Handling Multiple Imputation in Wilcoxon Signed-Rank Test: A Case Study Applied to a Hemophilia Clinical Trial

• Activity Number: 29 - Statistical Issues Specific to Therapeutic Areas, Power and Sample Size Calculations, and Trial Monitoring
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Eunhee Hwang* and Joseph Cappelleri and Satrajit Roychoudhury and Dan Meyer
• Companies: Pfizer Inc and Pfizer Inc and Pfizer Inc. and Pfizer
• Keywords: rare disease, hemophilia; multiple imputation; Rubin’s rule; Wilcoxon signed-rank test; inference; coverage probability; health-related quality of life endpoints

Abstract

In a rare disease clinical trial setting we often encounter regulatory requirements to employ analysis methods to account for the small sample size as well as for missing values. Non-parametric analyses and Multiple Imputation (MI) are widely accepted as measures to address each concern. However, there is a paucity of literature on valid analyses techniques that tackle both concerns. We present such analyses combining MI and the one-sample testing procedure using Wilcoxon signed-rank test and the accompanying Hodges-Lehman location parameter estimation that can be applied to a hemophilia clinical trial. We review asymptotic behavior of the Wilcoxon signed-rank sum statistics that can be extended to MI data sets via Rubin’s rule. We propose ways of deriving the statistical inference for multiple imputed data sets. The type I error rate, power, and coverage probability of these proposals are evaluated for general robustness assessment via simulations under various scenarios. Our proposal extends to cases with substantial proportion of tied observations as in Health-related Quality of Life endpoints which steadily gain importance in patient-centric drug development.