Abstract

A study of mean birth weight, small-for-gestational-age infants, and preterm birth was conducted at the US Marine Corps Base at Camp Lejeune, North Carolina, where drinking water was contaminated with volatile organic compounds. Tetrachloroethylene (PCE) was the predominant contaminant. The authors used multiple linear and logistic regression to analyze 1968–1985 data from 11,798 birth certificates. Overall, at most weak associations were observed between PCE exposure and study outcomes. However, associations were found between PCE exposure and birth-weight outcomes for infants of older mothers and mothers with histories of fetal loss. Adjusted mean birth-weight differences between PCE-exposed and unexposed infants were −130 g (90% confidence interval (CI): −236, −23) for mothers aged 35 years or older and −104 g (90% CI: −174, −34) for mothers with two or more previous fetal losses. Adjusted odds ratios for PCE exposure and small-for-gestational-age infants were 2.1 (90% CI: 0.9, 4.9) for older mothers and 2.5 (90% CI: 1.5, 4.3) for mothers with two or more prior fetal losses. These results suggest that some fetuses may be more vulnerable than others to chemical insult.

We retrospectively studied mean birth weight, small-for-gestational-age (SGA) birth, and preterm birth in infants who were exposed in utero to drinking water contaminated with a mixture of volatile organic compounds at the US Marine Corps Base at Camp Lejeune, North Carolina. Tetrachloroethylene (PCE) was the predominant drinking water contaminant.

PCE is a common and potentially important reproductive system toxin (1). It is lipophilic (1) and, in animal models, crosses the placenta (2). Trichloroacetic acid, a toxic metabolite of PCE, has been observed to persist in the rat fetus and may cycle from the fetus into the amniotic fluid and back again (2). PCE exposure has been found to reduce weight and delay development in fetal mice but not in fetal rats (3).

Although a number of studies have linked spontaneous abortion in humans with occupational exposure to solvents or employment in the dry-cleaning industry (47), two studies of occupational exposure to halogenated hydrocarbons did not observe associations with SGA, preterm delivery, or both (5, 8). In the single known previous study of PCE-contaminated drinking water and birth outcomes, PCE contamination at 10 ppb or more was associated with oral cleft defects but not with mean birth weight, SGA, or preterm delivery (9). Chloroform and trichloroethylene, two compounds chemically similar to PCE, have been more frequently studied in contaminated drinking water. Chloroform contamination at 50 ppb or more has been associated with SGA in some studies (911) but not others (12, 13). One study found that trichloroethylene exposure (6–267 ppb) during the third trimester of pregnancy was associated with SGA (14); a study of lower-level exposure to trichloroethylene (10 ppb or more) revealed no association with SGA (9).

Tarawa Terrace (TT) housing areas I and II are 2 of 15 housing areas for Navy and Marine Corps families stationed at the US Marine Corps Base at Camp Lejeune. Both receive water from the same water distribution system. As shown in table 1, drinking water supplied to TT contained PCE and lower levels of other volatile organic compounds (15, 16). The source of the volatile organic compounds was the ABC One-Hour Cleaners (ABC), a dry-cleaning business. PCE was released to the ABC septic system via a floor drain during routine storage, use, and recycling. The septic system discharged wastewater directly into subsurface soil. Supply well 26 (TT26) is located approximately 850 feet (259 m) from this septic field. Although TT26 is up-gradient of ABC, its cone of depression included the septic field.

TABLE 1.

Concentrations of volatile organic compounds in supply well #26 (TT26*) and finished water samples from the Tarawa Terrace distribution system, US Marine Corps Base at Camp Lejeune, North Carolina, 1982–1985

Date Contaminant Finished water concentration (ppb) TT26 concentration (ppb) 
May–June 1982 PCE 76–104  
 TCE ND  
January 16, 1985–February 5, 1985 PCE 215 1,580 
 TCE ND 57 
 DCE NA 92 
 VC NA 27 
February 12–19, 1985 PCE ND ND-64 
 TCE ND ND 
 DCE ND ND 
 VC ND ND 
Date Contaminant Finished water concentration (ppb) TT26 concentration (ppb) 
May–June 1982 PCE 76–104  
 TCE ND  
January 16, 1985–February 5, 1985 PCE 215 1,580 
 TCE ND 57 
 DCE NA 92 
 VC NA 27 
February 12–19, 1985 PCE ND ND-64 
 TCE ND ND 
 DCE ND ND 
 VC ND ND 
*

Supply well TT26 was shut off on February 8, 1985.

PCE, tetrachloroethylene; TCE, trichloroethylene; ND, not detected (detection limit, 10 ppb); DCE, 1,2-dichloroethylene; NA, not analyzed; VC, vinyl chloride.

Information regarding levels of volatile organic compounds in TT26 was not available prior to 1982, but we assumed that the supply well was contaminated soon after it was dug in 1958. ABC opened in 1954; according to the owners, business practices were essentially unchanged between 1960 and 1985, although volume decreased temporarily in 1975. TT26 and two other contaminated wells were permanently disconnected from the TT water system on February 8, 1985. After that date, TT26 samples were not indicative of previous contamination levels, because the pumping operation itself was responsible for drawing con-taminants into the well.

TT26 was the only one of six routinely used wells with detectable contamination. On any given day, use of five of the wells was essentially random. Therefore, exposure over time was intermittent and fluctuated depending on the proportion of water pumped from the contaminated well each day. PCE concentrations in finished water samples were consistent with what might be expected if water from TT26 had been diluted with a similar volume of water from four other wells.

A notable characteristic of the water distribution systems at Camp Lejeune is that all supply wells for a given system were mixed before distribution. Therefore, PCE concentrations were not related to the distance between particular housing units and TT26. Because the water was mixed prior to distribution, we expected that PCE concentrations in water delivered to each TT housing unit would have been similar on any given day (15).

MATERIALS AND METHODS

The study population consisted of singleton liveborn infants of 20 or more completed weeks of gestation born to Camp Lejeune residents in 1968–1985. We selected 1968 as the first year of study because it was the first year that that part of the North Carolina birth certificate was computerized.

Birth certificates were matched to base family housing records on mother's address and, in most cases, father's name. Housing records contained dates of occupancy and military pay grade for the family member assigned to the unit. Births to mothers who were living in base family housing when they delivered and who had lived there for at least 1 week prior to giving birth were included. TT residents were considered PCE exposed. Two groups of residents were exposed to trichloroethylene through a different water system and were excluded. Residents of base trailer parks were excluded because housing records were incomplete. The remaining base family housing residents were classified as unexposed based on water samples drawn from supply wells and finished distribution systems in 1984 and 1985.

Mean birth weight, SGA, and preterm birth were measured by using North Carolina birth records. We excluded less than 5 percent of the exposed and unexposed groups because of poor data quality. Premature births were defined as livebirths occurring at less than 37 completed weeks of gestation. Gestational age was calculated from the mother's last menstrual period. Approximately 1 percent of observations were deleted from the mean birth weight and SGA analysis, but not from the preterm birth analyses, because of extremely unlikely combinations of gestational age and birth weight (17); all of these observations were for preterm births. Twelve “preterm births” of infants weighing 3,600 g or more were recoded as full-term.

SGA is normally measured by comparing birth weight at specific gestational ages with a gestational-age-specific birth-weight distribution. Livebirths of infants weighing less than the 10th percentile are classified as SGA. Given the military's somewhat unique social and health care environment, we would ideally have used a standard birth-weight distribution for a military population during the years of study. Because such a standard was not available, we evaluated three different standards (9, 17, 18) and selected the one published by Williams et al. (18) because it came closest to classifying 10 percent of the non-PCE-exposed births as SGA. Of the three standards evaluated, this was the only one derived specifically from White births, but, when all races were included, it fit best.

Mean differences or odds ratios and 90 percent confidence intervals were computed. The covariates included in simple stratified analyses were infant's sex and year of birth; mother's race, age, educational level, parity, adequacy of prenatal care (19), marital status, and history of fetal death; and father's age, educational level, and military pay grade. Variable selection in regression models proceeded by backward elimination. Initial models considered main effects only, treating covariates as potential confounders. Subsequently, effect modification was also considered. For the analyses of SGA and preterm birth, covariates were retained as effect modifiers in regression models only if they were biologically plausible, described heterogeneous groups in which the odds ratios differed by more than 25 percent, and had p values of less than 0.20 (20, 21). For mean birth weight, covariates for which at least one stratum-specific estimate showed a mean difference between PCE-exposed and unexposed births of at least 50 g were examined for effect modification.

We adjusted for gestational age in mean birth-weight analyses to distinguish between associations with reduced fetal weight gain and those with early time of birth; failure to make this distinction can obscure associations between exposure to hazardous waste and delayed fetal growth (10, 22).

The influence of duration of exposure was also explored. For each household, the dates of occupancy were used to determine whether and when each family moved during the mother's pregnancy. Length of residence at TT prior to giving birth served as a surrogate for length of exposure. Following analysis of a sample of housing records, we assumed that length of exposure indicated the number of consecutive weeks prior to delivery that a mother lived at TT. For example, a mother residing at TT for 10 weeks lived at TT during the last 10 weeks of the pregnancy. Cutpoints for duration of exposure analyses were developed following a literature review to identify periods during gestation when a toxic agent might most interfere with fetal growth or timing of birth.

In the absence of more refined historical information, we assumed that PCE exposures were essentially constant throughout the years of study. However, we also conducted separate analyses of births occurring during the period of documented exposure—May 27, 1982, through February 7, 1985.

RESULTS

The distribution of demographic characteristics in the unexposed and PCE-exposed groups is presented in table 2. PCE-exposed mothers were less likely to be White, less likely to live in officer's housing, less likely to be college educated, and less likely to have a college-educated partner.

TABLE 2.

Characteristics of PCE*-exposed and unexposed births, US Marine Corps Base at Camp Lejeune, North Carolina, 1968–1985

Characteristic PCE-exposed (n = 6,117)
 
Unexposed (n = 5,681)
 
No. % No. % 
Mother's race     
 White 4,339 70.9 4,487 79.0 
 Black 1,415 23.1 1,006 17.7 
 Other 363 5.9 188 3.3 
Infant's sex     
 Female 3,057 50.0 2,778 48.9 
 Male 3,060 50.0 2,903 51.1 
Year of birth     
 1968–1970 1,046 17.1 1,000 17.6 
 1971–1975 1,556 25.4 1,507 26.5 
 1976–1980 1,859 30.4 1,639 28.9 
 1981–1985 1,656 27.1 1,535 27.0 
Mother's age (years)     
 <20 759 12.4 1,235 21.7 
 20–24 3,480 56.9 2,406 42.4 
 25–29 1,443 23.6 1,349 23.7 
 30–34 363 5.9 527 9.3 
 ≥35 72 1.2 164 2.9 
Military pay grade     
 Enlisted grades     
  1–3 633 10.3 1,408 24.7 
  4–5 3,875 63.3 1,896 33.4 
  6–9 1,011 16.5 1,177 20.7 
 Officer or warrant officer 507 8.3 1,064 18.7 
 Data missing 91 1.5 136 2.4 
Parity     
 Primiparous 1,861 30.4 2,223 39.1 
 Multiparous 4,251 69.5 3,454 60.8 
 Data missing 0.1 0.1 
Prenatal care     
 Less than adequate 3,846 62.9 3,731 65.6 
 Adequate or better 1,888 30.9 1,581 27.8 
 Data missing 383 6.3 369 6.5 
No. of past fetal deaths     
 ≥2 250 4.1 210 3.7 
 1 815 13.3 656 11.5 
 None 5,049 82.5 4,810 84.7 
 Data missing 0.0 0.1 
Mother's education (no. of years completed)     
 0–11 1,499 24.5 1,354 23.8 
 12 3,319 54.3 2,770 48.8 
 13–15 977 16.0 955 16.8 
 ≥16 319 5.2 600 10.6 
 Data missing 0.0 0.0 
Father's education (no. of years completed)     
 0–11 726 11.9 743 13.1 
 12 4,282 70.0 3,342 58.8 
 13–15 655 10.7 593 10.5 
 ≥16 447 7.3 997 17.5 
 Data missing 0.1 0.1 
Characteristic PCE-exposed (n = 6,117)
 
Unexposed (n = 5,681)
 
No. % No. % 
Mother's race     
 White 4,339 70.9 4,487 79.0 
 Black 1,415 23.1 1,006 17.7 
 Other 363 5.9 188 3.3 
Infant's sex     
 Female 3,057 50.0 2,778 48.9 
 Male 3,060 50.0 2,903 51.1 
Year of birth     
 1968–1970 1,046 17.1 1,000 17.6 
 1971–1975 1,556 25.4 1,507 26.5 
 1976–1980 1,859 30.4 1,639 28.9 
 1981–1985 1,656 27.1 1,535 27.0 
Mother's age (years)     
 <20 759 12.4 1,235 21.7 
 20–24 3,480 56.9 2,406 42.4 
 25–29 1,443 23.6 1,349 23.7 
 30–34 363 5.9 527 9.3 
 ≥35 72 1.2 164 2.9 
Military pay grade     
 Enlisted grades     
  1–3 633 10.3 1,408 24.7 
  4–5 3,875 63.3 1,896 33.4 
  6–9 1,011 16.5 1,177 20.7 
 Officer or warrant officer 507 8.3 1,064 18.7 
 Data missing 91 1.5 136 2.4 
Parity     
 Primiparous 1,861 30.4 2,223 39.1 
 Multiparous 4,251 69.5 3,454 60.8 
 Data missing 0.1 0.1 
Prenatal care     
 Less than adequate 3,846 62.9 3,731 65.6 
 Adequate or better 1,888 30.9 1,581 27.8 
 Data missing 383 6.3 369 6.5 
No. of past fetal deaths     
 ≥2 250 4.1 210 3.7 
 1 815 13.3 656 11.5 
 None 5,049 82.5 4,810 84.7 
 Data missing 0.0 0.1 
Mother's education (no. of years completed)     
 0–11 1,499 24.5 1,354 23.8 
 12 3,319 54.3 2,770 48.8 
 13–15 977 16.0 955 16.8 
 ≥16 319 5.2 600 10.6 
 Data missing 0.0 0.0 
Father's education (no. of years completed)     
 0–11 726 11.9 743 13.1 
 12 4,282 70.0 3,342 58.8 
 13–15 655 10.7 593 10.5 
 ≥16 447 7.3 997 17.5 
 Data missing 0.1 0.1 
*

PCE, tetrachloroethylene.

Some percentages do not total 100 because of rounding.

In models without interaction terms, the difference in mean birth weight between the PCE-exposed and unexposed groups was −26 g (90 percent confidence interval (CI): −43, −9). The odds ratio for PCE exposure and SGA was 1.2 (90 percent CI: 1.0, 1.3). The odds ratio for PCE exposure and preterm birth was 1.0 (90 percent CI: 0.9, 1.1). Adjustment for potential confounders did not tangibly affect these estimates.

Table 3 examines birth outcomes by length of exposure to PCE. Associations between PCE exposure and mean birth weight, SGA, and preterm birth followed no obvious pattern with duration of exposure.

TABLE 3.

Duration of PCE* exposure and birth-weight outcomes at the US Marine Corps Base at Camp Lejeune, North Carolina, 1968–1985

Duration of exposure No. Mean birth weight (g) (SE*) Mean difference 90% CI* Small for gestational age
 
Odds ratio 90% CI Preterm birth
 
Odds ratio 90% CI 
No. No. 
Never exposed 5,344 3,348 (7.7)   488 9.1   389 7.3   
1–3 weeks 189 3,345 (38.5) 18 −40, 76 15 7.9 0.9 0.5, 1.3 14 7.4 1.0 0.6, 1.6 
4–10 weeks 597 3,291 (25.3) −17 −51, 17 60 10.1 1.1 0.9, 1.4 55 9.2 1.3 1.0, 1.7 
11–20 weeks 915 3,298 (19.2) −31 −59, −3 84 9.2 1.0 0.8, 1.2 86 9.4 1.3 1.1, 1.6 
>20 weeks and < entire pregnancy 1,551 3,351 (13.9) −28 −50, −5 16 10.8 1.2 1.0, 1.4 94 6.1 0.8 0.7, 1.0 
Entire pregnancy and <1 year before LMP* 1,994 3,323 (12.8) −15 −35, 5 207 10.4 1.2 1.0, 1.3 158 7.9 1.1 0.9, 1.3 
Entire pregnancy and ≥1 year before LMP 605 3,349 (22.3) −18 −51, 16 61 10.1 1.1 0.9, 1.4 36 5.9 0.8 0.6, 1.1 
Duration of exposure No. Mean birth weight (g) (SE*) Mean difference 90% CI* Small for gestational age
 
Odds ratio 90% CI Preterm birth
 
Odds ratio 90% CI 
No. No. 
Never exposed 5,344 3,348 (7.7)   488 9.1   389 7.3   
1–3 weeks 189 3,345 (38.5) 18 −40, 76 15 7.9 0.9 0.5, 1.3 14 7.4 1.0 0.6, 1.6 
4–10 weeks 597 3,291 (25.3) −17 −51, 17 60 10.1 1.1 0.9, 1.4 55 9.2 1.3 1.0, 1.7 
11–20 weeks 915 3,298 (19.2) −31 −59, −3 84 9.2 1.0 0.8, 1.2 86 9.4 1.3 1.1, 1.6 
>20 weeks and < entire pregnancy 1,551 3,351 (13.9) −28 −50, −5 16 10.8 1.2 1.0, 1.4 94 6.1 0.8 0.7, 1.0 
Entire pregnancy and <1 year before LMP* 1,994 3,323 (12.8) −15 −35, 5 207 10.4 1.2 1.0, 1.3 158 7.9 1.1 0.9, 1.3 
Entire pregnancy and ≥1 year before LMP 605 3,349 (22.3) −18 −51, 16 61 10.1 1.1 0.9, 1.4 36 5.9 0.8 0.6, 1.1 
*

PCE, tetrachloroethylene; SE, standard error; CI, confidence interval; LMP, last menstrual period.

Mean birth-weight model adjusted for gestational age and maternal age.

Maternal age and maternal history of fetal loss seemed to interact with PCE for both birth-weight outcomes (table 4). These groups appear to be distinct; 17 percent of PCE-exposed older mothers had had two or more fetal losses, and 3 percent of PCE-exposed mothers with prior fetal losses were older. Compared with their unexposed counterparts, older PCE-exposed mothers were more likely to be non-White (36.1 vs. 20.8 percent), less likely to have college-educated husbands (1.4 vs. 28.0 percent), and less likely to have an officer's or warrant officer's household income (4.2 vs. 26.2 percent).

TABLE 4.

Mean birth weight and prevalence of small-for-gestational-age births for all births, by PCE* exposure and by mother's age and prior fetal loss, US Marine Corps Base at Camp Lejeune, North Carolina, 1968–1985

Characteristic PCE-exposed
 
Unexposed
 
Mean difference 90% CI* 
No. Mean (SE*No. Mean (SE) 
Mean birth weight       
 All births 6,039 3,326 (7.3) 5,630 3,337 (7.4) −26 −43, −9 
 Mother's age (years)       
  <35 5,968 3,326 (7.3) 5,472 3,348 (7.5) −2 −17, 13 
  ≥35 71 3,300 (71.5) 158 3,485 (45.2) −130 −236, −23 
 Mother's previous fetal losses       
  None 4,985 3,335 (7.8) 4,767 3,355 (7.8) −2 −17, 13 
  1 806 3,303 (21.0) 651 3,338 (23.4) −16 −79, 24 
  ≥2 245
 
3,219 (41.9)
 
207
 
3,327 (41.6)
 
−104
 
−174, −34
 
 No.
 
%
 
No.
 
%
 
Odds ratio
 
90% CI
 
Small for gestational age       
 All births 622 10.3 509 9.0 1.2 1.0, 1.3 
 Mother's age (years)       
  <35 611 10.3 501 9.2 1.1 0.9, 1.2 
  ≥35 11 15.5 5.1 2.1 0.9, 4.9 
 Mother's previous fetal losses       
  None 475 9.5 438 9.2 1.1 0.9, 1.2 
  1 104 12.9 56 8.6 1.5 1.1, 2.0 
  ≥2 43 17.6 14 6.8 2.5 1.5, 4.3 
Characteristic PCE-exposed
 
Unexposed
 
Mean difference 90% CI* 
No. Mean (SE*No. Mean (SE) 
Mean birth weight       
 All births 6,039 3,326 (7.3) 5,630 3,337 (7.4) −26 −43, −9 
 Mother's age (years)       
  <35 5,968 3,326 (7.3) 5,472 3,348 (7.5) −2 −17, 13 
  ≥35 71 3,300 (71.5) 158 3,485 (45.2) −130 −236, −23 
 Mother's previous fetal losses       
  None 4,985 3,335 (7.8) 4,767 3,355 (7.8) −2 −17, 13 
  1 806 3,303 (21.0) 651 3,338 (23.4) −16 −79, 24 
  ≥2 245
 
3,219 (41.9)
 
207
 
3,327 (41.6)
 
−104
 
−174, −34
 
 No.
 
%
 
No.
 
%
 
Odds ratio
 
90% CI
 
Small for gestational age       
 All births 622 10.3 509 9.0 1.2 1.0, 1.3 
 Mother's age (years)       
  <35 611 10.3 501 9.2 1.1 0.9, 1.2 
  ≥35 11 15.5 5.1 2.1 0.9, 4.9 
 Mother's previous fetal losses       
  None 475 9.5 438 9.2 1.1 0.9, 1.2 
  1 104 12.9 56 8.6 1.5 1.1, 2.0 
  ≥2 43 17.6 14 6.8 2.5 1.5, 4.3 
*

PCE, tetrachloroethylene; SE, standard error; CI, confidence interval.

Mean birth-weight model includes main effects terms for PCE exposure, mother's age ≥35 years, mother's history of one previous fetal loss, mother's history of two or more previous fetal losses, and interaction terms for PCE exposure and these three covariates. The model also adjusted for gestational age, mother's race, living in an officer's or warrant officer's household, year of birth, and sex of the infant.

The small-for-gestational-age model includes main effects terms for PCE exposure, mother's age ≥35 years, mother's history of one previous fetal loss, mother's history of two or more previous fetal losses, and interaction terms for PCE exposure and these three covariates. The model also adjusted for primiparity, living in an officer's or warrant officer's household, year of birth, and mother's education (no. of years completed).

For PCE-exposed infants of mothers less than 35 years of age with no history of fetal loss, the adjusted mean birth-weight difference was −2 g (90 percent CI: −17, 13), and the adjusted odds ratio for PCE exposure and SGA was 1.1 (90 percent CI: 0.9, 1.2). Among mothers 35 years of age or older, the adjusted mean birth-weight difference was −130 g (90 percent CI: −236, −23). For older mothers, the adjusted odds ratio for PCE exposure and SGA was 2.1 (90 percent CI: 0.9, 4.9). For PCE-exposed infants born to mothers who had one previous fetal loss, the adjusted mean birth-weight difference was only −16 g (90 percent CI: −79, 24) compared with unexposed infants in this category. However, the adjusted odds ratio for PCE exposure and SGA was 1.5 (90 percent CI: 1.1, 2.0). For infants of mothers with two or more prior fetal losses, the mean birth-weight difference rose to −104 g (90 percent CI: −174, −34), and the odds ratio for PCE exposure and SGA was 2.5 (90 percent CI: 1.5, 4.3).

When analyses were restricted to births in the period of documented exposure, the results did not change meaningfully. When all births in this time period were considered, the difference in mean birth weight was −23 g (90 percent CI: −72, 27), and the odds ratios for PCE exposure and SGA and preterm delivery were 1.5 (90 percent CI: 1.0, 2.2) and 0.9 (90 percent CI: 0.6, 1.3) respectively. Our study sample was too small to examine interaction during the period of documented exposure. However, when we restricted analyses to infants whose mothers were 34 years of age or younger, the difference in mean birth weight was 6 g (90 percent CI: −50, 61), and the odds ratio for SGA was reduced to 1.2 (90 percent CI: 0.8, 1.8). The odds ratio for preterm delivery was 1.0 (90 percent CI: 0.6, 1.6). Analyses for this time period were adjusted for mother's race and educational level and for pay grade. Mean birth-weight analyses were also adjusted for gestational age.

DISCUSSION

The main findings of this study were 1) no association between PCE exposure and preterm birth or mean birth weight and 2) a weak association between PCE exposure and SGA births for all infants combined (odds ratio = 1.2 (90 percent CI: 1.0, 1.3)). Stronger associations were observed between PCE exposure and both birth-weight outcomes for infants of mothers who were 35 years of age or older and for infants of mothers with a history of fetal death, especially those who had experienced two or more fetal deaths. For preterm birth, no biologically relevant interactions were found between PCE and covariates. These results indicate nothing about the potential effects of PCE on other pregnancy outcomes including spontaneous abortions and birth defects, which were more plausible given findings from previous literature (49) but could not be studied by using our records-based approach at Camp Lejeune.

The observation of at most a very weak effect for most infants must be tempered by an assessment of potential biases. Some potentially important confounders, such as maternal smoking habits and height, were not controlled. 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 exposure is a chronic problem in environmental epidemiology. In this study, water quality data were available for less than a 3-year period, although the study examined 28 years of birth-outcome data. However, the circumstances that led to the contamination were present throughout the study period. Moreover, to the limited extent that we could assess them, the results of analyses conducted during the period of known exposure were consistent with those for the entire study period.

Other sources of misclassification were likely to have been more relevant. Even during the known exposure period, exposure was intermittent. Nonetheless, given the practice of rotating use of wells, exposure probably occurred during a majority of the days in every month of the study period. The amount of exposure to PCE varied across the different persons studied, because women would have drunk different quantities of water and would have spent variable amounts of time showering. Unfortunately, we lacked information on variations in the personal habits of individual women. It is expected that these sources of misclassification would have reduced our ability to detect exposure-related effects. Extensive misclassification of exposure may also explain why analyses by duration of exposure yielded no additional insights.

There were two distinct groups, mothers aged 35 years or older and mothers with histories of fetal deaths, in which PCE exposure was associated with reduced fetal weight. These groups had limited overlap and seem to reflect two different associations. Limitations of the subgroup analyses include small sample sizes and the potential for residual confounding by socioeconomic status or confounding by unmeasured variables such as cigarette smoking. These issues are especially acute for older mothers.

Birth certificates do not enable distinction between prior spontaneous abortions, induced abortions, and stillbirths. Even among women who lost fetuses at similar gestational ages, there is considerable variability in baseline loss rates (2325). Therefore, mothers with histories of prior pregnancy loss could represent a heterogeneous group in which PCE exposure rendered some, but not all, mothers with previous fetal loss more susceptible to delivering an SGA infant.

The subgroup associations observed between PCE exposure and adverse outcomes are biologically plausible. Mothers aged 35 years or older are at higher risk for infertility, miscarriage, and chromosomal anomalies (26). Older maternal age is not always associated with decreased mean birth weight and increased SGA births (27), but a number of risk factors for SGA such as preeclampsia, pregnancy-induced hypertension, and chromosomal defects are associated with older maternal age (2830). The effects of maternal smoking on birth weight have been observed to increase with age (3133). As with older mothers, women with a history of fetal loss, especially late fetal loss, might represent a physiologically susceptible subgroup with poorer pregnancy outcomes than women without such a history (34, 35). Notably, young maternal age and maternal race are established maternal risk factors for SGA and were not effect modifiers of PCE in this study. Hence, simply being at high risk does not explain the interactions.

One potential mechanism for reproductive effects of PCE is central nervous system depression of the hypothalamus or pituitary glands resulting in hormonal changes in the mother, the fetus, or both (3638). Hormonal changes associated with the onset of menopause occur primarily in women in their forties (39). However, early menopausal changes such as increasing within-woman variation in menstrual cycle length have been observed in a substantial proportion of women aged 35–39 years (40, 41). In addition, an increase in chromosomally normal spontaneous abortions has been noted among women 37 years of age or older, an effect that might be related to a decline in uterine function at this age (42). In our study population, older women and women with a history of fetal death were distinct groups, but a changing hormonal environment or decreased uterine or placental function might have played a role in both risk groups.

As noted above, caution is needed when interpreting the findings regarding interaction. Nonetheless, if these interactions are observed in other populations, they may characterize fetuses that are especially vulnerable to reproductive system toxins.

Reprint requests to Dr. Nancy L. Sonnenfeld, Department of Epidemiology and Public Health, University of New England, 11 Hills Beach Road, Biddeford, ME 04005 (e-mail: nsonnenfeld@mailbox.

This research was partially supported by funds from the US Department of Defense, Environmental Restoration Account provided by the Department of the Navy and by the Agency for Toxic Substances and Disease Registry, Public Health Service, US Department of Health and Human Services.

REFERENCES

1.
Agency for Toxic Substances and Disease Registry. Toxicological profile for tetrachloroethylene. Update. Draft for public comment. Atlanta, GA: US Department of Health and Human Services, Public Health Service, 1995.
2.
Ghantous H, Danielsson BR, Dencker L, et al. Trichloroacetic acid accumulates in murine amniotic fluid after tri- and tetrachloroethylene inhalation.
Acta Pharmacol Toxicol (Copenh)
 
1986
;
58
:
105
–14.
3.
Schwetz BA, Leong BKJ, Gehring PH. The effect of maternally inhaled trichloroethylene, perchloroethylene, methyl chloroform, and methylene chloride on embryonal and fetal development in mice and rats.
Toxicol Appl Pharmacol
 
1975
;
32
:
84
–96.
4.
Taskinen HK. Effects of parental occupational exposures on spontaneous abortion and congenital malformations.
Scand J Work Environ Health
 
1990
;
16
:
297
–314.
5.
Windham GC, Shusterman D, Swan SH, et al. Exposure to organic solvents and adverse pregnancy outcome.
Am J Ind Med
 
1991
;
20
:
241
–59.
6.
Ahlborg G Jr. Pregnancy outcome among women working in laundries and dry-cleaning shops using tetrachloroethylene.
Am J Ind Med
 
1990
;
17
:
567
–75.
7.
Kyyronen P, Taskinen H, Lindbohm ML, et al. Spontaneous abortions and congenital malformations among women exposed to tetrachloroethylene in dry cleaning.
J Epidemiol Community Health
 
1989
;
43
:
346
–51.
8.
Savitz DA, Whelan EA, Kleckner RC. Effect of parents' occupational exposures on risk of stillbirth, preterm delivery, and small-for-gestational-age infants.
Am J Epidemiol
 
1989
;
129
:
1201
–18.
9.
Bove FJ, Fulcomer MC, Klotz JB, et al. Public drinking water contamination and birth outcomes.
Am J Epidemiol
 
1995
;
141
:
850
–62.
10.
Kramer MD, Lynch CF, Isacson P, et al. The association of waterborne chloroform with intrauterine growth retardation.
Epidemiology
 
1992
;
3
:
407
–13.
11.
Savitz DA, Andrews KW, Pastore LM. Drinking water and pregnancy outcome in central North Carolina: source, amount, and trihalomethane levels.
Environ Health Perspect
 
1995
;
103
:
592
–6.
12.
Gallagher MD, Nuckols JR, Stallones L, et al. Exposure to trihalomethanes and adverse pregnancy outcomes.
Epidemiology
 
1998
;
9
:
484
–9.
13.
Dodds L, King W, Woolcott C, et al. Trihalomethanes in public water supplies and adverse birth outcomes.
Epidemiology
 
1999
;
10
:
233
–7.
14.
Massachusetts Department of Public Health. Woburn Environment and Birth Study, synopsis. Boston, MA: Massachusetts Department of Public Health, 1994. (http://www.state.ma.us/dph/beha/reports/woburn/synwebs1.pdf).
15.
Agency for Toxic Substances and Disease Registry. Volatile organic compounds in drinking water and adverse pregnancy outcomes at the US Marine Corps Base Camp Lejeune, North Carolina. Final report. Atlanta, GA: US Department of Health and Human Services, Public Health Service, 1998. (NTIS publication PB 98-156540).
16.
Remedial investigation report. ABC One-Hour Dry Cleaners. Jacksonville, North Carolina, region IV. (US Government document obtainable from the Environmental Protection Agency, Washington, DC; refer to document control no. 4400-03-ACJH, November 1992, chap. 2).
17.
Zhang J, Bowes WA Jr. Birth weight for gestational age patterns by race, sex, and parity in the United States population.
Obstet Gynecol
 
1995
;
86
:
200
–8.
18.
Williams RL, Creasy RK, Cunningham GC, et al. Fetal growth and perinatal viability in California.
Obstet Gynecol
 
1982
;
59
:
624
–31.
19.
Kotelchuck M. An evaluation of the Kessner Adequacy of Prenatal Care Index and a proposed Adequacy of Prenatal Care Utilization Index.
Am J Public Health
 
1994
;
84
:
1414
–20.
20.
Pearce N. Analytical implications of epidemiological concepts of interaction.
Int J Epidemiol
 
1989
;
18
:
976
–80.
21.
Greenland S. Basic problems in interaction assessment.
Environ Health Perspect
 
1993
;
101(suppl 4)
:
59
–66.
22.
Berry M, Bove F. Birth weight reduction associated with residence near a hazardous waste landfill.
Environ Health Perspect
 
1997
;
105
:
856
–61.
23.
Hertz-Picciotto I, Pastore L, Beaumont J. Timing and patterns of exposures during pregnancy and their implications for study methods.
Am J Epidemiol
 
1996
;
143
:
597
–607.
24.
Hertz-Picciotto I. A new insight into pregnancy loss. (Editorial).
Epidemiology
 
1997
;
8
:
611
–12.
25.
Watier L, Richardson S, Hemon D. Accounting for pregnancy dependence in epidemiological studies of reproductive outcomes.
Epidemiology
 
1997
;
8
:
629
–36.
26.
Kline J, Stein Z, Susser M. Conception to birth. Epidemiology of prenatal development. New York, NY: Oxford University Press, 1989.
27.
Bianco A, Stoen J, Lynch L, et al. Pregnancy outcome at age 40 and older.
Obstet Gynecol
 
1996
;
87
:
917
–22.
28.
Prysak M, Lorenz RP, Kisly A. Pregnancy outcome in nulliparous women 35 years and older.
Obstet Gynecol
 
1995
;
85
:
65
–70.
29.
Savitz DA, Zhang J. Pregnancy-induced hypertension in North Carolina, 1988 and 1989.
Am J Public Health
 
1992
;
82
:
675
–9.
30.
Khoury MJ, Erickson JD, Cordero JF, et al. Congenital malformations and intrauterine growth retardation: a population study.
Pediatrics
 
1988
;
82
:
83
–90.
31.
Backe B. Maternal smoking and age. Effect on birth weight and risk for small for gestational age births.
Acta Obstet Gynecol Scand
 
1993
;
72
:
172
–6.
32.
Fox SH, Koepsell TD, Daling JR. Birth weight and smoking during pregnancy—effect modification by maternal age.
Am J Epidemiol
 
1994
;
139
:
1008
–15.
33.
Wen SW, Goldenberg RL, Cutter GR, et al. Smoking, maternal age, fetal growth and gestational age at delivery.
Am J Obstet Gynecol
 
1990
;
162
:
53
–8.
34.
Alberman E, Roman E, Pharoah POD, et al. Birth weight before and after a spontaneous abortion.
Br J Obstet Gynaecol
 
1980
;
87
:
275
–80.
35.
Meis PJ, Michielutte R, Peters TJ, et al. Factors associated with term low birthweight in Cardiff, Wales.
Paediatr Perinatal Epidemiol
 
1997
;
11
:
287
–97.
36.
Zielhuis GA, Gijsen R, van der Gulden JWJ. Menstrual disorders among dry-cleaning workers. (Letter).
Scand J Work Environ Health
 
1989
;
15
:
238
.
37.
van der Gulden JW, Zielhuis GA. Reproductive hazards related to perchloroethylene. A review.
Int Arch Occup Environ Health
 
1989
;
61
:
235
–42.
38.
Swaab DF, Boer GJ, Boer K, et al. Fetal neuroendocrine mechanisms in development and parturition.
Prog Brain Res
 
1978
;
48
:
277
–90.
39.
Harlow SD, Ephross SA. Epidemiology of menstruation and its relevance to women's health.
Epidemiol Rev
 
1995
;
17
:
265
–86.
40.
Harlow SD, Lin X, Ho MJ. Analysis of menstrual diary data across the reproductive life span. Applicability of the bipartite model approach and the importance of within-woman variance.
J Clin Epidemiol
 
2000
;
53
:
722
–33.
41.
Mitchell ES, Woods NF, Mariella A. Three states of the menopausal transition from the Seattle Midlife Women's Health Study: toward a more precise definition.
Menopause
 
2000
;
7
:
334
–49.
42.
Stein ZA. A woman's age: childbearing and child rearing.
Am J Epidemiol
 
1985
;
121
:
327
–42.