This project will develop statistical methods relevant to two common forms of environmental epidemiologic studies. The primary goal is to provide methods that extract a concise assessment of health risks associated with environmental exposures, supplemented by appropriate statistical inference. The first topic will evaluate the association between exposure and the risk of a health outcome using diagnosis data on a cohort of individuals supplemented with screening information on undiagnosed participants. The methods will be applied to data from the Seveso Women's Health Study which addresses health risks in women exposed to high levels of dioxin. It is intended that the statistical methods will apply generally to similar studies that include a combination of diagnostic and screening data. The second project concerns statistical techniques to investigate the effects of multiple environmental exposures on health and developmental outcomes. The ideas will be applied to data from the CHAMACOS study of Latino women and their children in California, where information has been collected on environmental (largely pesticide) exposures, in utero and in childhood, for a cohort of women and their infants. Statistical issues involve estimation and ranking-in importance-of suitable causal effects of each exposure, supplemented by a rigorous assessment of which of these represent real effects rather than spurious associations, allowing appropriately for multiple comparisons. Both studies involve the study of vulnerable populations exposed to above average environmental exposures with the potential for elevated risk for poor health outcomes. Statistical and computational algorithms will be developed and provided in an open source user-friendly format allowing their rapid dissemination and use by other investigators. The relevance to public health is two-fold: first, the research will allow environmental epidemiologists to accurately describe the effects of (i) acute dioxin exposure on the reproductive health of women, in particularly on the onset of fibroids, and of (ii) pesticide exposures on birth outcomes and subsequent neurodevelopment of children born to Latino women, in a farmworking community. Second, the proposed research will provide appropriate statistical tools and software to allow other investigators to apply these complex methods to similar studies of the effects of environmental exposures in a wide variety of settings.
