Control-Based Imputation and Delta-Adjustment Stress Test for Missing Data Analysis in Longitudinal Clinical Trials

Sensitivity analyses using multiple imputation (MI) provide a flexible approach to assess the impact of missing data on clinical trial results. An approach that imputes missing data in the test drug group using a model built from the control group has gained attention in recent research. This control-based imputation (CBI) approach typically provides a conservative point estimate for treatment difference. However, the combined variance using Rubin's rule may over-estimate the variability. In this article, we investigate the statistical properties for some specific CBI methods, and show the relationship between CBI and delta-adjustment tipping point analysis. This relationship helps us to understand and interpret the results from CBI using MI with Rubin's rule. In addition, we propose a new sensitivity measure for assessing the robustness of the result obtained under a missing at random (MAR) assumption. Results from simulation studies and applications to longitudinal clinical trial datasets are presented as illustration.