The strengths of our study include restricting to primiparous births to prevent confounding by birth order, a large sample size, detailed plant and time-specific JEMs grounded by air monitoring samples and corroborated with serum samples, and evaluating an occupational cohort with much higher PCB exposure than previous population-based samples. We did not find an association between cumulative maternal PCB exposure prior to conception and infant sex ratio among female primiparous workers in three electrical capacitor plants. This finding is consistent with a recent population-based study of parental PCB blood concentrations and offspring gender, which also noted no increase in the proportion of males born to mothers with the highest category of PCB exposure [52
]. A small increase in male births was suggested in that study, however, when both mothers and fathers were exposed to the highest category of PCBs--suggesting that paternal PCB exposure might be associated with the secondary sex ratio [52
Although our results conflict with two other recent studies of blood PCB concentration and secondary sex ratio, the discrepancy between these studies and ours may be explained, in part, by variations in study design and population. Hertz-Picciotto and colleagues reported a 7% decrease in the sex ratio of offspring per 1 ppb increase in PCB burden in a population-based study [53
], and Weisskopf and colleagues reported an 80% reduction in the secondary sex ratio among mothers in the highest quintile of PCB exposure from sport-caught fish [33
]. Both of these studies used one-time serum PCB measurements which tend to only capture PCBs with long half-lives, whereas our plant-specific JEMs were able to account for long-term patterns of PCB exposure.
Our study also restricted the analysis to primiparous births (to n = 2595 mothers), thereby eliminating both potential confounding and effect modification by birth order, pregnancy-related changes in BMI, or breast-feeding. Previous studies have indicated an association between sex ratio of infants and their birth order [21
]. Pregnancy-associated changes in BMI, and increased PCB clearance through breast-feeding, could result in associations between birth order and blood PCB concentrations. Hertz-Picciotto and colleagues did not adjust or stratify by birth order, possibly due to limitations in the sample size (n = 399 mothers) [53
]. Weisskopf and colleagues adjusted for parity and the presence of an older male sibling, but because of their small sample size (n = 173 mothers) were not able to examine the potential for effect modification by parity and BMI [33
Additionally, the route of PCB exposure varied across all these studies, which could impact absorption and metabolism of congeners. In our study, most exposure came from dermal absorption or inhalation of PCBs in an occupational setting; in the study by Weisskopf and colleagues, exposure primarily occurred through ingestion of sport-caught fish; and in the study by Hertz-Picciotto and colleagues, environmental exposure likely occurred through both ingestion and inhalation routes.
Perhaps the most important factor distinguishing our study from these previous studies, however, is that the magnitude of exposure in those study populations was substantially lower than in our occupationally exposed group. Weisskopf and associates reported a mean serum PCB concentration of 10.7 μg/L for mothers in the highest quintile of exposure [33
], while Hertz-Picciotto and associates reported a mean serum PCB concentration of 1.9 μg/L for mothers in the highest decile of exposure [53
]. By contrast, a study of the New York plant in 1976 found that workers (both male and female) in areas with high air PCB concentrations had serum L-PCB concentrations of 100 μg/L (geometric mean); female workers overall had serum H-PCB concentrations of 8 μg/L (geometric mean) [54
]. Average serum values were 502 μg/L for lower chlorinated PCBs (L-PCBs) and 44 μg/L for higher chlorinated PCBs (H-PCBs) among a sample of current workers in 1977 in the capacitor-production area of the Indiana plant, and 237 μg/L L-PCBs and 51 μg/L H-PCBs among workers in the maintenance areas of that plant [55
]. The PCB-exposed workers in our cohort therefore experienced PCB exposures substantially higher than the environmentally exposed individuals in other studies of maternal PCB exposure and secondary sex ratio. If a true association existed between PCB exposure and secondary sex ratio, we would expect to see that risk magnified in more highly exposed populations, such as we studied.
Additionally, no association between high-dose PCB exposure and secondary sex ratio has been observed in accidental contamination events. Rogan and colleagues found no change in the sex ratio of children born to women exposed to PCBs and polychlorinated dibenzofurans (PCDFs) from contaminated cooking oil in Taiwan in the 1978-79 Yucheng incident [56
], despite mean serum PCB concentrations of 49.3 μg/L [57
]. A significant decrease in the secondary sex ratio was later noted, however, among children born to exposed men - provided the men had been aged less than 20 at the time of exposure [30
]. It was postulated that the Yucheng exposure could have caused damage to the young men's developing reproductive systems. Similarly, an analysis of maternal PCB and PCDF exposure from the Yusho incident did not note a decrease in male births; however, the small number of births in this population precluded a detailed analysis [31
]. These studies might not be directly comparable with ours, however, due to their high accidental one-time exposures including both PCBs and PCDFs, with exposure occurring through ingestion; as opposed to this study, where the workers had long-term dermal and inhalation exposure.
PCBs have been suggested as a cause of altered secondary sex ratios due to their potential to disrupt normal endocrine function. Even the hypothesis that endocrine disruption could result in alterations in the secondary sex ratio, however, has inconsistent support in the scientific literature. Studies of children and grandchildren of women exposed to diethylstilbestrol (DES), an extremely potent endocrine disruptor, have failed to show perturbations in the proportion of male births [58
]. Other data show that the majority of PCB transfer from mothers to their offspring occurs during breastfeeding, not in utero
Besides consistency with other studies of high-dose PCB exposure and the ability to limit analysis to primiparous births, our study benefits from several unique strengths. The sample size is large, allowing us to have adequate power to detect associations. The estimated cumulative preconceptional PCB dose among exposed workers in our study covered a wider range than has been analyzed elsewhere, allowing us to examine whether a threshold effect or dose-response relationship existed. This study used JEMs to assign PCB exposure to workers at three electrical capacitor plants. Using a JEM, rather than direct monitoring for all workers, may introduce some exposure misclassification that would most likely dilute any observed association. Our JEMs, however, were based on industrial hygienist visits to the plants in question, detailed diagrams of the plant layout and processes, air samples, and some individual monitoring results. Using this sort of detailed information to inform a JEM improves the accuracy of exposure assessment [61
]. Though we relied on proxy respondents to complete questionnaires when workers were deceased, the outcome under study (gender of liveborn children) is not likely subject to misreporting as it is discrete and would be known to a proxy respondent. Exposure metrics did not rely on self-report, but on plant records.
Despite these strengths, our results are subject to several limitations. First is our inability to fully distinguish between types of PCBs that workers were exposed to. Though offspring of workers in the New York plant had a reversed sex ratio versus that observed in Indiana and Massachusetts, the PCB mixtures used in New York were similar to those used in Massachusetts; a more detailed analysis by type of PCB, however, cannot be performed based on these data.
Another limitation of our study is the potential for exposure misclassification of workers. A number of personal and workplace factors which we were unable to account for may influence worker exposure to PCBs. Most of the women in our study-whether occupationally exposed or not-- also lived near the factories where they worked, and may have had additional PCB exposure from environmental contamination; this further increases potential misclassification of PCB exposure. Repeated serum PCB measurements would have provided more accurate measures of absorbed PCB dose from all sources, but were only available from a small number of workers and not at the time of conception. In small biomonitoring studies at the Indiana and New York plants, however, serum PCB concentrations of exposed workers were 8-50 times higher than those of unexposed workers or people living in the surrounding communities for H-PCBs, and two to four times higher for L-PCBs [54
]. Serum PCB concentrations correlated fairly well with cumulative exposure estimates (combining both inhalation and dermal exposure) from the JEM in a sample of workers in the Indiana plant [34
]; this suggests that our results cannot be simply attributed to misclassification.
Because most of the women in our study were also exposed to PCBs by living in the communities around the plants, our study might have failed to detect a true relationship - but only if PCBs exerted a threshold (rather than dose-response) effect on secondary sex ratios at doses similar to environmental levels. We observed that the percentages of male infants born to workers in the Indiana and Massachusetts plants were slightly lower than the national average (~49% compared to ~51% for the same time periods), but the secondary sex ratio was similar between occupationally exposed and unexposed workers. The sex ratio observed in the New York plant, however, was similar to the national averages. This argues against a threshold effect on the secondary sex ratio caused by environmental exposure to PCBs among those who were not exposed to PCBs at work.
Unmeasured paternal exposure to PCBs might have contributed to the lower rates of male births we observed among mothers working in the Indiana and Massachusetts plant, which previous studies suggest might be related to the secondary sex ratio [30
]. Some fathers may have also worked in one of these plants, and could have been highly exposed to PCBs. Fathers living near the Indiana and Massachusetts plants might also have had higher environmental exposure to PCBs than fathers living near the New York plant. Unfortunately, we were unable to account for any paternal PCB exposure. Our finding of no association between maternal PCB exposure and the secondary sex ratio does not rule out an association between paternal PCB exposure, or combined parental PCB exposure, and the secondary sex ratio. We were also unable to obtain information on paternal age, which has been linked to the secondary sex ratio in previous studies.