Statistical power considerations in the use of win ratio in cardiovascular outcome trials

In cardiovascular outcome trials, the win ratio (WR) method models the composite endpoint under a hierarchical structure to account for clinical priorities. It also can be applied to both survival and nonsurvival outcomes.

In this article, we assess the performance of the WR method via extensive simulation studies and real data analyses and discuss power considerations of the method with respect to hierarchical order, variable type, magnitude of treatment effect, and event rates when applied to clinical studies.

In the hierarchy of the WR method, the first-ordered component (e.g., death) plays a dominant role in statistical power, especially when that component has a large treatment effect and a high event rate. This is in contrast with the score test of the Cox proportional hazards model in which the power is more likely affected by the nonfatal events that are usually observed earlier. Furthermore, when adding an additional component to the composite endpoint, the performance of the WR method varies depending on the treatment effect, event rate, and hierarchical position of the component. If the additional component has a relatively smaller or no treatment effect, the statistical power will decrease; if the additional component has a relatively larger treatment effect and higher event rate, the statistical power will increase. When adding a nonsurvival continuous outcome (e.g., 6-min walk distance) with even a tiny treatment effect, the statistical power could dramatically increase.