Veterans and Agent Orange: Herbicide/Dioxin Exposure and Acute Myelogenous Leukemia in the Children of Vietnam Veterans (2002)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

HERBICIDE/DIOXIN EXPOSURE AND ACUTE MYELOGENOUS LEUKEMIA 11 the age of 16. Leukemias in younger children are believed to have a different etiology from those of older children because the genetic abnormalities under- lying them are more likely to have been present at birth. Incidence rates through age 19 are similar in males and females, and in whites and African Americans (NCI, 2002). Research has identified parental exposure to pesticides (Buckley et al., 1989), in utero exposure to ethanol (Severson et al., 1993) and dietary topoisomerase II inhibitors (Ross et al., 1996), and parental use of marijuana during pregnancy (Robison et al., 1989) as potential environmental risk factors for childhood AML. However these findings require replication in other populations. Many other studies of environmental risk factors have examined childhood leukemia as one entity, failing to separate out the histologic subtypes. Chemotherapy may also be responsible for secondary AML. As in adults, the genetic disorders Down syndrome, Fanconi anemia, and Bloom syndrome are considered risk factors for the development of AML in children. Other illnesses associated with increased incidence include neurofibro- matosis Type I, Kostmann’s disease, Blackfin-Diamond anemia, paroxysmal noc- turnal hemoglobinuria, Shwachman-Diamond syndrome, and thrombocytopenia- absent radii syndrome (Langmuir et al., 2001). Chapter 7 of Update 2000 contains additional information on leukemia as part of the discussion of adult cancer outcomes. Chapter 8 of that report covers childhood cancers in general as part of a review of the scientific literature regard- ing herbicide and dioxin exposure and reproductive effects. SUMMARIES OF EPIDEMIOLOGIC EVIDENCE In seeking evidence for associations between health outcomes and exposure to herbicides and dioxin, many different kinds of epidemiologic studies must be considered. Each study has various strengths and weaknesses and contributes evidence regarding an association between exposure and the health outcome. The three main groups of individuals studied with respect to herbicide exposure are those with occupational, environmental, and military exposures. The historical basis for the groups studied was examined in Chapter 2 of VAO. A discussion of the criteria for inclusion in the review is detailed in Appendix A of that report. Epidemiologic studies and reports evaluated by the committee are summa- rized below. Each subsection begins with a recapitulation of relevant reviews originally presented in Update 2000. Information reviewed by the committee since the release of the Update 2000 report is then summarized under the heading New Studies. Table 1 lists the estimated risk, confidence interval, and number of observed cases for each of the epidemiologic studies considered by the com- mittee, where these data are available. 

12 VETERANS AND AGENT ORANGE Occupational Studies In a case-control study, Buckley and colleagues examined possible relation- ships between parental occupational exposures and acute myelogenous leukemia4 incidence in workers’ children in a case-control study (1989). This study was assembled by the Children’s Cancer Study Group and included cases diagnosed in North America from 1980 to 1984. Initial analyses focused on self-reported job titles, which were used as a proxy for exposures based on a previously developed job-exposure matrix (Hoar et al., 1980). One hundred seventy-eight case-control pairs provided information regarding paternal occupational pesti- cide exposures, including weed killers. Using the job-title linkage, pesticide exposures were associated with a 2.3-fold increased risk. Using self-reported information on workplace exposure to specific types of products and on duration of exposure, a 2.7-fold (1.0–7.0) increased risk of fathering a child who devel- oped AML was found for men exposed more than 1,000 days. Results were comparable when interviews conducted with surrogates for the fathers were excluded from the analysis. An elevated risk was seen for exposure before, during, and after the pregnancy, but since these were highly correlated, it was not possible to determine if preconception exposure alone would lead to increased risk of a child developing AML. This is an important consideration because the wartime exposure of male veterans to herbicides would have occurred prior to conception. For female veterans, it could have occurred during early pregnancy. A second study addressed childhood cancers in children born to male saw- mill workers in British Columbia, Canada (Heacock et al., 2000). The primary exposures in these plants were to chlorophenate fungicides, which are contami- nated with PCDDs and PCDFs formed during the production of these chemicals. Employees who worked at least 1 year at any of 11 such lumber mills in 1950– 1985 formed the cohort of 23,829. Estimates of exposure were made from job title in each mill using information from mill records as well as interviews with persons knowledgeable about technology and formulation changes; a validation study that compared urinary chlorophenates with job-based estimates yielded a correlation of 0.49. The cohort was linked to birth files and also to marriage files that were linked to birth files (since the mother’s name is on all birth certificates, whereas the father’s name is not always), in order to identify the cohort of workers’ children. The children’s cohort was then linked to the British Columbia Cancer Registry to determine cancer diagnoses by age 20. Cancer cases in chil- dren born to sawmill plant workers from 1952 through 1988 and diagnosed in 1969–1993 were included. There were a total of 40 such cases, of which 22 were female and 18 male. The initial analysis proceeded by calculation of a standardized incidence ratio (SIR), in which the cancer experience of these children was compared to 4Referred to as “acute nonlymphocytic leukemia” in the paper. 

HERBICIDE/DIOXIN EXPOSURE AND ACUTE MYELOGENOUS LEUKEMIA 13 that of the general population of the province of British Columbia, adjusted for age and sex of the children and calendar year. Results showed no elevation of risk for all cancers (SIR = 1.0, 0.7–1.4) or for leukemia (SIR = 1.0, 0.5–1.8). An internal comparison was conducted using a nested case-control design. For each case, five controls were selected from within the cohort matched on sex and year of birth. These results compared risks for differing cumulative exposure groups within four time windows: (1) more than 90 days prior to conception; (2) from 90 days before conception to conception; (3) from conception to birth; and (4) after birth. The risk for all cancers combined was greater in the high-exposure groups than in the low-exposure groups for all windows except the first (more than 90 days before conception); however, all confidence intervals included values con- sistent with no effect or even a lowered risk in those with high exposures. The risk for leukemia was not elevated; no specific information on AML risk was provided. The main limitations of the study were the small number of cases and the lack of quantitative data on chlorophenate or TCDD exposure. Environmental Studies A large population-based case-control interview study of several childhood cancers was conducted in Germany by Meinert et al. (2000). Information on exposures was gathered from interviews of parents and questionnaire responses, based on the methodology used by the Children’s Cancer Group. The children were diagnosed at less than 15 years of age, and there were 1,184 cases of leukemia. Parental occupational exposures to herbicides, insecticides, or fungi- cides were found to be related to childhood cancers regardless of the time period of exposure and the type of cancer (lymphoma or leukemia). Of particular note is the finding that there was a statistically significant association between paternal exposure in the year before pregnancy and leukemias (odds ratio [OR] = 1.5, 1.1– 2.2, based on 62 cases). Statistically significant associations were also found between leukemias and paternal exposure during pregnancy (1.6, 1.1–2.3) and “ever” (1.6, 1.1–2.3). However the strongest associations were in relation to maternal exposures during pregnancy (e.g., leukemia OR = 3.6, 1.5–8.8). No analyses of the separate types of leukemia were reported. The data provide some evidence of an increased leukemia risk for children exposed in utero to herbicides, insecticides, or fungicides, and also for children whose fathers were exposed prior to their conception. While this study may have been the largest ever to examine this hypothesis, the possibility of recall bias could not be ruled out, and there was some evidence suggesting that parents of cases consistently reported more occupational exposures than parents of controls. Neither study reported the effect of paternal preconception exposures indepen- dent of maternal exposures (or vice versa). The relevance of this study to expo- sures to the herbicides used in Vietnam and their contaminants is uncertain because the data did not permit analysis by specific chemicals. 

14 VETERANS AND AGENT ORANGE New Studies At the October 2001 workshop held by the committee (described in Appen- dix A), one of the authors of the Meinert et al. (2000) paper—Dr. Joachim Schüz—presented additional information on leukemia outcomes in the study cohort. This included the inclusion of additional observations and details on AML incidence. The committee took into consideration the fact that the data presented have not been subject to peer review. All of the cases in the expanded cohort were identified by the German Childhood Cancer Registry, which is believed to be greater than 95% complete. They were matched 1:1 with controls on gender, date of birth (within one year) and community of residence. The analyses included adjustment for socioeconomic status and whether the community of residence was rural, urban, or mixed. A self-administered questionnaire with telephone follow-up was used to obtain information on parental occupational history, chemical exposures, and the timing of exposure (prior to conception, during gestation, or after birth). A total of 167 AML cases were available for analysis. Among the subset of cases where the father had a preconceptional occupational exposure to pesticides/ herbicides, the adjusted OR for AML was 0.9 (0.3–2.4). Similar odds ratios were reported for paternal exposure during gestation (1.1, 0.4–3.1) and after birth (1.1, 0.4–3.1). The authors concluded that there was no evidence supporting an asso- ciation between preconceptional exposure to pesticides/herbicides and AML in this cohort. They noted that the study had a low statistical power to detect such an association and that there were few data available on exposure to specific substances. In a paper not previously reviewed in a Veterans and Agent Orange series report, Kristensen et al. (1996) analyzed cancer morbidity in children of agricul- tural workers in Norway. The cohort consisted of 323,359 offspring born between 1952 and 1991. Cancer diagnoses and outcome classifications were obtained from the Cancer Registry of Norway. Data from the Statistics Norway agricul- tural census closest to the year of birth were used to model parental exposure. Variables included type of agricultural activity, money spent on pesticides (in- cluding herbicides), and presence of spraying equipment. Reference rates were estimated using rural, non-farm Norwegian populations that were age- and gender- matched in 5-year strata. A total of 1,275 cases of cancer were identified. Of these, 36 were AML, 92 were ALL, and 35 were acute leukemias not classified by type. Overall, the rate of leukemia in the children of farmers was indistinguishable from the reference population (SIR = 1.0, 0.8–1.2;5 based on 113 cases). When the analysis was limited to the children of farming parents who had purchased pesticides, the rate ratio for AML, adjusted for year of birth and calendar year, was non-significantly 5These values are reported as SIR = 101; 95% CI 83–122 in the paper. 

HERBICIDE/DIOXIN EXPOSURE AND ACUTE MYELOGENOUS LEUKEMIA 15 elevated: 1.4 (0.6–2.9). The adjusted relative risks (RRs) for ALL and other acute leukemias were close to unity. Restricting the calculations to cases in which the child most likely spent the first few years of life on a farm did not affect the results. The researchers concluded that the data did not support an association between exposure to pesticides and AML. This study’s primary strengths are the completeness of the birth data and its linkage to census and cancer registry databases. Several weaknesses, however, limit its informativeness: the crudeness of the pesticide exposure proxies, the lack of adjustment for confounders other than farm-related exposures, and the small number of AML cases, which resulted in low statistical power and low precision for the effect estimate. Vietnam Veteran Studies In a large case-control study, Wen et al. (2000) examined service in Vietnam or Cambodia as a risk factor for childhood leukemia. The study included 1,805 cases of ALL and 528 cases of AML, including cases diagnosed through 17 years of age. It combined data from three studies conducted by the Children’s Cancer Group, which represents a consortium of hospitals and medical centers in the United States and Canada that pool their cases to enable large studies of child- hood cancers and thereby achieve sufficient statistical power. The cases were matched to controls on year of birth, location of residence, sex, and race. Controls were found through random-digit dialing, and cases were restricted to those with a telephone in the home. The overall response rates were 89 percent for cases and 77 percent for controls, and they were slightly lower for paternal interviews (83 percent for cases, 70 percent for controls). Analyses were conducted using conditional logistic regression for all leukemias combined, for ALL and AML separately, and stratified by age at diagnosis. Regression models were adjusted for potential confounders, including education, race, family income, smoking, exposure to X-rays, and paternal marijuana use. The results indicated no increased risk of either leukemia subtype associated with military service in general. However, for service in Vietnam or Cambodia, the risk of AML (OR = 1.7, 1.0–2.9), but not ALL (1.0, 0.8–1.4), was increased. Analyses examining tours in Vietnam and Cambodia were hampered by the small number of observations: the higher calculated risk was for offspring whose father served two or more tours there (OR = 5.0, 1.0–24.5; based on 8 cases). However, length of service produced a different pattern. For those serving one year or less in these countries, the OR was 2.4 (1.1–5.4; 21 cases), whereas the OR for those present more than one year was 1.5 (0.7–3.2; 16 cases). When stratified by years between service and conception of the child, the association was strongest in those who had served more than 15 years earlier; however, the numbers in this stratum were small. Self-reported exposure to Agent Orange showed no associa- tion. The strongest association was for cases diagnosed under the age of 2 years 

16 VETERANS AND AGENT ORANGE (OR = 4.6, 1.3–16.1). It is believed that childhood cancers at very young ages are more likely to be etiologically related to preconception or in utero exposures than those diagnosed at later ages. Limitations of the study include possible residual confounding from not having detailed exposure data on smoking and marijuana use; the unexplained stronger association with increasing interval between service and conception; and lack of adjustment for factors associated with service in Vietnam or Cambodia, including postwar exposures. The authors point out that the inconsistency in results for number of tours of duty versus number of years in Vietnam or Cambodia could have been related to exposures being correlated with movement in and out of these areas, rather than with duration, or to other exposures in Southeast Asian countries. Longer duration in Vietnam or Cambodia does not necessarily mean higher exposure to herbicides, since no information is available on the nature of these veterans’ activities during their service. In another study, close to 50,000 Australian Vietnam veterans were surveyed about their and their children’s health; an 80 percent response rate was achieved (Commonwealth Department of Veterans’ Affairs [CDVA], 1998). A follow-up validation study of selected conditions included children’s cancers (AIHW, 1999); a later supplement (AIHW, 2000) had, as one of its aims, the collection and analysis of data on specific subtypes of leukemia among children of veterans. Validation sources included pathology reports, doctor certifications, or records from a disease or death registry. Australia has had a cancer registry since 1982. Nine AML cases in the children of veterans were confirmed through clinical records. The investigators used various assumptions to adjust for nonrespondents in the validation study and for circumstances in which it was not possible to validate a reported case (e.g., physician could not be located, medical records were incomplete, etc.). Depending on the specific assumptions, up to nine addi- tional cases of AML were estimated for the full cohort (for a projected total of from 9 to 18 cases). The assumptions adopted by the study’s authors estimated 4 additional cases, for a total of 13 validated cases. The original (AIHW, 2000) report indicated that 3 (range 0–6) cases of AML were expected in the cohort based on incident rates in the community. Given this estimate, the authors reported a statistically significant 4.3-fold increased risk of AML. All of the alternative analyses, including those with the most restrictive assumptions (i.e., assuming zero cases among nonrespondents and no valid AML diagnoses among reported cases for which validation was not possible), also yielded large, precise, and hence statistically significant excesses of AML. These analyses did not adjust for any sociodemographic or life-style factors associated with increased risk of AML, although adjustment for age and gender was achieved through the methods used to derive expected numbers of cases in Australia’s community standard. No excess risk was observed for the other forms of childhood leukemia: ALL, chronic lymphocytic leukemia (CLL), or chronic myelogenous leukemia (CML).