Analysis of Exposure Concentration-Duration-Effect Relationships Using Categorical Regression. Daniel J. Guth, National Center for Environmental Assessment, U.S. EPA, Research Triangle Park, NC 27711
Exposure-response analysis typically focuses on exposure concentration to explain toxicity while duration is not explicitly treated as a variable, except by restricting the analysis to a specific duration range (e.g., acute, subchronic, chronic). Such restriction limits the data that can be included in the analysis as well as the applicability of the results, while extrapolation across durations may still be required to tailor the estimate to the intended human exposure scenario. These limitations are more severe for analysis of acute duration studies (defined as <24 h) because there may be multiple experiments of different duration and because the role of duration in toxicity is critical at short times. Explicit consideration of duration in the exposure-response model is preferable. Among 400 acute exposure-response experiments for 45 chemicals, only 6 experiments varied both concentration and duration. Thus, a general approach for exposure (concentration, duration)-response modeling requires combination of data. Categorical regression is a technique that addresses many of the difficulties in combination of dissimilar studies by analyzing responses on a common scale of severity categories. A stratified model, which allows some model parameters to vary across data subsets (strata) defined, e.g., by species, is useful to evaluate similarities in subsets of the data. Differences between strata may result from different kinetics or sensitivity, while similarities, e.g., between males and females, or rats and mice, would support data combination. A difficulty in combining dissimilar studies is the consistent interpretation of effect severity across different endpoints and experimental methods. The ability to treat partial or uncertain severity information with categorical regression allows evaluation of the sensitivity of the approach to the assignment of severity categories. Severity categories can be defined flexibly, based on the risk assessment application. Categorical regression is illustrated as a valuable statistical tool for analysis of exposure (concentration, duration) effect relationships. Results of analyses of methyl isocyanate and 1,1,1-trichloroethane are presented.