Society has long been concerned with protecting itself against exposures to chemicals, drugs or other substances that cause cancer, birth defects and reproductive problems. Limitations in the reliability and availability of epidemiological information has led to reliance on the use of controlled experiments in laboratory animals. The general purpose of this proposal is to address statistical issues in the design and analysis of rodent carcinogenicity and developmental and reproductive toxicity (teratology) experiments. Several of the specific aims are motivated by problems in both areas of application, while some are unique to carcinogenicity testing or developmental and reproductive toxicology.
Specific aims are: 1. Investigations of correlated multinomial models to analyze hierarchically related outcomes from developmental and reproductive toxicity. 2. Development of a multivariate dose response model for teratology that incorporates dose effects on fetal death, malformation and weight. 3. Evaluations and extensions of existing methods for fitting three-state models for carcinogenicity data. 4. Theoretical and simulation studies of recently proposed methods to analyze clustered and/or multiple outcome data. 5. Development of methods to incorporate historical control information into meta analysis for rare events. Empirical data analysis will play a central role in all five specific aims. The proposed research will make methodological contributions to the fields of survival analysis, analysis of 3-state illness-death models, generalized estimating equations for correlated data and the use of historical controls in laboratory studies and epidemiological studies.
| Cai, Tianxi; Parast, Layla; Ryan, Louise (2010) Meta-analysis for rare events. Stat Med 29:2078-89 |
| Horton, Nicholas J; Roberts, Kevin; Ryan, Louise et al. (2008) A maximum likelihood latent variable regression model for multiple informants. Stat Med 27:4992-5004 |
