Abstract

The relation between exposure to low levels of polychlorinated biphenyls (PCBs), a class of persistent organic pollutants, and cognitive and motor development in young children has been examined in several studies, and results have varied. The authors evaluated the association between prenatal exposure to PCBs and children’s neurodevelopment using data from the Collaborative Perinatal Project. Pregnant women were enrolled from 1959 to 1965 from 12 sites across the United States. PCBs were measured in maternal serum taken during pregnancy. To measure children’s mental and psychomotor development at 8 months of age, the authors administered the Bayley Scales of Infant Development (means, 87 (standard deviation, 15) and 88 (standard deviation, 18), respectively). Overall, they did not observe a relation between prenatal PCB exposure and children’s mental or psychomotor scores (n = 1,207; multivariate adjusted β = 0.1 point per µg/liter increase of PCB, p = 0.71, and β = 0.5, p = 0.14, respectively). The PCB-psychomotor score relation varied by study center (p < 0.05): The association was direct in some centers, inverse in others. This could not be attributed to variation in the timing or measurement of the child’s neurodevelopment or analysis of PCBs because these were standardized across centers. The reasons for variation in results within this study and across other studies remain unclear.

Received for publication February 7, 2002; accepted for publication October 1, 2002.

Polychlorinated biphenyls (PCBs) are chemical mixtures persistent in our environment and are detectable in most humans from developed countries (1, 2). Animal studies have found that certain PCB congeners and PCB commercial mixtures act as neurotoxic agents, altering synaptic transmission and reducing dopamine concentration (36). In humans, high-level exposure to thermally degraded PCBs in utero, through maternal intake of contaminated rice oil, has been associated with decreased neurodevelopment among children (7, 8).

Studies of low-level exposure to PCBs in utero in relation to neurodevelopment have produced various results. Among infants, the Bayley Scales of Infant Development have been among the most frequently used instruments for assessing mental and psychomotor development. Low-level exposure to PCBs in utero has been modestly associated with a decreased Bayley Psychomotor Development Index (PDI) among infants in studies from the United States and Europe (911). In the Netherlands, however, the inverse association between PCBs and the PDI observed among children at 3 months of age had attenuated by 7 months (11); and in Michigan, no association was observed among children 5 months of age (12). The Mental Development Index (MDI), measured by the Bayley Scales, has been inversely associated with PCBs in only one cohort in Germany (13). Delays in other aspects of development, such as infants’ visual recognition memory measured by the Fagan Test of Infant Intelligence and intelligence quotient measured among older children, have been detected in relation to PCB exposure in some studies (12, 14, 15) but not confirmed in others (16). Inconsistencies among results have been attributed to variation in the age at examination, examination protocol, variation in the total concentration or composition of PCB mixtures, the inability to distinguish acute high-level exposure from chronic low-level exposure in total PCB measures (which may present different associated risks), and limited control for social or environmental influences (14, 17).

We evaluated the relation between PCB exposure and mental and psychomotor development at 8 months of age, using the Bayley Scales of Infant Development in a cohort of children born to women participating in the Collaborative Perinatal Project. Participants from the 12 study centers offered a large demographically diverse population whose PCB levels varied within the background range of 40 years ago, and for whom the Bayley assessment instrument was administered to offspring in a standardized manner.

MATERIALS AND METHODS

Study population

The Collaborative Perinatal Project followed the growth and neurologic development of approximately 55,000 children born in 1959 through 1966 (18). Approximately 42,000 pregnant women were recruited into the Collaborative Perinatal Project from 12 hospitals in the United States (Baltimore, Maryland; Boston, Massachusetts; Buffalo, New York; Memphis, Tennessee; Minneapolis, Minnesota; New Orleans, Louisiana; New York, New York (two hospitals); Philadelphia, Pennsylvania; Portland, Oregon; Providence, Rhode Island; and Richmond, Virginia). Recruitment methods varied across study centers. In some centers, all women attending the prenatal clinic were invited to participate, while other centers used systematic sampling schemes based on patient identification numbers or presentation order in the clinic. Women were ineligible if they were incarcerated, if they were planning to move from the area or to give the child up for adoption, or if they gave birth on the day they were recruited into the study. Records were not kept for women who refused to participate at baseline. At registration, characteristics of the women in the sample were essentially the same as those in the sampling frame. Four percent of those enrolled were lost to follow-up before delivery. Eighty-two percent of those born during the study were followed through 8 months of age.

Children were eligible for the present study if they were a liveborn singleton, had a 3-ml third trimester maternal serum sample available, and had completed the Bayley Scales of Infant Development at the age of 8 months. From the 43,628 eligible children, 1,065 were selected at random. An additional 194 children were selected at random from among the children scoring more than 1 standard deviation below or above the mean Bayley Psychomotor Development Index in the total study population. If some or all subjects are selected using outcome-dependent sampling and the data are analyzed using a newly available statistical procedure, then greater statistical power can be obtained than if the total sample were selected using simple random sampling (19). In this report, however, we used standard, more straightforward, methods to account for the sampling, because additional power would not have altered our main conclusions.

Exposure assessment

PCB levels have been shown to be highly correlated between maternal and cord serum at birth (r > 0.7) (20). Maternal nonfasting blood was collected every 8 weeks during pregnancy, at delivery, and 6 weeks postpartum. Sera were stored in glass at –20°C, with no recorded thaws. Because third trimester samples were available for the greatest number of mothers, these samples were analyzed for PCBs. The correlations for lipid-adjusted PCBs among serial measures, such as first and third trimester specimens, were high (Pearson’s r ≥ 0.77) (21). Eleven PCB congeners (International Union of Pure and Applied Chemistry designations: PCBs 28, 52, 74, 105, 118, 138, 153, 170, 180, 194, 203) were analyzed in 1997–1999 using solid-phase extraction followed by dual-column gas chromatography with electron capture detection at the Centers for Disease Control and Prevention (22). Total PCB exposure represented the sum of the 11 measured congeners, expressed as µg/liter of serum (“wet weight”). Specimens were analyzed in batches of 10, each including an aliquot from a single large pool of serum to assess interbatch variation (19 percent at a total PCB concentration of 3.49 µg/liter). Serum cholesterol and tri-glyceride levels, which affect serum PCB levels, were analyzed using standard enzymatic assays.

Outcome measurement

The Bayley Scales of Infant Development were used to assess the infants’ mental and psychomotor development at 8 months of age. The Bayley MDI assesses age-appropriate cognitive, language, and social development. The Bayley PDI assesses fine and gross motor development (23). Testing procedures were standardized across centers and conducted by specially trained personnel. Dr. Bayley made regular site visits to monitor testing procedures (24). Tests were edited on site and again after the forms were received by the National Institutes of Health in Bethesda, Maryland. Raw scores were converted to age-standardized scores for these analyses (23). Most infants (93 percent) were tested between 7.5 and 9 months of age. Seven children (0.6 percent) were tested before 7.5 months; the remaining 6.4 percent were tested between 9 and 10 months of age.

Data analysis

The relations between maternal third trimester PCB levels and the child’s standardized MDI and PDI were evaluated using linear regression analysis with PCBs modeled as a continuous variable. The data met the assumptions necessary to support the use of linear models. The PCB-neurodevelopment relation was estimated for children from all centers combined and by each study center separately.

Total PCBs were also categorized by concentration intervals of 1.25 µg/liter to produce five groups. The interval for the highest exposure category was wider, however, in order to achieve ample representation for analysis. Adjusted mean developmental scores (MDI and PDI) were calculated for each category of maternal PCB using generalized linear models. Trend significance was also evaluated by linear regression of an ordinal variable that represented the five levels of PCB exposure. All primary analyses were weighted by the inverse sampling probability to account for the sampling scheme (25).

Covariates were evaluated as potential confounders in this analysis if they were important in the design of the study, were related to both PCBs and neurodevelopment in these data, or had been reported as confounders in other comparable studies. The covariates evaluated included study center; maternal race (White, Black, other); education (less than high school, high school, greater than high school); socioeconomic index (continuous scale); intelligence quotient (continuous); marital status (married, single, divorced, widowed); prenatal smoking (none, <10 cigarettes per day, ≥10 cigarettes per day); prepregnancy body mass index (underweight, normal, overweight, obese); third trimester serum triglyceride, total cholesterol, and dichlorodiphenyldichloroethylene (DDE) levels (continuous); the child’s birth order (first born compared with others); gestational age (continuous); and whether the child ever breastfed (yes/no). Only the design variable (center) and covariates that changed the effect estimate for the PCB-neurodevelopment relation by more than 10 percent in these data were included as confounders in the final models presented.

To assess heterogeneity in the PCB-neurodevelopment relation among centers, we added product terms for PCB (continuous) by each study center to the primary linear regression models. Estimates summarizing the PCB effects over all centers were calculated using both traditional linear regression and multilevel models (i.e., random-effects models). The random-effects models adjusted the covariance structure for the clustering of PCB levels and covariates within centers and were run in SAS software (SAS Institute, Inc., Cary, North Carolina). In addition, we fit a multilevel model that assessed the individual- and center-level influences of race, breastfeeding, and PCB congener ratios using BUGS software (26, 27). To determine whether the results were sensitive to the sampling scheme, we also conducted all analyses on the random sample only.

The ratio of PCB 118 to PCB 153 has been shown to vary considerably across the studies that have measured multiple PCB congeners (28). To determine whether the PCB-Bayley relation depended on congener mixture, we evaluated the interaction of total PCBs with the ratio of PCB 118 to PCB 153, and at the center level, we examined whether the ratio was related to center-specific differences in the PCB-Bayley relation. We also evaluated each of the 11 measured PCB congeners in separate models for the total study population. Individual assessment of all congeners in the same model was not possible because seven were highly correlated (Spearman’s correlation coefficient > 0.6 for PCBs 74, 105, 118, 138, 153, 170, and 180); however, we created one variable that reflected the sum of the concentration of the highly correlated congeners and ran a model that included the PCB-combination variable as well as the remaining four congeners (PCBs 28, 52, 194, and 203) separately to evaluate the independent effects of those congeners.

Our analysis included 1,207 children. Those excluded were missing information on maternal serum PCB concentration (n = 37) or cholesterol (n = 1), maternal education (n = 12), or parity (n = 2). Those missing PCB levels appeared demographically similar to those with measured levels.

RESULTS

Most of the mothers in this study were 21–30 years of age, Black or White, and nonsmokers during pregnancy, and they did not breastfeed their infants (table 1). Standardized MDI and PDI scores at 8 months of age were normally distributed (not shown). Demographic characteristics varied between those with low versus high PDI scores. Among children with low scores on the Bayley Scales of Infant Development, a greater proportion were low birth weight, preterm, not first born, and not breastfed and had mothers who did not finish high school and smoked during pregnancy.

Maternal serum PCB levels were detectable in 99.9 percent of the samples that met the quality control criteria. The PCB distribution had a long right tail with a mean of 3.1 µg/liter and a median of 2.7 µg/liter. Ninety-five percent of maternal total PCB concentrations were below 6.25 µg/liter; the highest concentration measured was 16.3 µg/liter.

We did not detect an association between maternal serum PCB level and the child’s MDI (β = 0.1, p = 0.71) or PDI (β = 0.5, p = 0.14) (table 2). The mean PDI scores appeared to increase slightly with increasing PCB levels (p = 0.09). Estimates of the relation between PCBs and Bayley scores were adjusted for the mother’s triglycerides, cholesterol, and education and for the child’s birth order (MDI crude β = 0.1, p = 0.62; or PDI crude β = 0.5, p = 0.13). Inclusion in the model of the other covariates evaluated, including maternal serum DDE, race, prepregnancy body mass index, and breastfeeding, did not affect the associations (data not shown). Inclusion of gestational age in the multivariate model strengthened the direct association between PCB levels and MDI (β = 0.2, p = 0.41) and PDI (β = 0.6, p = 0.06) (data not shown), but this was not considered further because it could act as an intermediate on the causal pathway. Results were similar when restricted to nonbreastfed children (n = 950; MDI (β = 0.2, p = 0.60) and PDI (β = 0.4, p = 0.21)), which excluded those with postnatal exposure through breast milk. Transforming the continuous PCB variable by log (dose + 1) did not alter results. The summary estimates from the multilevel models, which adjusted the covariance structure within and between centers, were similar to those from the fixed effects models. Results were also similar when restricted to the random sample, which excluded children sampled specifically because of their very low or very high Bayley scores.

The relation between total PCBs and Bayley scores varied among study centers (table 3). The inclusion of interaction terms for PCB and each center indicated that statistically significant heterogeneity existed among centers for the relation between total PCBs and PDI (p < 0.05) but not MDI. In most sites, PCBs were not associated with Bayley scores. Among the study population in New Orleans, PCBs were adversely associated with both the MDI (β = –6.3, p = 0.04) and the PDI (β = –5.1, p = 0.10) and in Baltimore with the PDI only (β = –2.9, p = 0.02); conversely, among the study population in the Richmond and Providence centers, the PDI increased with increasing PCB levels (β = 3.5, p = 0.03; β = 6.0, p = 0.01, respectively). Study sites varied by breastfeeding practices, race, mean baseline Bayley scores, and mean PCB levels; however, there was no evidence from the multilevel models that the direction of the relation between PCBs and Bayley scores was affected by these individual-level or center-level characteristics, nor was there other evidence of effect modification by other covariates such as maternal age, body mass index, socioeconomic status, or smoking status. When the analysis was restricted to subjects from the simple random sample, heterogeneity among centers was suggested but not statistically significant.

The ratio of PCB 118 to PCB 153 was 0.9 overall and ranged from 0.7 to 1.2 across centers. We detected no patterns in the PCB-neurodevelopment relation by the individual- or center-level ratio of PCB 118 to PCB 153 (data not shown). None of the individual congeners was related to the MDI or the PDI when evaluated in separate models that combined all study sites. In the models with the PCB-combination variable and four individual congeners (PCBs 28, 52, 194, and 203), none was related to the MDI or the PDI, but results were imprecise (data not shown).

DISCUSSION

Overall, this large, multicenter study did not demonstrate any relation between the maternal prenatal PCB level and the children’s mental or motor development in children at 8 months of age, as measured by the Bayley Scales of Infant Development. The procedures for measuring the children’s neurodevelopment and the analytic techniques for measuring PCBs were standardized across the 12 demographically diverse study sites. Analysis of this cohort also assessed many of the covariates suspected to affect estimates of the PCB-neurodevelopment relation, including race, breastfeeding, body mass index, and DDE, but found little evidence of confounding or effect modification. Our results are consistent with those of the many published studies of low-level PCB exposure that did not observe an association between PCBs and the Bayley MDI (911). We did not, however, confirm the adverse associations between prenatal PCB exposure and the PDI observed in some studies (911).

The baseline developmental scores, PCB levels, and the change in developmental score due to PCBs differed with respect to study center. Although maternal serum PCB levels were not related to Bayley scores among most centers, increased PCBs were associated with a decreased PDI in New Orleans and Baltimore and, contrary to expectation, an increased PDI in Richmond and Providence.

The statistical significance of the heterogeneity was not robust, with detection depending on the analytic method. Differences among centers were most pronounced for the PDI when all children, from the random sample and the outcome-dependent sample, were included. Although the overall results and interpretation did not differ on the basis of the source of the sample, the outcome-dependent sample increased the power to detect heterogeneity among centers. We did not anticipate heterogeneity when designing this study. By selecting subjects in our outcome-dependent sample according to their placement on the PDI distribution of the total study population, we found that these subjects tended to be from centers with high or low mean PDIs. Had we used center-specific sampling probabilities to select these and the simple random sample subjects, the sample could have been better balanced and more statistically efficient. Our sampling approach, however, combined with a weighted analysis caused no bias.

There was considerable demographic heterogeneity across centers. Both race and breastfeeding rates varied by center, but there was no pattern of variation in the relation between PCBs and neurodevelopment explained by individual- or center-specific race or breastfeeding rates in the multilevel models examined. In addition, results were similar when restricted to nonbreastfed children, who were the majority. Although the center effectively served as a proxy for characteristics such as race and breastfeeding, which did not prove to be important factors in this analysis, the center may have also served as a proxy for important unmeasured characteristics, such as diet, mercury, or lead exposure. The two sites with the most striking adverse associations were New Orleans and Baltimore. It is possible that women in these coastal cities consumed more fish, which if contaminated with mercury could have had a synergistic relation with PCBs in adversely affecting development. Yet, because no inverse relation was observed in Boston, Providence, or New York, which are also coastal cities where fish is consumed, such an explanation is not supported. Results from the multilevel models with random terms for PCBs reflected the variation in results by center but were not markedly different from the results from the more standard fixed effects models. In addition, adjusting for center as a random effect variable to account for clustering of the individuals’ characteristics within centers had little effect on results.

Exposure variability is another potential explanation for variation in results across centers. The sites showing adverse associations, however, were not those with the highest total PCB concentrations (table 3). Although unlikely, variation in the composition of congeners comprising the mixture could have affected our results. Our ability to evaluate the congener-specific effects was limited by the relatively small sample size for each center and the correlation among congener levels. Although the ratio of PCB 118 to PCB 153 ranged from 0.7 to 1.2 across centers, variation in the relation between PCB and neurodevelopment by the ratio of these two congeners was not statistically significant at the individual level or site level.

Variability in testing protocol and examiners among study sites could have also affected the PCB-neurodevelopment results; however, Bayley testing procedures were carefully standardized across study sites and regularly monitored by the test developer and other experts. The Bayley Scales performed as expected in relation to other characteristics known to influence children’s development, such as socioeconomic indicators, preterm birth, and birth weight (2931). We suspect that any differences in testing procedures across study sites were minimal and would not vary by PCB level within a site nor account for important differences in the PCB-Bayley relation across study sites.

The difference in the PCB-PDI relation among study sites remains unexplained. It is possible that the unknown factors responsible for heterogeneity among sites in this study may also be responsible for differences between our overall results and those of studies showing an inverse relation (911).

The PCB-Bayley relations reported in previous studies have varied by the timing of exposure (pre- vs. postnatal), the timing of the child’s examination, and the outcome associated with PCBs (usually PDI, not MDI). In the Netherlands, prenatal PCB exposure was associated with a decreased PDI at 3 months of age, but not at 7 or 18 months, yet postnatal PCB exposure was associated with a decreased PDI at 7 months of age (11). A larger cohort from North Carolina reported a subtle but consistent inverse association between prenatal PCB exposure and the PDI at 6, 12, 18, and 24 months of age but no association between postnatal PCB exposure and MDI scores (9, 10). Other studies have shown relations between PCBs and neurodevelopment among very young infants and among older children using assessments such as the Fagan Test of Infant Intelligence or McCarthy Scales (12, 14, 17); these instruments, compared with the Bayley Scales, may be more sensitive for detecting subtle neurodevelopmental effects that may be related to PCBs.

Most studies have accounted for socioeconomic influences, but the factors measured and methods used to adjust analyses have varied. Some have suggested that associations with PCBs may be more difficult to detect among children from lower socioeconomic households because such factors as poor parenting, nutrition, or lead exposure may have greater impact on neurodevelopment relative to PCBs (14, 17). Likewise, positive factors, such as stimulating home environment and breastfeeding, could counter any negative effects of PCBs in homes of higher socioeconomic status. Our study included children from a variety of socioeconomic backgrounds, and our analyses accounted for relevant differences by social factors, but the overall PCB-Bayley relation did not differ by the socioeconomic indicators in this study.

Other potential explanations for the differences among studies include the type and timing of the exposure and the biologic sample from which PCBs were measured. Breastfeeding confers postnatal PCB exposure. We did not have information on PCB levels in breast milk or breastfeeding duration. Yet, only 15 percent of the women in this study population ever breastfed their infant, limiting the proportion of children exposed to PCBs postnatally through breast milk, and results were similar when restricted to the nonbreastfed children. Some suggest breastfeeding as the most important source of exposure to PCBs among children (12); however, the critical time period for PCB activity in relation to a child’s neurodevelopment remains controversial (17).

The ratio of PCB 118 to PCB 153 has varied across similar studies and geographic regions. In this study, the PCB 118:PCB 153 ratios from all the centers were generally higher than those reported by other studies of PCBs. Although PCB levels have dropped in the United States since these data were collected, the levels in the Collaborative Perinatal Project from 40 years ago appear similar to those in some European countries today (28). For example, the median level of PCB 153 measured from a larger sample of this study population around 1960 was 140 ng/g of lipid. In the early 1990s, PCB 153 levels measured in a New York population were much lower (40 ng/g of lipid), but levels measured in Germany (140 ng/g of lipid) were similar (28). Thus, we did not find a relation between PCBs and neurodevelopment at PCB levels comparable with those of studies that reported adverse effects (911).

In summary, prenatal exposure was indicated by PCB levels in maternal third trimester sera. Despite standardization across study centers of 1) the timing of and procedures related to the child’s Bayley examination and 2) the analytic methods for quantifying PCBs, which are two important factors suspected to be responsible for variation among studies, we found variability in the association between the PCB and PDI across study centers. Thus, the variation in results across studies may not be attributable to differences in timing or measurement of the exposure or outcome. A more plausible explanation is that study populations vary with respect to other factors, such as nutrition or other environmental toxicants, which might confound or modify the effects of PCBs. Given the heterogeneity in results across the centers within this study and among other distinct studies, the relation between low-level prenatal PCB exposure and the child’s early mental and motor development remains unclear.

Correspondence to Dr. Julie Daniels, Department of Epidemiology, CB#7435, University of North Carolina, Chapel Hill, NC 27599-7435 (e-mail: Julie_Daniels@unc.edu).

TABLE 1.

Characteristics of mothers and their children born between 1959 and 1966 overall and by the distribution of Bayley psychomotor scores, Collaborative Perinatal Project, United States

 Total sample(n = 1,207) >1 SD* below the mean PDI*(n = 219) Within 1 SD of the mean PDI(n = 756) >1 SD above the mean PDI(n = 232) 
PCB* (µg/liter) (median (95% CI*)) 2.7 (1.8–3.7) 2.5 (1.8–3.3) 2.7 (1.8–3.7) 2.7 (1.9–3.7) 
Race (%)     
White 45 52 44 42 
Black 48 45 50 48 
Other 10 
Maternal age (%)     
11–20 years 32 26 33 35 
21–30 years 52 55 51 50 
≥31 years 16 19 16 15 
Maternal education (%)     
<High school 56 59 56 52 
High school 32 32 32 32 
>High school 12 12 16 
Socioeconomic index (%)     
≥1 SD below mean 14 16 14 14 
Within 1 SD of mean 67 70 66 63 
≥1 SD above mean 19 14 20 23 
Prenatal maternal smoking (%)     
None 55 51 55 57 
<10 cigarettes/day 20 21 20 23 
≥10 cigarettes/day 25 28 25 20 
Breastfed (%)     
Yes 15 17 19 
No 85 93 83 81 
First born (%)     
Yes 33 20 33 44 
No 67 80 67 56 
Preterm (%)     
Yes 12 24 11 
No 88 76 89 95 
Birth weight (g) (mean (SD)) 3,184 (540) 2,972 (644) 3,203 (509) 3,321 (468) 
Mental Development Index (mean (SD)) 87 (15) 71 (15) 87 (12) 97 (13) 
PDI (mean (SD)) 87 (18) 59 (8) 88 (10) 110 (5) 
Sample (% random/% from tails) 85/15 53/47 99/1 68/31 
 Total sample(n = 1,207) >1 SD* below the mean PDI*(n = 219) Within 1 SD of the mean PDI(n = 756) >1 SD above the mean PDI(n = 232) 
PCB* (µg/liter) (median (95% CI*)) 2.7 (1.8–3.7) 2.5 (1.8–3.3) 2.7 (1.8–3.7) 2.7 (1.9–3.7) 
Race (%)     
White 45 52 44 42 
Black 48 45 50 48 
Other 10 
Maternal age (%)     
11–20 years 32 26 33 35 
21–30 years 52 55 51 50 
≥31 years 16 19 16 15 
Maternal education (%)     
<High school 56 59 56 52 
High school 32 32 32 32 
>High school 12 12 16 
Socioeconomic index (%)     
≥1 SD below mean 14 16 14 14 
Within 1 SD of mean 67 70 66 63 
≥1 SD above mean 19 14 20 23 
Prenatal maternal smoking (%)     
None 55 51 55 57 
<10 cigarettes/day 20 21 20 23 
≥10 cigarettes/day 25 28 25 20 
Breastfed (%)     
Yes 15 17 19 
No 85 93 83 81 
First born (%)     
Yes 33 20 33 44 
No 67 80 67 56 
Preterm (%)     
Yes 12 24 11 
No 88 76 89 95 
Birth weight (g) (mean (SD)) 3,184 (540) 2,972 (644) 3,203 (509) 3,321 (468) 
Mental Development Index (mean (SD)) 87 (15) 71 (15) 87 (12) 97 (13) 
PDI (mean (SD)) 87 (18) 59 (8) 88 (10) 110 (5) 
Sample (% random/% from tails) 85/15 53/47 99/1 68/31 

* SD, standard deviation; PDI, Psychomotor Development Index; PCB, polychlorinated biphenyl; CI, confidence interval.

TABLE 2.

Bayley infant mental (MDI*) and psychomotor (PDI*) development scores at 8 months of age for children born between 1959 and 1966 in relation to maternal serum PCB* levels, Collaborative Perinatal Project, United States

 No. MDI  PDI 
Mean 95% CI*  Mean 95% CI 
Total sample       
PCB level (µg/liter)       
1 (0–1.24) 117 85.0 82.2, 87.8  85.0 81.4, 88.5 
2 (1.25–2.49) 426 86.5 84.9, 88.1  86.6 84.6, 88.6 
3 (2.50–3.74) 378 87.3 85.6, 89.0  88.9 86.8, 91.0 
4 (3.75–4.99) 145 87.0 84.4, 89.5  90.0 86.8, 93.2 
5 (5.00–16.50) 141 86.8 84.1, 89.4  88.7 85.4, 92.0 
Trend (ordinal) 1,207  p = 0.48   p = 0.09 
Total PCB (continuous)       
Fixed effects summary  β† = 0.10 (0.26)‡ p = 0.71  β = 0.47 (0.32) p = 0.14 
Random effect summary§       
R-PCB slope/F-center  β = –0.21 (0.52) p = 0.70  β = 0.39 (0.44) p = 0.40 
F-PCB slope/R-center  β = 0.13 (0.26) p = 0.62  β = 0.46 (0.32) p = 0.15 
R-PCB slope/R-center  β = 0.13 (0.26) p = 0.63  β = 0.41 (0.37) p = 0.30 
Random sample only       
PCB level (µg/liter)       
1 (0–1.24) 93 85.7 82.9, 88.6  87.5 84.2, 90.9 
2 (1.25–2.49) 359 87.8 86.2, 89.4  88.5 86.6, 90.3 
3 (2.50–3.74) 325 87.1 85.5, 88.7  88.7 86.8, 90.6 
4 (3.75–4.99) 126 87.9 85.4, 90.3  90.4 87.5, 93.3 
5 (5.00–16.50) 123 88.4 85.9, 90.9  89.2 86.2, 92.2 
Trend (ordinal) 1,026  p = 0.34   p = 0.31 
Total PCB (continuous)       
Fixed effects summary  β = 0.22 (0.25) p = 0.38  β = 0.29 (0.29) p = 0.32 
Random effect summary       
R-PCB slope/F-center  β = 0.05 (0.52) p = 0.92  β = 0.12 (0.36) p = 0.74 
F-PCB slope/R-center  β = 0.24 (0.25) p = 0.33  β = 0.24 (0.29) p = 0.40 
R-PCB slope/R-center  β = 0.31 (0.30) p = 0.32  β = 0.24 (0.29) p = 0.42 
 No. MDI  PDI 
Mean 95% CI*  Mean 95% CI 
Total sample       
PCB level (µg/liter)       
1 (0–1.24) 117 85.0 82.2, 87.8  85.0 81.4, 88.5 
2 (1.25–2.49) 426 86.5 84.9, 88.1  86.6 84.6, 88.6 
3 (2.50–3.74) 378 87.3 85.6, 89.0  88.9 86.8, 91.0 
4 (3.75–4.99) 145 87.0 84.4, 89.5  90.0 86.8, 93.2 
5 (5.00–16.50) 141 86.8 84.1, 89.4  88.7 85.4, 92.0 
Trend (ordinal) 1,207  p = 0.48   p = 0.09 
Total PCB (continuous)       
Fixed effects summary  β† = 0.10 (0.26)‡ p = 0.71  β = 0.47 (0.32) p = 0.14 
Random effect summary§       
R-PCB slope/F-center  β = –0.21 (0.52) p = 0.70  β = 0.39 (0.44) p = 0.40 
F-PCB slope/R-center  β = 0.13 (0.26) p = 0.62  β = 0.46 (0.32) p = 0.15 
R-PCB slope/R-center  β = 0.13 (0.26) p = 0.63  β = 0.41 (0.37) p = 0.30 
Random sample only       
PCB level (µg/liter)       
1 (0–1.24) 93 85.7 82.9, 88.6  87.5 84.2, 90.9 
2 (1.25–2.49) 359 87.8 86.2, 89.4  88.5 86.6, 90.3 
3 (2.50–3.74) 325 87.1 85.5, 88.7  88.7 86.8, 90.6 
4 (3.75–4.99) 126 87.9 85.4, 90.3  90.4 87.5, 93.3 
5 (5.00–16.50) 123 88.4 85.9, 90.9  89.2 86.2, 92.2 
Trend (ordinal) 1,026  p = 0.34   p = 0.31 
Total PCB (continuous)       
Fixed effects summary  β = 0.22 (0.25) p = 0.38  β = 0.29 (0.29) p = 0.32 
Random effect summary       
R-PCB slope/F-center  β = 0.05 (0.52) p = 0.92  β = 0.12 (0.36) p = 0.74 
F-PCB slope/R-center  β = 0.24 (0.25) p = 0.33  β = 0.24 (0.29) p = 0.40 
R-PCB slope/R-center  β = 0.31 (0.30) p = 0.32  β = 0.24 (0.29) p = 0.42 

* MDI, Mental Development Index; PDI, Pschomotor Development Index; PCB, polychlorinated biphenyl; CI, confidence interval.

† β = point increase in developmental score per mg/liter increase in PCB level adjusted for research center, mother’s education, triglycerides, cholesterol, and child first born.

‡ Numbers in parentheses, standard error.

§ In the random effect summary models, “R” refers to a random effect and “F” refers to a fixed effect.

TABLE 3.

Bayley infant mental (MDI*) and psychomotor (PDI*) development scores at 8 months of age for children born between 1959 and 1966 in relation to maternal serum PCB* levels, by study center, Collaborative Perinatal Project, United States

Study center No. PCB crude median, µg/liter (Q*1–Q3) MDI  PDI Breastfed (%) Race 
Crude score median (Q1–Q3) β† SE* p value  Crude score median (Q1–Q3) β SE p value  White (%) Black (%) Other (%) 
Portland, OR 62 1.6 (1.0–2.3) 81 (76–84) –1.6 1.5 0.31   81 (73–101) –1.7 2.3 0.46 26 65 31 
New York, NY, center 1 109 1.8 (1.3–2.8) 89 (83–95) –1.2 0.7 0.09   97 (83–105) –0.9 1.2 0.47 42 51 
New Orleans, LA 75 1.9 (1.4–2.5) 84 (73–95) –6.3 3.1 0.04   87 (77–101) –5.1 3.1 0.10 100 
Minneapolis, MN 63 2.0 (1.5–2.8) 84 (78–90) 0.8 2.1 0.72   83 (73–101) 0.6 2.5 0.80 46 95 
Memphis, TN 77 2.2 (1.7–3.0) 81 (76–89) 1.9 1.3 0.15   87 (77–94) 2.5 1.7 0.13 99 
New York, NY, center 2 43 2.4 (1.8–3.9) 98 (92–106) 1.5 1.3 0.26  100 (87–110) –0.2 1.9 0.92 15 24 44 32 
Providence, RI 87 2.8 (2.3–3.5) 84 (81–92) 2.7 1.5 0.08   93 (77–104) 6.0 2.4 0.01 74 26 
Boston, MA 286 2.9 (2.2–4.4) 84 (78–95) 0.1 0.5 0.81   83 (72–101) 0.8 0.6 0.15 22 88 11 
Philadelphia, PA  185 3.2 (2.3–4.5) 89 (78–96) 0.4 0.6 0.57   90 (77–104) 1.4 0.8 0.07 91 
Baltimore, MD 89 3.2 (2.5–4.2) 95 (84–106) –1.3 1.2 0.25   83 (73–97) –2.9 1.3 0.02 16 19 81 
Buffalo, NY 62 3.3 (2.5–4.5) 89 (78–95) 0.2 0.7 0.80   96 (83–106) –0.6 0.8 0.51 47 94 
Richmond, VA 69 3.7 (2.7–4.5) 81 (76–89) 1.6 1.2 0.20   84 (78–94) 3.5 1.6 0.03 15 19 81 
Study center No. PCB crude median, µg/liter (Q*1–Q3) MDI  PDI Breastfed (%) Race 
Crude score median (Q1–Q3) β† SE* p value  Crude score median (Q1–Q3) β SE p value  White (%) Black (%) Other (%) 
Portland, OR 62 1.6 (1.0–2.3) 81 (76–84) –1.6 1.5 0.31   81 (73–101) –1.7 2.3 0.46 26 65 31 
New York, NY, center 1 109 1.8 (1.3–2.8) 89 (83–95) –1.2 0.7 0.09   97 (83–105) –0.9 1.2 0.47 42 51 
New Orleans, LA 75 1.9 (1.4–2.5) 84 (73–95) –6.3 3.1 0.04   87 (77–101) –5.1 3.1 0.10 100 
Minneapolis, MN 63 2.0 (1.5–2.8) 84 (78–90) 0.8 2.1 0.72   83 (73–101) 0.6 2.5 0.80 46 95 
Memphis, TN 77 2.2 (1.7–3.0) 81 (76–89) 1.9 1.3 0.15   87 (77–94) 2.5 1.7 0.13 99 
New York, NY, center 2 43 2.4 (1.8–3.9) 98 (92–106) 1.5 1.3 0.26  100 (87–110) –0.2 1.9 0.92 15 24 44 32 
Providence, RI 87 2.8 (2.3–3.5) 84 (81–92) 2.7 1.5 0.08   93 (77–104) 6.0 2.4 0.01 74 26 
Boston, MA 286 2.9 (2.2–4.4) 84 (78–95) 0.1 0.5 0.81   83 (72–101) 0.8 0.6 0.15 22 88 11 
Philadelphia, PA  185 3.2 (2.3–4.5) 89 (78–96) 0.4 0.6 0.57   90 (77–104) 1.4 0.8 0.07 91 
Baltimore, MD 89 3.2 (2.5–4.2) 95 (84–106) –1.3 1.2 0.25   83 (73–97) –2.9 1.3 0.02 16 19 81 
Buffalo, NY 62 3.3 (2.5–4.5) 89 (78–95) 0.2 0.7 0.80   96 (83–106) –0.6 0.8 0.51 47 94 
Richmond, VA 69 3.7 (2.7–4.5) 81 (76–89) 1.6 1.2 0.20   84 (78–94) 3.5 1.6 0.03 15 19 81 

* MDI, Mental Development Index; PDI, Pschomotor Development Index; PCB, polychlorinated biphenyl; Q, quartile; SE, standard error.

† β = point increase in developmental score per mg/liter increase in PCB level adjusted for research center, mother’s education, triglycerides, cholesterol, and child first born.

References

1.
Laden F, Neas LM, Spiegelman D, et al. Predictors of plasma concentrations of DDE and PCBs in a group of U.S. women.
Environ Health Perspect
 
1999
;
107
:
75
–81.
2.
Longnecker MP, Rogan WJ, Lucier G. The human health effects of DDT (dichlorodiphenyltrichloroethane) and PCBS (polychlorinated biphenyls) and an overview of organochlorines in public health.
Annu Rev Public Health
 
1997
;
18
:
211
–44.
3.
Shain W, Bush B, Seegal R. Neurotoxicity of polychlorinated biphenyls: structure-activity relationship of individual congeners.
Toxicol Appl Pharmacol
 
1991
;
111
:
33
–42.
4.
Tilson HA, Kodavanti PR, Mundy WR, et al. Neurotoxicology of environmental chemicals and their mechanism of action.
Toxicol Lett
 
1998
;
102–103
:
631
–5.
5.
Bowman RE, Heironimus MP, Barsotti DA. Locomotor hyperactivity in PCB-exposed rhesus monkeys.
Neurotoxicology
 
1981
;
2
:
251
–68.
6.
Seegal RF, Bush B, Shain W. Lightly chlorinated ortho-substituted PCB congeners decrease dopamine in nonhuman primate brain and in tissue culture.
Toxicol Appl Pharmacol
 
1990
;
106
:
136
–44.
7.
Lai TJ, Guo YL, Guo NW, et al. Effect of prenatal exposure to polychlorinated biphenyls on cognitive development in children: a longitudinal study in Taiwan.
Br J Psychiatry
 
2001
;
178(suppl 40)
:
S49
–52.
8.
Rogan WJ, Gladen BC, Hung K, et al. Congenital poisoning by polychlorinated biphenyls and their contaminants in Taiwan.
Science
 
1988
;
241
:
334
–6.
9.
Gladen BC, Rogan WJ, Hardy P, et al. Development after exposure to polychlorinated biphenyls and dichlorodiphenyl dichloroethene transplacentally and through human milk.
J Pediatr
 
1988
;
113
:
991
–5.
10.
Rogan WJ, Gladen BC. PCBs, DDE, and child development at 18 and 24 months.
Ann Epidemiol
 
1991
;
1
:
407
–13.
11.
Koopman-Esseboom C, Weisglas-Kuperus N, de Ridder M, et al. Effects of polychlorinated biphenyl/dioxin exposure and feeding type on infants’ mental and psychomotor development.
Pediatrics
 
1996
;
97
:
700
–6.
12.
Jacobson SW, Jacobson JL, Fein GG. Environmental toxins and infant development. In: Fitzgerald HE, Lester BM, Yogman MW, eds. Theory and research in behavioral pediatrics. Vol 3. New York, NY: Plenum Press, 1986:96–146.
13.
Walkowiak J, Wiener J, Fastabend A, et al. Environmental exposure to polychlorinated biphenyls and quality of the home environment: effects on psychodevelopment in early childhood.
Lancet
 
2001
;
358
:
1602
–7.
14.
Jacobson JL, Jacobson SW. Sources and implications of interstudy and interindividual variability in the developmental neurotoxicity of PCBs.
Neurotoxicol Teratol
 
1996
;
18
:
257
–64.
15.
Darvill T, Lonky E, Reihman J, et al. Prenatal exposure to PCBs and infant performance on the Fagan Test of Infant Intelligence.
Neurotoxicology
 
2000
;
21
:
1029
–38.
16.
Winneke G, Bucholski A, Heinzow B, et al. Developmental neurotoxicity of polychlorinated biphenyls (PCBS): cognitive and psychomotor functions in 7-month old children.
Toxicol Lett
 
1998
;
102–103
:
423
–8.
17.
Shantz S. Developmental neurotoxicity of PCBs in humans: what do we know and where do we go from here?
Neurotoxicol Teratol
 
1996
;
18
:
217
–27.
18.
Niswander KR, Gordon M. The women and their pregnancies. Philadelphia, PA: Saunders, 1972.
19.
Zhou H, Weaver MA, Qin J, et al. A semiparametric empirical likelihood method for data from an outcome dependent sampling scheme with a continuous outcome.
Biometrics
 
2002
;
58
:
413
–21.
20.
Rogan WJ, Gladen BC, McKinney JD, et al. Polychlorinated byphenyls (PCBs) and dichlorodiphenyl dichloroethene (DDE) in human milk: effects of maternal factors and previous lactation.
Am J Public Health
 
1986
;
76
:
172
–7.
21.
Longnecker MP, Klebanoff MA, Gladen BC, et al. Serial levels of serum organochlorines during pregnancy and postpartum.
Arch Environ Health
 
1999
;
54
:
110
–14.
22.
Brock JW, Burse VW, Ashley DL, et al. An improved analysis for chlorinated pesticides and polychlorinated biphenyls (PCBs) in human and bovine sera using solid-phase extraction.
J Anal Toxicol
 
1996
;
20
:
528
–36.
23.
Bayley N. Manual for the Bayley Scales of Infant Development. New York, NY: The Psychological Corporation, 1969.
24.
Nichols PL, Broman SH. Familial resemblance in infant mental development.
Dev Psychol
 
1974
;
10
:
442
–6.
25.
Horvitz DG, Thompson DJ. A generalization of sampling without replacement from a finite universe.
J Am Stat Assoc
 
1952
;
47
:
663
–85.
26.
Higgins JPT, Whitehead A, Turner RM, et al. Meta-analysis of continuous outcome data from individual patients.
Stat Med
 
2001
;
20
:
2219
–41.
27.
Rutter CM, Gatsonis CA. A hierarchical regression approach to meta-analysis of diagnostic test accuracy evaluations.
Stat Med
 
2001
;
20
:
2865
–84.
28.
Longnecker MP, Wolff MS, Gladen BC, et al. Comparison of polychlorinated biphenyl levels across studies of human neurodevelopment.
Environ Health Perspect
 
2003
;
111
:
65
–70.
29.
Samson JF, de Groot L. Study of a group of extremely preterm infants (25–27 weeks): how do they function at 1 year of age?
J Child Neurol
 
2001
;
16
:
832
–7.
30.
Landry SH, Denson SE, Swank PR. Effects of medical risk and socioeconomic status on the rate of change in cognitive and social development for low birth weight children.
J Clin Exp Neuropsychol
 
1997
;
19
:
261
–74.
31.
Garcia CC, Vohr BR, Hoffman J, et al. Maternal and environmental factors affecting developmental outcome of infants of adolescent mothers.
J Dev Behav Pediatr
 
1986
;
7
:
230
–6.