Cocoa Flavanols and Cognitive Aging: The COSMOS Trial Results

The COcoa Supplement and Multivitamin Outcomes Study (COSMOS) was the first adequately powered randomised trial of cocoa flavanols for chronic disease outcomes. It enrolled 21,442 US adults over age 60 between 2015 and 2022 in a 2×2 factorial design comparing 500 mg/day cocoa flavanols (containing 80 mg of (-)-epicatechin) and a daily multivitamin against placebo over a median 3.6 years of follow-up. The cardiovascular and cognitive sub-studies have now both reported full results.

The cardiovascular primary outcome

The primary cardiovascular endpoint — a composite of myocardial infarction, stroke, coronary revascularisation, cardiovascular death, carotid surgery, peripheral artery surgery, and unstable angina — was not significantly different between cocoa flavanol and placebo arms in the full intention-to-treat analysis (HR 0.90, 95% CI 0.78–1.02) [1]. Secondary analyses showed a statistically significant 27% reduction in cardiovascular mortality (HR 0.73), but the prespecified primary endpoint was negative.

The COSMOS-Mind cognitive sub-study

COSMOS-Mind enrolled 2,262 participants from the parent trial for annual telephone cognitive testing. After three years, cocoa flavanols did not significantly slow decline in global cognition versus placebo. The multivitamin arm, in contrast, produced a small but statistically significant slowing of global cognitive decline equivalent to roughly 2 years of less aging [2]. The flavanol primary cognitive outcome was negative; the multivitamin was the headline.

The COSMOS-Web sub-study

COSMOS-Web (n=3,562) used a web-based cognitive battery and found a similar pattern — no significant flavanol effect on global cognition over 3 years [3]. A post-hoc analysis stratifying by baseline diet quality found suggestive flavanol benefits in the subgroup of participants with the lowest dietary flavanol intake, but this analysis was not prespecified and should be treated as hypothesis-generating.

How to interpret the negative cognitive findings

Three explanations are commonly offered. First, the trial population was relatively healthy and well-educated, leaving little room to detect a small effect. Second, the dose (500 mg total flavanols, 80 mg epicatechin) is modest compared to earlier mechanistic studies. Third, the trial measured cognition annually through brief tests rather than detailed neuropsychological batteries that captured the original signal in smaller MRI studies [4]. None of these excuses the negative primary result.

What the mechanistic literature still shows

Smaller short-term trials continue to show that cocoa flavanols acutely raise brachial flow-mediated dilation, improve dentate gyrus blood flow on MRI, and modestly improve some executive function tasks within 1–2 hours of consumption [4]. These acute findings did not translate into long-term cognitive benefit in COSMOS at the dose tested. The dissociation is common across the supplement literature.

How cocoa flavanols compare to dietary intake

The 500 mg/day flavanol dose used in COSMOS is roughly the amount in 4 servings of dark chocolate (≥70% cocoa) or 4 cups of well-prepared green tea, but the (-)-epicatechin content is concentrated in non-alkalized cocoa preparations and most commercial chocolate is alkalized to reduce bitterness, which destroys most of the flavanols. Anyone hoping to replicate the COSMOS exposure through diet should look for "non-Dutched" or "natural" cocoa rather than standard chocolate bars.

Practical takeaway

COSMOS was the well-conducted, adequately powered trial that the cocoa flavanol field needed, and it did not show what the marketing claims. Cocoa flavanols did not slow cognitive decline or prevent cardiovascular events in the prespecified primary endpoints. The 27% cardiovascular mortality signal is intriguing but secondary. For adults already eating a Mediterranean-style diet, the case for adding a cocoa flavanol supplement is weak. The multivitamin signal from the same trial is the more relevant cognitive finding of the decade.