This project is concerned with statistical methodology issues involved in the design, analysis, and interpretation of laboratory animal experiments, with particular emphasis on two-year carcinogenicity studies. Experimental design issues include examining potential sources of variability in tumor incidence. Inter-laboratory differences and time-related trends were found to be significant sources of variability. Future research plans include a re-evaluation of tumor diagnoses from early NCI and current NTP studies, to identify the specific factors responsible for this variability. Research in the area of data analysis has focused on the development of statistical procedures for evaluating tumor data that do not require cause of death information or numerous interim kills. Particularly promising is a procedure that weights the contribution of animals that die tumor-free based upon a hypothesized form of the tumor onset distribution. The resulting procedure is a simple modification of the Cochran-Armitage trend test and may ultimately result in an important advance in the statistical evaluation of tumor data. An evaluation of the relationship between toxicity, genotoxicity, and carcinogenicity in laboratory rodents indicated that only a relatively small proportion (10-15%) of positive carcinogenicity studies exhibited the types of target organ toxicity that could have been the cause of all observed carcinogenic effects. Furthermore, no apparent difference in mutagenicity was observed between """"""""high dose only"""""""" carcinogens and the entire set of carcinogens.
