Methods for Missing Data Handling in Randomized Clinical Trials With Nonnormal Endpoints With Application to a Phase III Clinical Trial

In randomized clinical trials, when the endpoint is the change from baseline at the last scheduled visit, various parametric, semiparametric, and nonparametric methods have been developed to handle the possible missing data due to dropouts. Although the last observation carried forward (LOCF) and the mixed-effects model for repeated measures (MMRM) have been extensively compared and widely used, they may lead to biased results when the required distributional or missing mechanism assumptions are not satisfied. Nonparametric missing data handling methods including the last rank carried forward (LRCF) and mean rank imputation (MRI) relax the underlying distributional assumptions; however, conditions for them to be valid have been investigated to a very limited extent. This article rigorously derives asymptotic properties of the MRI method and proves its validity to test the primary endpoint under certain mild distributional and missing mechanism assumptions. The test-based estimator for the location difference between the treatment and the control groups is also derived when the randomized clinical trial has two arms under a location shift assumption. The investigated methods are applied to an illustrative phase III clinical trial. Simulation studies based on the empirical distributions from the illustrative clinical trial and additional intensive simulation studies, based on various prespecified distributions and missing mechanisms, were conducted to compare the MRI method with selected traditional methods including LOCF, MMRM, and LRCF and they confirmed the better performance of the MRI method in terms of the Type I error rate control and the power under certain mild conditions.