You are here:
Lessons Learned - Modeling Cancer Data
Citation:
SETZER, R. W. Lessons Learned - Modeling Cancer Data. Presented at The International Life Science Institute Europe Workshop on the Application of Margin of Exposure (MoE) Approach to Compounds in Food which are both Geneotoxic and Carcinogenic, Rhodes, GREECE, October 01 - 03, 2008.
Impact/Purpose:
A novel approach (“average model”) was used for the modeling that evaluates the uncertainty in this benchmark dose (BMD) due to both statistical uncertainty and the fact that alternative models may adequately fit the data yet yield different BMD estimates. BMDs were derived for responses of 10%, 5%, and 1% risk, and relative uncertainties of each estimate compared. The question often asked by regulatory toxicologists, “Can the data be modeled?” can be turned on its head, to “What does modeling the data tell us about data quality?” since attempting to model a data set reveals its limitations quite sharply.
Description:
The primary goal of modeling in this exercise is to derive a single point of departure (POD) for use in calculating a margin of exposure (MOE) estimate, and evaluate its uncertainty. For the purpose, empirical models (models that capture the shape of the dose-response in the data range, but which are not derived from mechanistic biological understanding) are fit to the data using statistical methodology, and used to infer the dose associated with a predefined increase in tumor incidence. Multiple endpoints were modeled for each chemical, as long as there was a significantly increasing trend of tumor incidence with dose.
URLs/Downloads:
Lessons Learned - Modeling Cander Data (PDF, NA pp, 9 KB, about PDF)Lessons Learned - Modeling Cancer Data (slide) (PDF, NA pp, 243 KB, about PDF)