Developing dose-response assessment methods to inform environmental policy: An application of Byaesian hierarchical models using trihalomethanes.

摘要

Problem statement. Research is needed to develop and explore an alternative approach to current Environmental Protection Agency (EPA) risk assessment practices to better utilize available scientific data and account for inherent uncertainty. Aims. To (1) perform a literature review to identify alternative approaches; (2) develop a Bayesian hierarchical model (BHM) for an application to risk assessment; and (3) apply the BHM on a case study of a potential carcinogen commonly found in drinking water. Methods. Analyses consisted of (1) literature review of Bayesian model theory and application; (2) BHM development and application to a pilot data set; and (3) BHM application to risk assessment, using an EPA data set for setting regulatory standards for chloroform, and comparing results. Results. A literature review identified several BHMs with potential for incorporation into risk assessment, as well as limitations hindering their application. A BHM identified from the literature was modified to address these limitations. Analysis of a pilot data set revealed several issues which hindered model performance, attributed primarily to extrapolation required from observed data. Model results demonstrated that incorporating more information into prior distributions reduced the final estimate uncertainty. BHM results from analyzing the data set for chloroform were different compared to those estimated by the EPA. The former were often lower, reflecting the data's support for a lower exposure standard than that ultimately set by the EPA. Conclusions. This research develops and demonstrates a systematic and transparent approach for characterizing uncertainty in the data used for environmental risk assessment. Results demonstrate that failing to characterize this uncertainty can produce estimates that do not accurately describe what the scientific data informs about threats to human health from environmental exposures. This may underestimate risks posed to the public, leading to policy-setting that fails to provide adequate health protection. Further research is needed to explore alternatives for incorporating scientific information into these models, and implementing them into risk assessment. There may also be other aspects of risk assessment that may benefit from these methods, and in particular the development of informative uncertainty factors should be undertaken to inform risk assessments where data are limited.