Partly Conditional Regression to Inform Treatment Assignment Strategies

• Activity Number: 622
• Type: Invited
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Douglas E. Schaubel and Xu Shu*+ and John D. Kalbfleisch
• Companies: University of Michigan and University of Michigan and University of Michigan
• Keywords: Biased sampling ; Inverse weighting ; Subsampling ; Survival analysis ; Time-dependent treatment ; Time-varying covariates

Abstract

In clinical settings, the necessity of treatment is often measured in terms of prognosis in the absence of treatment. Along these lines, it is often of interest to compare subgroups of patients (e.g., based on underlying diagnosis) with respect to pre-treatment survival. The data structure of our interest is complicated by the following several factors, including the fact that treatment is not randomized and, rather, is assigned based on longitudinal measures strongly predictive of survival in the absence of treatment. In addition, subjects may have subintervals of time during which they are ineligible for treatment. We combine recently developed methods involving partly conditional regression, biased subsampling, and inverse weighting to evaluate various regulations governing the allocation of deceased-donor livers to patients on the liver transplant waiting list.