 mercury vapor effects, 11–13 metals, 234–248 methylmercury, 5–6 organic solvents, 83–84, 86–99 Parkinson’s disease risk, 345–346 Financial interests, 519–521 Finger tapping, 329, 332 First trimester, 290 Fish. See seafood FMRI. See functional magnetic resonance imaging Food retrieval behavior, 272 FR. See fecundability ratio Freshwater fish, 25–42 Frontal cortex, 423–424 FTII. See Fagan Test of Infant Intelligence Functional domains, 306–308 Functional form, 73–74 Functional magnetic resonance imaging (fMRI), 323, 327–332 Gasoline, 505, 507, 508 Gastrointestinal malformations, 108 GCs. See glucocorticoids

530 Gender-related effects, 84, 94, 138 Gene–environment interaction, 491 Genetic factors drug susceptibility, 106, 107, 124 PCBs effects, 455 polymorphisms as lead toxicity confounders, 77 smoking, 151 social-ecological perspective, 407–408 thyroid hormone, 452, 455 Genitourinary anomalies, 108 Geographic information system (GIS) technology, 403 German study, 44–52, 55 GIS. See geographic information system technology Global measures, 305–307 Glucocorticoids (GCs), 134 Glue sniffing, 96–97 Growth deficiency, 107–109, 116–121, 151–152, 169 Hair maternal drug testing, 293 mercury concentrations, 27–33 methylmercury levels, 6–7, 382–387, 389–390 neonatal drug testing, 295–296 neurotoxic metals, 235–237, 241–243, 245–247 structure differences, 33 Hair-to-blood ratio, 33 Halogenation, 88 Harding, Sandra, 513, 516 Hashish. See cannabis Hazard identification, 467–474, 491–492 Head circumference, 152–153. See also microcephaly Health Effects Test Guidelines for Developmental Neurotoxicity testing, 419 Health outcomes, 400–401 Heart defects, 108 Heterogeneous treatment protocols, 140 Hill’s criteria, 505–506 Hippocampal function, 420 Hispanic children, 309–310, 312, 315 Historical information, 3, 263, 502 Hog production, 521–522 Home environment, 365–366, 368 Homelessness, 396–399 Home Observation for Measurement of the Environment (HOME) scale, 309 HOME score, 48 Hooper visual organization test, 328–330

Index Hormone replacement therapy (HRT), 361 Housing, 397–399 HPT. See hypothalamic–pituitary–thyroid axis HRT. See hormone replacement therapy Hydantoins, 107, 116 Hygiene monitoring, 86–87 Hyperactivity disorders, 157, 176–178 Hypoplasia, 104, 107–109, 116–121 Hyporeflexia, 258 Hypothalamic–pituitary–thyroid (HPT) axis, 450, 451 Hypothyroidism, 453, 454 Hypoxia, 199 ‘Identical’ cognitive tasks, 435–436, 439 Immune modulators, 133 Independent variables, 363 Indoleamines, 200–201 Industrial interests, 503–505, 507 Industrial solvents, 83–84, 86–99 Infants characteristics, 365 dysmorphological characteristics, 104 neurobehavioral test batteries, 307 parental smoking effects, 154–155, 157 See also children; neonates Infectious diseases, 346 Infertility risk, 90 Inhalant abuse, 96–97 Inhaled mercury vapor, 1, 11–12 Inorganic mercury, 11–13, 232 Insecticides, 253–285 Integrated exposure uptake biokinetic (IEUBK) models, 238–239 Intellectual decrements, 104–105, 149, 156–158 International cooperation, 480–481 International harmonization, 470 Intertrial interval (ITI), 431 Interviews, 288–291 Intrathecal chemotherapy, 131–148 Inuit study, 52, 54 IQ animal cognitive tests, 422 fetal alcohol spectrum disorders, 174, 182 global measures, 305 human neurotoxicological study covariable, 366, 368, 372 intellectual functioning relationship, 182 lead toxicity studies in children, 71 methylmercury, 2–3, 39 parental smoking effects, 155, 157 PCB exposure studies, 49, 58–59, 472 preschool lead exposure, 403

Index Iraq, 3, 5–6 Isoforms, 451–453 ITI. See intertrial interval Japan, 43 Kaufman Assessment Battery for Children (K-ABC), 44, 48, 50, 57, 307 Kaufman Brief Intelligence Test (K-BIT), 313 Kennard, Margaret, 13–14 Knockout mice, 452–453, 456 Knowledge, 511–525 Lactational exposure, 488 Language issues, 174–176, 309–310, 314–315 Late neuropsychological effects, 135–138 Latent variables, 384–385, 388 Laties, Victor, 501 Lead age-related poisoning dosage, 343 Alzheimer’s disease relationship, 345 childhood poisoning/cognitive impairment in adults link, 346–347 children, 67–82 confounding variables, 75–77 critical outcomes choice, 71–73 critical period of exposure, 343 exposure assessment, 231–239 exposure biomarkers, 68–71 history of understanding, 502 neuroimaging, 326 parental smoking comparison, 157–158 PCB study confounder, 52 postnatal exposure, 370–371 preschool exposure, 404–405 reversibility of effects, 74–75 schizophrenia link, 344 social-ecological perspective, 402–403 Lead industry, 505, 507, 508 Learning difficulties animal cognitive tests, 419–421 antiepileptic drug teratology, 110, 115 fetal alcohol spectrum disorders, 175–176, 182 leukemia chemotherapy outcomes, 142 neurobehavioral tests for children, 315–316 phases, 433–435 public health burden of disabilities, 465–466 Legislation, 479–481. See also individual Acts Leukemia, 131–148 Leukoencephalopathy, 134 Liberal p-value criteria, 371–372, 374

531 Life-long studies, 14–15, 17–18 Ligand-dependent transcription factors, 451–452 Limitations, 69, 71, 89, 361–378 Linear models, 74 Lipids, 9–10, 84, 86 Lipopolysaccharides, 346 Log-transformed exposure data, 74 Longitudinal studies, 43–44, 171, 173, 315–316 Long-term outcomes, 181–182 Low birth weight, 151–152 Magnetic resonance imaging (MRI), 322–323, 325–332, 351 Magnetic resonance spectroscopy (MRS), 324–326, 351 Major malformations, 104, 108, 116, 117–119, 123 Maneb, 346 Manganese, 231, 232, 234 Marihuana. See cannabis Maternal aspects attachment in cocaine-using mothers, 214–215 biomonitoring of insecticide effects, 260–261, 264, 274 blood analysis, 241, 248 cannabis/cocaine testing, 291–294 hair methylmercury levels, 27–33, 382–387, 389–390 self reporting of cannabis/cocaine use, 288–291 serum levels, 477 smoking, 149–167 Maternal Health Practices and Child Development (MHPCD) project, 289–290 McCarthy Scales, 27, 44, 48, 53, 58 MDI. See Mental Development Index Measurement error, 385–386, 392, 502–505, 508, 518 pediatric neurotoxicity, 493–495 problems, 141 reliability, 367–369 Mechanisms of cocaine teratology, 195, 199–208 Mechanism versus outcome approach, 194–195 Meconium analysis, 264, 275, 294–295 Medial temporal lobes, 419, 430 Mediators of neurotoxic effects, 58 MeHg. See methylmercury Memory, 156, 157, 175–176, 419–421

532 Menstrual disorders, 90 Mental Development Index (MDI), 494 Mental retardation (MR), 110–116, 121–124, 419–421, 424–425 6-Mercaptopurine, 133 Mercury, 16–17, 231–248. See also methylmercury Mercury vapor, 11–13 Merton, Robert, 506, 512 Mesolimbic system, 200, 206 Meta-analysis, 93 Metals, 231–251. See also individual metals Methamphetamine exposure, 326 Methodological issues, 97–98, 309–310, 312–315 Methotrexate (MTX), 131, 133–137 Methylmercury, 1–23 cross-sectional studies, 29–32 delayed toxicity, 13–16 developmental tests, 35 exposure assessment, 231–248 Faroese Imaging Study, 327–332 inorganic mercury confusion, 11–13 Iraq, 3–6 neuroimaging, 326 neurological tests, 34 neuropsychological tests, 35–36 outcome variables, 34 PCB study confounder, 52–53 prenatal exposure, 381–394 prospective epidemiological studies, 25–29 seafood, 2–13 Mexico, 259 Michigan study, 43–49, 52, 55–56, 58 Micro-activity recording, 272 Microcephaly, 107–109, 116–121 Midface hypoplasia, 104, 107–109, 116–118, 121 Milk analysis, 235, 243, 258, 264 Minamata Disease, 3–6, 17–18 Minor malformations, 104 Misclassification errors, 68–71, 73 Misreporting, 291, 296 Mitosis, 6 Moderator variables on child performance, 107, 141 Modifications in structural equation models, 388 Monitoring workplace organic solvents, 86–87 Monkeys, 15, 436 Monoaminergic system, 197, 199–203 Monotherapy/polytherapy comparison, 107–116

Index Monte-Carlo studies, 370–371, 374 Motivation, 308, 313 Motor skills, 179–180, 329, 332 Mouse studies, 452–453 MR. See mental retardation MRI. See magnetic resonance imaging MRS. See magnetic resonance spectroscopy MTX. See methotrexate Multi-agent chemotherapy, 132–133, 135 Multigenerational testing, 488–489 Multi-level models, 408–409 Multi-level studies, 399–402 Multiple comparisons, 392 Multiple regression analysis, 383 Multivariate observational studies, 361–378 N-acetylaspartate (NAA), 70, 351–352 Nail analysis, 236, 237, 243–244 Narrow-band tests, 494 NAS committee, 389–390 National Health and Nutrition Examination Survey (NHANES), 244 National Human Activity Pattern Survey (NHAPS), 262 National Research Council Committee on Biological Markers, 76 National Survey on Drug Use and Health (NSDUH), 197–198 NBAS. See Neonatal Behavioral Assessment Scale NE. See norepinephrine Negative feedback, HPT axis, 450, 451 Neighborhood social context, 400–402 Neonatal Behavioral Assessment Scale (NBAS), 473 Neonates, 154, 258, 264, 266, 294–296 Nero, Emperor, 501 NES2 Finger Tapping test, 383, 387–389 Netherlands study, 44–52, 54–58 Neural tube defects, 122, 123 Neurobehavioral teratology, 103–130 Neurobehavioral tests, 303–320, 383, 387–389 Neurochemical studies, 201–209 Neurodevelopmental function tests, 71 Neuroimaging, 321–339 Neurological diseases of the elderly, 341–342, 344–353 Neurological Optimality Score (NOS), 32 Neurological tests, 34 Neuropathological changes, 139 Neuropsychological function, 30–31, 35–36, 121, 125 Neurotoxic metals, 231–251 New Zealand, 26, 28–29, 33, 35, 39, 245

Index NHANES. See National Health and Nutrition Examination Survey NHSDA (National Household Survey on Drug Abuse). See National Survey on Drug Use and Health Nicotine, 153. See also smoking; tobacco smoke Nicotinyls, 253, 256, 258 Nigrostriatal system, 200, 206 Nondysmorphic fetal alcohol spectrum disorders, 169–170 Non-experimental studies, 361–378 Non-linear dose–effect relationships, 74 Norepinephrine (NE), 200, 201, 206, 209 ‘Normal’ blood lead levels, 67–68 Normative context, 512–516 North Carolina (USA) study, 44–47, 56, 258 NOS. See Neurological Optimality Score NSDUH. See National Survey on Drug Use and Health Nun study, 344 Nutrition, 365, 407–408. See also diet Objectivity, science, 502, 512–514, 516 Observational studies, 272, 361–378 Occupational exposure, 83–84, 86–87, 90–96, 98–99 OECD. See Organization for Economic Cooperation and Development Offspring development, 83–84, 86–99 Oral fluid analysis, 291–292. See also saliva analysis Organic ligands, 231–232 Organic solvents, 83–99 Organization for Economic Cooperation and Development (OECD), 469–470 Organized skepticism, science, 512 Organochlorines (OC), 254, 257–259, 266–268, 273–274 Organohalogens, 10. See also PCBs Organophosphates, 253–257, 259, 269–271, 273–274, 476 Oswego birth cohort studies, 44–47, 51–52, 258–259 Outcomes, 30–31 Outcome variables, 34 Outcome versus mechanism approach, 194–195 Paraoxonase (PON), 491 Paraquat, 346 Parental smoking, 149–167 Parent–child relationship, 211–216 Parenting impairments, 211–215

533 Parent-rated behavior, 183–185 Parkinson’s disease, 342, 345–346 Passive smoking, 149–150, 155 Paternal smoking, 151 Path diagrams, 387 Patterns of environmental disease, 487–488 Patterns of minor malformations, 104, 107–109 PB. See phenobarbital PBPK. See physiologically-based pharmacokinetic models PCBs. See polychlorinated biphenyls PEA. See prenatal exposure to alcohol Pediatric Environmental Health (PEH), 485 Pediatric Environmental Health Speciality Units (PEHSUs), 485, 486 Pediatric Environmental Neurobehavioral Test Battery (PENTB), 306, 313 PEHSUs. See Pediatric Environmental Health Speciality Units PENTB. See Pediatric Environmental Neurobehavioral Test Battery Perceptual performance, 373 Perchlorate, 450 Perinatal period, 264, 453, 454 Peripheral neuropathy, 134 Perseverative phase, 434 Persistent deficits, 75 Pesticides exposure studies, 312–315 insecticide exposure, 253–285 mercury toxicity confounding, 38 Parkinson’s disease link, 346 PCB study confounders, 53–54 PET. See positron emission tomography Pharmacokinetic modeling, 234, 237–239, 247 Phases of learning process, 433–435 Phenobarbital (PB), 106, 107, 109, 115, 117–118, 121–122 Phenylketonuria (PKU), 343, 424–425 Phenylpyrazoles, 258 Phenytoin (PHT), 106, 109, 115–117, 120–121 Photic stimulation, 328, 329 PHT. See phenytoin Physical environment, 397–402 Physical properties of organic solvents, 84–86 Physiologically-based pharmacokinetic (PBPK) models, 237–239, 247 Pilot-revise-pilot process, 310, 315 Pink Disease, 12 PKU. See phenylketonuria Placental transfer, 87, 260–261

534 Plasma levels, 449–452, 456, 477 Plus maze, 426 Policy issues, 478–481 Policy-relevant science, 511–525 Polychlorinated biphenyls (PCBs) behavioral deficit types, 44–49 confounding variables, 51–54, 59 epidemiological study issues, 43–66 exposure–effect relationship, 57–58 Faroese Imaging Study, 327–332 IQ tests, 472 mercury toxicity confounding, 37 postnatal effects, 54–57, 59–60 prenatal studies, 373 seafood, 10–11 social-ecological perspective, 405 thyroid hormone signaling effects, 454–456 Polydrug effects, 135, 199 Polytherapy, 106–116 PON. See paraoxonase Population studies, 2–3, 29–32, 233, 244–245 Positron emission tomography (PET), 324, 325 Postnatal period, 16–17, 54–57, 59–60, 199, 370–371 ‘Post-normal’ science, 514–517 Potential pathways for adverse effects, 151–153 Potential risk assessment, 466–478 Poverty, 211, 396–401 Practice effects, 315–316 Preclinical models, 206–208 Predictors of adverse outcomes, 124–125 Prednisone, 133, 134 Prefrontal cortex, 424 Pregnancy acetyl cholinesterase activity, 256 antiepileptic medications, 103–130 cannabis/cocaine use assessment, 193–229, 287–302 creatine adjustment of urine samples, 274 drug use prevalence, 198 drug use reporting, 290 fetal alcohol spectrum disorders, 169–191 gestational exposure, 488 insecticide exposure, 256, 260–261, 264 mercury exposure, 240–248 neurotoxic metal biomonitoring, 236–237 organic solvent exposure, 83–84, 86–99 smoking, 149–167 Prenatal exposure, 6–8, 29–32, 43, 55–56, 95–96, 381–394

Index Prenatal exposure to alcohol (PEA), 170, 186. See also fetal alcohol spectrum disorders Prenatal hypothyroidism, 453 Prenatal PCB exposure studies, 373 Pre-School Activity Inventory, 48 Preschool children, 155, 157, 307–308, 312–315, 404–405 Preschool Language Scale Total Language and Auditory Comprehension, 28 Primates, 15, 436 Prognostic indicators, 493–495 Properties of organic solvents, 84–86 Prospective studies, 25–29, 43–44, 94, 170–171 Prospective versus retrospective design, 141 Protein binding, 232–233, 261 Protocol manuals, 311 Psychiatric disorders, 72–73, 150 Psychopathology, 184 Public health, 38–39, 67, 83–102, 395–416, 465–484 Public norms, 511–512 P-values, 371–372, 374 Pyrethrins, 257 Pyrethroids, 253–255, 257–258 Quality, 311–312, 515 Quality of life (QOL), 142 Questionnaires, 272, 288 Random assignment, 362–363 Reaction time (RT), 178–179 Recall, 239, 289 Red cell lifetimes, 69 Reference dose (RfD), 244–245, 247 Reflective functioning, 215–216 Regression coefficients, 386 Reliability, 368–369 ‘Representative’ learning tasks, 419–421 Reproductive outcome, 83–84, 86–99 Research, 485–499. See also individual study types Residential environmental risk factors, 397–402 Residuals, 363 Responsible scientific conduct, 512–514 Retrospective studies, 171–172 Retrospective versus prospective design, 141 Reversal learning tasks, 434 Reversibility of poisoning, 74–75 RfD. See reference dose

Index Right-to-know, 519–521 Risebrough, Robert, 3 Risk assessment, 466–478 characterization phase, 493 children, 491–493 methylmercury in seafood, 9–11 politics, 511–525 potential neurotoxins, 466–478 reproductive risk after organic solvent exposure, 90–96 SAB. See spontaneous abortion Safety standard setting, 389–392 Saliva analysis, 264, 265, 275, 291–292 Sample collection, 264–265 Sample sizes, 140 SCDS. See Seychelles Child Development Study Schizophrenia, 341, 344, 347–348 School-age children, 155–158, 174, 308 School performance, 158 Science, 512–516 Scientific paradigms, 516–517 Seafood, 2–13, 25–66, 239–240 Seattle Study on Alcohol and Pregnancy, 171 Secondary disabilities, 182 Seizures, 106–107, 109 Selection bias, 291, 349 Self reporting, 288–291, 296 SEM. See structural equation modeling Sensitivity of cognitive tests, 418–429, 438–439 Serotonin, 200, 201, 204–205 Serum levels, 449–452, 456, 477 Serum T4, 449–452, 456 Settings, 25–29 Seychelles Child Development Study (SCDS), 6–8, 16, 27–29, 32, 245, 246 Significance tests, 504–506 Signs and symptoms, 4 Single photon emission computerized tomography (SPECT), 324–325 Skeletal effects, 107–109 Small effects, 504 Smoking, 149–167, 405–406 Social/antisocial behavior, 49 Social construction, 513 Social context, 399–401 Social contextual units, 401–402 Social-ecological perspective, 395–416 Social exposures, 396 Socioeconomic factors, 37, 76, 365–366, 368, 494 Sodium channels, 257

535 Soil exposure, 233–234 Solvents, 83–99 Spanish-speaking children, 309–310, 312, 315. See also Hispanic children Species, 435–436, 439 Specific cognitive functions, 431–435 SPECT. See single photon emission computerized tomography Spontaneous abortion (SAB), 91–92 Spot urine samples, 274–275 SRO. See surprising reward omission Standards, 361–378, 389–392 Steady state pharmacokinetic models, 237–239 Sternberg Memory paradigm, 47 Stressors, 396–399, 406–407 Stress responsivity, 208, 209 ‘Strong objectivity’, 516 Structural equation modeling (SEM), 350, 381–394 Structural magnetic resonance imaging, 322–323, 327–328 Study methodology, 170–173 Subclinical Minamata Disease, 17–18 Substance abuse, 184–185, 211–216, 365–366 Surprising reward omission (SRO), 429 Survey data, 264, 266–272 Susceptibility, 303–304, 404–406, 489, 490 Sweat patches, 292–293 Syndromic actions, 104 Synergistic mechanisms, 404 T4. See thyroxine Taiwan, 43 Task parameters, 432–433 Task selection, 418–435 TCPY. See 3,5,6-trichloropyridinol TEFs. See toxic equivalency factors TEL. See tetraethyl lead TEQ. See toxic equivalency quotient Teratology alcohol, 169 antiepileptic medications, 103–130 cocaine, 193–229 definition, 193–194 inhalant abuse, 96–97 nicotine, 153–154 organic solvents, 89 Terminology, 169–170, 186, 289, 518–519 Test selection, 304–305 Tetraethyl lead (TEL), 505, 507, 508 TH. See thyroid hormone Theoretically-based approach, 369–370, 372–375

536 Thimerosal, 1 Thought Problems, 16 Thyroid hormone receptors (TRs), 455, 456 Thyroid hormone signaling, 58, 449–463 Thyroid receptor isoforms, 451–453 Thyrotropin (TSH), 451 Thyroxine (T4) suppression, 449–452, 455, 456 Time-activity diaries, 272 Time-dependent effects, 453–454 Time to pregnancy (TTP), 90 Timing, 453–454, 474–476 Tissue compartments, 49–51, 68–69 TLC. See Treatment of Lead-Poisoned Children trial Tobacco smoke, 149–167, 405–406 Tower of London, 50, 55 Toxic equivalency factors (TEFs), 50 Toxic equivalency quotient (TEQ), 49–51 Tracking techniques, 478 Traffic fumes, 405 Training neurobehavioral test examiners, 311–312 Transfer of learning. See cumulative learning Transforming data, 368 Translation, 309–310, 314–315 Treatment of Lead-Poisoned Children trial (TLC), 75 3,5,6-trichloropyridinol (TCPY), 274 Triggering of disease, 487–488 TRs. See thyroid hormone receptors TSH. See thyrotropin TTP. See time to pregnancy Twin studies, 345 Type I/II errors, 171, 502–505, 508, 518 Umbilical cord blood. See cord blood UNCED. See United Nations Conference on Environment and Development Underreporting, 150 Understanding child’s needs, 211–212, 215–216 Unemployment, 211 United Nations Conference on Environment and Development (UNCED), 479 Universalism (scientific), 512

Index Unsaturated bonds, 87 Upregulation, 455 Urine analysis insecticide exposure assessment, 264, 265, 273–275 maternal drug testing, 291 neonates, 294 neurotoxic metals, 235 U-shaped dose-response functions, 9 Valence states, 231–232 Validity, covariates, 368–369 Valproic acid (VPA), 106, 107, 109, 115, 119, 122–124 Variables, 36–38, 361–378. See also confounding variables VCR. See vincristine Verbal skills, 174–176 Videotaping, 471 Vincristine (VCR), 131–134, 138 Violence, 211 Visual attention tasks, 432–433 Visual functioning/acuity, 95–96 Visual recognition memory, 47 Visuospatial ability, 176 Volatility of solvents, 84 Voltage-gated sodium channels, 255, 257 VPA. See valproic acid Vulnerability. See susceptibility Weiss, Bernard, 501 Whole blood lead:plasma lead ratio, 69 Wing, Steve, 521–522 Wisconsin Card Sort test, 47 Workplace exposure, 83–84, 86–87, 90–99 X-ray fluorescence (XRF), 69–70, 234, 235 Young children behavior observation techniques, 272 creatine adjustment of urine samples, 274 insecticide exposure assessment, 254, 261–263, 265–272, 275 neurobehavioral test batteries, 307–308, 312–315. See also children; infants