Discussion

The main findings of this study were as follows:

1. No association was observed between PCE exposure and MBW, and a weak association was observed between PCE exposure and SGA infants, overall. Much stronger associations were observed between PCE exposure and both birth weight outcomes among the infants of mothers who were 35 years of age or older and among the infants of mothers who had histories of fetal deaths, especially mothers who had had two or more fetal deaths.

2. A modest association was observed between long-term TCE exposure and decreased MBW and increased SGA births, overall. In male infants, the association between long-term TCE exposure and these study outcomes was much more pronounced. No association was observed between long-term TCE exposure and decreased MBW and increased SGA births among female infants.

3. No association was observed between short-term exposure to TCE and MBW or SGA births.

4. No association was observed between any exposure group and preterm delivery.

5. Fetal death reporting was not complete enough to include in the analyses.

PCE Exposure

In the analyses of PCE exposure, a slight but statistically significant difference in MBW was observed between the exposed and comparison groups. However, the mean difference was so small as to be clinically negligible.

Given the lack of an association between PCE exposure and birth weight in the overall study population, and given the large number of persons in the study, it is unlikely that maternal exposure to the unique combination of contaminants in the Tarawa Terrace water system had much of an effect on birth weight except in some small subgroups. Sources of uncertainty remained, not only for the subgroups in which elevations were found, but also because some potentially important confounders (e.g., maternal smoking, alcohol consumption, and height) were not controlled for in the analysis. It seems unlikely that these factors could have totally obscured a strong effect, especially in a population as homogeneous as the one studied.

Misclassification of both exposure and outcome is a problem in almost every epidemiologic study. In this study, information on exposure was limited to water quality measurements taken over a 3-year period, while the study examined 28 years of birth weight data. It is, therefore, conceivable that no overall PCE effect was observed because the study population was primarily unexposed. However, this seems unlikely, because the activities that resulted in the PCE contamination occurred throughout the study period, and no large differences existed in the study findings for PCE exposure when the sample was restricted to years of known exposure. In addition, a variety of exposure categories were examined and at most a very weak association was observed among the births in which exposure occurred for the longest duration.

Other sources of misclassification might have been more relevant. Even during the known exposure period, exposure occurred intermittently because different wells were used on different days. However, although the exposure was intermittent, it probably occurred at least for some days over every month of the study period. Misclassification of gestational age was also possible, especially among the preterm births. These factors could be relevant to the PCE findings because it would be expected that these sources of misclassification would reduce the ability to detect exposure-related effects.

In addition, exposure to PCE probably did not occur consistently among the pregnant women included in the study because they would have drunk different quantities of water and would have spent variable amounts of time showering. Lack of information on the variation in the personal habits of individual women precluded quantification of the exposure dose each woman received. Such precise dose information is helpful in risk assessment; however, little is currently known about PCE exposure and birth weight, and such information is not necessary to advance understanding.

Groups Potentially Susceptible to PCE Exposure

Despite the overall finding of no association between PCE and birth weight outcomes, there were two clinically distinct subgroups in which PCE exposure was associated with birth weight outcomes--mothers 35 years of age or older and mothers with a history of fetal deaths. There are several reasons to question the meaning of the associations observed in these two subgroups. Of the mothers aged 35 years and older, the number of women studied was relatively small. This is reflected in the wide CIs, suggesting a variety of values that would be consistent with the observed data. Second, the exposed and unexposed women of older ages were not comparable in terms of important demographic characteristics. Given these important differences, there might have been some residual confounding by socioeconomic status, even after adjustment for measured risk factors. In addition, because many other maternal risk factors for SGA such as cigarette smoking and alcohol consumption are related to social factors, lack of comparability on demographic characteristics also may have introduced confounding by behavioral characteristics. However, adjustment for residence in an officer's household did not affect MBW, while adjustment for mother's race had a only a minor effect on MBW. For SGA, adjustment for mother's race produced negligible differences, while adjustment for residence in an officer's household actually increased the OR. The negligible impact of adjustment for these confounders suggests that residual confounding by socioeconomic factors or their behavioral correlates was unlikely.

The finding among mothers aged 35 years and older was unanticipated and not explainable. Nonetheless, the association between PCE exposure and adverse birth outcomes among mothers in this age group was sizeable and biologically plausible. As a general rule, older mothers are considered to be at higher risk for adverse reproductive outcomes, especially infertility, miscarriage, and chromosomal anomalies (⁽⁹²⁾). Older maternal age is not always associated with decreased MBW and increased SGA births (⁽⁹³⁾), but factors that have a clear role in reduced birth weight (e.g., pregnancy-induced hypertension) are associated with maternal age (⁽⁹⁴⁾,⁽⁹⁵⁾). Moreover, previously published reports indicate that the effects of maternal smoking on birth weight increase profoundly with age (⁽⁹⁶⁾, ⁽⁹⁷⁾, ⁽⁹⁸⁾, ⁽⁹⁹⁾).

The second potentially susceptible subgroup in which an association was observed between PCE exposure and birth weight was the group of women with a history of fetal deaths. This group included women who had had either early or late fetal deaths. As with older mothers, women with a history of fetal deaths might represent a more physiologically susceptible subgroup of women with poorer pregnancy outcomes, including low birth weight (⁽¹⁰⁰⁾). Unlike the older mothers, mothers with a history of fetal deaths composed a fairly large group, so that the effect estimates were statistically stable. In addition, there were differences in the strength of the association between PCE exposure and SGA among infants born to mothers who had had only one previous fetal death and to mothers with two or more previous fetal deaths. To the extent that the association increased with the severity of the medical history, it seemed less likely that this association occurred by chance.

Heterogeneity and data quality were a concern for reported history of fetal deaths. Each birth certificate contained information on previous fetal deaths at any gestational age. However, fetal deaths occurring at early gestational ages (miscarriages) have a different etiology than fetal deaths occurring at late gestational ages (stillbirths). Moreover, it is unclear whether fetal deaths at these different gestational ages were reported completely. In addition, differences in reporting of fetal deaths based on socioeconomic status might have occurred.

Analyses by length of exposure within this group provided no insight into the importance of the associations observed. A strict duration-response relationship would have reinforced concern about this association, while a duration-response in which the most exposed persons had the smallest (or no) effect would have lessened concern about this association. However, neither one of these patterns was detected, and a variety of other models would have been consistent with an effect of PCE on fetal weight gain. Moreover, the effect estimates within specific exposure-duration categories were based on relatively small numbers and, hence, would have fluctuated randomly.

One hypothesized mechanism for the reproductive effects of PCE is central nervous system depression of the hypothalamus or pituitary glands resulting in hormonal changes in the mother or the fetus, or both (⁽¹⁰¹⁾, ⁽¹⁰²⁾, ⁽¹⁰³⁾). An actual link between hormonal changes and older maternal age and mothers with a history of fetal deaths would be difficult to determine. Patterns of hormonal function and activity among women in their thirties have not been well studied. Hormonal changes associated with the onset of menopause occur primarily in women in their forties rather than in their late thirties (⁽¹⁰⁴⁾). However, an increase in chromosomally normal spontaneous abortions has been noted among women aged 37 years or older--an effect that could be associated with a decline in uterine function (⁽¹⁰⁵⁾). In the study population, older women and women with a history of fetal deaths were distinct groups, but it does seem plausible that changing uterine function could play a role in both risk groups.

Concentrations of PCE in the drinking water at Camp Lejeune might have been too low to influence birth weight within the overall study population, but the minor stress resulting from PCE exposure among groups that are known to have more reproductive problems could have been sufficient to disturb the developmental environment of the fetus. However, given that other known maternal risk factors for reduced MBW and SGA (i.e., young maternal age or maternal race) were not effect modifiers, mere vulnerability does not appear sufficient to result in an association between PCE exposure and delayed fetal growth.

The observed associations in these two potentially susceptible subgroups must be interpreted cautiously. Nonetheless, these findings suggest an important area for future research either at Camp Lejeune or in another PCE-exposed population. In addition, if these associations bear out under further scrutiny, they could influence general thinking about the groups that are especially vulnerable to toxic substances. As discussed previously, data quality for the reporting of previous fetal deaths is of considerable concern. Information regarding maternal medical conditions, such as diabetes and hypertension, and prepregnancy maternal weight would also greatly enrich existing data on this study population.

The pilot study summarized in Appendix B of this document indicates that it would be feasible to obtain medical record information for a sample of women in these susceptible subpopulations. However, it would not address concerns regarding maternal smoking and drinking habits.

TCE Exposure

Strong associations between long-term TCE exposure and birth weight were observed in male, but not female infants. Both the decrement in MBW and the increase in SGA among male infants exposed to TCE compared with unexposed male infants were large and were on the same magnitude of the effect of maternal cigarette smoking reported the general literature.

The finding of an association in male infants was unexpected, and it reduced the plausibility of a causal association. In studies of male-female differences in TCE metabolism, adult females were found to absorb TCE more completely than adult males and to metabolize TCE more slowly. The slower metabolism of TCE in females is due largely to the higher proportion of bodyweight contributed by fat in adult females. TCE, a lipophilic molecule, is more greatly diluted in females, and hence in females a greater proportion of the body's TCE is stored in fat and is not metabolized (23). Male-female differences in body fat concentration are already present at birth (⁽¹⁰⁶⁾). These are factors that would be expected to make female infants more--and not less--susceptible to TCE.

Solvents in general, and TCE in particular, are known to interfere with lipid metabolism in the liver and to affect lipid composition in the liver and brain (⁽¹⁰⁷⁾,⁽¹⁰⁸⁾). It is possible to speculate that the mechanisms that promote fat accumulation in female fetuses are sufficient to overcome small changes in lipid metabolism and composition, while male fetuses, lacking the same fat-accumulating hormones, do not have the capacity to overcome such small changes. In addition, it is possible that males, who have a higher mortality rate at birth and throughout infancy (⁽¹⁰⁹⁾), are slightly more susceptible to toxic insult, and they might respond to lower doses than females. Males also grow faster late in pregnancy compared with females; this higher growth rate might make them more vulnerable to interference.

In a study of occupational exposure to anesthetic gases and miscarriage, Askrog and Harvald reported a higher than expected proportion of female infants, suggesting a male-specific embryo lethality (⁽¹¹⁰⁾). In a similar study, Cohen et al. (⁽¹¹¹⁾) did not determine a sex-specific association between anesthetic gas exposure and miscarriage. However, there are a variety of different anesthetic gases, and Cohen et al. (108) did not study TCE. The English summary of the Askrog and Harvald study (107) did not indicate which anesthetic gases were evaluated.

Other studies of environmental exposure to toxic substances and birth weight have observed more pronounced effects in males than in females (63,⁽¹¹²⁾,⁽¹¹³⁾). The chemicals examined in two of these studies (109,110) were polycyclic aromatic compounds such as polychlorinated biphenyls and dioxins; in the third study (63), the responsible compounds were not identified. In other studies of polycyclic aromatic compounds, sex-specific interactions involving similar exposures either have not been observed (⁽¹¹⁴⁾) or were not evaluated (⁽¹¹⁵⁾). The differences reported by Rylander et al. (110) between the sexes were not nearly as profound as the differences observed in this study. In addition, polycyclic aromatic compounds might be expected to have more sex-specific effects because they bind to estrogen receptors (⁽¹¹⁶⁾). TCE has a very different chemical structure from these other compounds.

Given the small numbers of long-term TCE-exposed infants, it is also possible that the observed associations in this group occurred by chance or reflected bias. One potential source of confounding was that the long-term TCE-exposed group was a select subpopulation. Unlike the rest of the housing areas whose inhabitants had diverse occupations, residents of Hospital Point were primarily hospital workers. Therefore, the observed effect might have been associated with the presence of characteristics unique to health-care personnel or their spouses. Although possible, this seems unlikely because the effects of behavioral factors such as smoking, alcohol consumption, and patterns of use of medical care on birth weight have been studied much more frequently than the effects of TCE on birth weight. These factors have not been observed to have sex-specific effects. Contaminants other than TCE that the active-duty parents (which in all of these cases were the fathers) might have been exposed to at work were also possible sources of confounding. However, these contaminants would not necessarily have been more likely to have a sex-specific effect than TCE.

Chance might also have played a role in the sex specificity of the effects results of this analysis. Because there were only three SGA infants, the fact that they were all male could have been completely accidental. Arguing against this possibility is the generally decreased birth weight of all 12 TCE-exposed infants, none of whom had a birth weight above the average weight for their gestational age, and seven of whom fell below the 25th percentile in birth weight.

The study's failure to find any association between short-term exposure to TCE and birth outcomes was not inconsistent with findings in the long-term TCE-exposed group. It might be expected that a 12-day exposure would have less of an effect than a 40-week exposure. However, the short-term exposure findings also lend no support to the long-term exposure findings. The timing of the exposure was heterogeneous. One particularly relevant gestational time period is very early in gestation, during embryonic development, because cell death that occurs at this time can affect the development of entire organs. Another relevant gestational time period is shortly before delivery, because any effect of exposure earlier in pregnancy might be obscured by catch-up growth. However, the small number of short-term TCE-exposed births occurring within either of these two relevant periods would make it difficult to exclude the possibility of an effect specific to one or both of these particular times.

In short, the lack of additional studies to either support or refute the findings and the limitations in the size of the exposed population prevented stronger conclusions from being reached regarding the potential effects of long-term TCE exposure. Moreover, the male-only effect greatly weakened the biologic plausibility of the long-term TCE findings. However, it should be noted that many factors that adversely affect health have been identified and confirmed through epidemiologic analysis without a clear explanation as to the mechanism of action. TCE is an extremely common exposure at hazardous waste sites, and pregnancy outcomes are near or at the top of the list of community concerns when such exposures occur. Given these factors, and the magnitude of the association observed, the potential effects of exposure to TCE during pregnancy deserve further study. The best way to obtain more conclusive information on the effects of TCE exposure on birth weight is to repeat a similar analysis in a larger population of TCE-exposed pregnancies in which TCE concentrations occurred at similar or higher levels. In the meantime, it is far from certain that the long-term TCE exposure was actually responsible for the decreased MBW and increased SGA observed in male infants at Camp Lejeune. Still, prudence would dictate that the associations between TCE exposure and birth weight observed at Camp Lejeune be seriously considered both when identifying new avenues for research and when assessing the health impact of TCE exposure at Camp Lejeune and other hazardous waste sites.