BACKGROUND ON METHODS
Epidemiologic Study Designs
In light of the high priority of research on cancer and reproductive disorders, historic cohort studies and case-control studies are the research designs most likely to be used for epidemiologic studies of Vietnam veterans potentially exposed to herbicides. Either type of study would require retrospective assessment of herbicide exposure. However, the method of exposure reconstruction and the level of detail may differ depending on the design of the epidemiologic study in which it would be applied.
Cohort studies typically begin with estimation of the exposure status of every member of the population of interest and then proceed to ascertainment of health outcomes for each individual. This approach provides exposure estimates for the entire population. However, the need to obtain data for large groups (including many people who did not experience a health event) usually constrains the effort that can be devoted to estimating exposure for any individual. As a result, it is sometimes feasible to characterize exposures only crudely in large cohort studies.
Case-control studies may be conducted as an alternative to improve efficiency and validity by collecting more detailed exposure information for a smaller number of subjects. Because case-control studies include all of the cases of the health outcome of interest but only a sample of the "base" population that produced the cases, it is often feasible to utilize more costly, more time-consuming, or more labor-intensive techniques of estimating exposure than would be feasible in a cohort study including the entire base population. In an occupational study, for example, the exposures of an entire cohort of workers might be estimated from job titles and historic records, whereas a case-control study of selected workers from the same cohort might also utilize worker interviews and field measurements. A similar approach can be used in case-cohort, case-based, and other alternative study designs that involve cases and a sample from the base population.
Various exposure assessment approaches might seek to identify groups or individuals with qualitatively or quantitatively distinct exposures. Groups with qualitatively distinct exposures who served in Vietnam might include those who served in different branches of the military, or in different geographical areas, who had different job titles, or those who were in zones sprayed with herbicides versus those who were not. Alternatively, measuring herbicide or metabolite concentrations in the blood, fat, and tissues or organs of exposed individuals may provide the information needed to quantitate exposure(s). Some combination of environmental exposure patterns and biologic measures may be able to predict exposures with a better degree of confidence than either could alone.
Past Exposure Assessment
A variety of approaches have been developed for estimation of past exposures. The Appendix provides details on these approaches. In brief, previous studies of Vietnam veterans have used the simplest separation of job titles or service areas. These broad categorizations of exposed and unexposed individuals were based on judgment and in some cases spraying records, which may have produced a large degree of misclassification on exposure status. The goal of the RFP is to invite investigators to propose study designs that will improve the assessment of veterans' exposure to herbicides during their service in Vietnam. Regardless of the strategy, the common goal of such retrospective exposure assessments is to develop the most accurate and unbiased estimates of exposure possible within the limitations of the resources.
Quantitative dose-response epidemiologic studies try to approximate the dose to the target tissue as closely as possible because it is the presumptive cause of the adverse effect observed in the epidemiologic outcome. Although it is seldom possible to retrospectively generate precise estimates, these approximations may yield important information about the magnitude of exposures relative to other study subjects and unexposed individuals. A dose index is a single number calculated from (1) a subject's personal data and (2) an exposure model that is intended to summarize a subject's exposure history in a way that is relevant to the risk of an adverse outcome, such as total dose of the suspected agent received by the subject. The epidemiologic dose index that is closest to total administered dose is the cumulative exposure. Cumulative exposure is calculated as the mean exposure in a given category (in an occupational setting, this would typically be a job title) times the duration in the category, summed over all categories. This has been a useful measure in many studies of the relationship between disease incidence and exposure to asbestos, lead, cadmium, and other agents. Other dose indices may also be important for the risk of a particular disease, and the example given above should not be interpreted as precluding the use of other representations of exposure, such as categorical representations.
Responses to the RFP are expected to make clear the underlying relationship(s) hypothesized between exposure to herbicides and the dose indices chosen, including considerations of mechanism where appropriate. For example, if acute exposure is thought to be important, this should be stated and the means by which the proposed dose metric distinguishes and characterizes acute exposures should be detailed.
It is important to recognize that the steps used in extrapolation of past exposures have variable magnitudes of uncertainty. Quantitative estimates of exposure intensity for the distant past generally have the highest uncertainty. However, large uncertainty in the absolute magnitude of exposure may be acceptable if subgroups of subjects with different modes of exposure also have large relative differences in exposure. Respondents are encouraged to consider a
