This project is concerned with statistical methodology issues in the design, analysis, and interpretation of laboratory animal experiments, especially the long term rodent carcinogenicity studies carried out by the National Toxicology Program (NTP). One investigation used data from the NTP dietary restriction study to confirm body weight/ tumor incidence associations and to illustrate how these associations may reduce the sensitivity of the bioassay for detecting carcinogenic effects. Tumor incidence/body weight correlations were also shown to explain many of the chemically-related decreased tumor incidences observed in some NTP studies. Using NTP historical control data, mathematical models were developed to predict control site-specific tumor rates as a function of age at death and body weight at one year of age. The knowledge gained through this research will influence the design of future NTP bioassays. Another investigation developed new statistical methodology based on the constant risk difference (CRD) model to analyze tumor incidence data from animal experiments. One important advantage of this approach is that it focuses on the tumor onset distribution and yet does not require interim sacrifices, cause-of-death data, or lethality assumptions. The CRD analysis was applied to several NTP studies and it performed well relative to the standard analyses. Future research will deal increasingly with non-cancer endpoints and with mechanistic data.
