Assessing the Effects of Environmental Toxicant Exposure in Developmental Epidemiological Studies: Issues for Risk Assessment

NeuroToxicology 26 (2005) 483–489

Assessing the Effects of Environmental Toxicant Exposure in Developmental Epidemiological Studies: Issues for Risk Assessment Deborah C. Rice * Environmental Health Unit, Maine Bureau of Health, State House Station 11, Augusta, ME 04333, USA Received 20 September 2004; accepted 10 January 2005 Available online 12 March 2005

Abstract Epidemiological studies have been of critical importance to the understanding of the effects of environmental chemical exposure during development on the behavior of infants and children. The ultimate goal of these studies should be to provide information that may be used directly for the protection of public health. The strategies for the assessment endpoints include development of domain-specific tests based on knowledge concerning effects of the chemicals being assessed, or use of standard clinical instruments that sample a range of functions. Discussion of an overall strategy for choice of endpoints would allow more straightforward comparisons across studies. There is increasing recognition of the importance of measuring a number of chemicals relevant to the population under study; however, different investigators make different decisions concerning which and how many chemicals to measure, as well as how to include them in the statistical analysis, particularly when there is a high degree of collinearity. Chemicals that are highly correlated with the ‘‘chemical of interest’’ are sometimes not included in the statistical analysis, resulting in missed opportunity to derive important information from the study. In addition, the shape of the relationship between exposure and effect is usually not explored in epidemiological studies, even though such information is critical for risk assessment. Opportunity for discussion among investigators, statisticians, and risk assessors potentially would result in human developmental toxicity studies being maximally useful for public health decisions.

# 2005 Elsevier Inc. All rights reserved. Keywords: Toxicants; Epidemiology; Risk assessment

INTRODUCTION Epidemiological studies have been critically important to our understanding of the effects of exposure to environmental chemicals during development on the behavior of infants and children. Prospective longitudinal studies on the effects of lead, methylmercury, and PCBs have elucidated important, long-term effects of early exposure. These studies required enormous commitment and dedication over a number of years on the part of the investigative teams performing the * Tel.: +1 207 287 3365 E-mail address: [email protected]

studies, and have cost the governments of the countries in which they have been performed millions of dollars. A question that is central to this substantial effort is: What is the ultimate purpose of these studies? The premise of the present discussion is that the ultimate goal of neurodevelopmental epidemiological studies is the protection of human health. This requires that epidemiological studies be performed and analyzed in a way that is suitable for risk assessment: determination of the relationship between exposure to a chemical or chemicals and the behavior of the child. Risk assessment thus requires information in addition to identification of a statistical association. Promulgation of regulation also should be informed by an

0161-813X/$ – see front matter # 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neuro.2004.12.009

484

D.C. Rice / NeuroToxicology 26 (2005) 483–489

understanding of the monetary and other societal consequences of identified effects. To be able to use information from human studies to protect public health, we need to know three things: (1) what effect(s) are associated with chemical exposure, (2) what chemical(s) are associated with the effect(s), and (3) the shape of the relationship between exposure and effect. Each of these issues is discussed.

CHOICE OF TESTS A general dichotomy for choice of tests may be roughly (and perhaps somewhat artificially) divided between use of domain-specific tests and standard clinical instruments. Perhaps the optimal strategy is the use of well-chosen domain-specific tests and a measure of IQ. Domain-specific tests are designed or chosen to assess very specific aspects of performance: for example, fine motor control, attention, impulsivity, particular aspects of memory such as spatial or verbal, or specific skills related to language. Choice of such tests may be informed by previous epidemiological studies, the experimental behavioral literature, mechanistic studies, or neuropathological studies in humans or animals. For example, the choice of domains assessed in the Faroe Islands study was influenced by knowledge of the neuropathological consequences of high-dose methylmercury exposure (White, 2004). Similarly, the decision to assess control of inhibition in the Oswego study was influenced by effects of PCBs on such tasks in monkeys (P. Stewart, personal communication). Development of domainspecific tasks could also take advantage of information concerning the timing of various developmental processes in different brain areas, information on the timing of exposure, and knowledge of the behavioral processes subserved by specific brain regions, neuroanatomical networks, and biochemical systems. If the tests chosen assess functional domains affected by exposure to the neurotoxicant, domain-specific tests should provide maximal sensitivity. Conversely, if results are negative, it is impossible to know whether there are actually no effects, or if the tests chosen did not measure the affected domains. Domain-specific tests may be particularly recommended if there is enough information to generate hypotheses concerning the domains that may be affected. Even if that is not the case for a particular chemical, however, it may be reasonable to assess domains found to be sensitive to disruption by other chemicals. Lead, methylmercury, and PCBs have

undergone extensive characterization in epidemiological studies. Deficits in attention and failure of response inhibition have each been observed for at least two of these chemicals (Stewart et al., 2003a; Grandjean et al., 1997; Needleman et al., 1979; Jacobson et al., 1992; Winneke et al., 1989; Hatzakis et al., 1987), and perseverative behavior has been observed with lead in both humans (Stiles and Bellinger, 1993) and animals (Rice, 1996), as well as with PCBs in animals (Rice and Hayward, 1997). It may be that some neural circuits are particularly sensitive to disruption, and the behaviors subserved by them may be impaired by a number of chemicals. The alternate strategy is typified by the use of standardized clinical instruments (e.g. WISC, K-ABC, Bayley Scales). Such instruments assess a number of functional domains, to obtain a picture of the overall functioning of the child. An advantage of these instruments is that little or nothing need be known in advance about the neurotoxicity of the chemical(s) under study. A possible disadvantage is that real effects are not identified because they are ‘‘washed out’’ by assessment of domains that are not affected. Particular subtests of standard clinical instruments also may be administered or scored separately, if there is reason to hypothesize that a particular domain assessed by a subtest is sensitive to disruption by the chemical(s) under study. Despite the possible limitations of standard clinical instruments, it is recommended that IQ be tested, for a number of reasons. IQ has been found to be sensitive to disruption by lead (Schwartz, 1994), PCBs (Schantz et al., 2003), and perhaps methylmercury (Kjellstrom et al., 1989). In addition, IQ is relatively easy to discuss as an endpoint in the public arena; regulators, public officials, and citizens are familiar with the measurement of ‘‘how smart you are’’ (regardless of what IQ may actually be measuring). It is much more difficult, for example, to explain why increased reaction time on a continuous performance task is a negative outcome. This is not a trivial consideration, because if the goal of these studies is protection of public health, then the results must be understood and acted upon by risk assessors, regulators, lawmakers, public health professionals, and the public. A second important reason for the inclusion of some measure of IQ is that it is readily monetized. Although the fact that IQ is normalized is not important within any particular study, it nevertheless predicts specific patterns of behavior and societal success, such that the cost to society of a decrease in IQ of a certain magnitude across the population may be calculated in a relatively straightforward manner. Moreover, there is

D.C. Rice / NeuroToxicology 26 (2005) 483–489

precedent for doing so as a justification for legislation within the federal government. The U.S. Environmental Protection Agency monetized the costs in lost wages associated with a 1 mg/dl increase in blood lead concentrations, equivalent to the loss of less than one IQ point (U.S. EPA, 1985) as part of its evaluation of the phase-down of lead from gasoline. In addition, government is increasingly relying on such monetary exercises as part of its decision-making process: how much will it cost to clean up or stop further release into the environment, versus the monetary cost to human health and the environment of not doing so. The National Longitudinal Survey of Youth (NLSY) provides a robust database for calculation of the predictive power of IQ measured during youth on earning capacity in adulthood. This survey is a stratified random sample of 12,686 individuals recruited at ages 12–22 in 1970, with annual follow-up. It was estimated, based on this database, that a 1 mg/dl decrease in blood lead levels would result in increased earnings of US$ 7 billion (Salkever, 1995). More recently, it was estimated that each IQ point is worth $14,500 over a lifetime of earning (Grosse et al., 2002). Estimates have also been made concerning other societal benefits of a small increase in IQ using the NLSY data, including reduction in the high-school dropout and poverty rates (Weiss, 2000). In human neurotoxicity studies (other than case reports), it is not an advantage that clinical instruments are usually standardized whereas domain-specific tests typically are not. First, children are being compared to each other within the cohort, such that comparison to a ‘‘norm’’ is not necessary for the validity of comparisons within a study. Second, studies are sometimes performed in populations other than those used for standardization, such that normalized scores may not be useful. Third, normalizing scores may decrease sensitivity, since several raw scores are included in a normalized score, thereby rendering the measure of performance cruder than the original score. The NIEHS panel convened to review the methylmercury studies recommended that the Seychelles Islands investigators reanalyze their data using raw rather than normalized scores because of these concerns (NIEHS, 1998). (The reanalysis in this case did not change the conclusions (Davidson et al., 2001).)

EXPERIMENTAL DESIGN AND DATA ANALYSIS All of us, including the children in the studies of developmental neurotoxicity, are exposed to many

485

chemicals during development; the infants or children in these studies had measurable body burdens of multiple chemicals at the time of any assessment of exposure or effect. Chemicals that do not accumulate in the body, and yet may have effects if exposure occurs during a sensitive period, are difficult to measure and will probably remain unexamined. The focus of most studies has necessarily been on chemicals that are persistent in the environment and in the body. Early studies often only measured one contaminant, or failed to measure potentially important contaminants for the population under investigation. Some but not all more recent studies have measured more than one contaminant. The issue of measurement of multiple contaminants is still germane, however: specifically, how to choose which chemicals to measure, and how to include multiple chemicals in the statistical analysis. Early studies on the effects of lead measured body burdens of lead and no other environmental contaminants. This is also true even for recent lead studies (e.g. Canfield et al., 2003) presumably on the assumption that lead body burden is unlikely to be highly correlated with other contaminants. However, measurement of other chemicals might account for additional variance and thereby increase the power to detect an effect of lead exposure. For example, indoor pesticide use in inner city households may be high (Whyatt et al., 2002), and result in adverse health consequences for the fetus such as decreased birth size (Perera et al., 2003). Attention was focused on the issue of exposure to co-contaminants largely as a result of a study on the effects of developmental PCB exposure from maternal consumption of Lake Michigan fish (Jacobson et al., 1985, 1990, 1992; Jacobson and Jacobson, 1993; Fein et al., 1984). This study was criticized for not measuring methylmercury body burden (NIEHS, 1998), a known neurotoxicant present in Great Lakes fish. As a result of this lesson learned as the methodology of investigating the neuropsychological consequences of developmental chemical exposure matured, a subsequent study in Lake Ontario fish eaters measured multiple co-contaminants of PCBs, including methylmercury (Stewart et al., 2003a,b, 2000). In fact, PCBs and methylmercury may not be particularly highly correlated, so that it is often possible to differentiate effects produced by these contaminants (Stewart et al., 2003b, 2000). In contrast, highly lipid-soluble compounds such as PCBs, dioxins, DDE and other pesticides are usually highly correlated in human tissue, because they are ingested via the same fatty food sources, and are stored

486

D.C. Rice / NeuroToxicology 26 (2005) 483–489

in the same tissue (fat) in humans. Modern studies of the consequences of exposure to lipophilic contaminants have generally measured multiple contaminants that were likely to be highly correlated. The Oswego study measured cord blood concentrations of pesticides (DDE, mirex, hexachlorobenzene) found in Lake Ontario fish as well as multiple PCB congeners that were the focus of the study (Stewart et al., 2000). The Dutch study, in which exposure was through the general food supply, measured multiple PCB congeners, dioxins, and furans in breast milk in addition to four PCB congeners in maternal and cord blood (Patandin et al., 1999). Adverse effects have been found independently for PCBs and dioxins (Vreugdenhil et al., 2002). In contrast, the German study, which is a participant in the multinational European study that includes the Dutch study, measured only three PCB congeners and no dioxins or furans (Winneke et al., 1998). Pesticide exposure was not assessed in either European study, and dioxins were not measured in the Oswego study. Studies in the Arctic have measured multiple contaminants found in high concentrations in circumpolar populations, including PCBs, dioxins, DDT, DDE, mirex, and other pesticides (in addition to mercury and lead) (Muckle et al., 2001). It is entirely reasonable that the suite of contaminants for any study be chosen on the basis of sources of contamination and the resulting body burdens in the population under study. However, different study teams have made different decisions even with presumed similarity in study populations, making interpretation across studies difficult. In addition to different decisions regarding which contaminants to measure, investigative teams also make different decisions concerning the strategy for inclusion of contaminants in data analysis. Some studies have analyzed the effect of a single contaminant that is highly correlated with others. For example, the Faroe investigators determined the effects of PCBs on performance, but did not report effects of DDE, even though DDE concentrations in blood and the correlation with performance were analyzed (Grandjean et al., 2001). The rationale for failure to determine the effects of some of the chemicals analyzed in studies is often that there is a high degree of collinearity, such that including other chemicals along with a ‘‘favorite’’ runs a high risk of over-controlling. This may not be true if only one chemical is related to the effect. For example, the Oswego study found multiple associations between PCBs and performance after controlling for DDE, even though the two chemicals were highly correlated, presumably because DDE was not a good predictor of performance. For instances in which more than one

chemical is associated with an effect, and these chemicals are highly collinear, the only option may be to analyze each one separately without controlling for the other. It may not be possible under such circumstances to determine which chemical(s) is producing the observed effects. However, such a strategy is more informative (and honest) than failing to include chemicals in the statistical analysis that were not necessarily the main focus of the study. Investigative teams also develop different strategies for handling of covariates. There are numerous choices that must be made in model building, including inclusion criteria and at what point covariates are added to the model. Some studies use the same set of covariates for all analyses within a publication, others only include covariates that meet inclusion criteria for any particular analysis. Such choices should be made a priori based on the hypothesis of the study. A particularly problematic issue is lack of a consistent strategy between publications within the same study, with insufficient rationale for the change provided in the publication. There are also instances in which there is insufficient detail in a publication to determine how the model was built. The responsibility for this falls not only on the authors (sometimes it is difficult to recognize lack of clarity in one’s own writing) but also on reviewers and editors who do not require that sufficient detail be provided. Again, these issues are not trivial, or simple niceties. Consumers of this literature – risk assessors, regulators, and other stakeholders – need to be able to evaluate and compare across publications and across studies. Choices regarding the strategy for the main analyses of a study must be transparent, consistent, and made prior to data analysis. A posteriori analyses, to explore specific observations made during primary analyses, should be identified as secondary or exploratory analyses.

DOSE–EFFECT RELATIONSHIPS For epidemiological studies to be maximally useful for protection of public health, information must be available on the shape of the relationship between exposure and effect. In particular, risk assessment requires information on whether there is evidence for a threshold below which adverse effects would not be expected, as well as whether adverse effects are relatively greater or lesser at low compared to high exposures. Studies are typically analyzed by simple (usually linear) regression analysis or some other means of

D.C. Rice / NeuroToxicology 26 (2005) 483–489

determining whether there is a statistically-significant association between exposure and measures of performance. Such analyses are not performed to provide information on the shape of the relationship between effect and exposure (the latter usually measured as body burden). A few investigators have performed analyses regarding the shape of the exposure–effect relationship. Canfield et al. (2003) modeled the shape of the relationship between children’s IQ and blood lead concentrations. They found that the slope of the relationship was steeper below 10 mg/dl than above it. In other words, the function was supralinear. Subsequent similar analyses for a single study (Bellinger and Needleman, 2003), as well as a meta-analysis of several studies (Lanphear et al., 2004) also reported the shape of the body burden-effect function to be supralinear. The Faroe investigators modeled their methylmercury data, under contract to EPA, and found that supralinear models such as square root or logarithmic transformations provided a better fit than other models (BudtzJørgensen et al., 1999, 2000). The Seychelles Island study investigators also modeled the body-burden-effect relationship in their study (Axtell et al., 2000, 1998). However, such efforts are relatively rare. The reference dose (RfD) or reference concentration (RfC) derived by EPA for non-cancer effects of any chemical assumes that there is a threshold below which there are unlikely to be adverse effects. It is important to know, as part of the modeling exercise, whether there is evidence of a threshold within the range of exposures for any individual study. The fact that linear or supralinear models provide a better fit to the data than sublinear ones provides evidence against there being a threshold. Such seems to be the case for lead at least as low as 1 mg/dl (Schwartz, 1994), and for methylmercury within the range of body burdens of the Faroe Islands study (NRC, 2000). It is of critical importance to public health decisions that the relationship between exposure (or body burden) and effects be known as accurately as possible. The fact that the effect of lead on IQ is probably relatively greater below 10 mg/dl than above it may have implications for remediation of old housing stock containing lead-based paints, and treatment of children with elevated lead body burden. If such analyses had been performed in studies of the effects of lead on children’s development 15 or 20 years ago, would present public policy be different than it is currently? Modeling the exposure–effect function from epidemiological studies requires access to the raw data. This has proved a significant impediment to risk assessment. If investigators do not perform such analyses them-

487

selves, and are unwilling to make data available to others interested and able to perform such analyses, a critical piece of information necessary to make sound policy decisions with respect to public health will remain unavailable.

CONCLUSIONS AND RECOMMENDATIONS Considerable attention has focused over the years on exploring the strategies for choice of tests, data analysis, and interpretation of results from experimental neurotoxicology studies (c.f. EPA, 1998a,b; OECD, 2003, 1997; MacPhail et al., 1997; Moser et al., 1997a,b,c,d; Catalano et al., 1997; Tilson et al., 1997; Kimmel et al., 1985; Buelke-Sam et al., 1985; Nelson et al., 1985; Adams et al., 1985a,b). A comparable effort has not been mounted for neurodevelopmental epidemiological studies. If the ultimate purpose of these studies is that they be useful for making good decisions regarding public health, there are a number of issues that may be explored to make the literature optimally useful. (1) Investigators should take advantage of any known effects of the chemical or chemicals under study in choosing appropriate domain-specific endpoints. In addition, standard clinical instruments of IQ should be included, because IQ can be monetized in a straightforward manner. (2) The consequences of various choices of model building have not been systematically or publicly explored. Such an exercise by a group of epidemiologists and statisticians would provide information on the influence of various strategies on the outcome of the analyses. (3) The issue of how to include multiple contaminants in statistical analyses will undoubtedly become more relevant, as investigators increasingly recognize the importance of measuring a number of contaminants in their studies. This issue would also benefit from input from a group of experts. (4) Investigators should be strongly encouraged to explore exposure (or body burden)-effect relationships intheirstudies. Hopefully, collaborative efforts to provide the expertise to perform such modeling will be encouraged by funding agencies, who have as their mandate protection of public health. (5) Investigators are often unaware of the role of the risk assessment community in assessing their work, or indeed the close scrutiny that each decision in a study will receive from many interested parties. Yet, it was my experience as a risk assessor

488

D.C. Rice / NeuroToxicology 26 (2005) 483–489

at EPA that investigators want their studies to be useful, and to be used. Communication that would enable study investigators to understand how they can best provide information useful for risk assessment, and that would help risk assessors to understand real-world constraints (fiscal and otherwise) in performing these large and complex studies, would be advantageous. There may be instances, for example, when a particular decision matters little to an epidemiological researcher but matters a great deal to a risk assessor in terms of the interpretation or usefulness of a study. Better awareness and knowledge of how the data potentially will be used may help investigators in their decision-making process without compromising the goals of the study. The goal of these suggested activities is not to develop a set of rules or protocols that investigators are compelled in some sense to follow. The purpose is rather to explore the influence of various decisions on the results and interpretation of the studies. It may be that for some issues, various choices lead to more or less the same conclusion, so that different decisions are relatively unimportant. It may also be that, in some cases, the design or analytical strategy chosen has important consequences for the interpretation of the study. A greater understanding of these issues would benefit the entire field, including investigators, regulators, funders, lawmakers, and the public.

REFERENCES Adams J, Buelke-Sam J, Kimmel CA, Nelson CJ, Reiter LW, Sobotka TJ, et al. Collaborative behavioral teratology study: protocol design and testing procedures. Neurobehav Toxicol Teratol 1985;7:579–86. Adams J, Buelke-Sam J, Kimmel CA, Nelson CJ, Miller DR. Collaborative behavioral teratology study: preliminary research. Neurobehav Toxicol Teratol 1985;7:555–78. Axtell CD, Cox C, Myers GJ, Davidson PW, Choi AL, Cernichiari E, et al. Association between methylmercury from fish consumption and child development at five and a half years of age in the Seychelles child development study: an evaluation of nonlinear relationship. Environ Res Sect A 2000;84:71–80. Axtell CD, Myers GJ, Davidson PW, Choi AL, Cernichiari E, Sloan-Reeves J, et al. Semiparametric modeling of age at achieving developmental milestones after prenatal exposure to methylmercury in the Seychelles child development study. Environ Health Perspect 1998;106:559–64. Bellinger DC, Needleman HL. Intellectual impairment and blood lead levels. N Engl J Med 2003;349:500–2. Budtz-Jørgensen E, Grandjean P, Keiding N, et al. Benchmark dose calculations of methylmercury-associated neurobehavioral deficits. Toxicol Lett 2000;112–113:193–9.

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 Biostatstics, University of Copenhagen; 1999. Buelke-Sam J, Kimmel CA, Adams J, Nelson CJ, Vorhees CV, Wright DC, et al. Collaborative behavioral teratology study: results. Neurobehav Toxicol Teratol 1985;7:591–624. Canfield RL, Henderson CR, Cory-Slechta DA, Cox C, Jusko TA, Lanphear BP. Intellectual impairment in children with blood lead concentrations below 10 ug/dl. N Eng J Med 2003;348: 1517–26. Catalano PJ, McDaniel KL, Moser VC. The IPC collaborative study on neurobehavioral screening methods. VI: Agreement and reliability of the data. Neurotoxicology 1997;18:1057–64. 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;285:1291–3. EPA. Health effects test guidelines OPPTS 870.6200 neurotoxicity screening battery, EPA 712-C-98-238. Washington, DC; 1998a. EPA. Health effects test guidelines OPPTS 870.6300 developmental neurotoxicity study EPA 712-C-98-239. Washington, DC; 1998b. Fein GG, Jacobson JL, Jacobson SW, Schwartz PM, Dowler MA. Prenatal exposure to polychlorinated biphenyls: effects on birth size and gestational age. J Pediatr 1984;105:315. Grandjean P, Weihe P, Burse VW, Needham LL, Storr-Hansen E, Heinzow B, et al. Neurobehavioral deficits associated with PCB in 7-year-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol Teratol 2001;23:305–17. Grandjean P, Weihe P, White R, Debes F, Arak S, Yokoyama K, et al. Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol Teratol 1997;20:1–12. 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;110:563–9. Hatzakis A, Kokkevi A, Katsouyanni K, Maravelias K, Salaminios JK, Kalandidi A, et al. Psychometric intelligence and attentional performance deficits in lead-exposed children. In: Lindberg SE, Hutchinson TC, editors. Proceedings of the sixth international conference on heavy metals in the environment. Edinburgh: CEP Consultants; 1987. p. 204–9. Jacobson SW, Fein GG, Jacobson JL, Schwartz PM, Dowler JK. The effect of intrauterine PCB exposure on visual recognition memory. Child Dev 1985;56:853–60. Jacobson JL, Jacobson SW. A 4-year follow-up study of children born to consumers of Lake Michigan fish. J Great Lakes Res 1993;19:776–83. 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. Jacobson JL, Jacobson SW, Padgett RJ, Brummit GA, Billings RL. Effects of prenatal PCB exposure on cognitive processing efficiency and sustained attention. Dev Psychol 1992;28:297–306. Kimmel CA, Buelke-Sam J, Adams J. Collaborative behavioral teratology study: implications, current applications and future directions. Neurobehav Toxicol Teratol 1985;7:669–73. Kjellstrom T, Kennedy P, Wallis S, Mantell C. Physical and mental development of children with prenatal exposure to mercury from fish. Stage 2: Interviews and psychological tests at age 6.

D.C. Rice / NeuroToxicology 26 (2005) 483–489 National Swedish Environmental Protection Board, Report 3642. Solna, Sweden; 1989. Lanphear BP, Hornung R, Khoury J, Yolton K, Baghurst P, Bellinger D, et al. Low-level environmental lead exposure and intellectual impairment in children. An international pooled analysis. XXI international neurotoxicology conference. Honolulu, Hawaii; 2004. MacPhail RC, Tilson HA, Moser VC, Becking GC, Cuomo V, Frantik E, et al. The IPCS collaborative study on neurobehavioral screening. I: Background and genesis. Neurotoxicology 1997;18:925–8. Moser VC, Becking GC, Cuomo V, Frantik E, Kulig BM, MacPhail RC, et al. The IPCS collaborative study on neurobehavioral screening methods. V: Results of chemical testing. Steering group. Neurotoxicology 1997a;18:969–1055. Moser VC, Becking GC, Cuomo V, Frantik E, Kulig BM, MacPhail RC, et al. The IPCS collaborative study on neurobehavioral screening methods. IV: Control data. Steering group. Neurotoxicology 1997b;18:947–67. Moser VC, Becking GC, Cuomo V, Frantik E, Kulig BM, MacPhail RC, et al. The IPCS collaborative study on neurobehavioral screening methods. III: Results of proficiency studies. Steering group. Neurotoxicology 1997c;18:939–46. Moser VC, Tilson HA, MacPhail RC, Becking GC, Cuomo V, Frantik E, et al. The IPCS collaborative study on neurobehavioral screening methods. II: Protocol design and testing procedures. Neurotoxicology 1997d;18:929–38. 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; 109:1291–9. Needleman HL, Gunnoe C, Leviton A, Reed R, Peresie H, Maher C, et al. Deficits in psychologic and classroom performance of children with elevated dentine lead levels. N Eng J Med 1979;300:689–95. Nelson CJ, Felton RP, Kimmel CA, Buelke-Sam J, Adams J. Collaborative behavioral teratology study: statistical approach. Neurobehav Toxicol Teratol 1985;7:587–90. NRC (National Research Council). 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. NIEHS (National Institutes of Environmental Health Sciences). Scientific issues relevant to assessment of health effects from exposure to methylmercury. Workshop organized by Committee on Environment and Natural Resources (CENR), Office of Science and Technology Policy (OSTP), The White House; 1998. OECD (Organization for Economic Co-operation and Development). 424 Neurotoxicity study in rodents. Paris, France: Organization for Economic Co-operation and Development; 1997. OECD (Organization for Economic Co-operation and Development). 426 Developmental neurotoxicity study. Paris, France: Organization for Economic Co-operation and Development; 2003. Patandin S, Lanting CI, Mulder PGH, Boersma ER, Sauer PJJ, 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.

489

Perera FP, Rauh V, Tsai W-Y, Kinney P, Camann D, Barr D, et al. Effects of transplacental exposure to environmental pollutants on birth outcomes in a multiethnic population. Environ Health Perspect 2003;111:201–5. Rice DC. Behavioral effects of lead: commonalities between experimental and epidemiological data. Environ Health Perspect 1996;104:337–51. Rice DC, Hayward S. Effects of postnatal exposure to a PCB mixture in monkeys on nonspatial discrimination reversal and delayed alternation performance. Neurotoxicology 1997;8: 479–94. Salkever DS. Updated estimates of earnings benefits from reduced exposure of children to environmental lead. Environ Res 1995;70:1–6. Schwartz J. Low-level lead exposure and children’s IQ: a metaanalysis and search for a threshold. Environ Res 1994;65:42– 55. Schantz SL, Widholm JJ, Rice DC. Effects of PCB exposure on neuropsychological function in children. Environ Health Perspect 2003;111:357–76. Stewart P, Fitzgerald B, Reihman J, Gump B, Lonky E, Darvill T, et al. Prenatal PCB exposure, the corpus collosum, and response inhibition. Environ Health Perspect; 2003a [doi: 10.1289/ ehp.6173, online date 16 June 2003 @ http://dx.doi.org/]. Stewart PW, Reihman J, Lonky EI, Darvil TJ, Pagano J. Cognitive development in preschool children prenatally exposed to PCBs and MeHg. Neurotoxicol Teratol 2003;5525:1–12. Stewart P, Reihman J, Lonky E, Darvill T, Pagano J. Prenatal PCB exposure and neonatal behavioral assessment scale (NBAS) performance. Neurotoxicol Teratol 2000;22:21–9. Stiles KM, Bellinger DC. Neuropsychological correlates of lowlevel lead exposure in school-age children: a prospective study. Neurotoxicol Teratol 1993;15:27–35. Tilson HA, MacPhail RC, Moser VC, Becking GC, Cuomo V, Frantik E, et al. The IPCS collaborative study on screening methods. VII: Summary and conclusions. Neurotoxicol 1997; 18:1069. U.S. Environmental Protection Agency (U.S. EPA). Costs and benefits of reducing lead in gasoline: final regulatory impact analysis. Washington, DC; 1985. Vreugdenhil HJI, Slijper FME, Mulder PGH, Weisglas-Kuperus N. Effects of perinatal exposure to PCBs and dioxins on play behavior in Dutch children at school age. Environ Health Perspect 2002;110:A593–8. Weiss B. Vulnerability of children and the developing brain to neurotoxic hazards. Environ Health Perspect 2000;108:375–81. White R. Neurobehavioral and neuroimaging findings from the Faroese cohort. XXI international neurotoxicology conference. Honolulu, Hawaii; 2004. Whyatt RM, Camann DE, Kinney PL, Reyes A, Ramirez J, Dietrich J, et al. Residential pesticide use during pregnancy among a cohort of urban minority women. Environ Health Perspect 2002;110:507–14. Winneke G, Bucholski A, Heinzow B, Kraemer U, Schmidt E, Walkowiak J, et al. Developmental neurotoxicity of polychlorinated biphenyls (PCBs): cognitive and psychomotor functions in 7-month-old children. Toxicol Lett 1998;102–103:423–8. Winneke G, Brockhaus A, Collet W, Kraemer V. Modulation of lead-induced performance deficit in children by varying signal rate in a serial choice reaction task. Neurotoxicol Teratol 1989;11:587–92.