Nutrition · Polyphenols · Cardiovascular Health · Flavanols
The COSMOS randomised controlled trial found that 500mg of cocoa flavanols daily reduced cardiovascular disease mortality by 27% over five years. A subsequent analysis of two large cohorts — 30,663 adults with biomarker data — found that only 19% of people eating five or more servings of fruit and vegetables per day actually reach this threshold. Eating plants matters. Which plants, and how many, determines whether you're getting the compounds that have clinical evidence behind them.
The five-a-day recommendation was never about specific compounds. It was a population-level communication strategy — simplifying complex nutritional guidance into a single memorable number. The simplification was intentional and largely successful in communicating that eating more plants is better than eating fewer plants. But it obscures a more important truth: the value of plant foods is not uniformly distributed across all plants, and the compounds with the strongest cardiovascular evidence are not evenly present across all fruits and vegetables.
Flavanols are the clearest example of this gap. They are a subclass of polyphenol found in specific plant foods — cocoa, tea, apples, berries, grapes, and legumes being the primary dietary sources. Their cardiovascular benefit is supported by a substantial evidence base including a large RCT. And the amount present in a banana, a carrot, an orange, or a tomato — all of which count toward five-a-day — is negligible.
Two people eating five servings of fruit and vegetables per day can have flavanol intakes that differ by a factor of ten or more depending entirely on which five servings they choose. The recommendation counts servings. The biology runs on compounds.
Flavanols (also called flavan-3-ols) are a class of polyphenol found predominantly in cocoa, tea, apples with skin, berries, grapes, beans, and legumes. The principal bioactive forms are (-)-epicatechin and (+)-catechin — the most studied, with the most direct mechanistic evidence for cardiovascular benefit.
Their primary cardiovascular mechanism is nitric oxide (NO) production in vascular endothelium. Epicatechin activates endothelial nitric oxide synthase (eNOS) — increasing NO bioavailability, which drives vasodilation, reduces arterial stiffness, lowers blood pressure, and inhibits platelet aggregation. This mechanism was identified specifically in the landmark Schroeter et al. (2006) study in PNAS — demonstrating that epicatechin mediates the vascular effects of cocoa flavanols by increasing NO and improving flow-mediated vasodilation.
The COcoa Supplement and Multivitamin Outcomes Study (COSMOS) was an RCT involving older US adults randomised to 500mg/day cocoa flavanol extract or placebo for an average of 3.6 years. The primary finding — total cardiovascular events — did not reach statistical significance in the intention-to-treat analysis. The finding that produced the 27% figure was cardiovascular disease mortality specifically.
This distinction matters. Total cardiovascular events includes non-fatal events where ascertainment is imperfect in a study of this size and duration. Cardiovascular mortality is a hard endpoint — accurately recorded, not subject to the same ascertainment issues. The 27% reduction in cardiovascular deaths is the most reliable signal from the study.
A 2026 re-analysis of COSMOS using the "win ratio" method — which gives greater weight to more severe outcomes — found results more favourable to the flavanol arm, consistent with the cardiovascular mortality finding being the primary signal.
The honest framing is this: COSMOS provides strong evidence that cocoa flavanols at 500mg daily reduce cardiovascular mortality, with the total event reduction not reaching statistical significance in this sample size and duration. The cardiovascular mortality finding is robust.
The study that made this issue most concrete was an analysis of the COSMOS and EPIC-Norfolk cohorts (30,663 adults with urinary flavanol biomarker data) assessing whether people meeting standard fruit and vegetable guidelines were achieving the 500mg/day flavanol threshold.
The findings were stark. Only 19.2% of COSMOS participants and 17.9% of EPIC-Norfolk participants reached the biomarker-estimated 500mg/day threshold. Among those meeting or exceeding the five-a-day recommendation, only 20–21% reached the threshold — barely higher than the full cohort. In EPIC-Norfolk, the highest fruit and vegetable consumers were actually less likely to reach the threshold than those in the lowest quartile — because heavy vegetable consumption doesn't deliver flavanols if the vegetables chosen are low-flavanol.
The flavanol gap is a specificity problem. Total fruit and vegetable intake is a poor proxy for flavanol intake. The value of a dietary recommendation depends on the specific compounds it delivers — and generic plant guidance cannot substitute for knowing which plants contain which compounds at what concentrations.
| Food | Flavanol concentration | Notes |
|---|---|---|
| Cocoa powder (unsweetened) | High — 200–800mg per 10g | Highly variable by processing method — Dutch-processed cocoa loses 60–90% of flavanols |
| Dark chocolate (70%+) | High — but variable | Flavanol content is not labelled. Processing determines content, not cocoa percentage alone |
| Black tea (brewed) | High — 150–200mg per cup | The UK population's primary flavanol source in most dietary surveys |
| Green tea | High — 100–300mg per cup | Catechins including EGCG — overlapping but distinct flavanol profile from cocoa |
| Apples with skin | Moderate — 20–80mg per apple | Skin contains the flavanols — peeled apples are significantly lower |
| Berries (blueberry, blackberry, strawberry) | Moderate — 30–100mg per 100g | Anthocyanins also present — overlapping cardiovascular benefits |
| Grapes / red wine | Moderate — varies substantially | Procyanidins in grape seeds; resveratrol is a separate stilbene compound |
| Black beans, lentils | Moderate — 20–60mg per 100g | Often overlooked flavanol source — adds to diversity of intake |
| Banana | Very low | Excellent for potassium — not a flavanol source |
| Carrot, broccoli, tomato | Very low to negligible | Excellent for carotenoids, vitamin C, sulforaphane — not flavanol sources |
| Orange, grapefruit | Low | Hesperidin (flavanone) not flavanol — different compound class |
Flavanol bioavailability averages approximately 31% — but with standard deviation of 23%, meaning individual variation is enormous. The systematic review by Di Pede et al. (2022) identified the key drivers: the food matrix (intact plant cell walls vs. processed forms), gut microbiome composition (bacteria convert flavanol polymers to bioactive monomers), and cooking and processing (heat and alkaline processing destroy flavanols).
The gut microbiome connection is clinically important. Microbial catabolites from flavanol fermentation in the large intestine contribute over 20% of total flavanol bioavailability — meaning that gut dysbiosis directly impairs flavanol benefit. A person with significant gut dysbiosis eating flavanol-rich foods will have lower flavanol bioavailability than a person with a healthy microbiome eating the same foods. This is another reason why the OAT and GI-MAP findings matter beyond their direct clinical presentations.
Processing is the other critical factor. Dutch-processed (alkali-treated) cocoa loses 60–90% of its flavanols compared to natural cocoa. The flavanol content of dark chocolate cannot be inferred from the cocoa percentage — a 70% bar made with Dutch-processed cocoa may have lower flavanols than a 50% bar made with natural cocoa. This variability is the reason standardised supplementation has a legitimate clinical role for anyone targeting specific flavanol doses.
The recommendation to eat more fruit and vegetables is correct. The assumption that any five servings will deliver clinically meaningful flavanol intake is not. Dietary precision — knowing which foods deliver which compounds — is what bridges the gap between good advice and good outcomes.
The supplementation question divides people whose prior belief is food-only versus those who accept that some compounds require targeting. The flavanol data makes this a practical clinical question rather than a philosophical one.
If the target is 500mg/day reliably — the dose studied in COSMOS — dietary sources alone produce highly variable intake because flavanol content varies substantially between sources, between batches, and based on processing. For someone already eating a varied plant-rich diet who wants to ensure they are reliably reaching the cardiovascular-relevant threshold, a standardised cocoa flavanol extract is the practical solution.
This is not a universal supplement recommendation. It is a precision recommendation for a specific compound at a specific dose, for a specific clinical indication (cardiovascular risk reduction), where the evidence base is sufficiently robust to warrant it and dietary achievement is sufficiently variable to make supplementation clinically logical.
The supplement quality caveat applies as much here as anywhere: standardised cocoa flavanol products should specify the total flavanol content and the epicatechin content, not simply the cocoa percentage. CocoaVia (mentioned in the evidence literature) specifies 500mg total flavanols and 85mg epicatechin per serving — the dose used in COSMOS. That is the type of specification that matters.
The OAT (Organic Acids Test) reveals oxidative stress markers, mitochondrial function, and the nutritional deficiencies that diet quality alone doesn't reliably map. What you're eating tells you what you intend. The test tells you what's actually reaching your cells.
Talk to the Concierge