A hybrid approach of handling missing data under different missing data mechanisms: VISIBLE 1 and VARSITY trials for ulcerative colitis

Abstract

Missing data is common in clinical trials. Depending on the volume and nature of missing data, it may reduce statistical power for detecting treatment difference, introduce potential bias and invalidate conclusions. Non-responder imputation (NRI), where patients with missing information to determine endpoint status are considered as treatment failures, is widely used to handle missing data for dichotomous efficacy endpoints in inflammatory bowel disease (IBD) trials. However, it does not consider the mechanisms leading to missing data and can potentially underestimate the treatment effect. We proposed a hybrid imputation approach combining NRI and multiple imputation (MI) as an alternative to NRI to assess the impact of dropouts for different missing data mechanisms (categorized as "missing not at random [MNAR]" and "missing at random [MAR]"). Two phase 3 vedolizumab clinical trials under different study designs in patients with moderate-to-severe ulcerative colitis (UC), VISIBLE 1 and VARSITY, are presented to illustrate how the proposed hybrid approach can be implemented as a pre-specified sensitivity analysis in practice. The proposed hybrid imputation provided consistent efficacy results with those using NRI, and can serve as a useful pre-specified sensitivity analysis to assess the impact of dropouts under different missing data mechanisms and evaluate the robustness of efficacy conclusions.

Keywords: Inflammatory bowel disease; Missing data; Missing data mechanisms; Multiple imputation; Non-responder imputation; Ulcerative colitis.