Composite primary endpoints, when combining multiple time-to-event outcomes, are common in randomized clinical trials (RCT). Traditional time-to-first-event analyses are conducted; however, they may neglect the hierarchical importance of individual components and the correlations between events. Two alternative approaches—Win Ratio (WR) and joint frailty models (JFM)—can address these issues by improving the assessment of treatment effects in trials using composite endpoints. The WR is a recent nonparametric method that evaluates the treatment effect on a hierarchical composite endpoint and can be interpreted as the inverse of the hazard ratio under specific conditions. JFMs provide a more in-depth analysis by jointly modeling terminal and non-terminal events, while accounting for both individual variability and the correlation between event types. This work aims to give a perspective and comparison (with advantages and issues) on the use of the WR or an adaptation of the use of JFM for calculation of sample size or power when designing a study under certain assumptions.

First, simulation studies were conducted to compare the performance of WR and JFM in estimating treatment effects. The WR was applied to preserve clinical hierarchy when comparing treatments using composite endpoints. The WR approach was compared to the JFMs that were then fitted to jointly model non-terminal events and terminal outcomes, allowing estimation of correlated failure times and informing both on treatment effect estimation and sample size calculations. We applied these methods to two RCTs in oncology (GBC trial) and cardiology (PROGRESS trial). Finally, post-hoc power analyses and sample size estimations were performed using parameters estimated from these datasets as pilot data.

The WR yielded effect estimates comparable to standard Cox models while offering more nuanced clinical interpretations. JFMs enabled separate yet correlated treatment-effect estimations for non-terminal and terminal events, leading to refined sample size calculations. When applied to the GBC and PROGRESS trials, JFMs and WR both suggested that fewer participants could be required, compared with standard approaches under comparable settings.

WR and JFMs for composite endpoints can enhance the interpretability of treatment effects and produce more precise sample size estimates, while offering a more comprehensive view of clinical outcomes. These methods are promising for optimizing trial design and analysis, particularly when composite endpoints and correlated event structures are key to a study's objectives.