While environmental epidemiology aims to produce policy-relevant evidence for the protection of the public?s health, it requires consistent results across studies to create regulatory standards through the identification of dose-response curves. Current standard practices do not address how to compare estimates across studies when exposure distributions do not overlap, nor how to ascertain the dose-response curve of the complete exposure-disease relationship in such situations. This proposal was borne out of difficulties I encountered performing a systematic review, where the estimates I wanted to compare came from studies with partially- or non-overlapping exposure distributions. I brought the idea of utilizing causal inference methods to explore the issue to my sponsor, and she agreed that they could be applied to the issue to clarify the underlying causal structures that could both result in disparate effect estimates as well as complicate the ascertainment of accurate dose-response curves across the full range of exposure levels. This is the first study to use causal inference methods to articulate the assumptions individual studies must meet in order for risk estimates to 1) be comparable to other studies with different exposure distributions and 2) be utilized to create accurate dose-response curves that span the full range of range of exposure levels across populations. The results will allow us to better inform regulatory standards for polychlorinated biphenyls (PCBs), the exposure of interest in this proposal; the technique will also be applicable to other types of endocrine disrupting chemicals (EDCs). In order to carry out this research, I will analyze sample-specific and pooled data from three birth cohorts, analyzing the effect of prenatal PCB exposure on birth weight. I will also create simulations informed by the cohort data to examine the ways in which different underlying causal structure lead to different dose-response curves. Deepening our understanding of the ways in which chemicals impact our health is crucial both to inform current regulatory policy as well as for the future regulation of newer, structurally similar chemicals. Together with the proposed didactic instruction, this research will also serve as a training vehicle to meet my goals as an independent researcher in environmental epidemiology, utilizing causal inference methods to explicate exposure-outcomes relationships that will lead to a career conducting policy-relevant research to improve the public?s health.
Environmental epidemiology aims to produce policy-relevant evidence for the protection of the public?s health, but individual studies of environmental pollutants have limited exposure distributions, and the difficulty of making inferences across studies to inform regulatory standards is rarely addressed. Further, estimating dose-response curves for endocrine disrupting chemicals is notoriously difficult at low exposure levels, complicating the ascertainment of an accurate dose-response curve that spans the full range of exposure levels. This is the first study to use causal inference methods to articulate the assumptions individual studies need to meet in order for risk estimates to 1) be comparable with other studies of different exposure distributions and 2) be utilized to create accurate dose-response curves to inform regulatory standards.