Key result
Preference-informed win ratios yielded consistent results in 7 of 9 HF trials, but attenuated spironolactone benefit in TOPCAT-Americas (WR 1.03 [95% CI 0.92-1.15] vs 1.13 [1.01-1.26]).
Why the study?
Win statistics evaluate hierarchical composite endpoints simultaneously, but outcome hierarchies are typically determined by researchers without patient input. This study sought to develop a patient preference-informed approach to constructing hierarchical composite outcomes and apply it to re-analyze randomized controlled trials of HF pharmacotherapy.
Does incorporating patient preference-informed outcome hierarchies into win ratio analyses alter the estimated net clinical benefit of heart failure pharmacotherapies compared to conventional outcome hierarchies in adults with heart failure?
Does incorporating patient preference-informed outcome hierarchies into win ratio analyses alter the estimated net clinical benefit of heart failure pharmacotherapies compared to conventional outcome hierarchies in adults with heart failure?
Answers are grounded in this paper and cite their sources.
Integrating patient preferences into hierarchical composite endpoints for win ratio analysis is feasible and demonstrates that while most heart failure trial results are robust to outcome hierarchy changes, certain therapies with notable side-effect profiles or modest quality-of-life impacts exhibit preference-sensitive net benefits.
Background Win statistics allow simultaneous evaluation of multiple outcomes using hierarchical composite endpoints, but outcome hierarchies are typically determined by researchers without patient input. Purpose We sought to develop a patient preference-informed approach to constructing hierarchical composite outcomes for analysis with win statistics and apply it to re-analyze randomized controlled trials (RCTs) of heart failure (HF) pharmacotherapy. Methods We synthesized patient preference data from a discrete-choice experiment (DCE) of adults with HF and from three studies using simple rating or ranking exercises to derive two preference-informed outcome hierarchies. We then re-analyzed individual participant-level data from nine RCTs (n=24,725) using unmatched prioritized win ratios (WRs) with four outcome components: all-cause death, HF hospitalization, HF-specific quality of life (QoL) or function, and study drug discontinuation due to adverse events. Five WRs were estimated for each trial (outcomes prioritized in descending order of importance): (1) conventional two-component (death, HF hospitalization), (2) conventional 3-component (death, HF hospitalization, QoL); (3) conventional 4-component (death, HF hospitalization, QoL, side-effects), (4) preference-informed hierarchy based on DCE results (death, HF hospitalization, side-effects, QoL), and (5) QoL-prioritized hierarchy reflecting additional preference evidence (QoL, death, HF hospitalization, side-effects). Results Of nine re-analyzed trials, seven (DIG, EMPEROR-Preserved, EMPEROR-Reduced, EMPULSE, SOLVD-Treatment, FIGHT, LIFE) showed consistent results across conventional and preference-informed hierarchies, indicating preference-insensitive results. Two trials exhibited preference-sensitive findings. In TOPCAT-Americas, benefit with spironolactone over placebo was attenuated when side-effects were prioritized (WR 1.03 [95% confidence interval 0.92-1.15] versus conventional 4-component WR 1.13 [1.01-1.26]), reflecting large between-group differences in the side-effect component. In BEST, the benefit of bucindolol over placebo was attenuated and no longer statistically significant when QoL was ranked first (WR 1.07 [0.98-1.16] versus conventional 4-component WR 1.15 [1.05-1.26]), despite favourable effects on all four components. Conclusions Integrating empirically-derived patient preferences into hierarchical composite outcomes provides a pragmatic and generalizable approach to evaluating the net clinical benefit of HF pharmacotherapy using win statistics. These results highlight the value of incorporating patient priorities into clinical trial interpretation and therapeutic decision-making.
No takes yet. Share an insight, caveat, or question.
Turgeon et al. (2026) studied Heart failure (n=24,725). Heart failure pharmacotherapy (evaluated via patient preference-informed win statistics) vs. Placebo or active control (evaluated via conventional win statistics) was evaluated on Four-component hierarchical composite outcome (all-cause death, HF hospitalization, HF-specific quality of life or function, and study drug discontinuation due to adverse events) (WR 1.03, 95% CI 0.92-1.15). Preference-informed win ratios yielded consistent results in 7 of 9 HF trials, but attenuated spironolactone benefit in TOPCAT-Americas (WR 1.03 [95% CI 0.92-1.15] vs 1.13 [1.01-1.26]).
Synapse has enriched 10 related papers. Consider them for comparative context: