Evaluating Treatment Efficacy by Combining Multiple Measures in Clinical Trial Applications

A variety of clinical and laboratory measures can be used in clinical trials to assess the benefit of a new treatment over the standard of care. Data from clinical studies are often analyzed by combining individual outcomes into one primary outcome. That primary outcome is then referred to as a composite endpoint or a combined endpoint. We propose an analysis on the composite endpoint with Gehan's (1965) ranking approach where each subject in the treatment group is compared with each subject in the control group in a pair-wise manner. Our approach reduces computational time and complexity to construct a subject-level pairwise composite score. We develop a statistical testing procedure for the analysis of composite endpoints when using the hierarchical scores. In this article, we propose two tests (a parametric test and a non-parametric bootstrap procedure) for evaluating the effect of treatment. The proposed parametric test has an asymptotic F-distribution based on standard statistical assumptions. We conduct an extensive simulation study to assess the operating characteristics of the proposed methods and to compare them with an existing method. We illustrate the methods using publicly available data from two clinical studies.