The federal government's Agricultural Health Study: A critical review with suggested improvements

The Agricultural Health Study (AHS) has approximately 90,000 pesticide applicators and their spouses enrolled in a number of studies to determine whether exposures to specific pesticides are associated with various cancers and other adverse health outcomes. Although the AHS was intended to be an integrated program of studies, some significant difficulties have emerged. In this report, we examine the design of the AHS, identify important program strengths and flaws, suggest various improvements in the program, and recommend ancillary studies that could be undertaken to strengthen the AHS. Overall, the AHS is collecting a large amount of information on potential determinants of health status among farmers and farm families. A promising feature of the AHS is the prospective cohort study of cancers among farmers in which the research design determines exposures prior to the diagnosis of disease. More effort needs to be devoted to reducing selection bias and information bias. Success of the cohort study will depend in part on follow-up surveys of the cohort to determine how exposures and disease states change as the cohort ages. The cross-sectional and case-control studies planned in the AHS are less promising because they will be subject to some of the same criticisms, such as potentially biased and imprecise exposure assessment, that have characterized the existing literature in this field. Important limitations of the AHS include low and variable rates of subject response to administered surveys, concerns about the validity of some self-reported non-cancer health outcomes, limited understanding of the reliability and validity of self-reporting of chemical use, an insufficient program of biological monitoring to validate the exposure surrogates employed in the AHS questionnaires, possible confounding by unmeasured, nonchemical risk factors for disease, and the absence of detailed plans for data analysis and interpretation that include explicit, a priori hypotheses. Although the AHS is already well underway, most of these limitations can be addressed by the investigators if adequate resources are made available. If these limitations are not addressed, the large amounts of data generated in the AHS will be difficult to interpret. If the exposure and health data can be validated, the scientific value of the AHS should be substantial and enduring. A variety of research recommendations are made to strengthen the AHS. They include reliability and validity studies of farmer reporting of chemical use, biological monitoring studies of farmers and members of farm families, and validity studies of positive and negative self-reports of disease status. Both industry and government should consider expanded research programs to strengthen the AHS.