Estimating Treatment Effects from a Randomized Trial in the Presence of Post-Study Treatment

• Activity Number: 569
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Min Zhang*+ and Yanping Wang
• Companies: University of Michigan and Eli Lilly and Company
• Keywords: Causal inference ; Inverse probability weighting ; Covariate adjustment ; Cancer ; Marginal structural model ; Clinical trial

Abstract

In randomized clinical trials involving survival time, a challenge that arises frequently, for example in cancer studies, is that during follow up subjects may initiate post-study treatment (PST), which is not part of the study design and could potentially confound effect of the study drug. Marginal structural Cox's model and methods based on inverse probability weighting have been proposed to account for PST in the presence of time-dependent confounders. Inverse probability weighting methods tend to yield estimators that are of big variance and not stable. These methods were originally proposed for observational data and do not target the question of interest. In this paper, we adopt the marginal structural Cox's model and propose an augmented inverse probability weighting method. The proposed method improves the efficiency of the usual methods by exploiting the fact that the study treatment is independent of baseline covariates, guaranteed by randomization. The finite-sample performance of the proposed method is demonstrated via simulation studies and by application to data from a cancer clinical trial.