|
Abstract:
|
Missing data occur in almost all longitudinal clinical trials despite all the efforts made to retain patients in the trial. Properly analyzing trial data with appropriate methods to handle missing data is critical when evaluating the between-group treatment effect. Commonly used approaches for longitudinal data such as Mixed Model with Repeated Measures assume missing at random (MAR). However, the MAR assumption is often difficult to verify or justify in practice. Therefore, methods under missing not at random (MNAR) seem to be more reasonable and are frequently requested by regulatory authorities. In this presentation, we will evaluate two methods to estimate treatment effect under MNAR, the Jump to Reference (J2R) and Return to Baseline (RTB). Different approaches are proposed for J2R and RTB, such as multiple imputation (MI) and likelihood-based analytic approaches. Simulation studies will be presented to show the performances of the approaches.
|
