Causal Multiple Comparisons for Survival Data

• Activity Number: 244
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Hong Zhu*+ and Bo Lu
• Companies: The Ohio State University and The Ohio State University
• Keywords: Causal inference ; Multiple comparisons ; Propensity score stratification ; Simultaneous confidence intervals

Abstract

We consider the practical problem in non-randomized clinical trials and observational studies where multiple treatment or prognostic groups are compared and the observed survival data are subject to random right censoring. Two possible formulations of multiple comparisons are suggested. Multiple Comparisons with a Control compare every other group to a control group with respect to survival outcomes, for determining which groups are associated with less risk than the control. Multiple Comparisons with the Best compare each group to the truly most effective or minimum-risk group and identify the groups that are either with the minimum risk. Due to confounding effect, the data need to be adjusted to make causal comparisons. The causal multiple comparison procedures are developed based on a propensity-score-stratified Cox proportional hazards model. We extend the approaches of MCC test and MCB simultaneous confidence intervals for normal error general linear models to our setting. Tthe proposed methods are applied to two real datasets from leukemia and cancer studies for illustration.