This project addresses statistical issues in the design, analysis and interpretation of rodent tumorigenicity and teratology experiments, both of which play a prominent role in safety testing and risk assessment. Rodent tumorigenicity experiments involve a control group and several groups exposed to a suspected carcinogen. Special methods of analysis are needed sine most tumor types are undetectable prior to autopsy, and because many test substances have toxic as well as carcinogenic effects. The proposed research includes empirical and analytical investigations or methods that rely on cause-of-death data. In addition, analytic methods will be developed to assess the impact of multiple tumors on the reliability of cause-of-death determinations. A major component of the proposed work involves the development of a new approach that avoids the subjectivity of cause-of-death by incorporating pathology information directly into the analysis. Although detailed pathology data, such as tumor type and grade, are generally available, no methods currently exist to use such data in the statistical analysis. Attention will also focus on the development of simple, interpretable measures of carcinogenic potency, which are useful for comparing and ranking carcinogens. Teratology experiments provide valuable data for assessing the risk associated with suspected developmental toxins. Typically, female mice are exposed to the test substance early in pregnancy and their offspring are examined for defects. The statistical analysis of these experiments is complicated by intra-litter correlations and the many stages of fetal development that can be affected by exposure. An important component of the proposed research concerns the development of suitable dose response models. Related issues to be addressed include low dose extrapolation and estimation of teratogenic potency. Finally, multivariate models will be developed to analyze the multiple outcomes typically measured on each offspring. The project's broad aim is to develop practical and sound statistical designs and analyses, which lead to reliable risk assessment for carcinogens and teratogens.
| Cai, Tianxi; Parast, Layla; Ryan, Louise (2010) Meta-analysis for rare events. Stat Med 29:2078-89 |
