Use Multiple Imputation to Handle Missing Data in Longitudinal Clinical Trials with Multiple Correlated Endpoints

• Activity Number: 344
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yahong Peng and Lian Liu*+ and Ruifeng Xu
• Companies: Pfizer Inc. and Roche Product Development in Asia Pacific and Merck Research Laboratories
• Keywords: clinical trial ; missing data ; multiple imputation ; sensitivity analysis

Abstract

Missing data are very common in longitudinal clinical trials. EMEA Guideline on Missing Data in Confirmatory Clinical Trials (draft, April 2009) asks clinical trial sponsors to plan and describe sensitivity analysis in clinical study protocol/data analysis plan to assess the impacts of missing data. Multiple imputation (Rubin, 1987) has been broadly used as a powerful and yet flexible tool to handle missing data under sensitivity analysis framework in clinical trial. We explore the performance of multiple imputation method in longitudinal clinical trials with multiple endpoints moderately correlated with each other but with different missing rates, as well as trials with missing data in the item level of composite-score variables. Extensive simulation study and real data applications are used to demonstrate when multiple imputation method will be successful.