Win-ratios have become a vogue in contemporary clinical trial analysis. It involves a paired-comparison method for analyzing hierarchical composite endpoints derived from the Finkelstein–Schoenfeld approach, which has been used extensively in modern cardiovascular trials. It has gained momentum because it can prioritize clinically most important outcomes (e.g., mortality over nonfatal events) while still incorporating downstream events or patient-centered measures within a single framework. Yet despite this broader uptake across cardiology, its application in TAVI studies has been comparatively limited, appearing mainly in scattered, largely exploratory reports, rather than as a routine, prospectively powered primary endpoint strategy in TAVI programs.
The Edoxaban versus Standard of Care and Their Effects on Clinical Outcomes in Patients Having Undergone Transcatheter Aortic Valve Implantation Atrial Fibrillation (ENVISAGE-TAVI AF) trial was an international, open-label, adjudicator-blinded study that randomized 1,426 patients with prevalent or incident atrial fibrillation after successful TAVI to edoxaban or vitamin K antagonists (VKAs). Patients were elderly (mean age 85 years) and at intermediate to high baseline risk. The primary composite outcome of death, myocardial infarction, ischemic stroke, systemic thromboembolism, valve thrombosis, or major bleeding met the prespecified criterion for noninferiority of edoxaban versus VKA, although major bleeding, primarily gastrointestinal, was more frequent with edoxaban. Treatment discontinuation was high in both groups during follow-up (edoxaban 30.2%, VKA 40.5%).
A subsequent post-hoc analysis evaluated patient expectations, convenience, treatment burden, and satisfaction using the validated Perception of Anticoagulation Treatment Questionnaire (PACT-Q), which includes a baseline expectations module (PACT-Q1) and a follow-up module (PACT-Q2) assessing convenience, burden, and treatment satisfaction. In the original least-squares mean difference analysis, expectations were similar at baseline, but patients receiving edoxaban reported significantly higher convenience and satisfaction at 3 and 12 months. This method did not examine individual item behavior or patient-level directional differences.
In a recent issue of the Journal , Hengstenberg et al reanalyze these patient-reported outcomes using a win-ratio framework. Their rationale was that least-squares mean differences lack established thresholds for clinically meaningful change and may not capture patient-centered distinctions. Using patient-pair comparisons, they report that edoxaban was associated with significantly higher convenience and satisfaction scores at 3 and 12 months, with 18 of the 20 PACT-Q2 items favoring edoxaban.
There are certain advantages to using win-ratios in analyzing clinical trials. , Conventional composite endpoints and time-to-event methods do not account for the relative importance of individual endpoints (e.g., death is more important than non-fatal events). The win ratio supports a hierarchical system for analyzing outcomes and permits the inclusion of quantitative outcomes, such as surrogate measures or patient-reported outcomes. This allows substantially more patients to contribute to the analysis of efficacy and increases statistical power.
However, several caveats warrant cautious interpretation. First, when major clinical outcomes are not significantly different between treatments, as in ENVISAGE-TAVI AF, efficacy may be driven by effects on measures of lesser or questionable clinical importance. This concern is exacerbated by the open-label design, in which knowledge of assigned therapy may influence subjective perceptions of convenience, burden, or satisfaction.
Second, the win-ratio ignores ties, analyzing only the number of wins relative to the number of losses. Ties represent the absence of a “win”, yet their exclusion systematically inflates relative treatment effects. Misinterpretation is likely when ties are common, which is the case in ENVISAGE-TAVI AF, where almost a third of pairwise comparisons ended up in ties for overall PACT-Q2 and an even higher proportion of ties for some individual measures. Excluding these inflates the apparent advantage of edoxaban.
Third, the win-ratio is a purely relative measure. Methodological guidance recommends reporting the absolute win difference, also called net treatment benefit, defined as the percentage of wins minus the percentage of losses, to contextualize effect size and clarify item-level contributions alongside win-ratios. , It overcomes the aforementioned ties by displaying absolute effects. Without this information, a seemingly large win-ratio may reflect only a small net difference in patient experience.
Fourth, patient-reported outcome analyses are vulnerable to missing data, and only 81 percent of randomized participants had available PACT-Q2 assessments. Missingness was imbalanced and may not have been random. Using the latest shared observation for pair comparisons does not address potential systematic differences. Multiple imputation, now standard for patient-reported outcome analyses, is preferable. ,
Finally, applying a hierarchical win-ratio structure post hoc, after treatment effects are known, introduces substantial risk of bias, as ranking decisions may inadvertently prioritize domains favorable to 1 treatment arm. This limitation is especially relevant when outcomes are subjective and treatment is unblinded.
In summary, this post-hoc analysis using the win ratio provides useful insights into perceived anticoagulation convenience and satisfaction after TAVI, but several methodological considerations limit the strength of its conclusions. The open-label design, high prevalence of ties, absence of absolute effect reporting, missing patient-reported outcome data, and the post-hoc definition of hierarchical comparisons all weaken causal inference. While the win-ratio is a valuable tool for prospectively designed future studies that include patient-centered outcomes, interpretation of these findings should be cautious. Decisions regarding anticoagulation after TAVI should prioritize safety signals, particularly gastrointestinal bleeding, given their consistent clinical relevance.
CRediT authorship contribution statement
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
