The Benchmark Approach for Continuous Endpoints: A Proposal for a Convenient Class of Models. W. Slob, National Institute of Public Health and Environment (RIVM), Laboratory for Health Effects Research, P.O. Box 1, 3720 BA Bilthoven, The Netherlands
In the absence of biologically-based models, most observed dose-response data can only be described by arbitrary regression functions, the only requirement being that the model fits the data reasonably well. Although choosing a regression model seems largely an arbitrary matter, this is not entirely the case: to satisfy several desirable features, the family of suitable models appears to be restricted. A family of models will be discussed having these desirable features. For example, the proposed class of models is able to efficiently and transparently describe two (or more) populations (e.g. males and females in a rat study) that differ in background response (e.g. body weight), but are equally sensitive to the chemical. When the populations do differ in sensitivity to the chemical, this may be expressed by a parameter that is directly linked to the concept of uncertainty factor used for inter- and intraspecies extrapolation. A practical advantage of working with this class of models is that the choice of the regression model to be used for deriving the benchmark dose, can to a substantial degree be standardized: without that different risk assessors may easily derive different benchmark doses from the same data set. Several examples will illustrate how these models work in practice.
Go to . . .