All Times EDT
Keywords: Win ratio, Composite endpoint, Survival analysis, Subgroup, Heterogenous
In clinical studies, it is common to encounter a situation where the treatment effect is influenced by different subgroups defined by a specific baseline demographic or clinical characteristics. These subgroups characterize the heterogeneity of the disease and particularly its impact the endpoint of interest. For example, it was known that mortality (time to death) and disease progression of COVID-19 and cardiovascular disease varies between gender and age group (ref). Similarly, the recovery (time to recovery) in varies between milder symptoms group compared to the severe symptoms group. These observations create challenges in choosing the time-to-event endpoint and in making a reasonable assumption for sample size estimation. Furthermore, it may complicate the analysis method due to non-proportionality of the underlying hazard. In such cases, a composite endpoint could be considered which can provide a higher event rate and more statistical power resulting in smaller sample size and shorter study duration. Win ratio is a non-parametric methodology that allows handling of composite endpoint with consideration of the clinical importance of the endpoints. In this research work, we will evaluate the subgroup effect and its impact to analysis of study result. Both unmatched and matched pair win ratio approach will be explored to compare to clinical logrank test and Cox proportional model. The empirical power and type I error rate will be reported to compare each method.