Statistical Considerations and Challenges with Time-to-Event Analyses for Composite Endpoints in Clinical Trials

Abstract

The use of composite endpoint is a common strategy often employed to enhance statistical power and address the low incidence of individual outcomes, particularly in cardiovascular and kidney outcome studies. By merging multiple clinically relevant events into a single variable, these endpoints negate the need for multiple testing adjustments and augment the event rate, thus enabling studies of reasonable size and duration. However, as underscored by the FDA's guidance, a thorough evaluation of each component's impact is equally important to ensure the clinical relevance of these endpoints. This article delves into controversies surrounding the interpretation of hazard ratios derived from analyzing the composite endpoint and its individual components, exemplified by an observation from the CLEAR outcome trials. It highlights a paradoxical scenario where the combined treatment effect for the composite endpoint appeared less favorable than when assessing individual components separately. Moreover, we did a re-evaluation of the suitability of using Cox proportional hazards model in this context through theoretical investigation and simulation studies.

Keywords: Clinical trials; Composite endpoint; Proportional hazards; Time-to-event.