Estimate of Treatment Difference for Non-Normally Distributed Data in Clinical Trials – Comparison of Hodges-Lehmann Method and Quantile Regression

• Activity Number: 355 - Contributed Poster Presentations: Biopharmaceutical Section
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Youlan Rao* and Yonggang Yao and Lisa Edwards and Chunqin Deng
• Companies: United Therapeutics Corporation and SAS Institute Inc and United Therapeutics Corporation and United Therapeutics Corporation
• Keywords: Quantile Regression; Hodges-Lehmann’s method; Classic Rank Methods; Non-normally Distributed Data; Clinical Trials

Abstract

Currently, classic rank-based methods are used to analyze non-normal data collected in clinical trials for regulatory approval. While these rank-based methods, such as the Wilcoxon rank sum test and the Hodges-Lehmann’s estimator, have historically been used to detect location shift and estimate the median treatment difference, they may not detect or estimate differences between other quantiles of the individual treatment group distributions. For example, two groups may have similar medians but differ in other quantiles. Quantile regression is a modern statistical methodology for modeling quantiles. The benefit of median (quantile) regression is to provide a more accurate estimate for difference in medians (quantiles) between two treatment groups as evidenced by similar magnitude in the modeled difference in medians (quantiles) and the raw difference between group medians (quantiles). Classic rank-based methods and quantile regression are compared in the analyses of non-normally distributed continuous variables (6-minute walk test and NT-proBNP values) and ordinal categorical variables (BORG scores and WHO functional class) from Pulmonary Arterial Hypertension clinical trials.