A Comparison of MMRM and Other Methods for Handling Dropouts in a Cystic Fibrosis Clinical Trial

• Activity Number: 64
• Type: Contributed
• Date/Time: Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Tao Song*+ and Abdul J. Sankoh and Qunming Dong and Cynthia M. DeSouza
• Companies: Vertex Pharmaceuticals, Inc. and Vertex Pharmaceuticals, Inc. and Vertex Pharmaceuticals, Inc. and Vertex Pharmaceuticals, Inc.
• Keywords: Missing data ; Missing not at random ; Mixed model repeated measures

Abstract

A common problem in comparative longitudinal clinical trials is the occurrence of missing data due to premature patient withdrawal from the trial before its planned completion. The presence of missing data could lead to reduced power and bias in estimating treatment effect. Though there are no universally acceptable statistical methods for handling missing data, mixed model repeated measures (MMRM) is often preferred under random dropout assumptions. When dropout mechanism is assumed missing not at random (MNAR), joint modeling inference of the response and dropout mechanism to assess robustness of the analysis method is often used. We evaluate the performance of the MMRM and other commonly used statistical methods in the presence of MNAR in a Cystic Fibrosis trial. Simulation studies are conducted to investigate the properties of the selected methods under MNAR dropout assumption.