This project is concerned with statistical methodology issues in the design, analysis, and interpretation of laboratory animal experiments. One investigation evaluated the relationship between maximum tolerated dose (MTD) and study sensitivity for detecting rodent carcinogenicity for 216 rodent carcinogens. Approximately 2/3 of these carcinogens would have been detected even without the top dose, Nearly 80% of the remaining rodent carcinogens had numerically (but not significantly) elevated rates of the same site-specific tumors at lower doses, This evaluation indicates that most carcinogenic effects observed at the MTD in rodent studies are also present (with reduced incidence) at the lower doses (1/2 MTD, 1/4 MTD) typically employed. In another investigation, a large computer simulation was conducted to assess the operating characteristics of various trend tests for comparing incidence rates of occult tumors. Data were generated under 88 different scenarios in order to simulate a wide range of realistic animal carcinogenic experiments. Under a variety of experimental conditions and assumed underlying dose-response models, the procedure that consistently performed best (i.e., operated at the correct level of significance and had greatest power) was the survival- adjusted test based on risk differences, a procedure that does not require information on cause of death or tumor lethality. Future research efforts include an evaluation of the various factors affecting tumor incidence, with particular emphasis on associations with body weight and the effect of multiple caging. Future research will also be increasingly directed toward non-cancer endpoints.
