The distribution of demographic characteristics in the unexposed and PCE-exposed groups is presented in Table 8. PCE-exposed mothers were less likely to live in officer's housing (18% unexposed, 8% exposed), less likely to be college educated (11% unexposed, 5% exposed), and less likely to have a college-educated partner (18% unexposed, 7% exposed).
Table 9 contains results of the analyses of birth outcomes comparing PCE-exposed and PCE-unexposed residents. The difference in mean birth weight between the PCE-exposed and PCE-unexposed groups was -24 g (90% CL: -41, -7). The OR for PCE exposure and SGA was 1.2 (90% CL: 1.0, 1.3). The OR for PCE exposure and preterm birth was 1.0 (90% CL: 0.9, 1.2). None of the demographic characteristics examined met the criteria for confounding; therefore, adjusted estimates are not presented.
Year of Birth
When the analyses were restricted to data for 1982-1985, the effect estimates changed only slightly and without any pattern. The difference in MBW adjusted for residence in an officer's household was +5 g (90% CL: -34, +44) between PCE-exposed and PCE-unexposed births for these years, and the OR for PCE and SGA adjusted for father's level of education and military pay grade was 1.3 (90% CL: 1.0, 1.8). The OR for PCE and preterm birth was 0.7 (90% CL: 0.6, 0.9) adjusted for father's level of education and mother's race.
Duration of Exposure
Tables 10 and 11 present analyses based on the duration of exposure to PCE. Differences in MBW ranged from -31 g to +18 g for the various exposure categories. The ORs for SGA ranged from 0.9 to 1.2 for the various exposure categories. For preterm births, the ORs ranged from 0.8 to 1.3 and did not follow any clear pattern with duration of exposure.
Potential Interactions Between PCE and Demographic Characteristics
Crude stratum-specific estimates of the differences in MBW for PCE-exposed and PCE-unexposed infants were computed for each covariate of interest. For mother's race, military pay grade, mother's age, mother's history of fetal death, and father's level of education, at least one crude stratum-specific estimate had a mean difference of >50 g between PCE-exposed and PCE-unexposed infants. Table 12 contains adjusted stratum-specific estimates of the difference in MBW between exposed and unexposed infants for these potential effect modifiers. After adjustment for other covariates, only mother's age clearly modified the association between PCE exposure and MBW. In mothers younger than 35 years of age, the adjusted difference in MBW between PCE-exposed and PCE-unexposed mothers was -9 g (90% CL: -23, +6). The adjusted difference in MBW between infants of PCE-exposed and unexposed mothers 35 years of age or older was -205 g (90% CL: -333, -78). A much weaker effect modification was observed between PCE exposure and mother's history of fetal death. The adjusted difference in MBW between infants of PCE-exposed and PCE-unexposed mothers who had no history of fetal death was -21 g (90% CL: -40, +3). A similar estimate of -28 g (90% CL: -79, +24) was seen for PCE-exposed and PCE-unexposed infants born to mothers with one previous fetal death. The adjusted difference in MBW in PCE-exposed and PCE-unexposed infants born to mothers with two or more fetal deaths was -91 g (90% CL: -190, +8).
Crude stratum-specific estimates of the association between PCE exposure and SGA births were computed for each covariate of interest. In crude analyses, military pay grade, mother's age, and mother's history of fetal deaths contained at least one stratum that differed from the overall effect estimate by more than 25%. Table 13 contains adjusted stratum-specific estimates for these covariates. After adjustment for other confounders, mother's age and mother's history of fetal death were both effect modifiers of the association between PCE exposure and SGA. In infants born to mothers less than 35 years of age, the adjusted OR for the association between PCE exposure and SGA was 1.1 (90% CL: 1.0, 1.2). In infants born to mothers 35 years of age or older, the association between PCE exposure and SGA was 4.0 (90% CL: 1.6, 10.2). In infants born to mothers with histories of two or more fetal deaths, the OR for PCE exposure and SGA was 2.5 (90% CL: 1.5, 4.6). For infants born to mothers with one previous fetal death, the OR for PCE exposure and SGA was 1.4 (1.0, 1.9). In infants born to mothers with no history of fetal deaths, the OR was 1.0 (90% CL: 0.9, 1.1). The slightly lower prevalence of SGA in mothers with one fetal death compared with mothers with no fetal deaths reflected the greater parity of these mothers. (91) Because (1) the baseline rate of SGA in the group of unexposed infants of mothers who had had two or more fetal deaths was slightly lower than the prevalence of SGA among infants of mothers who had no histories of fetal deaths, (2) the total number of births to mothers who had two or more fetal deaths was small, and (3) because there was some evidence of effect modification even among mothers with only one previous fetal death, the categories for one previous fetal death and two or more previous fetal deaths were collapsed. The adjusted OR for PCE exposure and SGA among mothers with a history of one or more fetal deaths was 1.6 (90% CL: 1.2, 2.1).
Duration of Exposure to PCE in Mothers With Histories of Fetal Deaths
The influence of duration of exposure in the PCE-exposed and PCE-unexposed groups was explored for mothers with histories of one or more fetal deaths (Tables 14 and 15). Differences in MBW ranged from -121 g to +5 g, and ORs for SGA ranged from 1.4 to 2.1. These effect estimates did not follow a pattern of increasing effect with increasing duration. However, the CIs were sufficiently wide that it was difficult to conclude whether any real differences were present across the duration of exposure categories. There were too few mothers aged 35 years or older to permit analysis of duration of exposure in this group.
Long-Term TCE Exposure
The demographic characteristics for the long-term TCE-exposed group and an unexposed comparison group are described in Table 16. Because the housing area where long-term exposure to TCE occurred was for officers' families, a comparison group consisting of infants born to residents of unexposed officers' housing was used in all analyses of this group. It was felt that this restriction would make the exposed and unexposed groups more comparable in terms of demographic characteristics. Some differences between the two groups remained for the distribution of sex of infant, mother's age, military pay grade, parity, father's education level, and self-reported maternal history of fetal death.
Unadjusted measures of association between exposure and outcome are presented in Table 17. A -108 g (90% CL: -230, -13) difference in MBW and an OR of 1.5 for SGA (90% CL: 0.5, 3.8) were observed in the exposed group compare with the unexposed group. There were no preterm births in the long-term TCE-exposed group. After adjustment for gestational age, the difference in MBW was -139 g (90% CL: -277, -1) in the TCE-exposed group. Because there were only three TCE-exposed infants that were SGA, it was not possible to assess confounding for this outcome using multiple regression models. However, in simple stratified analysis, none of the covariates influenced the OR by more than 10%.
Interaction Between Long-Term TCE Exposure and Sex of Infant
Because there were so few observations in this category, no attempt was made to address the issue of interaction for most of the covariates. The interaction between long-term TCE exposure and sex of infant was examined because this interaction was so large that it was observed by simple inspection of the data. Adjusted models for the association between TCE exposure and MBW are presented by sex in Table 18. For females, there was almost no difference in MBW between the exposed and the unexposed groups. In exposed males, however, the difference in MBW was -312 g (90% CL: -632, -102) compared with their unexposed counterparts. Table 18 also examines the influence of duration of exposure on MBW. In models restricted to infants of mothers who resided in family base housing for 20 or more weeks during pregnancy, the difference in MBW between the exposed and unexposed groups was similar to that observed for the entire data set. However, the small number of exposed women residing in base housing for 20 weeks or more made this difficult to examine.
Interaction between sex and long-term TCE exposure was also present in the SGA analysis (Table 19). All three infants who were SGA were males. The OR for long-term TCE exposure among male infants was 3.9 (90% CL: 1.1, 11.8). No females in the TCE-exposed group were SGA, compared with 1.1 expected based on the prevalence of SGA in the unexposed group. The mothers of each of the three SGA infants in the long-term TCE-exposed group resided in an TCE-exposed housing area during the entire pregnancy.
Short-Term TCE Exposure
The distributions of demographic characteristics between short-term TCE-exposed and unexposed residents are compared in Table 20. By definition, all of the residents in the short-term TCE-exposed group were born in 1985. Because a trend of increasing birth weight with later year of birth was observed in this data set, only infants born during 1983-1985 were included in the unexposed group. Except for slight differences in parity and maternal education level, the exposed and unexposed groups were comparable demographically.
Table 21 presents crude analyses of birth outcomes among residents of unexposed and short-term TCE-exposed housing. MBW was slightly higher in the exposed group for both sexes combined. There were no differences in MBW in analyses restricted to males only. The prevalence of SGA was lower in this group compared with the unexposed group. Adjustment for potential confounders did not eliminate the differences between exposed and unexposed groups.
Within the course of each pregnancy, the timing of the short-term TCE exposure varied from the first week of gestation to the fortieth week. It was anticipated that the effect of exposure might be limited to a particular time in gestation, that is, during some critical period of organogenesis. Alternatively, the possibility was considered that the weeks closest to birth would be most relevant, because weight gain is greatest at the end of gestation and there would be limited time for catch-up growth. Examination of MBW based upon both (1) weeks elapsed between exposure and birth, and (2) gestational age at time of exposure revealed no pattern of decrement with exposure (see Figures 2 and 3). However, the number of observations within each time frame of exposure was quite small. The five SGA infants in the short-term TCE-exposed group did not share any distinct characteristics with regard to timing of exposure.