Placebo Multiple Imputation and Other Sensitivity Analyses for Incomplete Longitudinal Clinical Trial Data
View Presentation View Presentation
*Craig H Mallinckrodt, Eli Lilly and CO 

Keywords: Missing data, longitudinal data, sensitivity analyses

The National Research Council was commissioned by FDA to make recommendations on the prevention and treatment of missing data in clinical trials. The Panel’s report was made public in late 2010. This presentation will focus on a case study from a clinical trial in depression to illustrate the Panel’s recommendations for handling missing data from continuous endpoints. Specifically, the robustness of the a priori specified likelihood-based primary analysis will be assessed using a variety of MNAR analyses and an inclusive modeling multiple imputation approach, along with influence and residual diagnostics for the primary analysis. A novel method of imputing missing data for both the drug and placebo groups from the placebo group will be presented both as a likely conservative MNAR analysis and as an alternative to BOCF to assess effectiveness