PCB body burdens in US women of childbearing age 2001–2002: An evaluation of alternate summary metrics of NHANES data

ARTICLE IN PRESS Environmental Research 109 (2009) 368–378

Contents lists available at ScienceDirect

Environmental Research journal homepage: www.elsevier.com/locate/envres

PCB body burdens in US women of childbearing age 2001–2002: An evaluation of alternate summary metrics of NHANES data$ Daniel A. Axelrad a,, Stephanie Goodman b, Tracey J. Woodruff c a b c

US Environmental Protection Agency, Office of Policy, Economics and Innovation, Washington, DC 20460, USA Association of Schools of Public Health Fellow, assigned to US Environmental Protection Agency, Office of Policy, Economics and Innovation, Washington, DC, USA University of California, San Francisco, Program on Reproductive Health and the Environment; San Francisco, CA, USA

a r t i c l e in f o

a b s t r a c t

Article history: Received 19 June 2008 Received in revised form 22 December 2008 Accepted 15 January 2009 Available online 28 February 2009

An extensive body of epidemiologic data associates prenatal exposure to polychlorinated biphenyls (PCBs) with neurodevelopmental deficits and other childhood health effects. Neurological effects and other adverse health effects may also result from exposure during infancy, childhood, and adulthood. Although manufacture and use of PCBs were banned in the US in 1977, exposure to PCBs is a continuing concern due to the widespread distribution of these compounds in the environment and their persistence. The National Health and Nutrition Examination Survey provides PCB body burden measurements representative of the US population for the years 1999–2002. Interpretation of these data is challenging due to the large number of PCB congeners reported. We examined 6 PCB body burden metrics to identify an approach for summarizing the NHANES data and for characterizing changes over time in potential risks to children’s health. We focused on women of childbearing age, defined here as 16–39 years, because in utero exposures have been associated with neurodevelopmental effects, and used only the 2001–2002 data because of higher detection rates. The 6 metrics, each consisting of different combinations of the 9 most frequently detected congeners, were as follows: total PCBs (all 9 congeners); highly chlorinated PCBs (2 congeners); dioxin-like PCBs (3 congeners, weighted by toxic equivalency factors); non-dioxin-like PCBs (6 congeners); a 4-congener metric (PCBs 118, 138, 153, and 180); and PCB-153 alone. The PCB metrics were generally highly correlated with each other. There was a strong association of PCB body burdens with age for all metrics. Median body burdens of Mexican American women were lower than those of non-Hispanic White and non-Hispanic Black women for 5 of the 6 metrics, and there were no significant differences in body burdens between the latter two groups. Body burdens of women with incomes above poverty level were greater than those for lower-income women at the median and 95th percentiles, but the differences were not statistically significant for any metric. We conclude that the 4-congener and total PCBs metrics are the most promising approaches for tracking changes in body burdens over time and for comparing body burdens of different subgroups in NHANES. Published by Elsevier Inc.

Keywords: PCBs Biomonitoring Body burdens NHANES Neurodevelopment

1. Introduction Polychlorinated biphenyls (PCBs) are persistent and bioaccumulative environmental contaminants have been associated with a wide range of health effects. There are 209 different PCBs congeners that are defined by the number of chlorine atoms (1–10) and their positioning on the biphenyl rings. Manufacture of PCBs in the US was banned in 1977. Although levels of PCBs in environmental samples have declined from their peak, the rate of decline has slowed in recent years (Hickey et al., 2006; Sun et al.,

$ Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views of the US Environmental Protection Agency.  Corresponding author. US EPA, 1200 Pennsylvania Ave NW (1809T), Washington, DC 20460, USA. Fax: +1 202 566 2336. E-mail address: [email protected] (D.A. Axelrad).

0013-9351/$ - see front matter Published by Elsevier Inc. doi:10.1016/j.envres.2009.01.003

2007), and the persistent nature of PCBs and their distribution through the food chain has resulted in continuing human exposure. An important recent development for assessing human exposure levels in the US has been the publication of nationally representative body burden data for a number of PCB congeners, gathered through the National Health and Nutrition Examination Survey (NHANES) (Centers for Disease Control and Prevention, 2005). Much of the concern for health effects from PCB exposure is focused on the effects of in utero exposure. Children born to mothers with high exposures to a mixture of PCBs and polychlorinated dibenzofurans, in poisoning incidents in Taiwan and Japan, exhibited a number of adverse health effects, including neurodevelopmental effects such as cognitive deficits and developmental delays (Chen et al., 1992; Chen and Hsu, 1994; Lai et al., 2002; Rogan et al., 1988).

ARTICLE IN PRESS D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

369

Table 1 Key studies examining associations between prenatal PCB exposure and deficits in neurological development. Study Oswego

PCB metric (tissue analyzed)

 Sum of 68 congeners  Sum of 16 highly chlorinated congeners (cord plasma, placenta and breast milk)

Dutch

Sum of PCBs 118, 138, 153, and 180 (cord blood, maternal blood, and breast milka)

Key findings

 Cognitive deficits associated with cord blood PCBs in infancy (Darvill et al., 2000), and at age 3 years, 

 Cognitive effects associated with maternal blood PCBs at age 42 months (Patandin et al., 1999).  Cognitive and motor deficits associated with maternal blood and cord blood PCBs for those children 

German

Faroe Islandsd

Michigan

but not at 4.5 years (Stewart et al., 2003b). Cognitive deficits associated with placenta PCBs at age 9 years (Stewart et al., 2008). Behavioral deficits associated with cord blood PCBs in repeated testing up through age 9.5 years (Stewart et al., 2000, 2003a, 2005, 2006).

raised in less optimal home environments, but not in more advantageous home environments, at age 6.5 years (Vreugdenhil et al., 2002). Attentional deficits associated with maternal blood PCBs at age 9 yearsb (Vreugdenhil et al., 2004).

Sum of PCBs 138, 153, and 180 (cord blood and breast milk)

 Cognitive and motor development deficits associated with breast milk PCBs at ages 30 and 42 monthsc

Sum of PCBs 138, 153, and 180 (cord blood)

 Neurodevelopmental deficits not associated with cord blood PCBs at age 7 years, but there were some

e

Total PCBs (cord blood, maternal blood and breast milk)

(Walkowiak et al., 2001).

suggestions of interactions between PCBs and mercury (Grandjean et al., 2001).

 Cognitive deficits associated with cord blood and breast milk PCBs at ages 4f and 11 years (Jacobson et al., 1990b; Jacobson and Jacobson, 1996).

 Attention, behavior, and memory deficits associated with an index of maternal blood, cord blood, and breast milk PCBs at age 11 years, particularly among children who were not breast fed (Jacobson and Jacobson, 2003). North Carolina

Total PCBse (cord blood, maternal blood, and breast milk)

 Motor skills deficits associated with breast milk PCBs at ages 6, 12, and 24 months (Gladen et al., 1988; Rogan and Gladen, 1991).

 Cognitive skills not associated with breast milk PCBs at ages 3–5 years (Gladen and Rogan, 1991). a Breast milk samples in the Dutch study were analyzed for a large set of PCB congeners and dioxin-like chemicals, whereas only the four listed congeners were measured in the blood samples. Breast milk samples were available for only half of the cohort. Adverse neurological outcomes were more frequently associated with the blood PCBs than with breast milk PCBs (Schantz et al., 2003). b Attentional deficits at age 9 years were also associated with postnatal exposure via breast milk in the Dutch study (Vreugdenhil et al., 2004). c Cognitive deficits at age 42 months were also associated with postnatal exposure via breast milk in the German study (Walkowiak et al., 2001). d The primary goal of the Faroe Islands study is to evaluate the effects of mercury exposure on development, but effects of PCBs have also been evaluated. e Methods used in the Michigan and North Carolina studies did not measure individual congeners. f Cognitive deficits at age 4 years were also associated with postnatal exposure via breast milk in the Michigan study (Jacobson et al., 1990a).

Following the poisoning incidents, a number of epidemiologic studies have been conducted to examine the neurodevelopmental effects of PCBs at more typical exposure levels, including longitudinal cohort studies with assessments at several different ages conducted in Michigan, North Carolina, New York, the Netherlands, Germany and the Faroe Islands (Table 1). Other studies with neurodevelopmental evaluations only at a single age, or with neurodevelopmental results still forthcoming, include studies in Northern Quebec (Saint-Amour et al., 2006) and Massachusetts (Sagiv et al., 2008), and analyses of archived data from California (Hertz-Picciotto et al., 2005) and the Collaborative Perinatal Project (Gray et al., 2005). Overall, the epidemiologic studies indicate that prenatal PCB exposures are likely to be associated with deficits in cognition, attention, and behavior (Ribas-Fito et al., 2001; Schantz et al., 2003, 2004). In general, the studies do not provide any basis for distinguishing among the effects of different congeners, with the possible exception of the Oswego (New York) study, which conducted separate analyses for the sum of all PCBs measured in the study and for the sum of the more highly chlorinated congeners (Darvill et al., 2000; Schantz et al., 2003; Stewart et al., 2005). Most findings of associations between PCB exposure and neurodevelopment have been related to prenatal exposure, although some relationships with postnatal exposure have been detected (Jacobson et al., 1990a; Schantz et al., 2003; Vreugdenhil et al., 2004; Walkowiak et al., 2001); there is likely a bias to the null for assessments of postnatal exposure, due to exposure misclassification (Rice, 2004). Prenatal PCB exposures have also been associated with immunological endpoints (Dallaire et al.,

2004, 2006; Heilmann et al., 2006; Park et al., 2008; WeisglasKuperus et al., 2000, 2004) and with altered sex ratios (Karmaus et al., 2002; Taylor et al., 2007; Weisskopf et al., 2003). Because PCBs are widely distributed in the environment, are commonly found in many foods, and because they can adversely affect human health, it is important to monitor body burdens in the general population on a continuing basis. Beginning in 1999, NHANES has provided nationally representative PCB body burden data for the US population aged 12 years and older. Interpretation of these data is challenging due to the large number of congeners tested, the high frequency of non-detects for many congeners, and differences between the 1999–2000 and 2001–2002 data sets. Our goal is to develop summary quantitative measures of PCB body burdens in the US general population, relevant for characterizing changes over time in potential risks to children’s health. We conducted an analysis to identify preferred metrics for summarizing the NHANES PCB data. We focused on women of childbearing age (16–39 years) because of the scientific findings emphasizing the role of in utero exposures for neurodevelopmental effects, the lower frequency of non-detect values in this population compared with children aged 12 years and older, and the absence of data for younger children. 2. Methods 2.1. Data collection NHANES is a nationally representative survey of the health and nutritional status of the civilian non-institutionalized US population, conducted by the

ARTICLE IN PRESS 370

D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

Centers for Disease Control and Prevention (CDC). Interviews and physical examinations are conducted for approximately 5000 people each year. NonHispanic Blacks and Mexican Americans were oversampled in 1999–2002 to increase the reliability of estimates for these populations. Exposure to environmental chemicals is assessed through analysis of blood and urine samples. In 1999–2002, serum samples were analyzed for the presence of PCBs for one-third of participants aged 12 years and older (20 years and older for three coplanar PCB congeners in 2001–2002). Because of the complex stratified survey design, separate sample weights are assigned to each survey respondent; on average, each participant represents approximately 50,000 other US residents. Complete details regarding sampling procedures, survey questionnaires, as well as data files for download are available at the CDC NHANES website (http://www.cdc.gov/nchs/ nhanes.htm). In 1999–2000, 26 PCB congeners and in 2001–2002, 36 congeners were measured in NHANES. The limit of detection (LOD) was determined separately for each sample, due to differences by sample in the volume of blood available for analysis (Centers for Disease Control and Prevention, 2005). More sensitive methods were used for the analysis of many PCB congeners in the 2001–2002 survey, resulting in a greater proportion of samples with PCB levels above the LOD (Centers for Disease Control and Prevention, 2005). For example, among women aged 16–39 years, PCB-153 was detected in 78% of NHANES samples in 2001–2002, compared with only 14% detected in 1999–2000. Because of these differences between the 1999–2000 and 2001–2002 data sets, our analysis uses only the 2001–2002 PCB data. Although 36 PCB congeners were quantified in NHANES 2001–2002, only the 9 congeners with detectable levels in at least 20% of the samples for our target population were used for this analysis. This strikes a balance between the broader inclusion of congeners in the analysis and exclusion of congeners that may be less informative due to low detection frequencies. NHANES serum PCB measurements are reported in both wet weight and lipidadjusted terms. Because PCBs are lipophilic, PCB serum levels are strongly related to serum lipid concentrations. We used the NHANES lipid-adjusted values, as these are expected to reflect PCBs in body fat and thus better represent body burden than unadjusted values (Centers for Disease Control and Prevention, 2005). In epidemiologic studies, lipid adjustment may increase bias if factors such as laboratory methods, biological medium, and the true underlying relationships between PCBs, lipids, and health outcomes are not considered adequately, which can be accomplished through alternative methods of analysis such as including lipid levels as a covariate (Schisterman et al., 2005). For our analysis, which is only comparing exposure levels within a general population sample and is not evaluating associations with health outcomes, lipid adjustment is an appropriate method for accounting for variability in serum lipids. 2.2. PCB metrics We developed several exposure metrics, using different combinations of PCB congeners, based on a review of the scientific literature with particular emphasis on the longitudinal cohort studies of neurodevelopmental effects. The North Carolina and Michigan studies each used ‘‘total PCBs’’ as the exposure metric; methods used in these studies did not measure individual congeners, and are not comparable to more recent studies that sum the measurements of multiple congeners. The German and Faroe Islands studies summed together measurements of three congeners: PCBs 138, 153, and 180. The Dutch study took a similar approach, with a set of four congeners: PCBs 118, 138, 153, and 180. In the Oswego study, 68 individual PCB congeners or pairs of congeners were measured, and two exposure metrics were employed: the first sums the values of all 68 congeners and the second sums together 16 highly chlorinated congeners (7–9 chlorine atoms). Based on these studies, we designated the following exposure metrics for our analysis (Table 2): total PCBs (sum of 9 congeners); a 4-congener combination (the same congeners analyzed in the Dutch study); and highly chlorinated PCBs, as defined in the Oswego study. We also designated PCB-153 alone as an exposure metric, based on a study that used this congener to compare exposure levels across the different cohort studies (Longnecker et al., 2003).

We also considered the literature that identifies a subset of PCB congeners as possessing ‘‘dioxin-like’’ activity. These 12 congeners, with non-ortho- and monoortho-substituted structures, are included in risk assessments of dioxin (2,3,7, 8-tetrachlorodibenzo-p-dioxin) and dioxin-like chemicals (other chlorinated dioxins, chlorinated furans, and dioxin-like PCBs). The World Health Organization (WHO) has sponsored the development of toxic equivalency factors (TEFs) that represent the relative potency of each dioxin-like chemical (Van den Berg et al., 2006). For this analysis, we developed a dioxin-like PCBs metric that weights the relevant congeners by their WHO TEFs. We also constructed a ‘‘non-dioxin-like PCBs’’ metric consisting of the PCB congeners without TEFs; however, some congeners included in this metric may have some dioxin-like activity. The resulting 6 metrics are summarized in Table 2. All metrics are subject to the constraint that only congeners detected in at least 20% of the NHANES population of women aged 16–39 years are included. The indicated congeners are weighted equally in calculation of each metric, with the exception of the dioxinlike PCBs metric. 2.3. Statistical analysis The NHANES 2001–2002 lipid-adjusted serum concentrations for each PCB congener (ng/g), as well as demographic data and sample weights, were downloaded from the CDC website and imported into SAS version 9.1 for Windows (SAS Institute, Cary, NC). Demographic information such as race, sex, income, and age was linked to measured PCB levels by each participant’s unique identifier. We then calculated the distribution of PCB body burdens among women of childbearing age (16–39 years) for each of the 6 metrics, combining the NHANES measurements for each congener, the NHANES sample weight, and for the dioxinlike PCBs metric, the TEFs. For samples with levels below the LOD, we used the default value assigned by CDC of LOD/O2. We conducted a number of analyses to compare the PCB metrics. The correlation of the 6 metrics was calculated. For each metric, 50th and 95th percentile concentrations were computed for the entire group of women 16–39 years and for race/ethnicity and income sub-categories. Race/ethnicity sub-categories were defined as White non-Hispanic, Black non-Hispanic, and Mexican American. The poverty income ratio was used to define sub-categories of family income below the poverty level and greater than or equal to the poverty level. Differences in body burdens between subgroups were evaluated using logistic regression, and were considered statistically significant when the Satterthwaite-adjusted chi-square p-value was less than 0.05. We used SAS-Callable SUDAAN version 9.0.1 (Research Triangle Institute, RTP, NC) for these procedures to account for the complex survey design. To assess the influence of PCB measurements that were below the LOD, we recalculated the median and 95th percentile PCB concentrations for all PCB metrics with a value of zero for non-detects and compared these to the results derived using the value of LOD/O2. We also evaluated generational aspects of PCB exposure by comparing body burdens of PCBs across birth cohorts. This analysis considered both males and females of all ages in NHANES 2001–2002 (excluding the small number of births in the 1990s for which there are PCBs data) and compared the value of each metric across age groups defined by the decade in which the subject was born.

3. Results PCB body burden data were available for 496 women. All 9 congeners were detected in 45 women (9%); 166 women (32%) had detectable values of 5–8 congeners and 206 women (41%) had detectable values of 1–4 congeners. There were 79 women (17%) for whom all values were either non-detect or missing. PCBs 126 and 169 were each missing for 74 of these 79 women and for 238

Table 2 PCB metric definitions. PCB metrics

PCB 74

PCB 99

PCB 118

PCB 126

PCB 138a

PCB 153

PCB 169

PCB 170

PCB 180

Total PCBs Highly chlorinated PCBs Dioxin-like PCBsb Non-dioxin-like PCBsc Sum of 4 PCB congeners

X

X

X

X

X

X

X

X X

X X

X (3E-5)

X (0.1)

X

X

X X

X X

X

X X

a

X

X (0.03)

PCBs 138 and 158 could not be analytically separated; PCB-138 is expected to dominate in these measurements. Values in parentheses are toxic equivalency factors (TEFs) applied in calculation of the dioxin-like PCBs metric. c For purposes of this analysis, congeners that were not assigned a TEF in Van den Berg et al. (2006) are considered to be ‘‘non-dioxin-like.’’ PCB-74 shares some structural characteristics with the PCBs considered to be dioxin-like, and may have dioxin-like activity. PCB-118 has a TEF and is considered to be dioxin-like, but also has non-dioxin-like activity (Simon et al., 2007). b

ARTICLE IN PRESS D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

of the total 496 women. There were 178 participants aged 16–19 years; measurements of these two congeners were not attempted in NHANES 2001–2002 for these congeners in participants aged under 20. When measured, PCBs 126 and 169 were generally present at levels approximately 1000-fold lower than the measured levels of the other PCB congeners. Four other congeners had missing values for 1–6 women, and three congeners had no missing values. For women 16–39 years of age, our PCB metrics were generally highly correlated with each other, with r values being 0.84 and greater in most cases (Table 3). All exceptions were for the dioxinlike PCBs weighted by TEFs, which were more moderately correlated with other metrics (r ¼ 0.44–0.57). When the TEF weights were removed, the correlation of the summed dioxin-like PCBs with the other metrics increased to r ¼ 0.57–0.79. The median body burden for the total PCBs metric, considering all 9 congeners in our analysis, is 53 ng/g lipid, and the 95th percentile is 133 ng/g lipid (Fig. 1). The median for PCB-153 alone, which has been used as a marker to compare body burdens across studies, is 14 ng/g lipid, and the 95th percentile is 41 ng/g lipid. The 95th percentile PCB body burdens are generally about 3 times the median body burdens for each of our 6 metrics and across race/ethnicity and income groups. No significant body burden differences were detected between non-Hispanic Black women and non-Hispanic White women at either the 50th or 95th percentiles for any metric (Table 4). At the median, Mexican American women were found to have lower body burdens than both non-Hispanic White women and nonHispanic Black women for all metrics except for dioxin-like PCBs. Body burdens of Mexican American women at the 95th percentile were lower than those of non-Hispanic White women and nonHispanic Black women (Table 4), but these differences were not statistically significant. Among the Mexican American women, 77 were born in the US and 70 were born in Mexico. Median levels were 13–86% greater in US-born Mexican American women than in those born in Mexico; these differences were statistically significant only for PCB-153 (Table 5). At the 95th percentile, body burdens in USborn Mexican American women were 36–79% greater than in Mexico-born women for the dioxin-like PCBs and highly chlorinated PCBs; these differences were not statistically significant. Differences between the two groups were less than 10% for the remaining metrics. Median body burdens in US-born Mexican American women were significantly lower than those for nonHispanic Black and non-Hispanic White women for all metrics except for the dioxin-like PCBs and highly chlorinated PCBs. Calculated body burdens of women with incomes above poverty level were greater than those for lower-income women at the median and 95th percentiles, but the differences were not statistically significant for any metric. At median levels, the use of LOD/O2 resulted in metric values 8–48% greater than when non-detects were set equal to zero (Fig. 2). The difference was lower for metrics comprising congeners with lower rates of non-detects; there was a minimal

371

difference between the two methods for 4-congener metric. At the 95th percentile the methods resulted in the same values for each of the metrics. PCB metrics were also analyzed by birth cohort (men and women combined) to identify any changes in body burdens associated with birth decade. The data exhibit substantial differences, with those born more recently having lower PCB body burdens (Fig. 3). Body burdens of people born in the 1980s are generally 20–45% lower than those of people born in the 1970s, and 70–85% lower than those of people born in the 1940s. To evaluate the significance of birth cohort, we fit linear regression models for each PCB metric with decade of birth. We found that birth decade was highly significant (po0.0001) for all 6 metrics. Multivariate regression analysis was conducted to evaluate the association of the PCB metrics with race/ethnicity (non-Hispanic Black, Mexican American, other race/ethnicity), age, and poverty status (below poverty level). There was a negative and significant relationship between the PCB metrics and Mexican American, and a positive and significant relationship between the PCB metrics and age. All other variables, along with interaction terms, were not statistically significantly associated.

4. Discussion Regular and consistent measurements of the body burdens of environmental chemicals in the US population have become an important means for tracking changes in exposure. Body burdens of lead in children aged 1–5 years have been measured by CDC in NHANES since the 1970s; the decline over three decades has been used to illustrate the public health accomplishment (Meyer et al., 2003; US Environmental Protection Agency, 2003) and have become the basis for an important public health goal (US Department of Health and Human Services, 2000). More recently, body burdens of mercury in women of childbearing age have been used to characterize the potential risks of in utero exposures (Centers for Disease Control and Prevention, 2004; US Environmental Protection Agency, 2003). With NHANES data now available for PCBs, it is possible to use body burden data to track trends for another important threat to children’s neurodevelopment. In contrast to lead and mercury body burdens, which are characterized by one value per sampled individual, PCB body burdens involve a set of values per sampled individual, representing the many sampled congeners. Therefore, to track trends in PCB exposures, a method is needed to summarize the results for many PCB congeners into one (or a few) representative value. Our analysis proposes several options for aggregating the PCB data in existing and future NHANES data sets, based on evaluation of the NHANES 2001–2002 measurements. Our analysis is focused on this general population sample, and the comparison of body burden metrics may not be applicable to populations with unique exposure pathways that may involve different congener mixtures,

Table 3 Correlation of PCB body burden metrics among women aged 16–39 years.

Total PCBs Highly chlorinated PCBs Dioxin-like PCBsa Non-dioxin-like PCBs Sum of 4 PCB congeners PCB-153 a

Total PCBs

Highly chlorinated PCBs

Dioxin-like PCBs

Non-dioxin-like PCBs

Sum of 4 PCB congeners

PCB-153

1.000 0.896 0.793 [0.489] 0.989 0.994 0.971

1.000 0.570 [0.441] 0.837 0.898 0.870

1.000 0.771 [0.468] 0.765 [0.567] 0.691 [0.490]

1.000 0.985 0.974

1.000 0.981

1.000

Values in brackets are correlations for TEF-weighted dioxin-like PCBs. Other values shown for the dioxin-like PCBs are not TEF weighted.

ARTICLE IN PRESS 372

D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

or to studies in which detect frequencies are greater. We focused on exposures to women of childbearing age because, similar to mercury, the literature indicates greater concern for impacts on neurodevelopment from in utero exposure, rather than childhood exposure. In addition, there are fewer non-detect values in the samples from women than in the samples from children. The median level of PCB-153 for women of childbearing age (16–39 years) in 2001–2002 was approximately 14 ng/g lipid, and the 95th percentile was 41 ng/g lipid. Thus, the range of current US PCB exposures overlaps the observed range in several of the epidemiologic studies that have reported neurodevelopmental effects (Longnecker et al., 2003). For example, the median body burden of PCB-153 in the Oswego study (in serum, estimated from breast milk measurements) is 40 ng/g lipid (Longnecker et al., 2003). Although body burdens in the other epidemiologic studies were somewhat higher than in the Oswego study, the overlap of the NHANES data with the exposures in these studies indicates a need for continued monitoring of PCB body burdens in the US, continued evaluation of the effects of these exposures, and further efforts to reduce exposure. We specified 6 alternate PCB metrics for evaluation, based on review of the health effects literature. The 6 metrics were highly correlated with one another, and generally provided the same answers to the issues considered in our analysis. All metrics showed a significant relationship with age. Women in families

with income above poverty level had higher PCB body burdens than lower-income women, but these differences were not statistically significant. Mexican American women generally had lower PCB body burdens than non-Hispanic White and nonHispanic Black women, whereas there were no significant differences between the latter two groups. This part of the analysis considered only race/ethnicity, without controlling for covariates, to simply assess whether one group of women had higher or lower levels than another. A separate regression analysis accounting for race/ethnicity, age and income found that adjusting for other covariates did not change the basic findings regarding body burden differences by race/ethnicity. The observed differences in PCB body burdens by economic status and country of origin indicate a need for further research to identify factors that influence exposures. This issue illustrates a limitation of biomonitoring data, in that body burden measurements themselves currently provide no information on sources of exposure. Analysis of other data on NHANES participants, including dietary data, may provide some insights into differences by subpopulation. Several considerations may be relevant to selecting a preferred metric, including the relative risks of the congeners included in the different metrics, the performance of metrics in tracking body burdens over time, the breadth of coverage of each metric, and the impact of non-detects on each metric.

Total PCBs 95th Percentile

Total PCBs 50th Percentile White non-Hispanic

White non-Hispanic

Black non-Hispanic

Black non-Hispanic

Mexican American

Mexican American

Other Race/Ethnicity

Other Race/Ethnicity

All

All 0

10

20 All

30 40 ng/g lipid

> Poverty level

50

60

0

70

< Poverty level

All

White non-Hispanic

Black non-Hispanic

Black non-Hispanic

Mexican American

Mexican American

Other Race/Ethnicity

Other Race/Ethnicity

All

All

All

10 ng/g lipid > Poverty level

15

20

0

< Poverty level

White non-Hispanic

Black non-Hispanic

Black non-Hispanic

Mexican American

Mexican American

Other Race/Ethnicity

Other Race/Ethnicity

All

All 0.001

0.0015

ng/g lipid TEF weighted All

> Poverty level

80

100 120 140 160

< Poverty level

> Poverty level

< Poverty level

30 20 ng/g lipid > Poverty level

40

50

< Poverty level

Dioxin-like PCBs 95th Percentile

Dioxin-like PCBs 50th Percentile

0.0005

10

All

White non-Hispanic

0

60

Highly Chlorinated PCBs 95th Percentile

White non-Hispanic

5

40

ng/g lipid

Highly Chlorinated PCBs 50th Percentile

0

20

0.002

0 001 002 003 004 005 006 007 008 0. 0. 0. 0. 0. 0. 0. 0. ng/g lipid TEF weighted All

> Poverty level

Fig. 1. PCB body burden 50th and 95th percentiles for women aged 16–39 years, by race/ethnicity and income.

< Poverty level

ARTICLE IN PRESS D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

Non-Dioxin-like PCBs 50th Percentile

Non-Dioxin-like PCBs 95th Percentile

White non-Hispanic

White non-Hispanic

Black non-Hispanic

Black non-Hispanic

Mexican American

Mexican American

Other Race/Ethnicity

Other Race/Ethnicity

All

All 10

0

20 30 ng/g lipid

All

> Poverty level

40

< Poverty level

White non-Hispanic Black non-Hispanic

Mexican American

Mexican American

Other Race/Ethnicity

Other Race/Ethnicity

All

All 30

40

50

> Poverty level

60

0

20

40

> Poverty level

< Poverty level

All

> Poverty level

PCB 153 50th Percentile White non-Hispanic

Black non-Hispanic

Black non-Hispanic

Mexican American

Mexican American

Other Race/Ethnicity

Other Race/Ethnicity

All

All 5

10

15

0

20

10

20

ng/g lipid All

> Poverty level

120

100

120

140

< Poverty level

PCB 153 95th Percentile

White non-Hispanic

0

100

< Poverty level

60 80 ng/g lipid

ng/g lipid All

60 80 ng/g lipid

Sum of 4 PCB Congeners 95th Percentile

Black non-Hispanic

20

40

All

White non-Hispanic

10

20

0

50

Sum of 4 PCB Congeners 50th Percentile

0

373

< Poverty level

All

30 40 ng/g lipid

> Poverty level

50

60

< Poverty level

Fig. 1. (Continued)

Table 4 Tests for significance of differences in PCB body burdens between subgroups of women ages 16–39 yearsa. PCBs metric

Total PCBs Highly chlorinated PCBs Dioxin-like PCBs Non-dioxin-like PCBs Sum of 4 PCB congeners PCB-153

Non-Hispanic Whites vs. non-Hispanic Blacksb Odds ratio (p-value)

Non-Hispanic Whites vs. Mexican Americansc Odds ratio (p-value)

Non-Hispanic Blacks vs. Mexican Americansd Odds ratio (p-value)

Incomes above poverty level vs. incomes below poverty levele Odds ratio (p-value)

50th percentile

95th percentile

50th percentile

95th percentile

50th percentile

95th percentile

50th percentile

95th percentile

0.95 1.05 0.79 1.11 1.01 1.14

1.72 1.21 1.83 1.90 0.96 1.95

4.35 5.51 1.34 3.77 3.44 5.18

3.02 3.50 2.35 3.52 6.45 3.42

6.29 5.46 1.11 6.22 3.70 7.08

3.43 3.43 4.93 2.81 6.42 6.68

1.99 1.92 1.22 1.65 1.52 1.65

5.37 5.89 1.70 2.72 6.24 2.58

(0.85) (0.88) (0.53) (0.69) (0.98) (0.62)

(0.44) (0.77) (0.32) (0.37) (0.95) (0.26)

(0.0003) (0.0001) (0.31) (0.002) (0.0001) (0.0003)

(0.35) (0.23) (0.34) (0.27) (0.11) (0.29)

(0.0007) (0.002) (0.77) (0.001) (0.001) (0.001)

(0.19) (0.19) (0.17) (0.29) (0.12) (0.052)

(0.16) (0.17) (0.63) (0.29) (0.28) (0.31)

(0.14) (0.12) (0.56) (0.23) (0.11) (0.26)

a 496 women in total. 195 non-Hispanic White; 111 non-Hispanic Black; 147 Mexican American; 43 other race/multiracial. 132 with family incomes below poverty level; 338 with incomes above poverty level; 26 with unknown income. b Referent group: non-Hispanic Black women. c Referent group: Mexican American women. d Referent group: Mexican American women. e Referent group: women with family incomes below poverty level.

There is little conclusive information available to distinguish the relative risks of the different PCB congeners, apart from the identification of several congeners as possessing dioxin-like activity and the assignment of TEFs to these congeners. With the exception of the dioxin-like PCBs metric, each of the metrics

we examined weights each of the constituent congeners equally. A preferred approach would be to weight each of the congeners according to its toxic potency, in a manner similar to the TEF scheme but applying to other toxicological pathways and/or endpoints. Some limited data are available for comparing the toxic

ARTICLE IN PRESS 374

D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

Table 5 PCB body burdens (ng/g lipid) of Mexican American women aged 16–39 years, by country of birth (77 women born in the US, 70 women born in Mexico). PCB metric

Total PCBs Highly chlorinated PCBs Dioxin-like PCBs (TEF-weighted) Non-dioxin-like PCBs Sum of 4 PCB congeners PCB-153

50th percentile

95th percentile

Born in US

Born in Mexico

Born in US

Born in Mexico

41 10 0.0016 26 30 10

33 9 0.0013 20 21 5

95 34 0.0042 59 82 25

89 19 0.0031 55 75 24

 Significant difference at po0.05.

Fig. 2. PCB metrics in women aged 16–39 years with alternate assumptions for non-detect values: (a) 50th percentile; (b) 95th percentile. (Dioxin-like PCB values multiplied by 10,000 so that values appear in the same scale as other metrics.)

potency of PCB congeners from in vitro assays and animal experiments, and an approach to weighting PCB congeners by neurotoxic activity using available in vitro data has recently been proposed (Simon et al., 2007). We recalculated our total PCBs metric using the proposed weights, and found that the resulting values were very highly correlated with unweighted total PCBs

(r ¼ 0.99); thus, use of the proposed weights would appear unlikely to lead to any changes in our analysis. Performance of the PCBs metrics over time is difficult to project when considering only the 2001–2002 data, and this topic may warrant further examination when a time series can be evaluated. Given the persistence of PCBs, however, we found that

ARTICLE IN PRESS D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

375

Fig. 3. PCB body burdens by birth cohort (men and women combined): (a) 50th percentile; (b) 95th percentile. (Dioxin-like PCB values multiplied by 10,000 so that values appear in the same scale as other metrics.)

some analysis of time trends could be conducted using the 2001–2002 data by comparing body burdens for cohorts defined by decade of birth, considering men and women of all ages. This analysis further supports a conclusion that the various metrics provide a similar picture, as there was a significant decline in body burden by birth decade for each metric. The relationship between body burdens and age reflects two established characteristics of PCBs: first, environmental levels of PCBs have declined, so for example, 20-year olds today are exposed to lower concentrations of PCBs in food than were 20-year olds in past decades; and second, many PCB congeners are very persistent, and body burdens in the older portion of the population incorporate the accumulation of those PCBs over more years. Changes in diet may also play a role (Tee et al., 2003). The objectives of a greater breadth of coverage (i.e. number of congeners included) and reducing the impact of non-detects on the calculated metrics in large part pose a tradeoff in selection of a metric. Inclusion of more congeners may be desirable to best represent total PCB body burdens and to ensure that any congeners that may be important contributors to risk are

represented. However, broader metrics involve the inclusion of congeners with higher frequencies of non-detect values. There are two important drawbacks to the inclusion of congeners with more non-detects: increasing uncertainty in the calculated body burdens and reduced differentiation in body burdens across individuals. The latter problem arises because of assignment of an equivalent value (i.e. LOD/O2) to all non-detects; this could potentially be addressed by modeling the assignment of replacement values for each non-detect sample. After consideration of all the relevant factors, we believe that the 4-congener metric (PCBs 118, 138, 153, and 180) and the total PCBs metric are the most promising approaches for tracking changes in body burdens over time and for comparing body burdens of different subgroups. The 4-congener metric uses a combination of congeners that is well established in the epidemiologic literature; the four congeners are among those most frequently detected in the NHANES sample, and in other studies as well; and it is well-correlated with the other PCB metrics examined. A recent epidemiologic study concluded that the 4-congener combination was an appropriate representation of

ARTICLE IN PRESS 376

D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

exposure to all PCB congeners, based on comparison to a metric that summed together 51 congeners (Sagiv et al., 2008). The total PCBs metric has the broadest coverage and thus provides the strongest basis for detecting any unanticipated changes in body burdens. It may be useful to separately track changes in dioxinlike PCBs, as this metric may capture potential risks that are different in nature from those of the other PCB congeners, and incorporates weighting by potency of two congeners (PCBs 126 and 169) that are present in comparatively low concentrations, but that have high biological activity. However, these risks may be most usefully considered along with the broader set of dioxinlike chemicals measured in NHANES. Among women aged 20–39 years, the dioxin-like PCBs account for 13–20%, depending on race/ethnicity, of the geometric mean total toxic equivalent (TEQ) body burden of dioxin-like chemicals in NHANES 2001–2002 (Patterson et al., 2008). Other combinations of congeners could have been considered in our analysis. We conducted some evaluation of variants on our 4-congener measure, which adopted the 4 congeners measured in the Dutch study. One alternative would be to remove PCB-118 and use a 3-congener metric. PCB-118 has the lowest detection frequency of the 4 congeners; the remaining 3 congeners would be the same set measured in the German and Faroe Islands studies. However, we found that PCB-118 makes a considerable contribution to the metric value for most individuals (greater than 10% of value for 90% of population); we also found that this relative contribution varies among the NHANES participants (as high as 40% of the value of this metric for some individuals). Thus, we concluded that PCB-118 should be retained in this metric. We also considered expanding the 4-congener metric to a 6-congener metric by the addition of PCBs 126 and 169. The resulting metric would include the 6 congeners with the highest rates of detection. However, the two additional congeners are generally found at levels 100- to 1000-fold lower than PCBs 118, 138, 153, and 180 (Centers for Disease Control and Prevention, 2005). Thus, there was only negligible difference between the 4-congener and 6-congener metrics, and we concluded that there was little value in expanding the 4-congener metric by adding PCBs 126 and 169. The latter two congeners may be considered important due to dioxin-like activity, but including them in a metric with other congeners present in human samples at much higher levels will not capture their potential contributions to the health risks of PCBs. PCB-118 is also a dioxin-like PCB, but is found at levels only two-fold lower than PCBs 138, 153, and 180 in the NHANES samples. We also considered possible advantages and disadvantages of narrowing the dioxin-like PCBs metric to just PCB-126, which has a larger TEF than PCBs 118 and 169 (Table 2). We found that whereas PCB-126 generally accounts for most of the total value of the TEF-weighted dioxin-like PCBs metric, PCBs 118 and 169 make non-negligible contributions to the metric for many individuals (more than a quarter of the metric value for about 40% of the population). We therefore concluded that it is valuable to include all three congeners in this metric. However, the substantial number of sampled women with missing data for PCBs 126 and 169, particularly those aged 16–19 years, is an important limitation to this metric. In addition, this metric will not necessarily correspond well to total TEQ exposure, considering exposure to dioxins and other dioxin-like chemicals, as body burdens of dioxin-like PCBs will not necessarily be well-correlated with body burdens of the other contributors to total TEQ. The effects of PCBs may result from different mechanistic pathways for different PCB congeners. Our analysis incorporates one approach to categorizing PCBs by mechanism by specifying metrics for dioxin-like PCBs and non-dioxin-like PCBs. The total PCBs metric and the 4-congener metric each include both

dioxin-like and non-dioxin-like PCBs. A recent study found that rats exposed to a mixture of dioxins, furans, dioxin-like PCBs and non-dioxin-like PCBs displayed additive or greater-than-additive effects on thyroid hormone levels (Crofton et al., 2005). This provides further support for inclusion of both dioxin-like and nondioxin-like PCBs in a single metric, as PCBs may operate by multiple modes of action that contribute to common outcomes. Other PCBs literature suggests additional metrics that could be considered. Several recent publications have proposed and/or applied groupings of PCB congeners. For example, several authors have proposed groups of congeners based on mechanistic data, considering estrogenic, anti-estrogenic, and enzyme-inducing activity (Cooke et al., 2001; DeCastro et al., 2006; Moysich et al., 1999; Wolff et al., 1997). Some of these categories have been applied in epidemiologic studies of PCBs (Buck Louis et al., 2005; Chevrier et al., 2007; Denham et al., 2005; Hauser et al., 2005; Herbstman et al., 2008; Taylor et al., 2007) and some associations with relevant effects have been found, lending support to the proposed groupings. However, the different proposed grouping schemes are not fully consistent with one another, and further evaluation may be needed to develop consensus approaches. Further, many of the congeners considered in these classifications had high rates of non-detects in NHANES 2001–2002, and thus were not included in our analysis. In addition, our analysis looks at a reasonably broad range of metrics, and results are similar for each metric (in terms of differences/similarities in body burdens across race/ethnicity, income and age). It is unlikely that other metrics would have different results. Specification of PCB metrics is constrained by the high frequency of non-detects for most sampled congeners; further examination of alternate metrics may be useful if future versions of the NHANES survey have lower rates of non-detects. This would occur if more sensitive analytic techniques are applied, as it is expected that PCB body burdens overall will be declining in the coming years. More sensitive measurement techniques have been applied in the PCB epidemiology studies; however, measurement sensitivity for the NHANES samples is constrained by the volume of blood available and the need to analyze samples for the presence of a large number of chemicals. Previous analyses of NHANES mercury body burden data defined women of childbearing age to be in the range of 16–49 years. Women aged 40–49 years accounted for 30% of the female population aged 16–49 years in the US in 2000 (US Census Bureau, 2008), but only 2.5% of births in the US were to women in this age range (Centers for Disease Control and Prevention, 2008). Women aged 40–49 years also have higher PCB body burdens than younger women; thus application of the NHANES population weights would increase the influence of women aged 40–49 years on the calculated body burden statistics for women of childbearing age, relative to the number of births to women these ages. By narrowing the age range to 16–39 years, we still include women who give birth to 97.5% of babies each year. The choice of age range substantially affected our results: body burdens for all metrics and all strata considered were lower with the narrower age range, and differences between subpopulations defined by race/ethnicity and income were smaller. In many cases, subgroup differences that were statistically significant for women aged 16–49 years were not significant when only ages 16–39 were included. A more refined analysis could consider re-weighting the data using fertility rates. Although our analysis focuses on a subset of 9 PCB congeners, continued measurement of a broader set of congeners in NHANES is important, since the 2001–2002 data cannot be presumed to predict future findings. It is conceivable that body burden levels or rates of detection for some congeners could increase in future NHANES samples. In the absence of continued monitoring,

ARTICLE IN PRESS D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

unanticipated changes of potential public health significance would go undetected. In addition, further study of the health effects of PCBs may determine that some of these less-frequently detected congeners present greater health risks than those in our subset. An important implication of the NHANES body burden data is that the general population of the US is continually exposed to a complex mixture of environmental chemicals, including metals, persistent and non-persistent pesticides, dioxins and furans, PCBs, phthalates, and environmental tobacco smoke. Ultimately, it would be informative to have a method for considering the combined significance of these multiple exposures for childhood neurological development and other types of outcomes. At this time, an important step is to consider the combination of chemicals within a single class. This analysis of PCBs proposes alternatives for condensing data for multiple PCB congeners into a single summary measure, recognizing that any particular method will have limitations. Aspects of this approach may be applicable to other categories of chemicals reported in NHANES, including phthalates, polybrominated diphenyl ethers, and perfluorinated compounds.

Acknowledgments We appreciate the many helpful suggestions from Linda Birnbaum, Danelle Lobdell, Marian Olsen, Margaret McDonough, Pauline Mendola, and Deborah Rice, as well as the statistical advice from Jonathan Cohen. References Buck Louis, G.M., Weiner, J.M., Whitcomb, B.W., Sperrazza, R., Schisterman, E.F., Lobdell, D.T., et al., 2005. Environmental PCB exposure and risk of endometriosis. Hum. Reprod. 20, 279–285. Centers for Disease Control and Prevention, 2004. Blood mercury levels in young children and childbearing-aged women—United States, 1999–2002. Morb. Mortal. Wkly. Rep. 53, 1018–1020. Centers for Disease Control and Prevention, 2005. Third National Report on Human Exposure to Environmental Chemicals. CDC, Atlanta, GA. Centers for Disease Control and Prevention, 2008. CDC Wonder natality public-use data for 2000, /http://wonder.cdc.gov/natality-v2002.htmlS. Chen, Y.C., Guo, Y.L., Hsu, C.C., Rogan, W.J., 1992. Cognitive development of Yu-Cheng (‘‘oil disease’’) children prenatally exposed to heat-degraded PCBs. JAMA 268, 3213–3218. Chen, Y.J., Hsu, C.C., 1994. Effects of prenatal exposure to PCBs on the neurological function of children: a neuropsychological and neurophysiological study. Dev. Med. Child Neurol. 36, 312–320. Chevrier, J., Eskenazi, B., Bradman, A., Fenster, L., Barr, D.B., 2007. Associations between prenatal exposure to polychlorinated biphenyls and neonatal thyroidstimulating hormone levels in a Mexican American population, Salinas Valley, California. Environ. Health Perspect. 115, 1490–1496. Cooke, P.S., Sato, T., Buchanan, D.L., 2001. Disruption of steroid hormone signaling by PCBs. In: Robertson, L.W., Hansen, L.G. (Eds.), PCBs: Recent Advances in Environmental Toxicology and Health Effects. The University Press of Kentucky, pp. 257–263. Crofton, K.M., Craft, E.S., Hedge, J.M., Gennings, C., Simmons, J.E., Carchman, R.A., et al., 2005. Thyroid-hormone-disrupting chemicals: evidence for dosedependent additivity or synergism. Environ. Health Perspect. 113, 1549–1554. Dallaire, F., Dewailly, E., Muckle, G., Vezina, C., Jacobson, S.W., Jacobson, J.L., et al., 2004. Acute infections and environmental exposure to organochlorines in Inuit infants from Nunavik. Environ. Health Perspect. 112, 1359–1365. Dallaire, F., Dewailly, E., Vezina, C., Muckle, G., Weber, J.P., Bruneau, S., et al., 2006. Effect of prenatal exposure to polychlorinated biphenyls on incidence of acute respiratory infections in preschool Inuit children. Environ. Health Perspect. 114, 1301–1305. Darvill, T., Lonky, E., Reihman, J., Stewart, P., Pagano, J., 2000. Prenatal exposure to PCBs and infant performance on the fagan test of infant intelligence. Neurotoxicology 21, 1029–1038. DeCastro, B.R., Korrick, S.A., Spengler, J.D., Soto, A.M., 2006. Estrogenic activity of polychlorinated biphenyls present in human tissue and the environment. Environ. Sci. Technol. 40, 2819–2825. Denham, M., Schell, L.M., Deane, G., Gallo, M.V., Ravenscroft, J., DeCaprio, A.P., 2005. Relationship of lead, mercury, mirex, dichlorodiphenyldichloroethylene, hexachlorobenzene, and polychlorinated biphenyls to timing of menarche among Akwesasne Mohawk girls. Pediatrics 115, e127–e134.

377

Gladen, B.C., Rogan, W.J., 1991. Effects of perinatal polychlorinated biphenyls and dichlorodiphenyl dichloroethene on later development. J. Pediatr. 119, 58–63. Gladen, B.C., Rogan, W.J., Hardy, P., Thullen, J., Tingelstad, J., Tully, M., 1988. Development after exposure to polychlorinated biphenyls and dichlorodiphenyl dichloroethene transplacentally and through human milk. J. Pediatr. 113, 991–995. Grandjean, P., Weihe, P., Burse, V.W., Needham, L.L., Storr-Hansen, E., Heinzow, B., et al., 2001. Neurobehavioral deficits associated with PCB in 7-year-old children prenatally exposed to seafood neurotoxicants. Neurotoxicol. Teratol. 23, 305–317. Gray, K.A., Klebanoff, M.A., Brock, J.W., Zhou, H., Darden, R., Needham, L., et al., 2005. In utero exposure to background levels of polychlorinated biphenyls and cognitive functioning among school-age children. Am. J. Epidemiol. 162, 17–26. Hauser, R., Williams, P., Altshul, L., Calafat, A.M., 2005. Evidence of interaction between polychlorinated biphenyls and phthalates in relation to human sperm motility. Environ. Health Perspect. 113, 425–430. Heilmann, C., Grandjean, P., Weihe, P., Nielsen, F., Budtz-Jorgensen, E., 2006. Reduced antibody responses to vaccinations in children exposed to polychlorinated biphenyls. PLoS Med. 3, e311. Herbstman, J.B., Sjodin, A., Apelberg, B.J., Witter, F.R., Halden, R.U., Patterson, D.G., et al., 2008. Birth delivery mode modifies the associations between prenatal polychlorinated biphenyl (PCB) and polybrominated diphenyl ether (PBDE) and neonatal thyroid hormone levels. Environ. Health Perspect. 116, 1376–1382. Hertz-Picciotto, I., Charles, M.J., James, R.A., Keller, J.A., Willman, E., Teplin, S., 2005. In utero polychlorinated biphenyl exposures in relation to fetal and early childhood growth. Epidemiology 16, 648–656. Hickey, J.P., Batterman, S.A., Chernyak, S.M., 2006. Trends of chlorinated organic contaminants in great lakes trout and walleye from 1970 to 1998. Arch. Environ. Contam. Toxicol. 50, 97–110. Jacobson, J.L., Jacobson, S.W., 1996. Intellectual impairment in children exposed to polychlorinated biphenyls in utero. N. Engl. J. Med. 335, 783–789. Jacobson, J.L., Jacobson, S.W., 2003. Prenatal exposure to polychlorinated biphenyls and attention at school age. J. Pediatr. 143, 780–788. Jacobson, J.L., Jacobson, S.W., Humphrey, H.E., 1990a. Effects of exposure to PCBs and related compounds on growth and activity in children. Neurotoxicol. Teratol. 12, 319–326. Jacobson, J.L., Jacobson, S.W., Humphrey, H.E., 1990b. Effects of in utero exposure to polychlorinated biphenyls and related contaminants on cognitive functioning in young children. J. Pediatr. 116, 38–45. Karmaus, W., Huang, S., Cameron, L., 2002. Parental concentration of dichlorodiphenyl dichloroethene and polychlorinated biphenyls in Michigan fish eaters and sex ratio in offspring. J. Occup. Environ. Med. 44, 8–13. Lai, T.J., Liu, X., Guo, Y.L., Guo, N.W., Yu, M.L., Hsu, C.C., et al., 2002. A cohort study of behavioral problems and intelligence in children with high prenatal polychlorinated biphenyl exposure. Arch. Gen. Psychiatry 59, 1061–1066. Longnecker, M.P., Wolff, M.S., Gladen, B.C., Brock, J.W., Grandjean, P., Jacobson, J.L., et al., 2003. Comparison of polychlorinated biphenyl levels across studies of human neurodevelopment. Environ. Health Perspect. 111, 65–70. Meyer, P.A., Pivetz, T., Dignam, T.A., Homa, D.M., Schoonover, J., Brody, D., 2003. Surveillance for elevated blood lead levels among children—United States, 1997–2001. MMWR Surveill. Summ. 52, 1–21. Moysich, K.B., Mendola, P., Schisterman, E.F., Freudenheim, J.L., Ambrosone, C.B., Vena, J.E., et al., 1999. An evaluation of proposed frameworks for grouping polychlorinated biphenyl (PCB) congener data into meaningful analytic units. Am. J. Ind. Med. 35, 223–231. Park, H., Hertz-Picciotto, I., Petrik, J., Palkovicova, L., Kocan, A., Trnovec, T., 2008. Prenatal PCB exposure and thymus size at birth in neonates in eastern Slovakia. Environ. Health Perspect. 116, 104–109. Patandin, S., Lanting, C.I., Mulder, P.G., Boersma, E.R., Sauer, P.J., Weisglas-Kuperus, N., 1999. Effects of environmental exposure to polychlorinated biphenyls and dioxins on cognitive abilities in Dutch children at 42 months of age. J. Pediatr. 134, 33–41. Patterson Jr., D.G., Turner, W.E., Caudill, S.P., Needham, L.L., 2008. Total TEQ reference range (PCDDs, PCDFs, cPCBs, mono-PCBs) for the US population 2001–2002. Chemosphere 73, S261–S277. Ribas-Fito, N., Sala, M., Kogevinas, M., Sunyer, J., 2001. Polychlorinated biphenyls (PCBs) and neurological development in children: a systematic review. J. Epidemiol. Community Health 55, 537–546. Rice, D.C., 2004. Evidence for neuropsychological effects of PCBs in environmental studies: getting better all the time. Psychol. Schools 41, 693–707. Rogan, W.J., Gladen, B.C., 1991. PCBs, DDE, and child development at 18 and 24 months. Ann. Epidemiol. 1, 407–413. Rogan, W.J., Gladen, B.C., Hung, K.L., Koong, S.L., Shih, L.Y., Taylor, J.S., et al., 1988. Congenital poisoning by polychlorinated biphenyls and their contaminants in Taiwan. Science 241, 334–336. Sagiv, S.K., Nugent, J.K., Brazelton, T.B., Choi, A.L., Tolbert, P.E., Altshul, L.M., et al., 2008. Prenatal organochlorine exposure and measures of behavior in infancy using the Neonatal Behavioral Assessment Scale (NBAS). Environ. Health Perspect. 116, 666–673. Saint-Amour, D., Roy, M.S., Bastien, C., Ayotte, P., Dewailly, E., Despres, C., et al., 2006. Alterations of visual evoked potentials in preschool Inuit children exposed to methylmercury and polychlorinated biphenyls from a marine diet. Neurotoxicology 27, 567–578.

ARTICLE IN PRESS 378

D.A. Axelrad et al. / Environmental Research 109 (2009) 368–378

Schantz, S.L., Gardiner, J.C., Gasior, D.M., McCaffrey, R.J., Sweeney, A.M., Humphrey, H.E.B., 2004. Much ado about something: the weight of evidence for PCB effects on neuropsychological function. Psychol. Schools 41, 669–679. Schantz, S.L., Widholm, J.J., Rice, D.C., 2003. Effects of PCB exposure on neuropsychological function in children. Environ. Health Perspect. 111, 357–376. Schisterman, E.F., Whitcomb, B.W., Louis, G.M., Louis, T.A., 2005. Lipid adjustment in the analysis of environmental contaminants and human health risks. Environ. Health Perspect. 113, 853–857. Simon, T., Britt, J.K., James, R.C., 2007. Development of a neurotoxic equivalence scheme of relative potency for assessing the risk of PCB mixtures. Regul. Toxicol. Pharmacol. 48, 148–170. Stewart, P., Fitzgerald, S., Reihman, J., Gump, B., Lonky, E., Darvill, T., et al., 2003a. Prenatal PCB exposure, the corpus callosum, and response inhibition. Environ. Health Perspect. 111, 1670–1677. Stewart, P., Lonky, E., Reihman, J., Pagano, J., Gump, B., Darvill, T., 2008. The relationship between prenatal PCB exposure and intelligence (IQ) in 9-year-old children. Environ. Health Perspect. 116, 1416–1422. Stewart, P., Reihman, J., Gump, B., Lonky, E., Darvill, T., Pagano, J., 2005. Response inhibition at 8 and 9 1/2 years of age in children prenatally exposed to PCBs. Neurotoxicol. Teratol. 27, 771–780. Stewart, P., Reihman, J., Lonky, E., Darvill, T., Pagano, J., 2000. Prenatal PCB exposure and neonatal behavioral assessment scale (NBAS) performance. Neurotoxicol. Teratol. 22, 21–29. Stewart, P.W., Reihman, J., Lonky, E.I., Darvill, T.J., Pagano, J., 2003b. Cognitive development in preschool children prenatally exposed to PCBs and MeHg. Neurotoxicol. Teratol. 25, 11–22. Stewart, P.W., Sargent, D.M., Reihman, J., Gump, B.B., Lonky, E., Darvill, T., et al., 2006. Response inhibition during differential reinforcement of low rates (DRL) schedules may be sensitive to low-level polychlorinated biphenyl, methylmercury, and lead exposure in children. Environ. Health Perspect. 114, 1923–1929. Sun, P., Basu, I., Blanchard, P., Brice, K.A., Hites, R.A., 2007. Temporal and spatial trends of atmospheric polychlorinated biphenyl concentrations near the Great Lakes. Environ. Sci. Technol. 41, 1131–1136. Taylor, K.C., Jackson, L.W., Lynch, C.D., Kostyniak, P.J., Buck Louis, G.M., 2007. Preconception maternal polychlorinated biphenyl concentrations and the secondary sex ratio. Environ. Res. 103, 99–105. Tee, P.G., Sweeney, A.M., Symanski, E., Gardiner, J.C., Gasior, D.M., Schantz, S.L., 2003. A longitudinal examination of factors related to changes

in serum polychlorinated biphenyl levels. Environ. Health Perspect. 111, 702–707. US Census Bureau, 2008. Census 2000 Summary File 1, Matrices P13 and PCT12, /http://factfinder.census.gov/servlet/QTTable?_bm=y&-geo_id=01000US&-qr_ name=DEC_2000_SF1_U_QTP1&-ds_name=DEC_2000_SF1_US. US Department of Health and Human Services, 2000. Healthy People 2010, /http://www.healthypeople.gov/Document/HTML/Volume1/08Environmental. htm#_Toc490564710S. US Environmental Protection Agency, 2003. America’s Children and the Environment: Measures of Contaminants, Body Burdens, and Illnesses, /http://www. epa.gov/envirohealth/childrenS. Van den Berg, M., Birnbaum, L.S., Denison, M., De Vito, M., Farland, W., Feeley, M., et al., 2006. The 2005 World Health Organization reevaluation of human and mammalian toxic equivalency factors for dioxins and dioxin-like compounds. Toxicol. Sci. 93, 223–241. Vreugdenhil, H.J., Lanting, C.I., Mulder, P.G., Boersma, E.R., Weisglas-Kuperus, N., 2002. Effects of prenatal PCB and dioxin background exposure on cognitive and motor abilities in Dutch children at school age. J. Pediatr. 140, 48–56. Vreugdenhil, H.J., Mulder, P.G., Emmen, H.H., Weisglas-Kuperus, N., 2004. Effects of perinatal exposure to PCBs on neuropsychological functions in the Rotterdam cohort at 9 years of age. Neuropsychology 18, 185–193. Walkowiak, J., Wiener, J.A., Fastabend, A., Heinzow, B., Kramer, U., Schmidt, E., et al., 2001. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on psychodevelopment in early childhood. Lancet 358, 1602–1607. Weisglas-Kuperus, N., Patandin, S., Berbers, G.A., Sas, T.C., Mulder, P.G., Sauer, P.J., et al., 2000. Immunologic effects of background exposure to polychlorinated biphenyls and dioxins in Dutch preschool children. Environ. Health Perspect. 108, 1203–1207. Weisglas-Kuperus, N., Vreugdenhil, H.J., Mulder, P.G., 2004. Immunological effects of environmental exposure to polychlorinated biphenyls and dioxins in Dutch school children. Toxicol. Lett. 149, 281–285. Weisskopf, M.G., Anderson, H.A., Hanrahan, L.P., 2003. Decreased sex ratio following maternal exposure to polychlorinated biphenyls from contaminated Great Lakes sport-caught fish: a retrospective cohort study. Environ. Health 2, 2. Wolff, M.S., Camann, D., Gammon, M., Stellman, S.D., 1997. Proposed PCB congener groupings for epidemiological studies. Environ. Health Perspect. 105, 13–14.