Can Interim + Imputation = Final?

• Activity Number: 179
• Type: Contributed
• Date/Time: Monday, August 4, 2003 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Susan Y. J. Zhou*+ and Guoxing Soon
• Companies: Food and Drug Administration and Food and Drug Administration
• Keywords: clinical trials ; HIV drugs ; interim analysis ; simulation study ; imputation

Abstract

To obtain FDA traditional approval for HIV drugs, data at or beyond 48 weeks in double-blind randomized clinical trials have been required. For the purpose of obtaining accelerated approval, an interim analysis was performed when all patients had completed 24 weeks of therapy. In a recent FDA submission for an accelerated approval, Week 48 results from an additional analysis were also reported for a fraction of subjects who had reached the final stage. Findings for Week 48 cases alone, a subset that accounts for only half the subjects, were consistent with findings for Week 24 using data from all treated patients. Based on favorable results, a question was asked whether or not these data were sufficient to make a regulatory decision for a traditional approval. Several approaches to this problem were examined. A simulation study based on empirical Bayesian method was employed to impute success rates for subjects who had not completed the 48 weeks of study. Alternative approaches, such as multiple imputations, were also utilized to assess the suitability for a regulatory decision. Assumptions concerning different approaches will be discussed, and results presented.