Win ratio: A new technique for the analysis of multiple events in clinical trials

The win ratio is a relatively recent statistical technique introduced with the aim of better managing the analysis of clinical trials involving more than one clinical event as an endpoint in the evaluation of a treatment. The calculation of the win ratio begins by defining a “hierarchy” of clinical events on the basis of their severity, e.g. death, followed, for example, by the number of hospitalizations for heart failure, followed, for example, by softer endpoints including functional or laboratory changes. The analysis begins by comparing each patient in a hypothetical “Group A” with each patient in a hypothetical “Group B” on the hierarchically most important clinical event only. If the patient in Group B dies and the one in Group A does not, or if the one in Group B dies before the one in Group A, then that particular comparison is “won” by Group A. If the patient in Group A dies and the one in Group B does not, or if the one in Group A dies before the one in Group B, then that comparison is “won” by Group B. If nobody dies, or if they die on the same day, then that specific comparison ends in a tie. All comparisons of each patient in Group A with each patient in Group B are then performed. On the comparisons that end in a tie, we move on to the analysis of the endpoint hierarchically in second place, using the same technique. Then we proceed to the analysis of the endpoint hierarchically in third place, and so on down to the event in the lowest position in the hierarchy. The win ratio represents the total sum of comparisons in which, for example, Group A wins, divided by the total sum of comparisons in which Group A “loses”. The absolute difference, rather than the ratio, between these two sums indicates the “win difference”. Compared to the conventional “time-to-event” statistical techniques including, for example, the Cox model, the win ratio calculation has advantages, but also potential disadvantages. This review aims to offer a summary of the advantages and potential disadvantages of the win ratio, with some practical examples derived from the use of the win ratio in the analysis of important trials in the cardiovascular area.