We observed a reduction in the ratio of male:female births in relation to total maternal PCB concentrations in this population. Inverse associations between all nine PCB congeners and secondary sex ratio were also observed, though not all of these were statistically significant at the 0.05 level. In models adjusted for hormonal medications taken in the periconceptional and early pregnancy period, higher concentrations of total PCBs were significantly associated with male birth ratio (p = 0.02). As PCB blood concentrations have been shown to remain very stable during pregnancy [37
], these findings suggest that high maternal PCB concentrations, at least for some congeners, may either favor the fertilization by female sperm, or result in greater male embryonic or fetal losses. Sex ratio alterations or decreased survival of wildlife in association with organochlorine exposures have been well documented [38
], although not all studies have confirmed such findings [42
Only a handful of human studies of PCBs with sufficient sample sizes have reported data on sex ratio. At environmental exposures, several provide some suggestion of a reduced proportion of males [43
]. Weisskopf and colleagues [46
] examined consumers of sport-fish caught in the Great Lakes and observed a reduction in male births in association with maternal but not paternal serum PCB concentration. In contrast, another investigation found increased male offspring in relation to paternal PCB exposures, though this analysis may have over-adjusted for multiple exposures correlated with paternal PCBs [47
]. Reports from the Yucheng [48
] and Yusho [16
] accidents showed no alteration in sex ratio, however, serious shortcomings in these studies should be noted. The Yucheng group consisted of a select set of mothers who had registered with the health department as having been affected, and who had at least one live child in 1985. Not all affected subjects were registered, and some of the most severely affected children may not have survived: since either of these factors could have been related to the sex of the child, the sex ratio of the survivors could be highly confounded. Moreover, both the Yucheng and the Yusho studies were small: a total of 137 births in the former and only 67 in the latter. These sample sizes would not be adequate to detect small variations in sex ratio. Analysis of a more complete cohort from Yucheng, numbering 469 births with paternal data and 902 with maternal data revealed reduced male:female sex ratio when fathers were exposed at a young age [15
], results similar to those previously reported for dioxin exposures from the Seveso accident. Analyses for maternal exposures also showed a lower than expected proportion of males, but the findings were not statistically significant. The difference in findings using maternal vs. paternal exposure appeared to be due, at least partly, to a higher percentage of male offspring among unexposed fathers than among unexposed mothers, or, than expected in the population. Since every child has both a mother and a father, but fewer fathers were tested, this raises the question of selection bias for children whose fathers participated in the study.
In addition to studies of secondary sex ratio, research on the proportion of Y-bearing spermatozoa in relation to paternal PCBs have produced mixed results, i.e., both positive and negative associations [49
]. This would suggest that increased paternal PCB concentrations may also alter secondary sex ratio, but other factors may modify this effect. It is not clear from these studies – due to inadequate sample size – whether changes in Y-bearing spermatozoa resulted in more or fewer male births.
The possibility of bias deserves consideration. Selection factors would have to be associated with sex ratio and PCBs, jointly to have produced artifacts for the relationship of these variables of primary interest. Since demographic variables are not generally associated with sex ratio, they would not bias the results. The slight preponderance of girls followed up at age 5 years (50.5%) as compared to the original cohort (48.8% female), combined with our deliberate oversampling of children with low cognitive scores or hearing difficulties increased the percentage of females to 53.4%, however, our sampling weights accounted for these differences, comparable to their standard use in survey sampling. It is possible that because a small number of deaths occurred between birth and five years of age, part of the shift in sex ratio might be due to far higher deaths among boys with higher PCBs, or among girls with lower PCBs. At these exposure levels, however, which were far lower than occurred in the Yusho or Yucheng incidents, it seems unlikely that PCBs would have had a sufficiently strong association with child mortality to produce the results observed and yet not be related to fetal mortality. If PCBs were strongly associated with unknown or gestational age below 35 weeks, some selection bias could have occurred by our exclusion of these births. However, the literature does not support an association of preterm delivery with PCBs, nor did we find any trend of higher PCBs with earlier gestational age among births at weeks 35–45 [23
]. In regard to other variables, the sample was representative of the original cohort. Thus, selection bias seems unlikely to be of sufficient magnitude to explain the findings.
Validity of our data is supported by the confirmation of known risk factors for other outcomes, such as maternal smoking, short stature, and low body mass index in relation to intrauterine growth restriction [23
]. High quality PCB determinations were ensured by use of within-batch replicates, standards across batches, and surrogate standards. Thus, in comparison with previous work, our study is unique in that it thoroughly addresses maternal exposures, uses an adequate sample, has high quality direct measurements of PCB exposures close in time to the events of interest, and represents exposures at the upper end of today's environmentally exposed populations.