Benchmark Dose Calculation for Ordinal Responses Using Categorical Regression

• Activity Number: 72
• Type: Topic Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): James Chen*+ and Chu-Chih Chen
• Companies: National Center for Toxicological Research, FDA and NHRI
• Keywords: aldicarb exposure ; Bayes risk ; model uncertainty ; reference dose

Abstract

Categorical regression (CatReg) provides an approach for risk assessment of ordered (severity) responses such as no effect, mild effect, adverse effect, and severe effect. The standard CatReg can estimate the probability of response severity of a specific category for any given concentration level. However, the estimated probability of the adverse effect, and hence benchmark dose (BMD), depends on the how the adverse effect is defined. This paper proposes an integrated risk calculation method to estimate the overall adverse effect (summary risk) as a weighted combination of the probabilities of the different levels of severity. The summary risk is used to derive the BMD corresponding to a given benchmark response (BMR); the reference dose (RfD) of the CATREG approach can then be obtained accordingly. The procedure is illustrated by the analysis of toxicity data on aldicarb and phosgene exposure.