We propose to study key issues in the design and analysis of the carcinogen bioassay in small rodents using empirical evidence and methodologic development. We focus on 5 key dimensions: prechronic studies, determining the experimental unit, determining when to sacrifice, determining and use of cause of death, and combining evidence over tumor types, sites, across genders and strains, and over chemicals. For the prechronic studies we plan to improve determination of the (MTD) by developing sequential dosing schemes and models of bioequivalence for single and repeated doses. Most analyses of the bioassay use the single rodent as unit. If the cage effect is large (rodents in the same cage responding similarly), then the unit should be the cage. We shall estimate random effects models to produce cage effects and make recommendations on single or multiple caging and methods of analysis. This approach will be extended to litter and multigenerational effects. We shall study the benefits and drawbacks of using cause of death information, make recommendations on its determination, and develop dose group comparisons that do not depend strongly on cause of death. We propose to study combining evidence over tumor types and sites for a single rodent, and over gender and strain for a single experiment using the analysis of covariance. The former will control the problem of multiple comparisons, and the latter has potential to increase statistical power over a stratified analysis. We shall employ empirical Bayes methods to combine evidence over chemicals. This research will improve the validity, efficiency, and credibility of the bioassay, thus improving the risk assessment process and the public health.
