In most randomized clinical trials evaluating intervention effects on time-to-event outcomes, the log-rank test combined with Cox's hazard ratio (HR) estimation has been the standard test/estimation approach for decades. While this traditional framework offers several advantages, concerns have been raised about HR's limitations as a summary measure of the magnitude of intervention effects. To address these concerns, an alternative approach based on the average hazard (AH) has been proposed. AH is defined as a functional of the survival function of the event time distribution and is interpreted as an average incidence rate over a given time window. Although its theoretical properties and nonparametric inference procedures have already been established, practical guidance for its use in the primary analysis of randomized clinical trials (RCTs) remains limited. In this paper, we address two key aspects critical to the implementation of AH in RCTs: (1) design considerations when the AH-based analysis is used for the primary analysis, and (2) tools for identifying the appropriate timing of the analysis. We provide a framework for study design using AH and introduce a new R package, survAHtools, to support its practical application in RCTs.
Horiguchi et al. (Mon,) studied this question.