In developed countries, polychlorinated biphenyls (PCBs) are ubiquitous and persistent contaminants of the environment, including food.1 Within the range of the resulting low-level exposure, associations of PCBs with lower birth weight have been observed in several studies.2–5 Among the adverse health effects potentially caused by low-level PCB exposure, lower birth weight is especially interesting because for one PCB mixture, Aroclor 1016, current regulations (oral reference dose) are based on birth weight effects in primates.6 Associations between PCBs and birth weight in humans, however, have not been shown in many studies,7–11 and whether the preponderance of evidence supports an association remains an open question.
Studying the association of exposure to environmental contaminants with birth weight has been a traditional approach in reproductive epidemiology.12–14 Disaggregating effects on birth weight, if any, into those on preterm birth and intrauterine growth may provide greater insights into biologic mechanisms and later consequences of early exposures.
As part of a larger project,15–18 we measured PCBs in 1168 maternal serum samples from pregnancies selected at random from the Collaborative Perinatal Project. Most mothers were enrolled in the early 1960s, before PCB production peaked in the United States.19 The level of exposure to PCBs was approximately 3-fold greater than in more recent studies of U.S. mothers20 and was comparable to that in U.S. studies from the 1960s and 1980s in which an association with birth weight was found.2,4 Although our dataset was rather small for an analysis of risk factors for preterm birth and small-for-gestational-age (SGA), it was relatively large compared with previous studies of PCBs and birth weight, and allowed us to explore possible effect modification by sex4 and other factors.
The subjects were enrolled in the Collaborative Perinatal Project, a prospective study of neurologic disorders and other conditions in children.21,22 Pregnant women were recruited from 1959 to 1965 at 12 U.S. study centers (in Baltimore, Boston, Buffalo, Memphis, Minneapolis, New Orleans, New York [2 centers], Philadelphia, Portland, Providence, and Richmond). Eleven study centers recruited patients from the prenatal clinics of a university hospital, and one study center (Buffalo) recruited patients from 13 private practices. Subject selection method varied among the study centers. For example, at Columbia-Presbyterian Medical Center in New York City, every sixth woman who was potentially eligible was invited to participate; at Charity Hospital in New Orleans, potentially eligible subjects were selected if their patient identification number ended in zero; and at Boston Lying-in Hospital, all potentially eligible women were invited to participate. Women were ineligible if they were incarcerated, if they were planning to leave 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 of the number of potential subjects who declined to participate at baseline were not kept, although there were probably few such women.16 The characteristics of women in the sample were, at registration, essentially the same as those in the sampling frame.21 Four percent of subjects who enrolled were lost to follow up before delivery. Once enrolled, the mothers’ nonfasting blood was collected approximately every 8 weeks for the remainder of the pregnancy, at delivery, and 6 weeks postpartum. Sera were stored in glass at −20°C with no recorded thaws. Approximately 42,000 women were enrolled, with 53,000 children born in the study.
We measured serum organochlorine levels in a subset of these mothers. Eligibility criteria were delivery of a live-born singleton and availability of a 3-mL aliquot of third-trimester maternal serum. The eligibility criteria were met for 44,075 pregnancies. From these, we took a simple random sample (n = 1200). This research was approved by the National Institute of Environmental Health Sciences Institutional Review Board.
Serum levels of 11 polychlorinated biphenyls (IUPAC congener numbers 28, 52, 74, 105, 118, 138, 153, 170, 180, 194, and 203),23p,p′-DDT, p,p′-DDE, and 7 other organochlorines (β-hexachlorocyclohexane, dieldrin, hexachlorobenzene, heptachlor epoxide, mirex, oxychlordane, and transnonachlor) were measured at the Centers for Disease Control and Prevention in 1997 to 1999 after solid-phase extraction, cleanup, and dual-column gas chromatography using electron capture detection.24 The proportion of PCBs in each sample recovered by extraction was approximately 60%.24 The results shown are not recovery-adjusted.16 The between-assay coefficient of variation was 19% at 3.49 μg PCB/L (n = 291). Of the 1200 pregnancies selected for inclusion in the study, a laboratory result for PCB was not obtained for 3% (n = 32), mainly because the measured value did not meet the quality-control standards for acceptance.16 Thus, PCB results were available for 1168 subjects. Because PCBs are lipid-soluble and serum contains a variable amount of lipid, serum cholesterol and triglycerides were measured, using standard enzymatic assays.
For 6 of the 11 congeners measured, all subjects had a measured level. For the remaining 5 congeners, measured levels were available for more than 95% of subjects. For those subjects (<5%) with a missing congener level, we imputed a value as follows. Using the data for subjects with complete data for all congeners, we modeled the level of one congener based on the levels of the others (data for 10 other congeners were nearly always available). For the subjects missing data for the value of the congener modeled, the level was predicted based on the levels of the other congeners. Models were fit using a natural logarithm transformation of all levels, and predicted levels were reexpressed on an untransformed scale. Logs were started with a 0.001, if needed; demographic data were not used in the imputation. The median level of total PCBs that included the imputed values was 1% higher than when the missing values were assigned as zero.
We determined PCB levels in serum from the third trimester because these samples were the most complete. However, maternal PCB levels appear to be relatively stable throughout pregnancy. In 67 women selected at random from the study, the Pearson's correlation coefficient between lipid-adjusted PCB levels measured in first and third trimesters was 0.77.25 PCBs cross the placenta. A Dutch study found that levels of the main congeners in the mother's serum during the third trimester and the child's cord serum were correlated at r = 0.52–0.74.26
Length of gestation was date of delivery minus the date of last menstrual period. Delivery was classified as preterm if it occurred before 37 completed weeks of gestation. Small-for-gestational-age was defined as birth weight less than the 10th percentile at each week of gestation using all live births in the larger study as the standard.
We divided subjects into 4 categories of PCB exposure that contained approximately equal numbers of cases. Our primary analyses were based on the total of measured PCBs, expressed on a wet-weight (per liter serum) basis for reasons discussed elsewhere.16 We included a priori the following covariates in all multivariate models: center (11 indicator variables), serum triglycerides and cholesterol (both continuous), race (white, black, other), sex, smoking (yes, no), maternal age (continuous), and serum dichlorodiphenyldichloroethene level (DDE, as 5 categories). We also considered as potential confounders the continuous variables of maternal height (m), maternal prepregnancy body mass index (BMI, kg/m2), maternal pregnancy weight gain (g/wk), parity, socioeconomic index, prenatal care index, maternal education, maternal age squared, and 7 additional organochlorines (β-hexachlorocyclohexane, dieldrin, heptachlor epoxide, hexachlorobenzene, mirex, transnonachlor, and oxychlordane); and the categorical variables of season of birth, use of estrogens or progestins during pregnancy, and marital status. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using logistic regression models.
We evaluated the effect on the OR of adjustment for potentially confounding factors, one at a time, in a model that included the a priori covariates. If addition of a factor changed the OR by ≥15% for preterm birth or SGA, it was considered a confounder. The change in estimate was evaluated both in models in which the OR for the high-to-low PCB category contrast was examined and in models with a single coefficient for PCBs. To fit the latter (trend) models, among subjects in a given exposure category, we calculated the median PCB level and assigned this median to the subjects in that category; then, with data for subjects from all categories combined, we fit a coefficient to the assigned level.27 Oxychlordane was the only confounding factor so identified. Effect modification was evaluated for the following variables: center, maternal age and smoking, and infant race and sex. The improvement in fit was evaluated with and without crossproduct terms in models in which exposure was represented as a single variable (see previously). An improvement in model fit with an associated P value of ≤0.15 triggered further evaluation of whether the magnitude of modification was noteworthy.
Sensitivity analyses were performed to check the effect of basing the primary analyses on the dataset version that included imputed values of some specific congeners for some subjects, to examine evidence of congener-specific associations (without using imputed PCB values), and to see if expressing PCBs on a lipid-weight basis had any effect on the results. We also conducted secondary analyses in which the outcomes were birth weight and length of gestation; in these models, ordinary least squares models were fitted.
Of the 1168 pregnancies with known gestational length, birth weight, and a PCB measurement, we excluded 129 from the analysis because of missing data on covariates. Of the remaining pregnancies, for the 5 women who had 2 pregnancies in the study, we excluded the second pregnancy. Thus, in the final analyses, we included 1034 subjects with 132 preterm births and 101 SGA births.
Almost half of the subjects were black, nearly as many were white, and a small proportion of others, primarily Puerto Ricans, comprised the balance (Table 1); this distribution reflected the larger study overall.21 Approximately one third of subjects had had no previous pregnancies. Most of the subjects had a socioeconomic index that was below the contemporaneous U.S. median of 5. The median level of the DDT metabolite DDE in this population was approximately 5 times higher than among U.S. residents at the present time. The median level of total PCBs in this population (2.8 μg/L) was approximately 3-fold higher than in contemporary U.S. populations but similar to levels in recent studies of Europeans.20
The overall prevalence of preterm birth in this population was 13%, greater than in the United States at present.29 The greater prevalence of preterm birth among boys, blacks, those with a lower socioeconomic index, and smokers (Table 1) was consistent with data on established risk factors.30 The greater prevalence of preterm birth among younger mothers, mothers with short stature, lower prepregnancy BMI, or lower rate of pregnancy weight gain has been reported previously but less consistently.30
The greater prevalence of SGA among girls (Table 1) was expected because birth-weight percentiles were calculated for both sexes combined. Mothers who were black, shorter, slimmer, had lower rate of pregnancy weight gain, or who were nulliparous or smokers had a greater proportion of SGA babies, consistent with previous reports.31
For preterm birth, the crude OR suggested an association with maternal serum PCB levels, but after adjustment, the OR for the highest compared with lowest exposure category was 1.1 with a wide confidence interval (Table 2). The variable for which adjustment caused the largest decrease in the PCB estimate was DDE. For example, in a model with center, triglycerides, and cholesterol, the addition of DDE dropped the PCB trend OR by 30%. The P value for effect modification by sex was 0.51 (1 df test). The adjusted OR for those in the highest exposure category compared with those in the lowest was 1.5 for boys (95% CI = 0.5–4.0) and 0.9 for girls (0.3–2.6).
PCB levels were directly associated with odds of SGA, especially after adjustment, but the estimates were imprecise. The variable for which adjustment caused the largest increase in the PCB estimate was oxychlordane (median, 0.3 μg/L). For example, in a model with center, triglycerides, cholesterol, and DDE, the addition of oxychlordane increased the PCB trend OR by 30%. The P value for effect modification by sex was 0.24 (1 df test). For those in the highest exposure category compared with those in the lowest, the adjusted OR was for boys, 6.6 (1.4–30.0), and for girls, it was 0.9 (0.3–2.6).
When we repeated the analyses shown in Table 2 using only the measured PCB levels (without any imputed values), the findings were essentially as shown. We also evaluated whether the level of specific PCB congeners was related to preterm birth or SGA birth; in no instance was there a suggestion that congeners differed in their relation (not shown). With total PCBs expressed on a per-unit serum-lipid basis (without triglycerides or cholesterol in the model), the findings were not materially different (not shown). Additional sensitivity analyses were done to evaluate homogeneity of findings across subjects from different study centers, and according to age, race, and smoking status. For SGA, the association with PCBs was stronger among blacks. The P value for effect modification was 0.04 (2 df test). The adjusted OR for those in the highest exposure category compared with the lowest was for blacks, 3.7 (1.1–12.3), and for whites, it was 0.6 (0.2–2.4).
In a model of birth weight adjusted for the same variables as for the preterm and SGA analyses, an association with PCB levels was not seen (Table 3). When this analysis was repeated among term births the results were similar. The P value for effect modification by sex was 0.58 (1 df). For those in the lowest exposure category, the adjusted mean birth weight was for girls, 3014 g; for those in the highest, the mean was 3152; and for boys, the corresponding values were 3230 g and 3335 g. The P value for effect modification by age was 0.08 (1 df). Among the mothers <20 years and those 20 to 29 years, little association was present; in the mothers ≥30 years old, birth weight tended to increase with exposure (results not shown).
Similarly, length of gestation showed little relation with level of PCB exposure (Table 3). The P value for effect modification by race was 0.08 (2 df). Among blacks, the adjusted mean gestation in weeks among those in the lowest exposure group was 38.7, and among those in the highest exposure group, it was 39.9; among whites, the corresponding values were 40.1 and 40.1. The P value for effect modification by smoking was 0.08 (1 df). Among nonsmokers, the adjusted mean gestation among those in the lowest exposure group was 39.4, and among those in the highest exposure group, it was 40.5; among smokers, the corresponding values were 39.7 and 39.6.
Our results provide little support for the hypothesis that low-level exposure to PCBs is associated with preterm birth, birth weight, or length of gestation. PCB levels were directly associated with odds of SGA, but the estimates were imprecise.
Although the relation of low-level PCB exposure with birth weight has been reviewed recently elsewhere,10,32 several aspects of these data are worth noting here. The other larger studies of this relation also reported no association.7,33 A study among women with relatively high levels of exposure to PCBs, through diet, also reported no association.34 In most studies reporting an inverse association of low-level PCB exposure with birth weight,2–4 results adjusted for the potentially confounding effect of DDE were not presented, and DDE could be an important confounding factor.15 Alternatively, maternal fish intake, an important source of PCB exposure in some populations,34,35 may affect birth weight36 and could confound the results to varying degrees across studies. Furthermore, differences in proportions of susceptible subgroups, if any, could account for variation in results. The increased odds of SGA among boys more exposed to PCBs in our data is reminiscent of the association with lower birth weight among boys reported by Hertz-Picciotto et al,4 although the birth weight finding was not present in our male infants. To our knowledge, the association of SGA with PCBs has been examined in only one small study of whites, and no relation was found.11 Previous studies on PCBs and length of gestation, like with birth weight, have given mixed results.2,4,5,33
PCB congeners vary widely in their toxicity.19 Another potential explanation for why the association of PCBs and birth outcomes in the present data differs from some other studies is that the mixture of PCBs present in this population was richer in PCB 118 (designated by its IUPAC number23) than in the populations in which an association was found.14 On the other hand, other studies showing no relation with birth weight have not had an unusual PCB 118 content.7,10 Other differences in the congener mix across studies were likely present and are difficult to characterize completely because the specific congeners quantitated varied across studies.
Apart from the findings among populations with relatively low-level PCB exposure are those from highly exposed groups. In 405 women who were exposed occupationally,37 a large increase in exposure (20 μg/L serum of “high homolog” [highly chlorinated] PCBs) was associated with a small decrease in birth weight (33 g) and gestational length (1.1 d). In that population, the mixture of PCBs used in the plants (Aroclors 1254, 1242, and 1016) resulted in a lot of “low homolog” (less chlorinated) PCBs in the serum of workers, and the effects of exposure to that mixture might be different than effects from the mixture encountered in the low-level setting. Poisonings during pregnancy with a high dose of a PCB mixture also reduced birth weights38; whether the mixture, which included dioxin-like polychlorinated dibenzofurans, has effects like the mixture encountered in the low-level setting is not clear.
A comparison of the exposure level in our population with those in the studies of primates5 used to identify an oral reference dose for the mixture of PCBs in Aroclor 1016 might be of interest if the distribution of congeners present in Aroclor 1016 and human serum were similar—but there was little overlap in the congeners present.39
If PCBs degraded in storage, nondifferential misclassification of exposure could have obscured a true relation with outcomes of pregnancy. PCBs levels in serum measured after storage at −20°C for 15 years, however, showed no evidence of instability,40,41 consistent with the properties of these compounds that made them so useful in many applications. We have little reason to suspect errors in birth weight; some error in length of gestation was certainly present, as would be expected, especially in a study conducted before ultrasound was available. Nonetheless, as noted here, the predictors of preterm birth and SGA in this population were largely as expected, and thus we have no reason to suspect that random errors in the outcomes assessed accounted for much imprecision in the associations measured.
In conclusion, our data from the Collaborative Perinatal Project showed little evidence to support an association of low-level PCB exposure with preterm birth, birth weight, or length of gestation. For SGA, the data were inconclusive but suggested a direct relation that was stronger among boys and blacks. Why results for birth weight vary across studies in populations with similar levels of exposure remains unclear, but may be due to the specific mixture of PCBs involved or to the effects of other contaminants or nutrients.
1. Robertson LW, Hansen LG, eds. PCBs: Recent Advances in Environmental Toxicology and Health Effects
. Lexington, KY: The University of Kentucky Press; 2001.
2. Fein GG, Jacobson JL, Jacobson SW, Schwartz PM, Dowler JK. Prenatal exposure to polychlorinated biphenyls: effects on birth size and gestational age. J Pediatr
3. Patandin S, Koopman-Esseboom C, de Ridder MA, Weisglas-Kuperus N, Sauer PJ. Effects of environmental exposure to polychlorinated biphenyls and dioxins on birth size and growth in Dutch children. Pediatr Res
4. Hertz-Picciotto I, Keller J, Willman E, James R, Teplin S, Charles MJ. Fetal and early childhood growth in relation to prenatal PCB and organochlorine pesticide exposures. Organohalogen Compounds
5. Karmaus W, Zhu X. Maternal concentration of polychlorinated biphenyls and dichlorodiphenyl dichlorethylene and birth weight in Michigan fish eaters: a cohort study. Environ Health
7. Rogan WJ, Gladen BC, McKinney JD, et al. Neonatal effects of transplacental exposure to PCBs and DDE. J Pediatr
8. Vartiainen T, Jaakkola JJ, Saarikoski S, Tuomisto J. Birth weight and sex of children and the correlation to the body burden of PCDDs/PCDFs and PCBs of the mother. Environ Health Perspect
9. Bjerregaard P, Hansen JC. Organochlorines and heavy metals in pregnant women from the Disko Bay area in Greenland. Sci Total Environ
10. Gladen BC, Shkiryak-Nyzhnyk ZA, Chyslovska N, Zadorozhnaja TD, Little RE. Persistent organochlorine compounds and birth weight. Ann Epidemiol
11. Ribas-Fito N, Sala M, Cardo E, et al. Association of hexachlorobenzene and other organochlorine compounds with anthropometric measures at birth. Pediatr Res
12. Wilcox AJ. On the importance—and the unimportance—of birthweight. Int J Epidemiol
13. Boy E, Bruce N, Delgado H. Birth weight and exposure to kitchen wood smoke during pregnancy in rural Guatemala. Environ Health Perspect
14. Hopenhayn C, Ferreccio C, Browning SR, et al. Arsenic exposure from drinking water and birth weight. Epidemiology
15. Longnecker MP, Klebanoff MA, Zhou H, Brock JW. Association between maternal serum concentration of the DDT metabolite DDE and preterm and small-for-gestational-age babies at birth. Lancet
16. Longnecker MP, Klebanoff MA, Brock JW, et al. Maternal serum level of 1,1-dichloro-2,2-bis(p-chlorophenyl)ethylene and risk of cryptorchidism, hypospadias, and polythelia among male offspring. Am J Epidemiol
17. Daniels JL, Longnecker MP, Klebanoff MA, et al. Prenatal exposure to low-level polychlorinated biphenyls in relation to mental and motor development at 8 months. Am J Epidemiol
18. Longnecker MP, Hoffman HJ, Klebanoff MA, et al. In utero exposure to polychlorinated biphenyls and sensorineural hearing loss in 8-year-old children. Neurotoxicol Teratol
19. Hansen LG, ed. The Ortho Side of PCBs: Occurrence and Disposition
. Boston: Kluwer Academic; 1999.
20. Longnecker MP, Wolff MS, Gladen BC, et al. Comparison of polychlorinated biphenyl levels across studies of human neurodevelopment. Environ Health Perspect
21. Niswander KR, Gordon M. The Women and Their Pregnancies. Pub. No. (NIH) 73-379. Washington, DC: US Government Printing Office, US Department of Health, Education, and Welfare; 1972.
22. Broman SH. The Collaborative Perinatal Project: an overview. In: Mednick SA, Harway M, Finello KM, eds. Handbook of Longitudinal Research
, vol I. New York: Praeger Publishers; 1984:185–215.
23. Ballschmiter K, Bacher R, Mennel A, et al. The determination of chlorinated biphenyls, chlorinated dibenzodioxins, and chlorinated dibenzofurans by GC-MS. J High Resolut Chromatogr
24. 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
25. Longnecker MP, Klebanoff MA, Gladen BC, et al. Serial levels of serum organochlorines during pregnancy and postpartum. Arch Environ Health
26. Koopman-Esseboom C, Huisman M, Weisglas-Kuperus N, et al. Dioxin and PCB levels in blood and human milk in relation to living areas in The Netherlands. Chemosphere
27. Greenland S. Avoiding power loss associated with categorization and ordinal scores in dose-response and trend analysis. Epidemiology
28. Myrianthopoulos NC, French KS. An application of the US Bureau of the Census socioeconomic index to a large, diversified patient population. Soc Sci Med
29. Hamilton BE, Martin JA, Sutton PD. US Department of Health and Human Services Centers for Disease Control and Prevention. Births: preliminary data for 2002. Natl Vital Stat Rep
30. Berkowitz GS, Papiernik E. Epidemiology of preterm birth. Epidemiol Rev
31. Lang JM, Lieberman E, Cohen A. A comparison of risk factors for preterm labor and term small-for-gestational-age birth. Epidemiology
32. Longnecker MP, Korrick SA, Moysich KB. Human health effects of polychlorinated biphenyls. In: Schecter A, Gasiewicz TA, eds. Dioxins and Health
, 2nd ed. Hoboken, NJ: Wiley-Interscience; 2003:679–728.
33. Schade G, Heinzow B. Organochlorine pesticides and polychlorinated biphenyls in human milk of mothers living in northern Germany: current extent of contamination, time trend from 1986 to 1997 and factors that influence the levels of contamination. Sci Total Environ
34. Grandjean P, Bjerve KS, Weihe P, Steuerwald U. Birthweight in a fishing community: significance of essential fatty acids and marine food contaminants. Int J Epidemiol
35. Fitzgerald EF, Hwang SA, Langguth K, et al. Fish consumption and other environmental exposures and their associations with serum PCB concentrations among Mohawk women at Akwesasne. Environ Res
36. Thorsdottir I, Birgisdottir BE, Halldorsdottir S, Geirsson RT. Association of fish and fish liver oil intake in pregnancy with infant size at birth among women of normal weight before pregnancy in a fishing community. Am J Epidemiol
37. Taylor PR, Stelma JM, Lawrence CE. The relation of polychlorinated biphenyls to birth weight and gestational age in the offspring of occupationally exposed mothers. Am J Epidemiol
38. Guo YL, Lambert GH, Hsu CC. Growth abnormalities in the population exposed in utero and early postnatally to polychlorinated biphenyls and dibenzofurans. Environ Health Perspect
. 1995;103(suppl 6):117–122.
39. Dobson S, van Esch GJ. Polychlorinated Biphenyls and Terphenyls
. International Program on Chemical Safety, United Nations Environment Programme, International Labour Organisation, Geneva: World Health Organization; 1993.
40. Noren K. Changes in the levels of organochlorine pesticides, polychlorinated biphenyls, dibenzo-p
-dioxins and dibenzofurans in human milk from Stockholm, 1972–1985. Chemosphere
41. Lunden A, Noren K. Polychlorinated naphthalenes and other organochlorine contaminants in Swedish human milk, 1972–1992. Arch Environ Contam Toxicol