Lit Matters 2: Evaluating the Evidence

Composite endpoints can hide as much as they reveal, and the win ratio is a prime example. This trial-analysis method is increasingly common in cardiology and MACE literature, but its apparent treatment effect can look stronger than the actual probability of benefit.

Win Ratio in Clinical Trials

• Pairwise hierarchical comparisons: Win ratio matches one treated patient to one control patient, then compares them across a prespecified hierarchy such as death, heart failure events, and quality of life until one side records a win.
• Ties dropped from analysis: Pairs with no winner are excluded rather than counted as non-wins, a design choice that can materially inflate the apparent treatment effect when many patients have similar outcomes.
• Outcome ordering problem: The hierarchy is a researcher value judgment, and an early-tier event like death can be outweighed in the final headline by many late-tier wins from less important outcomes. We get into why that framing matters in the episode.
• Short-term versus long-term mixing: Win ratio can combine early symptomatic or biologic changes with later hard outcomes, making a treatment look broadly effective even when the signal is concentrated in short-term or surrogate measures.
• Seductive headline numbers: A reported win ratio of 1.36 can be misread as a 36% greater chance of benefit, when the actual win probability may be much closer to a single-digit absolute advantage.
• Retrospective use caution: The method is most defensible when the hierarchy is prospectively designed; retrospective win-ratio analyses are especially vulnerable to cherry-picked components and overstated conclusions.

Misleading Trial Examples

• EMPULSE effect inflation: In EMPULSE, the published win ratio favored empagliflozin at 1.36, but the estimated win probability was only 58%, showing how the ratio can overstate the clinical impression.
• Ignored ties magnify benefit: When ties were treated as absences of wins rather than discarded, the apparent EMPULSE advantage dropped further to about 54%, underscoring how much the denominator matters.
• PARAGLIDE surrogate dominance: In PARAGLIDE, the headline advantage for sacubitril-valsartan was driven largely by natriuretic peptide changes, with roughly 36% of pairwise comparisons ending in ties.
• TRILUMINATE patient-reported skew: TRILUMINATE reported a win ratio of 1.48 favoring TEER, but much of that signal came from KCCQ improvement while heart failure hospitalizations actually leaned the other direction.
• Bottom-line appraisal: Win ratio is a potentially useful tool, but when lower-tier surrogate or symptom outcomes drive the wins, the final result may look more convincing than the long-term clinical benefit really is. That distinction is worth hearing in the chapter.