Cocoa flavanols may reduce cardiovascular disease (CVD) risk, yet large randomized trials remain inconclusive. The COcoa Supplement and Multivitamin Outcomes Study (COSMOS) suggested a modest, nonsignificant benefit using Cox models, which do not account for event severity in composite outcomes. To address this, we applied generalized pairwise comparison (GPC), or "win ratio" (WR), to assess cocoa flavanols versus placebo on hierarchical CVD outcomes among healthy older US adults. This secondary analysis of COSMOS, a randomized, placebo-controlled, 2 × 2 factorial trial of cocoa extract and multivitamins for preventing CVD and cancer, included 21,442 adults (women ≥ 65, men ≥ 60 years) followed for a median of 3.6 years. The primary outcome was a hierarchical composite of total CVD, prioritizing: fatal CVD, non-fatal myocardial infarction (MI), non-fatal stroke, coronary revascularization, carotid surgery, peripheral artery surgery, and hospitalized unstable angina. Analyses followed the intention-to-treat principle. GPC estimated WRs and net treatment benefits (NTBs) for cocoa flavanols versus placebo. GPC analyses showed cocoa flavanol wins of 3.41% and placebo wins of 2.87%, yielding a reciprocal WR of 0.84 (95% CI 0.72-0.99) and negative NTBs of - 0.54% (- 1.04 to - 0.03), p = 0.037. Sensitivity analyses prioritizing stroke over MI produced similar findings. By contrast, Cox regression of the same composite yielded a nonsignificant hazard ratio of 0.90 (95% CI 0.79-1.03), suggesting standard time-to-first-event models underestimated benefit. GPC "WR" analyses showed cocoa flavanols significantly reduced CVD events by accounting for event severity in the composite CVD outcome, whereas Cox regression marginally missed these effects.
Keywords: Cardiovascular disease; Cocoa flavanol supplementation; Coronary heart disease; Hierarchical composite outcome; Stroke; Win ratio.
© 2026. Springer Nature B.V.