Issues Related to Benchmark Dose Modeling. L. T. Haber and B. C. Allen, ICF Kaiser International, 9300 Lee Highway, Fairfax, VA 22031; and C. A. Kimmel, USEPA, 401 M St. SW, Washington, DC 20460
A series of case studies were conducted to investigate the application of the benchmark dose (BMD) methodology to individual data sets. Methods and implications of incorporating biological significance into the definition of the benchmark response (BMR) for modeling of continuous endpoints were considered. As an example, different definitions of the BMR for changes in body weight were compared. BMRs were defined as (1) a 10% decrease in mean body weight, and (2) 10% increase in the probability of an adversely low weight (with adverse defined as 10% below the concurrent control mean, and the BMD calculated using the hybrid continuous model). BMDs calculated using the latter definition are lower than those calculated using the former definition. This is because approximately 50% of the animals will be affected (approximately 50% increased risk) in order for a 10% decrease in the mean to result, compared with a 10% increased risk for the latter definition. Furthermore, BMDs calculated based on a probability of response were closer to the corresponding NOAELs, while the BMDs calculated based on change in the mean were closer to the corresponding LOAELs. This example illustrates the ambiguity inherent in even moderately well-accepted definitions of biological significance. The definition of an adverse response for this and other endpoints, and the most appropriate definition of the BMR, are key issues for the application of BMD methodology to continuous endpoints. For other endpoints, such as changes in pulmonary function, different estimates of the background rate of response were explored.