Statistical Methods in Evidence-based Risk Assessment

点击翻译按钮获取中文摘要

The major research theme underlying the proposal is value-of-information (VOI) analysis. In preliminary work, a new Bayesian framework for VOI analysis has been developed, allowing trade-offs between the cost, uncertainty, and timeliness of data collected in support of risk assessment. An important component of this new framework is the explicit characterization of the cost of delay in acquiring information on critical population health risks, which can result in a significant public health burden, pending identification and elimination or reduction of those risks. The framework considers two risk decision making paradigms: a benefit-risk decision maker (BRDM), who seeks to minimize the total social cost associated with the risk issue of concern (with adverse population health impacts valued in economic terms), and a target-risk decision maker (TRDM), who seeks to ensure risk is below a pre-determined level. Initial applications of the VOI framework in testing large numbers of chemicals to which humans are exposed suggests that the use of so-called new approach methodologies (NAMs) based on high-throughput in vitro screening (HTS) tests will provide a greater return on investment (ROI) than traditional animal-based toxicity tests, which, although subject to less uncertainty, are much more expensive and time-consuming to conduct. Two additional research themes will also be pursued as part of this proposal. First, novel points of departure (PoD's) on the exposure-response curve will be examined as the basis for establishing human exposure guidelines, specifically the signal-to-noise crossover dose (SNCD) corresponding to the point where the biological signal cannot be statistically distinguished from the background response rate. Second, optimal experimental designs for estimating the SNCD will be developed in order to obtain the best possible human exposure guidelines. Collectively, this research will greatly strengthen evidence-based risk assessment practices through the application of VOI analyses to identify the most informative sources of data to support human health risk assessment; by developing improved human exposure guidelines through the use of more refined exposure-response analyses; and by developing optimal experimental designs to maximize the statistical information about the shape of the exposure-response curve.