Cocoa Flavanols

Also sold as: Cacao, Cocoa Extract

The polyphenol fraction of cocoa, and the subject of one of the largest supplement outcome trials ever run.

How Cocoa Flavanols actually works

Epicatechin and related flavanols increase nitric oxide bioavailability via eNOS, producing measurable improvements in flow-mediated dilation within hours of a dose. The effect on blood pressure and endothelial function is one of the better-replicated findings in nutritional science, and the COSMOS trial found reduced cardiovascular mortality as a secondary endpoint.

Where to get Cocoa Flavanols

The evidence for Cocoa Flavanols

Graded by what exists behind each claim.

✅ Clinically validated

• COSMOS — 21,442 participants, randomized, 3.6 years — found cocoa flavanols did NOT reduce the primary composite cardiovascular endpoint. It DID reduce cardiovascular DEATH by 27% in a secondary analysis.
• Meta-analyses consistently show improved flow-mediated dilation and modest blood pressure reduction (~2 mmHg).
• Trials show improved insulin sensitivity and small cognitive improvements in older adults.

📊 Correlative data

• Cocoa-consuming populations, notably the Kuna, show unusually low age-related blood pressure rise.

🧪 Theoretical / extrapolated benefits

• Epicatechin increases nitric oxide bioavailability, which explains the endothelial and blood pressure findings cleanly.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What Cocoa Flavanols actually does

The active is one molecule and somebody proved it. (−)-Epicatechin is a flavan-3-ol monomer, 2,3-cis, and it is the smallest unit of the procyanidin chemistry cocoa is full of. In healthy men, flavanol-rich cocoa acutely raised circulating nitric oxide species, improved flow-mediated dilation and improved microcirculation; (−)-epicatechin and its metabolite epicatechin-7-O-glucuronide were independent predictors of the vascular effect; a flavanol mixture reproduced acetylcholine-induced relaxation in rabbit aortic tissue; both the ex vivo relaxation and the in vivo rise in flow-mediated dilation were abolished by inhibiting nitric oxide synthase; and administering pure (−)-epicatechin on its own reproduced the effect of the whole cocoa drink Schroeter 2006.

Read that chain again, because it is unusually complete. Candidate identified, present in plasma, correlated with effect, pathway named, pathway blocked, effect abolished, pure compound given, effect reproduced. That is a full pharmacological argument for a food constituent, and almost nothing else in this category has one. The enzyme at the end of it is endothelial nitric oxide synthase, the product of NOS3, and the read-out is the diameter of a brachial artery.

And here is what is different about this molecule, against every other polyphenol on this site. The standing objection to polyphenol pharmacology is that the parent compound is nearly absent from plasma and what circulates is a glucuronide or sulfate conjugate Manach 2005, so the dish experiment is testing a molecule the body never sees. On this page that objection does not land, because the conjugate is not a caveat here — epicatechin-7-O-glucuronide was itself an independent predictor of the vascular response Schroeter 2006. The metabolite tracks the effect. That is the strongest form this argument takes anywhere in the cohort.

The stereochemistry is a manufacturing variable, not a footnote. Cocoa's native monomer is (−)-epicatechin. Alkalization — Dutching, the process that makes cocoa powder dark and mild — both destroys flavanols and epimerizes (−)-epicatechin toward (−)-catechin, a different stereoisomer with different absorption and different activity. So the processing step that makes cocoa taste like cocoa is the step that removes the molecule the trials are about.

Cell, rodent, human — and where it stops

Ex vivo and acute human, in one paper. Rabbit aortic rings relaxing to a flavanol mixture, that relaxation abolished by nitric oxide synthase inhibition, and in healthy male volunteers an acute rise in circulating nitric oxide species and flow-mediated dilation after flavanol-rich cocoa, reproduced by pure (−)-epicatechin Schroeter 2006.

Chronic human, on blood pressure. The Cochrane review of cocoa and blood pressure is the class-level synthesis, and it finds a small reduction over short trials Ried 2017. Small is the operative word and it is stated here rather than buried: this is a few millimetres of mercury in trials lasting weeks.

And then the largest randomized trial of any supplement on this site, reported exactly. COSMOS randomized 21,442 older US adults to cocoa extract delivering 500 mg of cocoa flavanols per day, including 80 mg of (−)-epicatechin, or placebo, with a median follow-up of 3.6 years. The primary endpoint was negative. Confirmed total cardiovascular events occurred in 410 participants on cocoa extract and 456 on placebo, hazard ratio 0.90 with a 95% confidence interval of 0.78 to 1.02 and P = 0.11 Sesso 2022.

The secondary endpoints, all of them, in order. Cardiovascular death: hazard ratio 0.73, 95% confidence interval 0.54 to 0.98 — a 27% relative reduction, and the number the product is now sold on. Myocardial infarction: 0.87, 0.66 to 1.16. Stroke: 0.91, 0.70 to 1.17. Neither of the last two reached significance. In a per-protocol analysis restricted to adherent participants, total cardiovascular events came out at 0.85, 0.72 to 0.99 Sesso 2022. The same trial's cognitive ancillary is the second place COSMOS has been read for a benefit Baker 2022.

The specific obstacle here is not bioavailability. It is multiplicity, and it deserves to be spelled out rather than resolved. A trial declares one primary endpoint because that is the single comparison its statistics are calibrated for. COSMOS's primary did not reach significance at P = 0.11 Sesso 2022. Cardiovascular death was one of several pre-specified secondaries, and its confidence interval runs to 0.98 — it clears 1.0 by two hundredths. When several secondary endpoints are examined, one of them landing just inside significance is the expected behavior of chance as much as it is the signature of a drug. It generates a hypothesis. It does not settle one.

And the per-protocol number needs the same honesty. Hazard ratio 0.85 for total cardiovascular events among adherent participants Sesso 2022 is a real analysis and a legitimate one to report. It is also an analysis that discards randomization: people who keep taking a study pill for three and a half years differ from people who stop, in health, in motivation and in everything else that predicts not having a heart attack. Adherence analyses are biased toward benefit in essentially every trial that runs them, including trials of placebo. So the honest reading of COSMOS is: a null primary, a promising secondary, and a supportive analysis that cannot be taken at face value. Anyone who reports only one of those three is selling something.

Cocoa Flavanols — which form, and does it matter

The only number that matters is milligrams of cocoa flavanols, and inside that, milligrams of (−)-epicatechin. COSMOS used 500 mg of cocoa flavanols per day containing 80 mg of (−)-epicatechin Sesso 2022. That is the specification. A label reading cocoa extract 500 mg is describing the mass of the powder, not its flavanol content, and the two can differ by an order of magnitude.

Chocolate is not a smaller version of this. It is a different product. Alkalization strips flavanols and epimerizes the monomer; roasting and conching remove more; and flavanol content is not on a confectionery label, so a dark chocolate bar's actual content is unknowable from the wrapper and varies enormously between brands and batches. Then the arithmetic finishes the argument: reaching a COSMOS-equivalent flavanol intake from ordinary dark chocolate means eating a substantial daily quantity of a food that is mostly cocoa butter and sugar, which pushes the metabolic numbers in the direction you were trying to fix.

Cocoa powder, drink mix and extract are three different things. A natural, non-alkalized cocoa powder retains far more flavanol than a Dutched one. A branded flavanol-standardized drink or capsule declares its flavanol content because that is the point of it. The trials, both the mechanistic Schroeter 2006 and the outcome Sesso 2022, used defined, characterized preparations, and the whole reason those trials mean anything is that somebody measured what was in them.

What the site's own card says, and it is right. 500–750 mg of cocoa flavanols providing about 80 mg of epicatechin, with the COSMOS dose noted as 500 mg. That maps cleanly onto the trial Sesso 2022 and it is the rare case where a card dose and a trial dose are the same number.

What would have to be true, and how you would know it was not

1. Blood pressure, systolic (mmHg), seven-day home average — down 1–3 mmHg, and you will not be able to detect it. The class evidence supports a small reduction Ried 2017. Say the uncomfortable part: the standard deviation of a seven-day home average, between one week and another, in the same untreated person, is larger than the effect being predicted. This is a population-level number that a single reader cannot verify, and a reader who measures 3 mmHg lower has measured noise that happened to point the right way.

2. The two-hour window, which is the mechanistic test. The flow-mediated dilation effect is acute: it follows the dose by hours and it is gone by the next day Schroeter 2006. So predict that any perceived benefit which builds steadily over a fortnight and persists after stopping is not this mechanism. Nothing in the epicatechin story predicts an effect that accumulates.

3. The prediction that cuts against the product: Apolipoprotein B, mg/dL — unchanged. COSMOS's own primary cardiovascular endpoint did not reach significance Sesso 2022, and there is nothing in the eNOS pathway that touches hepatic apolipoprotein B secretion or LDL receptor expression. Predict less than 5% change at 12 weeks. A reader with an ApoB of 120 mg/dL who buys this as a lipid product has bought a vasodilator.

4. hs-CRP, mg/L — unchanged. Same argument. The demonstrated pathway is nitric-oxide-dependent vasodilation Schroeter 2006, not suppression of hepatic acute-phase synthesis. Predict no movement at 12 weeks in someone starting under 3 mg/L.

What will fool you, and it is the headline: a 27% reduction in cardiovascular death sounds like the strongest claim in this whole category. It came from a secondary endpoint of a trial whose primary endpoint returned a hazard ratio of 0.90 at P = 0.11 Sesso 2022. If you buy this product, buy it knowing the primary was null. That is still a better evidence base than almost anything else on this shelf, which says more about the shelf than about the cocoa.

What nobody has tested yet

Nobody has stratified COSMOS by plasma epicatechin metabolites, and the samples exist. This is the most obviously available experiment in the whole cohort. The mechanism paper says the predictors of vascular response are plasma (−)-epicatechin and epicatechin-7-O-glucuronide Schroeter 2006. COSMOS randomized 21,442 people and followed them for 3.6 years Sesso 2022. Measure those two metabolites in stored baseline and on-treatment samples, split the treated arm into high and low converters, and re-run the primary endpoint within each stratum. If the null primary is a diluted positive, that is where it would show up.

Nobody has run a dose as a randomized variable. COSMOS used one dose, 500 mg of flavanols with 80 mg of epicatechin Sesso 2022. Nobody knows whether that sits on the steep part of the curve, the flat part, or past a plateau. A four-arm outcome trial at 250, 500, 1,000 and 2,000 mg would answer it and nobody will fund one, because the first attempt cost a decade.

Nobody has tested it in midlife. COSMOS enrolled older adults Sesso 2022. A mechanism that acts on endothelial nitric oxide availability Schroeter 2006 should plausibly matter most before endothelial dysfunction is established, which is decades earlier. That trial would need to run for twenty years and has never been proposed seriously.

And the cardiovascular-death finding has never been tested as a primary endpoint. That is the experiment that would settle this product's status, and it would have to be as large as COSMOS again, powered on cardiovascular mortality rather than on total events Sesso 2022. Until somebody runs it, a 27% relative reduction remains the most interesting unconfirmed number in supplement cardiology.

Cocoa Flavanols — its own safety story, not its category's

This is close to the safest product in the cohort, and the useful thing a safety section can do here is name the two real questions rather than manufacture a warning. Across 21,442 people for 3.6 years, cocoa extract's tolerability was unremarkable Sesso 2022. There is no hepatic signal, no renal signal, no withdrawal and no established upper limit.

Cadmium is the real one, and it is a sourcing question rather than a pharmacology question. Cacao is a documented accumulator of cadmium from soil, which is why chocolate products carry heavy-metal warnings in California and why cocoa cadmium limits exist in European food law. A concentrated extract concentrates the plant; whether it also concentrates the metal depends entirely on the manufacturing process and the bean origin, and no supplement label states it. This is reasoning from the botany rather than from a trial of these products, and it is labeled as such — but it is the question to put to a manufacturer, by lot, in writing, if you intend to take this for years.

Theobromine, which is why some people cannot sleep on it. Cocoa carries theobromine and a smaller amount of caffeine. Theobromine is an adenosine receptor antagonist with a half-life considerably longer than caffeine's, so an evening dose of a theobromine-rich extract has a long tail. Trial-grade cocoa flavanol extracts are generally prepared to be low in methylxanthines; retail cocoa powders are not. If a product keeps you awake, that is the molecule, and the fix is to take it in the morning.

Blood pressure additivity, small and real. The class lowers blood pressure slightly Ried 2017. On top of two antihypertensives and a diuretic that is additive, and the person who notices is the older reader who stands up quickly. A couple of millimetres is not dangerous alone; it is dangerous as the fourth thing doing the same.

Antiplatelet potential, stated at its real size. Flavan-3-ols have measurable antiplatelet activity in vitro. At 500 mg of flavanols a day this is a theoretical consideration rather than a clinical one, and the outcome trial reported no safety signal of that kind Sesso 2022. Mention it before surgery and stop worrying about it otherwise.

And migraine, because it is the complaint people actually have. Chocolate is a commonly reported migraine trigger, and whether that is the methylxanthines, the biogenic amines or a craving that precedes the attack is genuinely unsettled. If you get migraines and start a concentrated cocoa product, keep a simple diary before concluding either way.

Sources read for this page

• Sesso HD, et al. Effect of cocoa flavanol supplementation for the prevention of cardiovascular disease events: the COcoa Supplement and Multivitamin Outcomes Study (COSMOS) randomized clinical trial.. American Journal of Clinical Nutrition 2022 · PMID 35294962
• Schroeter H, et al. (-)-Epicatechin mediates beneficial effects of flavanol-rich cocoa on vascular function in humans.. Proceedings of the National Academy of Sciences 2006 · PMID 16418281
• Ried K, et al. Effect of cocoa on blood pressure.. Cochrane Database of Systematic Reviews 2017 · PMID 28439881
• Baker LD, et al. Effects of cocoa extract and a multivitamin on cognitive function: a randomized clinical trial.. Alzheimer's & Dementia 2022 · PMID 36102337
• Manach C, et al. Bioavailability and bioefficacy of polyphenols in humans. I. Review of 97 bioavailability studies.. American Journal of Clinical Nutrition 2005 · PMID 15640486

How you would know if it worked

Cocoa Flavanols — safety & side effects

Bloodwork to run alongside Cocoa Flavanols

Baseline first, then again at 8–12 weeks.

The Real Cardiovascular Risk panel covers these in one order — 9 markers, $187.60 with the discount applied.

Check results you already have → · All 103 markers A–Z

Cocoa Flavanols — frequently asked questions

Cocoa Flavanols inside a finished plan

One arm of 2 Protocol Blueprints, free to read in full.

What Cocoa Flavanols is used for

Cocoa Flavanols appears under 3 goals in the goal router.

Related Cardiovascular supplements

← Browse the full Supplement Vault

↑ Back to on this page