Recent Development of Missing Data Imputation for Longitudinal Binary Outcomes

• Activity Number: 646
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Huyuan Yang*+ and Serap Sankoh and Jing Xu and Ming-Xiu Hu
• Companies: Takeda and Takeda and Takeda and Takeda
• Keywords: multiple imputation ; longitudinal binary outcome

Abstract

Missing data issue has been more and more important in regulatory submissions. It is very common that sponsors got questioned through regulatory requests or at adversary committee meetings regarding how to deal with missing data in the analysis. Besides posthoc sensitivity analysis, missing data imputation started to be applied as part of pre-planned analyses in industry.

How to reassess the efficacy and demonstrate the robustness of treatment effect with scientific sound methods is a challenging task for statisticians. Examples have been given in our recent regulatory submissions. For binary outcomes, in pre-planned analyses, ITT principle as observed or single imputation are more often used in clinical trials. In this presentation, different methods including multiple imputations methods were exercised to conduct as sensitivity analyses to single imputation. Results show similar treatment effect and demonstrate the drug efficacy. Statistical inference based on p-values and confidence intervals of adjusted clinical endpoints take into consideration both within-imputation and between-imputation variance. Analyses for other endpoints will also be discussed.