| Summary: | Humans are invariably exposed to a myriad of synthetic chemicals on a daily basis through their diet, consumer products, and via the ambient environment. Exposure also occurs in early life as many chemicals transfer from maternal stores into breast milk and across the placenta. However, the health risks of the majority of environmental chemicals are poorly characterized or unknown. In this thesis, we used emerging multi-pollutant statistical methods to evaluate associations between multiple chemical exposures and aspects of reproductive and child health. These methods, namely elastic net penalized regression and sparse partial least squares regression, allow for simultaneous variable selection and effect estimation. In a simulation study of 270 data structures typical of environmental epidemiological analyses, we demonstrated that multi-pollutant approaches, notably penalized regression, yielded fewer false positive discoveries and effect estimates less biased from co-exposure confounding compared to standard assessments of a single exposure. In European and Arctic study populations, we characterized the exposure levels of multiple chemical classes: brominated flame retardants, phthalates, perfluoroalkyl and polyfluoroalkyl substances (PFAS), heavy metals, polychlorinated biphenyls (PCBs), and organochlorine pesticides. In an analysis of chemical exposures in relation to 22 biomarkers of male reproductive function, we detected 10 associations. These included robust associations between a metabolite of diisononyl phthalate (DiNP), which is abundant in polyvinyl chloride materials, and lower testosterone levels; and between PCB-153, a banned yet environmentally persistent chemical, and reduced sperm motility. In another analysis, we found that prenatal exposure to several chemicals was independently associated with birth weight in infants born at term. Low birth weight is a risk factor for chronic disease later in life. The largest effect estimate was observed for perfluorooctanoate (PFOA) and reduced birth ...
|
|---|