Cocoa Flavanols

Full profile

Evidence, honestly graded

Downgraded from B to C on 2026-09-14. Acute cerebral-blood-flow RCTs (Scientific Reports, 2020) found a single dose improved blood flow and attention measures, and that acute signal is real. But the COSMOS program (the large, multi-year parent trial and its cognitive sub-studies, the most substantial test of cocoa flavanols to date) found no overall cognitive benefit from 500 mg/day: no effect on global cognition in COSMOS-Mind (doi:10.1002/alz.12767) or the COSMOS clinic sub-cohort, and no overall effect on the primary memory outcome in COSMOS-Web, where memory improved only in participants with the lowest baseline diet quality or habitual flavanol intake (Brickman 2023). That is a repletion effect in a subgroup, not a general one. Under our standard, null results in the most substantial trial program hold the grade at C, the same basis used for resveratrol, where blood flow rose and cognition did not. Funding transparency: COSMOS ran on an investigator-initiated grant from Mars Edge, with study pills and packaging donated by Mars and partly by Pfizer Consumer Healthcare (now Haleon), and NIH grants supported parts of the program including the COSMOS-Mind cognitive sub-study. Much other cocoa-flavanol research, and the CocoaVia-branded material, is also tied to Mars. The null results therefore come from trials the ingredient's maker helped fund, which cuts against the ingredient rather than for it. Re-examined 2026-09-17 against a newer GRADE-assessed meta-analysis (Arabpour 2026) that reports faster Trail Making Test times and better verbal fluency with chronic cocoa intake, strongest at 500 mg flavanols/day or more for eight weeks or longer, at moderate-to-high certainty. It is the strongest positive synthesis this ingredient has, and it is the reason this grade is now flagged for monitoring rather than settled. What it does not do is answer COSMOS: the paper is paywalled, so we could not read its list of included trials, and we do not know whether the COSMOS cognitive sub-studies are among its 16. C holds on COSMOS's own result either way. A positive pooled estimate does not displace a null finding in the largest and longest trial programme, and if COSMOS is inside the pool then the pooled estimate is being carried by the smaller trials.

See the full grading rubric (study type, replication, population match, and dose adequacy) in The Evidence Standard.

Side effects

• Generally well tolerated
• Excellent overall safety profile

Interactions

• No material known interactions at studied doses

Stacks well with

• L-Theanine

What to look for on a label

• Specify a de-theobrominated or low-theobromine extract to keep the stimulant-free positioning honest; regular cocoa extract carries meaningful theobromine.
• Frame as acute blood-flow support, not a cognition benefit. The COSMOS trials found no overall cognitive effect, and the memory signal was confined to people with low dietary flavanol intake.

References

• Arabpour Z, Arabpour Z, Jazayeri S (2026). Nutrition Reviews (online ahead of print)GRADE-assessed systematic review and meta-analysis, 16 RCTs, 2,847 participantsStudied dose: Subgroup benefit concentrated at ≥500 mg cocoa flavanols/day for ≥8 weeksTrail Making Test A −4.51 s and B −13.1 s, verbal fluency +3.15, at moderate-to-high certainty; no benefit on the Mini-Mental State Examination and inconsistent results for working and verbal memory.
• Acute cocoa flavanol cerebral blood flow RCTs, Scientific Reports 2020. Single dose improved cerebral blood flow and attention measures in healthy adults versus placebo. PMID 33235219. Educational, not a product claim.
• COSMOS 2023: chronic cocoa flavanol trial, null primary endpoint. Large, multi-year RCT in older adults; null on its primary global-cognition endpoint. doi:10.1002/alz.12767. Stated plainly as the honest limit on the chronic claim.
• Vyas CM, Manson JE, Sesso HD, et al. (2024). The American Journal of Clinical Nutrition, 119(1):39–48Randomized controlled trial (COSMOS clinic sub-cohort), older adults, 2 yearsStudied dose: 500 mg cocoa flavanols dailyNo significant effect on 2-year change in global cognition, episodic memory, or executive function; uncorrected subgroup analyses suggested possible benefit with lower baseline diet quality.
• Brickman AM, Yeung LK, Alschuler DM, et al. (2023). Proceedings of the National Academy of Sciences, 120(23):e2216932120Randomized controlled trial (COSMOS-Web), n=3,562 older adultsStudied dose: 500 mg cocoa flavanols dailyNo significant overall effect on the primary memory outcome; memory improved in participants with lower baseline diet quality or habitual flavanol intake.

Grades and studied doses are our conservative reading of the human research, shown for education. They are not product claims, and a studied dose is not a recommended dose.

Revision history

Every change to this ingredient's evidence status, dated and explained. The full cross-library log lives at Evidence Updates.

• 2026-09-17Citation review

  Added the strongest positive synthesis this ingredient has — a 2026 GRADE-assessed meta-analysis reporting faster Trail Making Test times and better verbal fluency at 500 mg flavanols/day or more — and flagged this grade for monitoring rather than moving it. The paper is paywalled, so whether the COSMOS cognitive sub-studies are among its 16 trials is unknown. C holds either way: a positive pooled estimate does not displace the null result in the largest and longest trial programme.

• 2026-09-14Grade changeGrade B → Grade C

  Downgraded from B to C. The COSMOS cognitive sub-studies found no overall cognitive benefit from 500 mg/day of cocoa flavanols, and the memory signal was confined to people with the lowest baseline diet quality — so the acute blood-flow finding no longer carries a B.

• 2026-07-07New profile

  Initial profile published at Grade B (qualified): acute cerebral-blood-flow and attention effects are demonstrated, while the large chronic trial (COSMOS) was null on its primary global-cognition endpoint.

See how Cocoa Flavanols compares on grade, dose, and goal in the Evidence Explorer.