Advanced Methods for Dose-Response Assessment: Bayesian Approaches—Final Report

Resources for the Future (RFF), in conjunction with the U.S. Environmental Protection Agency, the Society for Risk Analysis, and the Electric Power Research Institute, held a workshop September 18–20, 2000, at the RFF Conference Center in Washington, D.C. The intent was to discuss how Bayesian approaches could be useful in improving techniques for estimating exposure–response functions. Ten distinguished scholars from a range of fields (medical biostatistics, decision sciences, environmental engineering, and toxicology) served as faculty. Approximately 80 people attended the workshop. Bayesian methods have been applied to a variety of problems in biomedical research and environmental risk analysis, including design of clinical trials, estimation of exposures to humans and local environments, and, in a few cases, estimation of exposure–response functions. Bayesian methods offer two signal advantages: their use requires careful analysis of problem logic, which has intrinsic utility, and disparate data can be incorporated into calculations. Although application of formal Bayesian analysis can be computationally challenging, widely available computer programs now greatly reduce this burden. Participants identified several factors that may impede the dissemination of Bayesian approaches among practitioners of dose–response assessment and made some recommendations for overcoming these hurdles. EPA, other regulatory agencies that use dose–response assessment as part of their processes, and the private sector all should take steps to foster the use of Bayesian approaches. EPA and other agencies should work to persuade professional societies (for example, Society for Risk Analysis, Society of Toxicology) to seek out and recognize meritorious analyses that use Bayesian approaches. EPA and private-sector organizations should consider sponsoring research into using Bayesian approaches, demonstration analyses that use them, and using the results of this work to help educate peers in the risk analysis and toxicology professions. EPA should request all staff and contractor scientists who develop mathematical models to use Bayesian techniques to calibrate models. EPA should consider ways to inform its staff, contractors, and the research community as to the utility of Bayesian analyses. EPA should consider improving its research planning by making use of Bayesian techniques (including value-of-information analyses).