Background

Background

Environmental exposure to hazardous substances and the adverse health effects that can result are increasing in public health importance. In 1981, the U.S. Environmental Protection Agency (EPA) estimated that 264 million metric tons of hazardous wastes were produced. By 1988, this estimate had risen to 5.5 billion1. In 1990, an estimated 4 million people in the United States lived within 1 mile of one or more of the 1,135 hazardous waste sites then on the National Priorities List (NPL) (1). By 1994, there were more than 1,400 sites on the NPL 2 . These sites represented a small fraction of the estimated 439,000 hazardous waste sites that might be present in the United States (1). The number of people actually exposed to toxic substances either at NPL sites or at hazardous waste sites in general cannot be estimated accurately.

ATSDR is required by law to conduct a public health assessment at each NPL site. The aim of each assessment is to determine whether the population residing around a particular site might have been exposed to any toxic substances and to assess whether adverse health effects possibly resulted from this exposure. Known health effects are documented in these assessments, and public health recommendations are made accordingly. Potential health effects are also identified and referred to ATSDR scientists for additional investigation. As part of this health assessment process, Camp Lejeune personnel provided ATSDR with drinking water monitoring records indicating that two drinking water supplies at Camp Lejeune were contaminated over a period of 34 months. Included in the population supplied with this water were slightly more than half of all residents in family base housing. Because this population consisted of a large proportion of young married women, concern was raised about potential health effects on fetuses exposed to toxic substances in utero.

Site Description and Exposure History

Camp Lejeune is a military base that comprises approximately 233 square miles in Onslow County on the coast of North Carolina. It is one of 123 federal facilities on the NPL, and it is included because of the presence of contaminants in the environment originating at the facility. The military base consists of six Marine Corps commands and two Navy commands. Almost 130,000 people have access to the base. The population includes active military personnel (43,000) and their dependents (52,000). Base housing for enlisted personnel, officers, and their families are located in 15 different areas on the base. An average of 8.3 million gallons of water is distributed daily at Camp Lejeune. More than 100 wells have been drilled to supply this water. Almost all of these wells use a sand aquifer that is permeable to contamination (3).

Personnel at Camp Lejeune first detected VOC contamination in drinking water in April 1982. This coincided with a change in the laboratory that conducted routine water-quality testing and was unlikely to have been related to the onset of first exposure. Because test results of water samples obtained in April were anomalous, samples were collected in May and July and analyzed for a limited number of VOCs. PCE and TCE were found in two drinking water systems, the Tarawa Terrace system and the Hadnot Point system. However, the source of the contamination was not identified. Although officials at the base contacted the state for advice, no further action was taken because water quality standards had not been established for these VOCs in 1982.

In July 1984, Camp Lejeune began sampling wells in the Hadnot Point area as part of the base's environmental restoration program. As a result of this sampling, seven contaminated wells were closed in November and December 1984. Tap water sampling conducted in December after the closure of these seven wells showed no additional evidence of contamination. However, on January 27, 1985, a fuel pump broke at the Holcomb Boulevard water system. Water from Hadnot Point was supplied to the Holcomb Boulevard service area while repairs were conducted. Tap water samples taken from buildings temporarily supplied by Hadnot Point contained high levels of TCE, which prompted additional tap and finished water sampling for VOCs at Hadnot Point and Tarawa Terrace. Contaminated wells in both water systems were closed soon after they were identified in January and February 1985, and routine sampling was implemented at all distribution systems on the base. Notable contamination has not been detected in Camp Lejeune's drinking water systems since February 1985.

The Hadnot Point system has been used primarily for industrial purposes, but the Hospital Point housing area also receives water from the Hadnot Point system. This small housing area was populated by hospital personnel and their families until 1983, when the area became housing for a more diverse group of officers' families. It is not known when the Hadnot Point supply wells first became contaminated, but VOCs were present for at least 2½ years. Industrial activity on the base began in the 1940s. No records indicate when the VOC plumes that contaminated supply wells in the Hadnot Point system originated. A chronology of these events is included in Table 1.

At Tarawa Terrace, the highest concentrations of contaminants measured in tap water samples were 215 parts per billion (ppb) PCE, 8 ppb TCE, and 12 ppb 1,2-DCE. This distribution system continued to serve base family housing until 1986. The highest contaminant levels measured in tap water samples from Hadnot Point were 1,400 ppb TCE and 407 ppb 1,2-DCE.

Contamination at Tarawa Terrace probably occurred many years before it was first documented in 1982. The source of the PCE at Tarawa Terrace was the ABC One-Hour Cleaners, a dry-cleaning establishment near Tarawa Terrace3. PCE leaked into the groundwater from the company's septic system. According to EPA records, the septic system was in operation from 1954 through 1985. In 1958, military personnel dug a supply well for the Tarawa Terrace system approximately 900 feet from the dry cleaners. Because this supply well was near the contaminated septic system, because few changes were made in the dry-cleaning operation after 19604, and because of the very permeable aquifer at Camp Lejeune, the Tarawa Terrace well probably was contaminated soon after it was built. Human exposure to PCE and other contaminants through this well could have occurred for as long as 30 years(3).

The housing areas that received contaminated water in each exposure group, the contaminants, and the estimated contaminant levels are summarized in Table 2. Each of the affected housing areas received water containing a mixture of many contaminants, a phenomenon noted with almost every population exposed to contaminants released from hazardous waste sites. For simplicity, each group of exposed housing areas is referred to by the predominant contaminant in the mixture. Residents of Tarawa Terrace are referred to as the PCE-exposed group, and residents of Hospital Point are referred to as the long-term TCE-exposed group. The short-term TCE-exposed group comprises residents of Berkeley Manor, Midway Park, Paradise Point, and Watkins Village during the 12-day period from January 27 through February 7, 1985, when these residents received water from the same supply as Hospital Point residents.

The exposure data, summarized in Tables 3 and 4, are limited. Water samples were collected on three different dates; the May 1982 samples, however, were preserved for several months before they were analyzed, which might have decreased the observed concentration of VOCs. Moreover, the 1985 sampling at Hadnot Point was conducted after seven of eight contaminated wells were closed. Hence, the expected contamination levels in the Hadnot Point distribution system before 1985 would have been higher than the concentrations measured in 1985. In addition, one supply well for the Hadnot Point distribution system contained concentrations of benzene as high as 700 ppb. Because the 1982 analyses were limited to TCE and PCE, and because the well containing benzene was shut off before the distribution system was sampled again, benzene was never detected in Hadnot Point tap water. Nonetheless, low-level exposure (an estimated 35 ppb) would have been expected among women receiving Hadnot Point water before December 1984.

An important feature of the exposure at Camp Lejeune was its intermittent nature. Each of the contaminated systems had more wells than were necessary to supply water on any one day. Contaminant levels have been noted to differ with the supply wells in service. The process by which a particular well was selected for use was essentially random, but all wells presumably were used in a given month unless they were out of service for mechanical failure or contamination. Daily or monthly well-usage logs were not available for evaluation. Despite these variations, on any specified day, VOC concentrations were probably distributed uniformly to all residential units because the water from all wells was mixed before treatment and distribution. For example, on January 31, 1985, VOC concentrations were similar in tap water samples obtained from several different buildings (Table 4).

Human Health Effects of Concern

Human gestation is a time of great vulnerability to environmental and pharmacologic agents. Environmental exposure to mercury has been shown to cause adverse effects in utero even though the pregnant woman is unaffected5. The outcomes evaluated (i.e., decreased MBW, SGA, and preterm delivery) are several of the many possible adverse pregnancy outcomes that might be associated with exposure to environmental toxins6. These outcomes are important because of their contribution to infant mortality and morbidity; moreover, they are among the most practical outcomes to study near hazardous waste sites because they are common, well-ascertained, and reported in a standardized fashion on birth records7. Birth records also include information on maternal residence. These practical aspects of the study outcomes are important in situations, such as that at Camp Lejeune, in which exposure ceased almost 10 years before the study and most of the exposed population had moved in the intervening period.

Intrauterine growth retardation (measured as decreased MBW and SGA) and preterm delivery are two conditions with distinct pathogeneses that are usually grouped together and measured as low birth weight. In 1989, 7.0% of infants had low birth weight, weighing <2,500 grams (g) at birth8. Low birth weight is the third most important predictor of infant mortality in the United States and the most important predictor of infant mortality among blacks in the United States9. In 1980, the risk for infant mortality for singleton infants with very low birth weight (i.e., <1,500 g at birth) was 94 times higher than for infants of normal birth weight (>2,500 g at birth)10. Low birth weight and very low birth weight infants are also at greater risk for neurodevelopmental handicaps (e.g., cerebral palsy and seizure disorder), lower respiratory tract conditions, and complications from neonatal care11.

Distinguishing between effects on fetal growth and effects on gestational age at delivery is often difficult because growth and maturity of an infant are both highly dependent on gestational age. Infants who are born small because they are born at <37 weeks of gestation are considered to be preterm. Approximately 10% of all infants born during 1988 in the United States were preterm12. Approximately 40% of these preterm infants weighed <2,500 g (12). Such infants are clearly at higher risk for morbidity and mortality. The risk for fetal death is three times higher for infants surviving to 26 weeks than for infants surviving to 40 weeks13. Factors predictive of preterm delivery include maternal socioeconomic status, race-ethnicity, cigarette smoking, stress, nutrition, past pregnancy history, access to prenatal care, and medical complications such as sexually transmitted diseases, infection, hypertension, and preeclampsia14.

Infants who have sufficient time to grow and mature but have low birth weight often are less viable because of intrauterine processes that delayed their growth. In general, whether preterm or full-term, growth-retarded infants are at greater risk for antenatal and neonatal mortality than full-term infants who are at the appropriate weight for their gestational age (14,15). SGA infants are those within the bottom tenth percentile of the birth weight distribution at any given gestational age. As with other population-based measures, some SGA infants will be healthy and simply smaller than average, but many will be growth retarded. At present, SGA is the only marker for intrauterine growth retardation that is readily available for population-based studies.

Biologic factors reducing growth include young maternal age, low maternal prepregnant weight, short maternal height, insufficient maternal weight gain during pregnancy, maternal alcohol consumption, and anoxia resulting from cigarette smoking and altitude (14,16). Maternal medical complications, such as hypertension, can also produce anoxic conditions resulting in SGA infants17. Plurality, the sex of the infant, and maternal parity also influence birth weight. Important social determinants of SGA infants in the United States are maternal race, education, socioeconomic status, and utilization of prenatal care (14).

Late fetal deaths (i.e., stillbirths) occur more rarely than preterm birth and SGA but account for a greater proportion of perinatal mortality. Late fetal deaths, defined as fetal deaths occurring after 20 weeks of gestation, account for approximately 80% of all perinatal deaths. In 1989, the fetal death rate after 20 weeks of gestation was 7.5 deaths per 1,000 births, with a rate of 6.4 per 1,000 births in whites and 13.1 per 1,000 births in nonwhites18. These rates probably underestimate the actual numbers of fetal deaths because of underreporting (7,19). Despite the importance of late fetal death, the causes of fetal death have not been widely studied. Important maternal risk factors for fetal death are maternal age, race, education, parity, body mass, cigarette smoking, hypertension and hypertensive disorders, diabetes, and previous adverse pregnancy outcome. Risk factors inherent to the specific infant or pregnancy include sex, congenital anomalies, plurality, and cord and placental complications (15,2021).

Routes of TCE and PCE Exposure and Metabolism

PCE and TCE, the predominant contaminants in the chemical mixtures studied, are structurally similar chemicals with many common toxicologic properties. Both compounds are lipophilic2223 and readily cross the placenta242526. Compared with other solvents, both TCE and PCE have relatively long half-lives in the human body27. PCE is retained in the body three to four times longer than TCE, and females retain both compounds longer than males (27). A number of models have been developed to estimate the distribution of PCE and TCE within the human body after exposure to contaminated air or groundwater282930. In general, ingesting contaminated drinking water is not an efficient way to deliver these toxic chemicals to the fetus (26). Activities that cause VOCs in household water supplies to evaporate include bathing; showering; cooking; and operating toilets, washing machines, and dishwashers (2931). Inhalation of TCE and PCE that have evaporated from household water is likely to result in higher exposures than ingesting water from the same water supply (29). Larger fractions of PCE and TCE are metabolized after ingestion than after inhalation (26). Moreover, trichloroacetic acid (TCA), a biologically active metabolite of PCE and TCE, has been observed to persist in the rat fetus after exposure to either PCE or TCE has stopped; TCA can cycle from the fetus into the amniotic fluid and back into the fetus (25). Therefore, the relative contributions of inhalation and ingestion of PCE and TCE depend on whether the primary toxicants are the chemical(s) or the chemical metabolites.

One potential mechanism for reproductive toxicity of PCE is a generalized central nervous system depression that suppresses the hypothalmus and pituitary in the mother, the fetus, or both32. A similar mechanism might operate for TCE because it has similar chemical properties. Although this hypothesis remains untested, central nervous system depression after PCE and TCE exposure is well-established (22,23), fatty acid composition changes in the brains of fetal guinea pigs have been observed after in utero exposure to PCE33, and suppression of the fetal hypothalmus would affect fetal growth3435. Suppression of the maternal hypothalmus probably does not affect fetal growth (34), but the interactions between the maternal hormonal environment, the placental hormonal environment, and the fetal hormonal environment are complex.

Metabolites of TCE are possibly responsible for the developmental defects observed in laboratory animals exposed to TCE in drinking water363738. Infants with birth defects are often SGA. An association between SGA or reduced MBW and exposure to PCE or TCE might also be a marker for birth defects. However, SGA is only a weak surrogate for birth defects39. Therefore, an association between exposure and birth defects would have to be very strong to be detected in this study of SGA. It is not known whether metabolites of TCE or PCE might affect fetal growth through a mechanism independent of birth defects.

Toxicologic Literature

The association between low level environmental exposure to PCE or TCE and adverse pregnancy outcomes has not been determined. In one controlled clinical trial40, pregnant mice exposed to 300 parts per million (ppm) PCE delivered litters that had an average birth weight that was 9% lower than the normal average; these litters also were twice as likely to have subcutaneous edema than unexposed mice, and the increase in the number of litters with delayed ossification of skull bones was statistically significant. There was a 60% increase in the number of mice that were runts (defined as weighing less than three standard deviations below average) among the exposed litters, but this difference was not statistically significant. Fetal rats exposed to the same regimen did not have lower birth weight or excessive delays in ossification. However, there were a greater proportion of fetal resorptions among exposed rats than among unexposed rats. This effect was not observed among mice. Maternal toxicity resulting from PCE exposure was manifested by decreased maternal weight gain and increased maternal liver weight in rats and mice, respectively. However, it seems unlikely that the developmental effects of PCE were the result of maternal toxicity because pregnant rodents exposed to other solvents in this investigation experienced similar toxicity but their litters were unaffected.

TCE exposure has not been associated with measured adverse pregnancy outcomes in the late stages of gestation except with severe maternal toxicity (40). However, both developmental and behavioral effects in laboratory animals after exposure to TCE have been noted (384142). The timing of the development of human and rat brains is different. Neonatal development of the rat brain corresponds to development of the human brain during the third trimester of pregnancy43; therefore, behavioral effects observed in neonatal rats might be of significance to the developing human fetus.

Although useful in generating hypotheses regarding the developmental hazards of specific contaminants, toxicologic studies are complicated by the need to extrapolate from animal species and high doses. In addition, laboratory studies do not adequately capture the complex personal and environmental contexts in which human exposures to VOCs occur44.

Epidemiologic Literature

Several studies have examined the issue of late pregnancy outcomes and occupations in which women might have been exposed to VOCs45464748495051525354555657. However, fewer of these studies have examined exposure to specific chemicals or chemical classes. Two studies of maternal occupational exposure to solvents (47) and degreasing agents (52) noted small decreases in birth weight (-41 g ±124 g and -16 g ±75 g, respectively), but these decreases were not statistically significant. A small case-control study (26 cases) of birth outcomes among female workers in Sweden, Finland, and Norway found no association between very low birth weight, congenital malformations, or stillbirths and working in the dry-cleaning or laundry industry58. However, in addition to the limited number of cases, all three outcomes were combined into a single case definition; this categorization did not account for the different times at which developing organisms are vulnerable to stillbirth or very low birth weight and when they are vulnerable to congenital malformations.

Only two studies have evaluated the possible association between halogenated hydrocarbons and late pregnancy outcomes. Savitz et al. (54) noted no association between exposure to halogenated hydrocarbons and SGA (OR: 0.6 [95% CL: 0.2, 1.4]), preterm delivery (OR: 1.1 [95% CL: 0.5, 2.4]), and stillbirth (OR: 1.0 [95% CL: 0.7, 1.5]). Windham et al. (55) noted no association between SGA and maternal exposure to halogenated solvents during the first 20 weeks of pregnancy (OR: 1.1 [95% CL: 0.4, 2.9]); however, fetal growth is considered most vulnerable to environmental insults during the third trimester of pregnancy. Therefore, this latter study might have focused on exposures that occurred at a time when the fetus was relatively invulnerable to effects on birth weight.

Limitations common to many of these occupational studies included: indirect estimates of exposure derived from job titles rather than measured exposure in the work place, the small numbers of women in specific exposure categories, and differential participation and recall by underlying maternal risk. In addition, because exposure to many different substances occurs in the same work place59, the relevant hazards could be difficult to identify. Many of these factors are more likely to introduce bias toward the null hypotheses than they are to introduce associations where none exist, although an upward bias could be introduced by differential participation or recall.

Environmental exposures to toxic substances occur at lower concentrations relative to the occupational setting. However, environmental exposures often occur through contaminated drinking water, while occupational exposures usually occur through inhalation or skin contact. As discussed previously (see "Routes of TCE and PCE Exposure and Metabolism"), PCE and TCE metabolism differ depending on whether these compounds are inhaled or ingested. Environmental exposures are not limited to the 40-hour work week and can occur in populations that are not represented in the work force. Women who are less likely to work include those who cannot find work, those who already have children, and those without economic incentive to work60. In addition, women who are at high risk for adverse pregnancy outcomes might be instructed by their physicians to cease work during pregnancy (60). These factors all limit the generalizations that can be made from studies of occupational populations to residential populations exposed to environmental contaminants.

The earliest report of a relationship between environmental exposure to toxic substances at hazardous waste sites and late pregnancy outcomes was based on an investigation at Love Canal in Niagara Falls, New York, a former dump site where 248 different chemicals were identified. The prevalence of low birth weight was elevated in two different studies of area residents6162. Home ownership among whites in the area of Love Canal where contaminants had seeped into the basements of several homes was associated with a 60% increase in low birth weight compared with all white residents of upstate New York (OR: 1.6 [95% CL: 1.0, 2.3]). Both the rate of low birth weight and the rate of prematurity were higher among Love Canal homeowners compared with rates among homeowners in neighboring areas of Niagara Falls (low birth weight OR: 3.1 [95% CL: 1.3, 7.1], prematurity OR: 1.4 [95% CL: 0.8, 3.5]). However, no increases in low birth weight (OR: 1.1 [95% CL: 0.5, 2.3]) or preterm delivery (OR: 1.1 [95% CL: 0.6, 2.2]) were observed among renters at Love Canal when compared with rates among renters in neighboring areas.

More recently, an increased prevalence of term low birth weight (another index of intrauterine growth retardation) was found among residents near Lipari landfill in Gloucester County, New Jersey. During the years when odors were greatest at the site, the OR was 5.1 (90% CL: 2.5, 10.6)63. Moreover, a strong correlation was observed between 3-year weighted averages of excess term low birth weight around the landfill and the timing of dumping and odors throughout the 25-year study period. A cohort study conducted near the Stringfellow hazardous waste site in Riverside County, California64, and an ecologic study conducted of hazardous waste sites in five counties in the San Francisco Bay area of California65 reported no associations between proximity to site and low birth weight (OR: 0.9 [95% CL: 0.3, 2.7]) or MBW (0.6 g ±12.3).

Each of the studies of birth weight around hazardous waste sites, summarized in Table 5, had methodologic problems. One problem faced at Love Canal was that the families living closest to Love Canal were relocated before the study was conducted (61), and the remaining families were evacuated selectively, beginning with those families with pregnant women and young children (6166). Hence, selective migration could have introduced bias in the association between exposure and outcome. Selective migration also is likely to be a problem at other hazardous waste sites, especially when strong odors reduce the quality of life in a neighborhood. In such a situation, residents with the highest incomes (who would be at the lowest risk (10)) and residents who were most sensitive to the exposures would be most likely to leave the vicinity. Although the effects that selective migration had on the results cannot be predicted, it seems reasonable that the results of the Love Canal evaluation might have been biased toward the null hypothesis. The women who were most likely to have been exposed had already been evacuated and were not included in the study.

The conflation of preterm delivery and SGA births in all but two of the studies probably reduced the observed effect measures. Failure to account for such etiologic heterogeneity has been discussed in detail elsewhere67. The study conducted at Lipari landfill demonstrates this problem: the association found between residence near the landfill and low birth weight in full-term infants was stronger than the association found between residence near the landfill and low birth weight in all infants (63).

Small numbers were also a problem, especially at the Stringfellow site (64), limiting the precision of the observed effect estimates. In most cases however, it would not have been appropriate to increase the sample size because this would have created a more heterogeneous exposure and, therefore, would have diluted the observed association between exposure and outcome. Control for most major risk factors was addressed in the studies summarized in Table 5, except for smoking, which was not measured in the studies in San Francisco, California (65), or Gloucester County, New Jersey (63). However, both the San Francisco and Gloucester County studies controlled for demographic variables that would have minimized bias from smoking.

The most important limitation to the studies summarized in Table 5 was misclassification of exposure. In all these studies, proximity to the hazardous waste site was the primary classification of exposure. Although there was some evidence of population exposure to VOCs based on reports of odors, and at both Love Canal and Lipari minimal air measurements were taken, it was difficult to determine if the persons included in the studies had been exposed and, if so, to which substances and at what concentrations. Because both the probability of exposure and the chemical mixture at each site differ, it was difficult to evaluate, on the basis of consistency across these studies, whether exposure to hazardous waste reduces birth weight. However, in general, it seems reasonable to infer that environmental exposure to some compounds or combinations of compounds found at hazardous waste sites might decrease birth weight at least at some sites.

Studies of populations consuming contaminated drinking water, although still imprecise, are a substantial improvement over studies based on proximity to site. Even when exposure data are limited and multiple contaminants are detected in the same water distribution system, studies that focus on drinking water provide a well-defined route of exposure, leaving less uncertainty as to whether there is an exposed population, how many persons might be exposed, which chemicals are present, and at what concentrations exposures are occurring or have occurred in the past. Moreover, exposure is defined in a manner that can be directly compared with exposure in other studies. Only three analytic studies have investigated the relationship between TCE, PCE, or DCE in drinking water and late pregnancy outcomes686970. As summarized in Table 6, contaminant levels measured in these studies were comparable to, or lower than, those observed at Camp Lejeune (68,69,70).

Two studies were conducted in Woburn, Massachusetts, where two wells that supplied the town were found to be contaminated with TCE, PCE, and chloroform. In the first of the Woburn studies (68), self-reported outcomes were examined for 4,396 pregnancies from 1960 through 1982. An important feature of the study was that the investigators used information about the municipal use of supply wells in different areas of Woburn to characterize exposure. Based on a logistic regression model that examined exposure as a continuous variable, women who received 100% of their water from contaminated wells had a tenfold increase in risk for perinatal mortality relative to women who received no water from contaminated wells. Although this risk estimate is impressive, it was based on small numbers (four exposed cases). Furthermore, only two women whose infants died could have received 100% of their water supply from contaminated wells. This effect was noted only among pregnancies occuring after 1970, with no increase in perinatal mortality noted among women exposed during the first 10 years of the study. Because exposure was not measured before 1979 (i.e., when tests first became available) it is possible that there was less or no contamination during the earlier study period. Despite the magnitude of the observed association, the small number of exposed cases and inconsistency across time periods raises the possibility that this finding was artifactual (i.e., arising through chance or confounding). No association was noted between exposure to contaminated well water and low birth weight, but birth weight was not adjusted for gestational age. Moreover, low birth weight was reported by each mother, and a nonstandard definition of low birth weight was used. Other limitations included the sample selection process, which was based on residence in Woburn at the time the study began, and the convenient selection process which could have resulted in selection bias.

The second study conducted at Woburn addressed a number of the methodologic limitations of the first study by examining birth weight as recorded on birth certificates of infants born to residents of Woburn during the time of exposure (69). The most relevant comparisons in this study of SGA were those between birth weights of live-born infants of East Woburn residents who were exposed to high or moderate levels of contaminants during 1975 through 1979 and live-born infants of East Woburn residents who were not exposed to contaminants. For the approximately 3,000 live births, the prevalence of SGA was not elevated among live-born infants of women who were highly exposed (OR: 1.1; 95% CL: 0.5, 2.4) or moderately exposed (OR: 0.7; 95% CL: 0.4, 1.4) when exposure was classified on the basis of the entire pregnancy. However, when the exposure classification was restricted to the third trimester, the ORs were 1.6 (95% CL: 0.9, 2.8) and 1.3 (95% CL: 0.8, 2.1) for highly and moderately exposed births, respectively. Despite the authors' conclusions that the study "was unable to detect an adverse effect of exposure to Wells G and H on the reproductive health of exposed subgroups of Woburn residents," (69) the specificity of these findings–that is, increasing ORs with more refined classification of exposure and outcome–provides some evidence for an association between TCE exposure and SGA.

The relationship between exposure to TCE, PCE, and DCE71 and late pregnancy outcomes in drinking water was also examined for the entire state of New Jersey (70). Information was obtained from birth certificates and fetal death certificates; exposure levels were based on semiannual, quarterly, or monthly monitoring of drinking water. Maternal residence on the birth certificate was used to assign exposure and was assumed to be the residence throughout pregnancy. Although no associations were found between TCE, PCE, or DCE exposure and SGA, preterm birth, or fetal death, the median exposures evaluated were 200-1,000 times lower than the exposures evaluated in this study.

Overall, knowledge about the potential relationship between PCE, TCE, and 1,2-DCE exposure and late pregnancy outcome is limited; although the results of several studies indicate that environmental exposure to these VOCs might affect late pregnancy outcomes, literature on this topic is limited and equivocal. Maternal occupational exposure to solvents and other VOCs has been associated with increases in stillbirths (51,53,54) and decreases in birth weight (53). However, two occupational studies that focused specifically on halogenated hydrocarbons reported no associations between these exposures and stillbirths, preterm deliveries, or birth weight (54,55). Low birth weight has also been associated with residence near two different hazardous waste sites containing large quantities of VOCs and other chemicals (61,62,63), but the exposures were too poorly defined and too complex to permit generalizations from these hazardous waste sites to others. Only the investigations (69,70) conducted in Woburn, Massachusetts, examined the relationship between perinatal mortality and SGA and a chemical exposure at concentrations similar to those at Camp Lejeune. These investigations found increased rates of perinatal mortality and moderate excesses in SGA, but both associations were based on small numbers.

In addition to this direct, albeit limited, evidence that one or more of the solvents studied are associated with adverse late pregnancy outcomes, two studies have noted associations between term low birth weight and SGA and exposure to carbon tetrachloride (70) and trihalomethanes, including chlorinated compounds of similar chemical structure (7072). Other reports suggest that occupational or environmental exposures to solvents in general, and to TCE or PCE in particular, can cause other adverse pregnancy outcomes, such as spontaneous abortion, cardiac anomalies, oral clefts, and neural tube defects (42,55,707374).

Finally, there may be an association between solvent exposure and maternal complications of pregnancy. In a small prospective study of women occupationally exposed to solvents, Eskenazi et al. found increased rates of preeclampsia (OR: 3.9; 95% CL: 2.4, 5.4) and hypertension (OR: 3.0; 95% CL: 0.9, 9.9) (47). Moreover, these complications were restricted to women who worked during their second trimester of pregnancy. In a small case-control study of 130 pregnant residents of an industrial area of Bulgaria, Tabacova et al.75 found substantially increased odds of exposure to styrene among pregnant women with anemia (OR: 2.4; 95% CL: 0.5, 13.8), proteinuria (OR: 7.4; 95% CL: 1.7, 37.2), hyperemesis (OR: 13.1; 95% CL: 1.4, 165.9), arterial hypertension (OR: 26.4; 95% CL: 2.2, 1266.8), and nephropathy (OR: 30.8; 95% CL: 2.6, 1448.0). However, as one might gather from the wide confidence intervals76, that study was extremely small. Although the literature relating VOC exposure to medical complications of pregnancy is only suggestive, it provides a biologically plausible mechanism by which exposure to VOCs might affect fetal growth.

In summary, information is sparse regarding the relationship between exposure to organic solvents, such as PCE and TCE in drinking water, and late pregnancy outcomes; and PCE and TCE frequently occur in the environment. Only three studies have examined directly the relationship between PCE or TCE in drinking water and late adverse pregnancy outcomes. Only two of those studies, both analyzing data from the same city, observed exposures of similar concentration to the exposures experienced at Camp Lejeune. This study at Camp Lejeune should add to the existing body of knowledge, providing more information on a topic of great public health concern.