8
Studies of the Camp Lejeune Population

This chapter summarizes research that directly addresses the potential impact of contaminated water supplies on the health of Camp Lejeune residents. Although there is indirect evidence on the chemicals of concern from laboratory research and epidemiologic studies of other populations (Chapters 4-7), such information must be extrapolated to the Camp Lejeune setting and population, and extrapolation carries the potential for incorrect inferences. To the extent that scientifically valid epidemiologic research has been conducted directly on Camp Lejeune residents, extrapolation is unnecessary. Thus far, the research on the Camp Lejeune population has been limited with respect to the scope of health outcomes considered and the quality of exposure assessment.

COMPLETED STUDIES

The Agency for Toxic Substances and Disease Registry (ATSDR) is the only agency to have performed epidemiologic studies of the Camp Lejeune population exposed to water supplies contaminated with volatile organic compounds (VOCs). In a public health assessment, ATSDR (1997a) judged that exposure to VOCs in drinking water did not pose health risks to adults but raised questions about risks to children who may have been exposed via their mothers while in utero. Thus, the first study was a case-control study of pregnancy outcomes. Two published analyses resulted from that effort: ATSDR (1998), which focused on trichloroethylene (TCE) and perchloroethylene (PCE) exposures at Tarawa Terrace; and Sonnenfeld et al. (2001), which considered only PCE exposure at Tarawa Terrace. Both analyses focused on pregnancy outcomes regarding live-born infants, including mean birth weight, small for gestational age (SGA), and preterm delivery. ATSDR initially planned to evaluate fetal deaths, also, but that plan was abandoned because of the small number (83) of fetal deaths identified with the computerized state database and because the cause of fetal death was missing from death certificates in most cases (ATSDR 1998). The study methods used in the two analyses will be presented here first, followed by the results of each.

Outcome Measures

Birth weight and pregnancy duration were derived from North Carolina birth records. Preterm birth was defined as a live birth occurring before completion of 37 weeks of gestation. SGA, defined as below the 10th percentile of weight for gestational age, was calculated by using published sex-specific growth curves for white newborns in California (Williams et al. 1982) because a standard birth-weight distribution for the military population was not available. According to Sonnenfeld et al. (2001), of the three standards considered for use, the California standard was the one that fit best when all races were included.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 184
8 Studies of the Camp Lejeune Population This chapter summarizes research that directly addresses the potential impact of contaminated water supplies on the health of Camp Lejeune residents. Although there is indirect evidence on the chemi- cals of concern from laboratory research and epidemiologic studies of other populations (Chapters 4-7), such information must be extrapolated to the Camp Lejeune setting and population, and extrapolation car- ries the potential for incorrect inferences. To the extent that scientifically valid epidemiologic research has been conducted directly on Camp Lejeune residents, extrapolation is unnecessary. Thus far, the re- search on the Camp Lejeune population has been limited with respect to the scope of health outcomes considered and the quality of exposure assessment. COMPLETED STUDIES The Agency for Toxic Substances and Disease Registry (ATSDR) is the only agency to have per- formed epidemiologic studies of the Camp Lejeune population exposed to water supplies contaminated with volatile organic compounds (VOCs). In a public health assessment, ATSDR (1997a) judged that ex- posure to VOCs in drinking water did not pose health risks to adults but raised questions about risks to children who may have been exposed via their mothers while in utero. Thus, the first study was a case- control study of pregnancy outcomes. Two published analyses resulted from that effort: ATSDR (1998), which focused on trichloroethylene (TCE) and perchloroethylene (PCE) exposures at Tarawa Terrace; and Sonnenfeld et al. (2001), which considered only PCE exposure at Tarawa Terrace. Both analyses fo- cused on pregnancy outcomes regarding live-born infants, including mean birth weight, small for gesta- tional age (SGA), and preterm delivery. ATSDR initially planned to evaluate fetal deaths, also, but that plan was abandoned because of the small number (83) of fetal deaths identified with the computerized state database and because the cause of fetal death was missing from death certificates in most cases (ATSDR 1998). The study methods used in the two analyses will be presented here first, followed by the results of each. Outcome Measures Birth weight and pregnancy duration were derived from North Carolina birth records. Preterm birth was defined as a live birth occurring before completion of 37 weeks of gestation. SGA, defined as below the 10th percentile of weight for gestational age, was calculated by using published sex-specific growth curves for white newborns in California (Williams et al. 1982) because a standard birth-weight distribution for the military population was not available. According to Sonnenfeld et al. (2001), of the three standards considered for use, the California standard was the one that fit best when all races were included. 184

OCR for page 184
Studies of the Camp Lejeune Population 185 The study considered a base population of 12,493 singleton live births delivered after at least 20 weeks gestation to women residing in base housing during the period 1968-1985 who were identified through birth records (ATSDR 1998). That population did not include births to mothers who resided on the base during pregnancy but were no longer residents of Onslow County at the time of delivery. Resi- dential mobility may be substantial: according to ATSDR, “approximately one-third of the women who sought prenatal care at the Navy Regional Medical Center at Camp Lejeune moved or were transferred before they delivered” (ATSDR 1998, p. 16). Although exposures were presumed to have occurred before 1968, a starting date of January 1, 1968, was chosen because electronic files of North Carolina birth cer- tificates began that year. The analyses assumed delivery of contaminated water via the water-distribution system through February 1985 (ATSDR 1998; Sonnenfeld et al. 2001). ATSDR documented that 523 (4%) of the 12,493 live births were excluded because exposure to contaminated water supplies was for less than 1 week or exclusively before conception (44), or because data were missing, inconsistent, or insufficient (479), leaving 11,970 live births for the mean-birth-weight analyses. Of the11,970 live births, 6,117 (51%) were to women who resided at Tarawa Terrace at the time of birth, 31 (0.26%) were to women who resided at Hospital Point (which received water from Hadnot Point), 141 (1.2%) were to women who resided in housing units temporarily supplied by Hadnot Point during a fuel-pump failure, and 5,681 (47%) were to women who resided in housing supplied by the Hol- comb Boulevard system, were considered to be unexposed, and served as a comparison group. Additional exclusions were made for the SGA analyses (eight births with gestational age under 22 weeks) and the preterm-birth analyses (the eight births excluded from the SGA analyses plus 101 births classified as im- plausibly heavy preterm births). Exposure and Confounder Data Exposure was defined by linking birth records to the base’s family housing records according to the mother’s address at delivery and the father’s name. The housing records, which contained dates of residence, were used to estimate the dates when the mother resided in base housing units. The study “as- sumed that each family resided in only one base housing unit during pregnancy” (ATSDR 1998, p. 21). A residential-history substudy indicated that about 55% of mothers in the study moved during their preg- nancies, and 3.5% of them moved between base housing units (ATSDR 1998). The 1998 ATSDR study included all identified births regardless of exposure, whereas the 2001 Sonnenfeld et al. study limited the exposed population to residents of Tarawa Terrace. The Tarawa Ter- race residents were considered exposed to PCE from water contaminated by an off-base dry-cleaning es- tablishment (ABC One-Hour Cleaners). ATSDR’s analysis also included births to two groups of residents who were exposed to TCE and other VOCs through the Hadnot Point water system on either a long-term or a transitory basis. Transitory exposure (called short-term in the ATSDR report) covered all births to residents who received drinking water from the Holcomb Boulevard water system and who were pregnant for at least 1 week of the 12-day period during January-February 1985 when Hadnot Point water served the Holcomb Boulevard system. In both studies, residents of the base trailer park were excluded because housing records were incomplete, and, as noted above, a few births to mothers residing on base for a very short time or during ambiguous exposure periods were excluded. The remaining births to mothers resid- ing on the base were considered unexposed, including births to all residents of the Marine Corps Air Sta- tion, Rifle Range, and Courthouse Bay and the remaining residents of Berkeley Manor, Midway Park, Paradise Point, and Watkins Village. Exposure was categorized further by length of residence as a proxy for duration of exposure. Du- ration of exposure was defined as length of time before the birth that the mother lived at the residence specified on the birth certificate. Because inclusion in the study was based on maternal residence at the time of birth, exposure duration was relative to the end of pregnancy. Duration-of-exposure analyses ex- cluded births that occurred after exposure ended in 1985. In analyses, duration of exposure was catego- rized as never, 1-3 weeks, 4-10 weeks, 11-20 weeks, over 20 weeks and less than the entire pregnancy,

OCR for page 184
Contaminated Water Supplies at Camp Lejeune—Assessing Potential Health Effects 186 the entire pregnancy and less than 1 year before the last menstrual period, and the entire pregnancy and at least 1 year before the last menstrual period. The covariates available for analysis were limited to information that could be obtained from the birth certificates and military records. They included infant’s sex, year of birth, and gestational age; ma- ternal age, race, parity, education level, military pay grade, adequacy of prenatal care, marital status, and history of fetal death; and paternal age, education level, and military pay grade. Gestational age was cal- culated from the date of the last menstrual period reported on the birth certificate. Women with records showing a month and year of last menstrual period but missing information on the day had their day in- terpolated to 15. Women with records missing the month of the last menstrual period were excluded. In the remaining data, there was evidence of gestational-age misclassification in that 17% of preterm infants of gestational age less than 28 weeks had birth weight above the 90th percentile of the distribution for the standard population (ATSDR 1998). Preterm infants above the 90th percentile for birth weight at 36 weeks of gestation were excluded from the preterm-delivery analysis but not the birth-weight or SGA analysis. Results of the Sonnenfeld et al. Study Exposure was not equally distributed across various demographic groups. Exposed women were less likely to be white, less likely to live in officers’ housing, less likely to be college-educated, and less likely to have a college-educated partner (Sonnenfeld et al. 2001). Those differences raise questions about whether any observed differences in reproductive outcomes by exposure status were confounded by so- ciodemographic factors because not all the variables were examined as potential confounders or included in the adjusted analyses that were reported. The overall results of the study indicated that “long-term” PCE exposure from the Tarawa Ter- race water system was not strongly associated with reduced birth weight, preterm birth, or SGA. The mean birth weight in the PCE-exposed group was 26 g less than that in the PCE-unexposed group (90% confidence interval [CI], -43 to -9) (note use of 90% CI rather than 95% CI). The unadjusted odds ratio (OR) for PCE exposure and preterm birth was 1.0 (90% CI, 0.9-1.1) and for PCE exposure and SGA 1.2 (90% CI, 1.0-1.3). It was noted that adjustment for potential confounders had little effect on the results. The authors reported no consistent patterns in the associations between PCE exposure and mean birth weight, preterm birth, or SGA by duration of exposure. In subgroup analyses, Sonnenfeld et al. reported that long-term exposure to PCE from the Tarawa Terrace water system was marginally associated with lower mean birth weight and an increase in risk of SGA but only in newborns of mothers more than 35 years old and mothers who had already had more than two fetal losses. The birth-weight analysis was adjusted for mother’s age, history of fetal loss, race, and residence in officers’ housing and infant’s gestational age, year of birth, and sex. The SGA analysis was adjusted for mother’s age, history of fetal loss, parity, residence in officers’ housing, and education and infant’s year of birth. The authors noted that older PCE-exposed mothers were different from their unexposed counterparts in race, college education of husbands, and household income (defined by the father’s rank). However, not all those variables were included in the analyses. Specific subgroups showed statistically significant effects, but no formal hypothesis test for the presence of interaction between sub- groups defined by maternal age or history of fetal loss was mentioned. The authors concluded that there was no association between PCE exposure and mean birth weight or preterm birth and that there was a weak association between PCE exposure and SGA in all groups. In subgroup analyses, they observed stronger associations between PCE exposure and low birth weight and SGA of infants of mothers who had a history of fetal death and mothers more than 35 years old.

OCR for page 184
Studies of the Camp Lejeune Population 187 Results of the Agency for Toxic Substances and Disease Registry Study This section focuses on aspects of the ATSDR results that are distinctive from the Sonnenfeld et al. results. ATSDR reported analyses of PCE exposure at Tarawa Terrace that were unadjusted, and this may have contributed to the slight differences from Sonnenfeld et al. in birth-weight results (-24 g; 90% CI, -41 to -7), but the SGA and preterm delivery results were identical. In spite of the reported difference, the birth-weight results were said to show no association, because the magnitude of the difference was viewed as clinically negligible. The duration-of-exposure analyses were identical, but the effect- modification results were slightly different because of different exclusion of data and more limited control for confounding. In particular, the OR for PCE exposure and SGA in women more than 34 years old was 4.0 (90% CI, 1.6-10.2) after adjustment only for officers’ housing. No exposure-response patterns were observed for PCE exposure and mean birth weight or SGA in women who had had fetal deaths. The much smaller population of TCE-exposed births was analyzed with stratification by resi- dence. Births in the long-term TCE-exposed group were to mothers living in housing ordinarily served by the Hadnot Point water-distribution system. Overall, there was limited evidence of a reduction in mean birth weight (reduction by 108 g; 90% CI, -230 to 13) or of increased risk of SGA (OR, 1.5; 90% CI, 0.5- 3.8), interpreted by ATSDR as modest associations. The reported results were unadjusted despite differ- ences between the two groups in the distribution of infant sex; mother’s age, pay grade, history of fetal death, and parity; and father’s education. Few analyses of interaction were conducted because of the small sample. TCE effects were found to be modified by infant sex for both birth weight and SGA. The study reported an increased risk of SGA in TCE-exposed male infants (OR, 3.9; 90% CI, 1.1-11.9) on the basis of three exposed cases. According to a rate estimated from the female control group, one exposed SGA female infant was expected; none was observed. No risk of any of the outcomes was found in the tempo- rarily exposed population with a maximum exposure duration of 12 days. Review and Evaluation Retrospective case-control studies can be extremely difficult to conduct when historical informa- tion on exposure, outcome, and covariates—challenges applicable to the study of birth outcomes at Camp Lejeune—is scarce. This section discusses limitations in identifying the study population, assignment of exposure, confounder control, and analytic approach. Exposure misclassification is a major limitation of the ATSDR and Sonnenfeld et al. analyses. A number of exposed births were misclassified as unexposed because of incorrect assumptions about the water-delivery system, which ATSDR later identified. Both studies assumed that all mothers who resided in family housing in the Holcomb Boulevard system service area from 1968 through 1984 were unex- posed. In the course of exposure reconstruction of the Tarawa Terrace system, it was learned that the Hol- comb Boulevard plant came on line in June 1972 and that before then the housing now served by Hol- comb Boulevard was served by the Hadnot Point water-supply system. Thus, any mothers who resided in family housing in the Holcomb Boulevard system service area in 1968-1972 were actually exposed. That is an important (and correctible) source of misclassification that has the potential to alter study results dramatically because a sizable number of pregnancies will be reclassified from unexposed to exposed. Other limitations in exposure classification in these studies are more difficult to correct. Aspects of residential-history assignment would have caused exposure misclassification of unknown magnitude. First, all mothers were assumed to have had only one residence on the base and to have been unexposed at all other residences. The residence-history validation study estimated that a sizable proportion of mothers changed housing on the base during their pregnancies. Second, the contaminant exposure and its variation over time are impossible to quantify accurately. As reviewed in Chapter 2, water-supply measurements of contaminant concentrations are sparse, and the data were collected only in the 1980s. Third, there is no information about individual behaviors that affect exposure (such as water consumption and frequency

OCR for page 184
Contaminated Water Supplies at Camp Lejeune—Assessing Potential Health Effects 188 and duration of bathing and showering). Fourth, exposure was determined exclusively by place of resi- dence, excluding workplace and other locations in which exposure may have occurred. The studies relied on North Carolina birth-certificate data from Onslow County linked to base housing records. That was a feasible and efficient approach to conducting a study, but the information- retrieval process and restricted data sources have implications for population selection, outcome defini- tion and quality, and confounder control. In particular, the base population used in the studies does not represent the entire population of live births to all women who resided at Camp Lejeune in 1968-1985. Infants whose mothers were transferred or moved away from Camp Lejeune before giving birth were not included. In addition, because residence at birth determined inclusion, all exposure-duration analyses were relative to the end of pregnancy. For instance, nearly all infants who were exposed only during the first trimester were excluded. Beyond its obvious impact on interpretation of the exposure-duration analy- ses, the effect of a selection approach based on location at the time of delivery is unknown. Outcome variables were based on information included on birth certificates, and there are known limitations in the quality of some items (Wingate et al. 2007). In particular, accurate estimates of the date of the last menstrual period are critical for defining SGA and preterm birth. The ATSDR study found a disproportionate number of heavy liveborn infants relative to a standard population of the same gesta- tional age—a reminder of the fallibility of birth-certificate-based gestational-age estimates. Outcome- based exclusions varied among the three outcomes; preterm birth outcome was related to the largest num- ber of exclusions. Control for confounding is another challenge. Because of reliance on birth-certificate data on the period of the exposure episode, such key confounders as maternal smoking and alcohol use were not available. In addition, in reported analyses, control for confounding was not often done even for variables that were available. The ATSDR report gives unadjusted estimates of the primary results even though the exposed and unexposed populations differed in important respects and the study protocol (ATSDR 1994) stated that all analyses would be adjusted for race. The sensitivity of results to potential confounders should be examined more thoroughly. The implications of the results of subgroup analyses are unclear. The interactions of exposure with maternal age, history of fetal loss, and infant sex do not appear to be based on strong assumptions but instead resulted from exploratory statistical analysis. Although such interactions cannot be discounted, they should not be taken as evidence of an important effect of exposure. But these results are often cited as the primary study findings (for example, ATSDR 2005a). It is well known that overinterpretation of subgroup analyses can be misleading; such analyses typically suffer from low power and higher than nominal probability of reporting false positive effects (for example, Stallones 1987; Brookes et al. 2004; Weiss 2008). In addition, the various subgroup analyses used different numbers of observations and dif- ferent adjustment variables, depending on the report, outcome, and exposure variable. Subgroup member- ship should be described, and the sensitivity of results to data exclusions and more thorough confounder adjustment should be examined. CURRENT STUDIES Study Methods ATSDR’s 1997 public-health assessment for Camp Lejeune led to a recommendation that an epi- demiologic study be performed to evaluate whether mothers exposed to chlorinated solvents in drinking water, particularly TCE and PCE, during pregnancy have a higher risk of giving birth to a child with a birth defect or cancer, given the recognition of the limited scientific information on how those chemicals might affect a fetus or child (ATSDR 1997a). (ATSDR withdrew this report on April 28, 2009.) ATSDR later began a multistep process to determine the appropriateness of such a study. First, the childhood health problems to study were identified. On the basis of its review of the scientific literature, ATSDR decided to focus on specific childhood cancers and birth defects: childhood leukemia, childhood non- Hodgkin lymphoma, spina bifida, anencephaly, cleft lip, and cleft palate (ATSDR 2005a). The rationale

OCR for page 184
Studies of the Camp Lejeune Population 189 for focusing on those particular outcomes given the prior epidemiologic and toxicologic research and con- siderations of feasibility (specifically, statistical power) is discussed later in this chapter. The second step was to identify the children eligible for the study by conducting a telephone sur- vey. The survey, conducted from September 1999 to January 2002, built on the database initially con- structed for the two case-control studies of preterm birth and fetal growth (ATSDR 1998; Sonnenfeld et al. 2001). The survey sought information on all children who were born in 1968-1985 to mothers who resided on the base at any time during their pregnancies. Births in Onslow County were included, as were births that occurred after mothers were transferred off the base. ATSDR attempted to locate and contact the parents of each eligible child to elicit information on the child's health, to confirm that the mother was a Camp Lejuene resident during the pregnancy, and to collect data on potential confounders. It identified eligible children in multiple ways. Initially, it used the birth-certificate information from the previous Camp Lejeune study of SGA (ATSDR 1998) that included only women who were residents on the base at the time of their deliveries. Next, children born in 1968-1985 to mothers whose pregnancies occurred while they lived in base housing but who delivered after moving off the base were identified by word of mouth (for example, in parent groups), by referrals from other parents during their interviews, or by pub- lic requests (via the mass media, e-mails from the Marine Corps, and notices) that parents contact ATSDR. ATSDR surveyed the parents of 12,598 eligible children of an estimated 16,000-17,000 eligible births, representing an overall participation rate of 74-79%, depending on the estimated number of births that occurred off the base (ATSDR 2003). Parents were asked if their children had had birth defects or childhood cancer. A total of 106 cases that fit the case definition of parent-reported birth defect or child- hood cancer were reported in the survey: 35 neural-tube defects, 42 oral clefts, and 29 childhood cancers. The third step was to confirm the children’s health problems by reviewing their medical records. As of June 23, 2008 (Bove and Ruckart 2008), of the 35 reported or potential cases of neural-tube defects, 15 were confirmed (six anencephaly and nine spina bifida), 13 were ruled out, two had no medical records for confirmation, three were ineligible, and the parents of two potential cases refused to participate. For children who had parent-reported oral clefts without medical records, a dental examination was used to confirm that surgery was performed as a result of a cleft lip or palate. Of the 42 children who were re- ported to have oral clefts, 24 were confirmed (11 cleft palate and 13 cleft lip with or without cleft palate), 11 were ruled out, four had no medical records for confirmation and dental examinations could not con- firm the conditions, and the parents of three potential cases refused to participate. Of the 29 reported childhood leukemia or non-Hodgkin lymphoma cases, 13 were confirmed (11 leukemia and two non- Hodgkin lymphoma), eight were ruled out, one had no medical records for confirmation, four were ineli- gible, and the parents of three potential cases refused to participate. The parents of 15 children with neu- ral-tube defects, 23 children with oral clefts, and 13 children with leukemia or non-Hodgkin lymphoma were successfully interviewed. The fourth and final step of the process is to conduct a case-control analysis that incorporates wa- ter-system modeling; that work is under way. The primary hypotheses concern the association between drinking TCE- or PCE-contaminated water during the first trimester and specific birth defects and the as- sociation between drinking TCE- or PCE-contaminated water during pregnancy and childhood cancers. The hypotheses are extended to incorporate contaminant concentration and personal exposure (taking into account the amount of water consumed by the mother or used in showering, hand-washing dishes, and so on). The base population for the case-control study consists of all live births to mothers residing at Camp Lejuene in 1968-1985 who participated in the survey. Cases are confirmed birth defects (diagnosed by the age of 5 years) or childhood cancers (diagnosed by the age of 20 years). (Planned sensitivity analy- ses will also include unconfirmed cases.) Controls will be randomly selected from all other births in- cluded in the survey to attain a target of 10 controls for each case. Exposure assessment will be based on the ATSDR water-distribution system modeling (see Chap- ter 2). That includes a protocol for modeling the present water-distribution system and then developing historical distribution-system models for the study period and generating estimates of contaminant con- centrations in the water supply by year and housing complex. The stated exposure variables will be “ex-

OCR for page 184
Contaminated Water Supplies at Camp Lejeune—Assessing Potential Health Effects 190 posure status, concentration level, and/or percent of water from a contaminated source during the specific time periods of interest in the 1-year period before the child’s birth. TCE and PCE will be evaluated sepa- rately” (ATSDR 2005a, p. 29). Categorization of exposure is planned to be collapsed into ever vs never and into more refined exposure categories. Cut points will be determined from the contaminant- concentration distributions. Water use and consumption will be incorporated into the exposure metrics. According to the ATSDR protocol (ATSDR 2005a, page 25), with alpha set at 0.10, 80% power, and an exposure prevalence of 40%, minimum detectable ORs are as follows: 4.3 for 15 cases and 2.9 for 28 cases of neural-tube defects, 3.6 for 20 cases and 2.5 for 36 cases of oral cleft, and 5.2 for 14 and 4.3 for 19 cases of childhood cancer. Even with the uncertainty about the total number of cases that will even- tually be included in the analysis and even under the more optimistic scenario, statistical power is low. Review and Evaluation Owing to the paucity of measurements of PCE and TCE concentrations in contaminated water at Camp Lejeune during the period of interest, exposure assessment is a major limitation of the current birth- defect and childhood-cancer study. ATSDR has proposed to use water-system modeling as a way to im- prove the quantification of exposure. As indicated in Chapter 2, exposure estimates based on water- system modeling require a number of assumptions, and the validity of many of the assumptions is impos- sible to evaluate in light of the historical measurement data. Given the lack of information on which wells were used to supply water on any particular day, the quality of exposure estimates based on water-system modeling is highly uncertain, especially for the quantification of PCE and TCE concentrations over the short periods of interest for the study of birth defects. In addition, historical information about water be- havior will be available in two pregnancy-related periods (the mother’s questionnaire asks about only two periods: before and during the first trimester and during the second and third trimesters), and that infor- mation will be obtained only if the mother can be interviewed. Recall of such information over periods of decades is of questionable accuracy. Although the study-protocol data-analysis plan appropriately ad- dresses exposure-assessment limitations by proposing that exposure be categorized in analysis, the pro- posed analytic approach calls into question the need for complex water-system modeling. To the extent that simple categories of exposure will be used in the final analysis, the rationale for waiting for complex water modeling to be completed is unclear. Another major limitation of the study is the inadequate statistical power to detect associations in a plausible range. The selection of specific health end points is the primary reason that power is so limited, so the question arises as to whether they were the most informative outcomes to study. There is some ba- sis for speculating that those outcomes are associated with the solvents of interest largely on the basis of prior epidemiologic studies of water-contamination episodes, but the evidence is not compelling, and there is no reason to believe that these are the “best” choices, given their rarity. The committee’s review of the literature on the epidemiology of populations exposed to TCE and PCE (Chapter 5) and the toxi- cology of the compounds (Chapter 4) did not identify birth defects or childhood cancers as among the outcomes more plausibly related to exposure. For each of the three outcomes (neural-tube defects, oral clefts, and childhood cancers), there is adequate power only for markedly increased odds ratios (larger than 3). Given current knowledge about the etiologies of these conditions, it is highly unlikely that the exposures that occurred at Camp Lejeune would have increased risk to that degree, regardless of uncertainty about exposure magnitudes. Further- more, because the investigators also proposed to conduct multivariate analyses to control for the potential impact of other factors on the risk of the conditions, it is important to note that the power of a multivariate analysis will probably be even lower than the estimate for the unadjusted associations. The data-analysis plan in the protocol is very general and leaves room for the possibility of a pro- liferation of analyses that will make it more difficult to assess the meaning of any associations that are identified. A detailed written analysis plan specifying primary exposure metrics and key confounders should be prepared in advance of the analysis and should consider alternative approaches to controlling

OCR for page 184
Studies of the Camp Lejeune Population 191 confounders. The planned secondary and sensitivity analyses should be discussed more fully in the analy- sis plan. Because there is interest in multiple exposure periods (for various durations before, during, and after pregnancy), in different approaches to estimating exposure, and in different exposure categories, it is necessary to distinguish the primary exposure metric (such as peak exposure) from those to be evaluated in secondary and sensitivity analyses. FUTURE STUDIES An expert panel convened by ATSDR in 2005 judged that additional studies of the Camp Lejeune population would be challenging, perhaps requiring medical evaluation of hundreds of people from widely scattered locations. However, the panel concluded that it might be feasible to conduct a study of mortality outcomes and a study of cancer incidence. Before performing such studies, it recommended that their feasibility be assessed (ATSDR 2005b). ATSDR has prepared a report on the feasibility of conducting epidemiologic studies to address exposures that occurred at Camp Lejeune (Bove and Ruckart 2008). The report proposed a study of all- cause mortality and a study of cancer incidence by using Department of Defense (DOD) personnel data- bases to identify a cohort of active-duty marines and Navy personnel who were assigned to Camp Lejeune at any time from June 1975 through December 1985 and a cohort of civilians who worked at the base at any time from June 1974 through December 1985. The agency also proposed to include as a comparison population a sample of active-duty marines and civilians stationed at Camp Pendleton at any time during 1975-1985 who started duty on or after June 1975 and were never stationed at Camp Lejeune during the period of drinking-water contamination. The three cohorts would be considered for inclusion in an all- cause mortality study and a cancer-incidence study, and the Camp Pendleton cohort would serve as an external comparison group for the analysis of civilian and military personnel at Camp Lejeune. ATSDR proposed to link study participants’ residence history on the base with housing records (family housing unit or barracks) to identify participants’ drinking-water supply-system history. That would allow inclusion of monthly estimates of water contamination from the water-distribution system in individual-level exposure assessment. For civilian workers, the occupation code and information on the location of each occupation obtained from base staff (such as base industrial hygienists) would be used to link the workplace with the appropriate drinking-water system. Information on length of service on the base obtained from computerized personnel data would be used to estimate the duration of exposure. Ma- rines and civilians assigned to Camp Pendleton would be considered unexposed. ATSDR’s feasibility assessment included a literature review of the health effects of VOCs, par- ticularly TCE and PCE. The review concluded that previous studies supported evaluation of a variety of health effects, predominantly cancers, in future studies at Camp Lejeune. ATSDR’s review relied on pre- vious reports by the National Toxicology Program and the National Research Council, occupational stud- ies, and community drinking-water exposure studies. The review identified more health outcomes than described in Chapter 7 of this report, and this suggested a lower threshold for inclusion than applied by the present committee. Both reviews identified kidney cancer, lung cancer, breast cancer, scleroderma, hepatic disease, renal disease, and spontaneous abortion as being of interest. The ATSDR review also suggested that the following outcomes may be important: liver cancer, leukemias, cervical cancer, bladder cancer, esophageal cancer, soft-tissue sarcoma, skin disorders, aplastic anemia, non-Hodgkin lymphoma, multiple myeloma, Hodgkin disease, pancreatic cancer, brain cancer, Parkinson disease, and lupus. The present committee and ATSDR took different approaches to assessing the epidemiologic literature. ATSDR focused on previous reviews and studies that yielded positive results, especially community stud- ies of drinking-water contamination. The committee used an approach developed by the Institute of Medicine (IOM 2003) for reviewing the epidemiologic literature, including consideration of individual study characteristics and biases, synthesis of the available studies, and consideration of evidence from the toxicology literature. Only outcomes that were corroborated and single, very strong studies were flagged as deserving of consideration.

OCR for page 184
Contaminated Water Supplies at Camp Lejeune—Assessing Potential Health Effects 192 Health Survey In January 2008, Congress mandated a Navy-Marine Corps health survey to be conducted in 2009. The survey will be mailed to the active-duty and civilian cohorts at Camp Lejeune, the Camp Pen- dleton sample, the 12,598 respondents in the 1999-2002 ATSDR survey, and anyone who has registered with the Marine Corps or provided contact information to ATSDR. Items on the survey will include in- formation about any cancer diagnoses (such as type of cancer, date of diagnosis, and state and hospital of diagnosis), residential history, residences on the base, occupational history, and several risk factors (such as socioeconomic status, demographics, smoking, and alcohol consumption). Permission to gain access to medical records will be requested from those reporting cancer diagnoses. The health survey has the potential to improve future studies of Camp Lejeune residents. For ex- ample, the survey would enhance the collection of relevant covariates and expand the potential scope of nonfatal disease and disability beyond what can be addressed in a typical mortality study or in a cancer registry. The health survey would also demonstrate that the health concerns of Camp Lejeune residents are being investigated to the extent feasible. Nevertheless, the committee has several concerns about the health survey as a source of scientifically useful information for assessing the impact of water-supply contamination at Camp Lejeune. First, the statistical power for evaluating relevant outcomes appears to be low and incompletely addressed in the feasibility study. Second, there may be a bias in disease reporting and participation; a person who has a disease or disability may be more likely to participate. ATSDR has determined that for the health survey to be successful, and therefore useful for the proposed studies de- scribed below, a participation rate of at least 65% would be necessary. Even with that level of response, there is much potential for participation to be influenced by exposure or disease history. Third, the health survey would include only active-duty personnel and civilians who lived on the base after 1975, not those who were present and exposed before then. Fourth, as previously noted, the quality of exposure data would remain uncertain for the same reasons noted above in connection with the completed and current studies. All-Cause Mortality Study The purpose of the mortality study is to evaluate all causes of death in the three cohorts—Camp Lejeune military, Camp Lejeune civilian, and Camp Pendleton military. Followup would begin at the start of known assignment at Camp Lejeune or at the start of active duty for the Camp Pendleton cohort and continue to the end of the study period (December 31, 2007) or death. Cause-specific mortality in the cohorts would be compared with national rates by using standard- ized mortality ratios and standardized mortality ORs. ATSDR also proposes to compare those exposed to contaminated drinking water at Camp Lejeune with those unexposed at Camp Pendleton to minimize bias due to the healthy-veteran effect caused by differences in underlying mortality between veterans and the general public (Bove and Ruckart 2008). ATSDR considered conducting internal comparisons between exposed and unexposed groups at Camp Lejeune but rejected such analysis because of the small number of subjects at Camp Lejeune who were free of exposure. Finally, the agency proposed to consider lagging exposures in the analyses to account for a latent period. Because individual-level information on potential confounders is not available in the computer- ized databases used to identify study subjects, ATSDR proposes two approaches to consider potential confounders. If the Navy-Marine Corps health survey is deemed successful, it will use information from the survey participants to adjust for confounding in a two-stage approach, extrapolating the information from the health survey for application to the mortality study. If the survey does not generate an adequate response, consideration will be given to nested case-control sampling with interviews of decedents’ next of kin to determine information on risk factors. Those are reasonable strategies but are of unknown feasi- bility.

OCR for page 184
Studies of the Camp Lejeune Population 193 Cancer-Incidence Study The cancer-incidence study would evaluate all confirmed cancers diagnosed in the active-duty and civilian worker cohorts at Camp Lejeune and Camp Pendleton and the cohort of survey participants. Because the number of women in the active-duty cohort is small, an additional 2,900 women who lived on the base and were identified through their participation in the birth-defects and childhood-leukemia study would be added to the Camp Lejeune active-duty cohort. To identify cancer cases, ATSDR pro- posed to match each cohort member’s personal identification information to the available data on cancers in all 50 state cancer registries (or at least the cancer registries from the 25-30 states with the highest per- centages of known retirees), the DOD, and Department of Veterans Affairs (VA) cancer registries, the Naval Health Research Center’s Career History Archival Medical and Personnel System (CHAMPS), death certificates, and the National Death Index. Followup would begin with the start of each registry’s operation or 1975, whichever is later, and continue until December 31, 2007. If the Navy-Marine Corps health survey is successful, the cancer-incidence study would also include participants in the survey. Per- sonal identification information on the survey participants will be matched to available data on cancer in the state, DOD, and VA cancer registries. Therefore, like the mortality study, the incidence study will use a two-stage approach in which information on exposure and cancer would be available on everyone in the study who is not lost to followup, but information on individual-level potential confounders will be avail- able only on those who complete the health survey. That information will be used to adjust for confound- ing in the analyses of the entire study population. Because all state cancer registries have data available from 1997 on, cancer incidences in the Camp Lejeune and Camp Pendleton cohorts will be compared with national incidences for the period 1997-2007. Comparisons between the exposed and unexposed participants stationed at Camp Lejeune and comparisons between Camp Lejeune and Camp Pendleton would use all cancers identified from 1975 to 2007—the entire study period. Other Future Studies ATSDR will also consider studying nonfatal, noncancer diseases. The Navy-Marine Corps health survey would include questions on nonfatal diseases and symptoms that are known to be or suspected of being associated with solvent exposure. Such diseases as Parkinson disease, renal failure and other severe renal diseases, severe hepatic diseases, lupus, and scleroderma will be asked about directly, and space will be provided so that respondents can report other disease conditions. Symptom ascertainment may include questions on skin disorders and neurologic disorders. All those diseases and conditions can be confirmed by using medical records. The CHAMPS database can also be used to identify and confirm diseases oc- curring in marines on active duty from 1980 on. However, ATSDR states that a study using that database would probably have insufficient statistical power and therefore the study is of very low priority. Review and Evaluation ATSDR proposed to conduct morbidity and mortality studies that would address some of but not all the questions that have been raised by the affected community. The health end points to be considered would include fatal conditions that are sufficiently common for analysis (depending on the success of the mortality study), incident cancers (depending on the success of the cancer-incidence study), and nonfatal diseases of interest other than cancer, such as scleroderma and neurologic deficits (depending on the suc- cess of the health survey). The mortality study is very likely to be feasible, given the documentation of data sources in the ATSDR feasibility assessment, whereas it is not clear that the cancer-incidence study would be successful in engaging and linking with all 50 state registries. The health survey is subject to uncertain response, as noted by ATSDR, which may limit its value.

OCR for page 184
Contaminated Water Supplies at Camp Lejeune—Assessing Potential Health Effects 194 ATSDR recognized that it is necessary to focus on health conditions that are sufficiently common to allow useful epidemiologic evaluation. It conducted a series of sample-size calculations to ensure that there would be sufficient statistical power to evaluate associations of exposure with prevalent cancers and all-cause mortality with a 10-year lag in exposure (Bove and Ruckart 2008). It is not clear whether there is sufficient power for comparisons of the Camp Lejeune and the Camp Pendleton cohorts, nor is it clear whether outcomes of particular interest to ATSDR and to the committee (such as kidney cancer) can be evaluated with adequate power. ATSDR has begun to consider the adequacy of statistical power, but the information and interpretation fall short of making a clear case that the study methods, even if successful, would generate adequate power for the comparisons of interest. ATSDR recognized the potential for confounding due to unmeasured risk factors in both the mor- tality and cancer-incidence studies. With the exception of age, sex, and race, individual-level factors in the populations of Camp Lejeune and Camp Pendleton are not available. However, some information on the population that completes the health survey would be available. ATSDR proposes a two-stage ap- proach, using the survey data to estimate the effects of confounding with reference to the cohort as a whole. How that would be performed is not described in detail (that is, on an individual basis or by apply- ing patterns of confounding from the health survey to the mortality and cancer-incidence studies). It also is not clear whether the survey will be adequately designed to provide information on the Camp Pendleton cohort that is comparable with that on the Camp Lejeune residents. As ATSDR notes, the value of those data is contingent on generating an adequate response. The use of nested case-control studies of deaths from causes of interest with interviews of next of kin to assess confounding is an alternative approach that is feasible but quite demanding in that it will be necessary to locate, recruit, and interview the next of kin after identification of deaths or incident cancers. ATSDR recognized the potential for bias in the assessment of exposures because of uncertainties in identifying locations on the base where cohort members were stationed and because of possible expo- sure to drinking-water contaminants at other than primary residences or work locations. The agency sug- gested that such bias would tend to underestimate the disease risk associated with exposure if exposure actually causes the disease. ATSDR was confident that the extensive water modeling that is being done at the base would reduce the effect of exposure-misclassification bias that might occur. The committee has less confidence in the certainty of the modeling efforts, given the small number of water-supply meas- urements available for validating the models (see Chapter 2). ATSDR has discussed basing the exposure assessment on the monthly concentrations of contaminants in the drinking water at either the residences or the workplace locations, as appropriate. However, there has been no discussion of the exposure metric that would be calculated and linked with outcomes. For example, it was unclear whether ATSDR would assess the effect of cumulative exposure or of peak exposure. Advantages of the cancer-incidence study over the mortality study, as described by ATSDR, are the higher number of cancer cases and the ability to assess etiology independently of survival. Several female cancers (breast, ovarian, cervical, and uterine) could be evaluated with adequate statistical power (Bove and Ruckart 2008). However, there are concerns about the comparability of the women at Camp Lejeune, who include spouses of workers and women identified because of having given birth, compared with those identified at Camp Pendleton. The cancer-incidence study would also have greater power to detect associations with a broader array of cancers of interest (such as kidney, non-Hodgkin lymphoma, and leukemia) and would eliminate potential effects of differential survival. ATSDR discussed the possi- bility of missed cancers in the incidence study due to incomplete coverage of the study period by the indi- vidual state cancer registries. As it noted, there should be no bias in the internal comparisons, because missing cases are unlikely to be associated with exposure status. However, the comparison between Camp Lejeune and Camp Pendleton could be affected if there are differences between the bases in the percent- age of retired marines migrating to states whose cancer registries are older, and there are broader concerns about the constitution of the study populations and the multiple ways in which the Camp Lejeune cohort would be assembled. In summary, although the major issues bearing on the feasibility of the proposed studies have been considered by ATSDR and the approach has some strengths, notably inclusion of a comparable ma-

OCR for page 184
Studies of the Camp Lejeune Population 195 rine base, there are serious unresolved questions about the feasibility and ultimate value of the studies. It is not clear that the cancer-incidence study or the health survey would be successful; success in the former would be contingent on the cooperation of many cancer registries, and success of the latter on generating an adequate response. The statistical power to compare groups of interest across the array of outcomes of interest was not provided. The ultimate ability to measure and adjust for potential confounding factors is not certain, nor is it clear how the information from the health survey would be applied to the study co- horts. With those concerns layered on the previously noted problems regarding the accuracy of exposure assessment, it is not clear what the scientific value of additional studies would be. FINDINGS OF COMPLETED, CURRENT, AND FUTURE STUDIES The committee considered the value of completed, current, and planned studies of the Camp Le- jeune population in light of the information available on assessing exposure, health end points of primary concern, and what is known about the potentially affected population from previous studies and work in progress. Review of data and modeling efforts pertaining to exposure provided clear documentation that contaminants were present but provided little basis for suggesting that exposures of the population can be reconstructed with much precision. The literature on potential health effects of the agents of primary con- cern, TCE and PCE (see Chapters 4 and 5), indicates an array of possible health effects, including can- cers, reproductive effects, neurobehavioral effects, immunologic effects, and renal and hepatic toxicity, possibly affecting both children and adults. Completed and current research at Camp Lejeune has been limited to particular end points and focused on pregnancy outcomes—including fetal growth, preterm birth, and birth defects—and childhood cancers. Those studies have not distinguished and are unlikely to be able to distinguish between an ab- sence of adverse effects and the presence of modest effects that fall below the limits of what can be iden- tified in light of exposure misclassification and low statistical power. A broader consideration of health effects would be needed to provide scientific evidence to answer questions regarding the possible effects of water-supply contamination. For new studies to make a substantial contribution to evaluating whether exposure to contaminated water resulted in adverse health effects, an array of feasibility considerations needs to be addressed and resolved favorably. ATSDR has made a reasonable effort to evaluate those is- sues in the study of the feasibility of future work, but structural problems make it difficult to show that such research will be of high scientific merit. Key feasibility considerations that apply to all environ- mental epidemiology studies, including the evaluation of water contaminants and health at Camp Lejeune, are listed below.  Study population. The residents of Camp Lejeune potentially exposed to the contaminated water supplies of concern need to be enumerated for study, with inclusion of exposed people and comparable unexposed people identified from elsewhere on the base, from periods beyond the years of contamination, or from other military bases.  Exposure. The water serving the homes of the individual residents at specific times would need to be identified to assess potential exposure to specific toxicants. There would need to be an independent process of exposure assessment that allows estimation of concentrations of specific pollutants going from the source to the tap and related to specific time and places. It would then be necessary to reconstruct residential histories in Camp Lejeune to link people to estimated water concentrations of pollutants in their homes. Ideally, studies would consider water sources at the locations of work, day care, and schools and consider individual behavior, including water consumption and bathing.  Statistical power. The health outcomes of interest vary greatly in frequency of occurrence. For re- search results to be informative, sufficient numbers of exposed and unexposed people are needed to gen- erate stable estimates of rates of diseases and to make comparisons. Disease latency—the time between exposure and development and manifestation of disease—is important. The Camp Lejeune population was generally young, so even with the passage of 20 or more years since exposure onset, they are still not

OCR for page 184
Contaminated Water Supplies at Camp Lejeune—Assessing Potential Health Effects 196 at the ages at which some of the specific diseases of concern are commonly observed. Given the size and age distribution of the population, it may be infeasible to focus on such end points as kidney cancer, al- though it is justified on the basis of independent research as reflected in the toxicology and epidemiology literature. Furthermore, given the brevity of many people’s residence on the base, realistic effect sizes would need to be considered in assessing adequacy of statistical power.  Potential confounders. The potential for confounding of the observed effects of water exposure by other factors that affect disease incidence would need to be addressed. Because residence or workplace on the base is a primary determinant of exposure and may be related to rank, seniority, or job duty, which themselves may be markers of disease risk, they would need to be measured and adjusted for in the analy- sis. More direct markers of disease risk—such as tobacco and alcohol use history, body-mass index, and diet—would also need to be addressed for selected health end points, including those of primary concern (such as renal disease).  Time and cost. Realistic estimates of the time required to conduct the study are needed, particu- larly in light of the long history of concerns regarding contaminated water and health at Camp Lejeune. The financial cost is also a key consideration in that studies that require generating large volumes of new data through individual contact and advanced water modeling are expensive and time-consuming.  Credibility of findings. It is important not only that the research be scientifically rigorous but that the results be fully and widely accepted. That issue would need to be addressed from the outset in framing the question, the mechanism of funding, the selection of the researchers, the conduct of the study, and the interpretation, evaluation, and dissemination of results. For structural reasons, meeting the criteria above is problematic. One major problem is that the number of people available for the study may be too small to generate statistically meaningful results re- lated to rare outcomes of greatest interest (such as kidney cancer). Historical contaminant-exposure esti- mates are difficult to construct and might be impossible to quantify with any confidence in the absence of contaminant measurements taken during the period of concern, no matter how elaborate the water models are. Many residents were exposed for relatively short periods; most lived in the affected areas for only a few years (2-3 years was typical for marines stationed at the base), and it is difficult to know what types of exposures they had before or after they lived at Camp Lejeune. We know that there were some highly contaminated wells for some periods, but their operations were cycled with those of uncontaminated wells, so exposure to water contaminants was intermittent and cannot be determined on an individual ba- sis or for time frames of weeks as required to assess the occurrence of reproductive health end points. Even if all the information on the population, exposure, and health outcomes could be obtained, consid- eration should be given to whether the cost and time required to conduct more definitive studies justify the likely delay in or distraction from resolving the public-health concerns and the controversy that has developed around the issue. The costs and benefits of such efforts need to be reconciled. Finally, the long- standing controversy over this episode is apparent, and some question the objectivity of the Marine Corps in generating valid, objectively interpreted scientific data on the topic. Future research needs to be both scientifically informative and credible to the multiple target audiences. CONCLUSIONS AND RECOMMENDATIONS The scope of health outcomes addressed in completed and current studies of the Camp Lejeune population is limited and driven, to a large extent, by the types of diseases that are feasible to measure with available surveillance data and a health survey. They are not necessarily the conditions or diseases that would be considered of highest priority on the basis of the committee’s review of the literature of epidemiology and toxicology. There are serious limitations in the quality of existing studies of the Camp Lejeune population. Consequently, those studies provide little information to assess directly whether the population exposed to water contaminants has suffered adverse health effects of them. Completion of the studies in progress will provide only a marginal improvement in understanding.

OCR for page 184
Studies of the Camp Lejeune Population 197 Recommendations:  The planned reanalyses of the preterm-birth and fetal-growth study should be completed as soon as possible, taking advantage of the corrected exposure information that is available but not awaiting more extensive water modeling. Reanalyses should include development of a detailed written analysis plan (for example, Sheppard 2008). Careful attention should be paid to confounding, given the associa- tions between residence and indicators of risk. Given the inherent limitations of birth-certificate data, sen- sitivity analyses to address gestational-age misclassification, subgroup analyses, and confounding should be incorporated. Finally, future reports should provide full details of the approach, results, and sensitivity analyses; the STROBE (strengthening the reporting of observational studies in epidemiology) guidelines (Vandenbroucke et al. 2007) would be suitable for such documentation. Despite the limited scientific benefit of this effort, the modest cost justifies its prompt completion.  The current case-control study of birth defects and childhood cancer should be completed, given the effort already invested, despite severely limited statistical power. The same recommendations noted for the study of preterm birth and fetal growth apply here as well, including careful planning of analytic methods and full documentation. Relative to the overall effort expended thus far, the committee recog- nizes the need for completion of this study. It could be argued that additional studies of the potential health effects from the historical con- tamination of drinking water at Camp Lejeune could help guide decisions on how to resolve the claims of former residents. Beyond its scientific merit, a more thorough evaluation of health patterns of former Camp Lejeune residents could be seen as providing a valuable public-health service in providing docu- mentation of the experience of former residents and perhaps characterizing the population better. How- ever, on the basis of what is known about the contamination of water supplies at Camp Lejeune; the size, age, and residential mobility of the residents; and the availability of records, the committee concludes that it would be extremely difficult to conduct direct epidemiologic studies of sufficient quality and scope to make a substantial contribution to resolving the health concerns of former Camp Lejeune residents. Con- duct of research that is deficient in those respects not only would waste resources but has the potential to do harm by generating misleading results that erroneously implicate or exonerate the exposures of con- cern. Recommendations:  New studies should be undertaken only if their feasibility and promise of providing substan- tially improved knowledge on whether health effects have resulted from water exposure at Camp Lejeune are established in advance.  Decisions regarding the appropriate policy response to health concerns about exposure to contaminated water at Camp Lejeune should not be delayed or await the results of epidemiologic studies that are in progress or planned inasmuch as those studies are unlikely to provide definitive information on potential health effects.