Subgroup Analyzes from Multiple Imputed Data Sets

• Activity Number: 505 - Missing Data and Multiple Imputation in Clinical Trials
• Type: Contributed
• Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Tianyue Zhou* and Zhiying Qiu and Meehyung Cho and Hui Quan
• Companies: and Sanofi and Sanofi and Sanofi
• Keywords: subgroup analyses ; multiple imputation ; interaction test ; simulation

Abstract

For handling missing data in confirmatory trials, recently a method using multiple imputation under the assumption of missing not at random has been more increasingly implemented as the primary method. When the primary method is based on a multiple imputation approach, for conducting subgroup analyses we often use the same imputation model and imputed data as the one for the primary endpoint analysis. This may introduce bias to the estimate of treatment effect among the subgroups and affect the test of the treatment by subgroup interaction. We will examine the robustness of this approach and compare several methods for testing the treatment by subgroup interaction. We consider different missing data scenarios with various missing data mechanisms and missing rate for the subgroups for a continuous variable in simulated data sets and apply to a real data example.