The applicant plans to study four statistical problems frequently encountered in cancer clinical trials and observational studies. These are: 1. nonproportional hazards models for failure time data, 2. analysis of competing risks failure time data, 3. analysis of multivariate survival time data, and 4. analysis of repeated measures in the presence of dependent censoring For item 1., the applicant notes that the Cox model is the most popular model for analyzing censored observations. However, it often does not fit the data well. The applicant will continue to develop alternatives to the Cox model. Under topic 2., the applicant notes that in the presence of dependent competing risks, the Cox model can be used to examine the covariate effects on the cause-specific hazard function. It is noted, however, that in this setting very little has been done on predicting survival probabilities for patients with specific covariates. The applicant plans to work on this problem and to develop nonproportional hazards models to handle competing risks failure time data. Under item 3., the analysis of multivariate survival time data, the applicant plans to develop robust methods for analyzing recurrent event time and multi-state data. He also plans to investigate analyses for interval-censored count data. For item 4., the analysis of repeated measures in the presence of dependent censoring, the applicant notes that repeated cancer marker measurements have been used to identify and/or define disease progression in modern cancer studies. However, if the patient s follow-up time depends on the observed or unobserved response variables, commonly used methods will not be applicable. The applicant plans to develop robust methods to handle such incomplete repeated measurements data.
Showing the most recent 10 out of 12 publications
