Estimation of Subgroups in Randomized Clinical Trials

• Activity Number: 442
• Type: Invited
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Jeremy Michael George Taylor*+ and Jared Foster
• Companies: University of Michigan and University of Michigan
• Keywords: Subgroups ; RCT ; Potential Outcomes

Abstract

We consider the problem of identifying subgroups of patients who may have an enhanced treatment effect in a randomized clinical trial, and it is desirable that the subgroup be defined by a small number of covariates. For this problem, the development of a standard, pre-determined strategy may help to avoid the well-known dangers of subgroup analysis. A two stage method is developed to find a subgroup. This method involves considering what the outcome would have been if the subject had taken the other treatment. In stage 1 the expected response is estimated under both treatment and control scenarios for each subject. In stage 2 the difference between these predicted outcomes is used as the response variable in a regression tree, which can potentially include any set of the covariates. The terminal nodes of the tree with large predicted values define the subgroup. The difference between the treatment effect in the estimated subgroup and the marginal treatment effect is a measure of the quality of the subgroup. Several resampling based methods are investigated to give honest estimates of this measure and p-values for whether a subgroup exists.