COSMOS randomized 21,442 US women aged at least 65 years and men aged at least 60 years to a cocoa extract supplement providing 500 mg/day of cocoa flavanols (including 80 mg/day of epicatechin) or placebo. After a median 3.6 years, the primary composite of total cardiovascular events was not significantly reduced (hazard ratio 0.90, 95% CI 0.78–1.02). Cardiovascular death, a prespecified secondary outcome, had a hazard ratio of 0.73 (95% CI 0.54–0.98); it was the only secondary cardiovascular endpoint to reach nominal statistical significance. An adherence-censored per-protocol analysis favored supplementation for total cardiovascular events (hazard ratio 0.85, 95% CI 0.72–0.99), but it was not the primary intention-to-treat result. The cardiovascular-death finding is therefore a secondary signal that warrants confirmation, not proof that cacao foods or cocoa extract prevent cardiovascular disease or reduce mortality.
A 2025 systematic review identified 19 randomized, placebo-controlled studies of cocoa flavanols and microvascular function. Across 14 acute studies, vote counting favored improved vasodilator responses in 12 (85.7%; p=0.013), and all eight studies rated at low risk of bias favored the intervention. Acute effects at rest were not detected, and vote-counted results from seven chronic studies did not reach statistical significance. The small studies used heterogeneous populations, interventions, and outcome measures, preventing meta-analysis. The review discusses nitric oxide as one plausible contributor, but the included evidence does not establish direct eNOS activation, reduced arterial stiffness, reversal of vascular aging, or an optimal epicatechin dose.
A later COSMOS analysis included 8,905 older adults who were free from hypertension at baseline and tested the same cocoa extract for a median 3.4 years. Cocoa extract did not significantly reduce incident hypertension overall (hazard ratio 0.96, 95% CI 0.88–1.05). In a subgroup with baseline systolic blood pressure below 120 mmHg, incident hypertension was lower (hazard ratio 0.76, 95% CI 0.64–0.90), whereas no benefit appeared among participants with baseline systolic pressure of 120–139 mmHg (hazard ratio 1.05, 95% CI 0.93–1.18; interaction p=0.002). This subgroup result supports a prevention hypothesis in older adults with baseline systolic blood pressure below 120 mmHg; it does not show treatment of established hypertension or establish a blood-pressure-lowering dose.
A 2016 systematic review and meta-analysis pooled 19 randomized trials with 1,131 adults. Interventions provided 166–2,110 mg/day of cocoa flavanols for 2–52 weeks. Compared with placebo, pooled results favored lower fasting insulin and HOMA-IR and higher QUICKI and insulin-sensitivity-index values. Among lipid markers, triglycerides fell by 0.10 mmol/L and HDL cholesterol rose by 0.06 mmol/L on average; other tested biomarkers did not all improve. These are selected biomarker findings, not evidence that cocoa prevents diabetes, and the review did not establish the AMPK, oxidative-stress, alpha-glucosidase, or postprandial-glucose mechanisms previously asserted here.
A small 2014 randomized, double-masked trial assigned 41 healthy, sedentary adults aged 50–69 to a high-flavanol diet (900 mg/day of cocoa flavanols, including 138 mg/day of epicatechin) or a low-flavanol diet (10 mg/day, with less than 2 mg/day of epicatechin), with or without an exercise program, for three months; 37 participants completed the study. High flavanol intake improved reaction time on the study’s specific ModBent pattern-separation task and increased cerebral blood volume in the dentate gyrus. The trial does not establish broad memory rejuvenation, performance equivalent to people decades younger, reversal of age-related decline, or improved blood flow as the causal mechanism.
Theobromine is a methylxanthine found in cocoa and other foods. A 2025 observational analysis found that circulating theobromine was associated with lower GrimAge epigenetic-age acceleration and a higher DNA-methylation estimate of telomere length in 509 women from TwinsUK; both associations were replicated in 1,160 participants from KORA. The authors noted possible unmeasured confounding and reverse causality. Serum-metabolite associations do not show that consuming cacao slows biological aging.
Quality and caveats: The interventions differed widely in form, flavanol content, dose, duration, and population. COSMOS tested a standardized cocoa extract that contained flavanols, epicatechin, theobromine, and other cocoa constituents, so it could not isolate the effect of any one component. These sources do not establish equivalent effects for raw cacao, cocoa powder, dark chocolate, or milk chocolate.