The COSMOS Trial: What 21,000 People Revealed About Cocoa Extract
Roon Team
The COSMOS Trial: What 21,000 People Revealed About Cocoa Extract
Cocoa got its biggest test yet, and the headline most people remember is wrong.
The COSMOS trial cocoa study followed more than 21,000 older adults for roughly three and a half years to answer a simple question: does a daily cocoa extract supplement actually protect your heart? The answer depends entirely on which line of the results table you read. Skim the top, and cocoa looks like a miss. Read carefully, and a more interesting story shows up.
This is a case study in how one trial can be honest and confusing at the same time. Here is what the data really said.
Key Takeaways
- COSMOS randomized 21,442 older adults to a cocoa extract supplement or placebo and tracked cardiovascular events for a median of 3.6 years.
- The primary endpoint, total cardiovascular events, dropped 10% but missed statistical significance.
- A secondary endpoint, death from cardiovascular disease, fell 27%, which is the number most headlines ran with.
- Among people who actually took their pills consistently, the effects looked stronger, but that analysis comes with caveats.
- One trial rarely settles a question. COSMOS narrowed it.
What the COSMOS Trial Cocoa Study Actually Tested
COSMOS stands for the Cocoa Supplement and Multivitamin Outcomes Study, and it was built to be taken seriously. Participants were randomly assigned to a cocoa extract supplement of 500 mg flavanols per day, including 80 mg of (–)-epicatechin, or placebo.
The design was rigorous. The study used a randomized, double-blind, placebo-controlled, 2 × 2 factorial trial of a cocoa extract supplement and a multivitamin supplement to reduce the risk of CVD and cancer. That 2×2 setup let researchers test cocoa and a multivitamin at the same time in the same people.
The population was specific and large. The nationwide study population included 21,442 U.S. women aged 65 years and older and men aged 60 years and older without baseline myocardial infarction, stroke, or a recent cancer diagnosis.
A cocoa extract supplement is not a chocolate bar. The dose here was standardized cocoa flavanols, with epicatechin as the bioactive that researchers care about most. That epicatechin supplement angle matters, because it is the compound thought to support blood vessel function.
The Primary Endpoint: A Near Miss
Here is the headline number, stated plainly: the cocoa group had 10% fewer total cardiovascular events, but the result was not statistically significant.
During a median follow-up of 3.6 years, 410 participants taking cocoa extract and 456 taking placebo had confirmed total cardiovascular events, with a hazard ratio of 0.90 and a P value of 0.11. In trial language, a P value above 0.05 means researchers cannot rule out chance as the explanation.
So on its main question, COSMOS came up short. That is the version of the story you will see in a one-line summary.
It is also the version that misses the most clinically interesting finding.
The Secondary Endpoint Everyone Quoted
The number that traveled the furthest was the drop in cardiovascular death. The large-scale randomized trial found signs of preventive cardiovascular effects for cocoa flavanols, including a 27 percent reduction in the secondary endpoint of cardiovascular death.
A 27% reduction in dying from heart disease sounds dramatic, and it is meaningful. But the word doing heavy lifting in that sentence is secondary.
Secondary endpoints are real measurements, but they are not what a trial is designed to prove. When the primary endpoint misses and a secondary one hits, scientists treat the secondary result as a strong hint, not a verdict. It generates the next hypothesis. It does not confirm the last one.
This is the single most important reading skill the COSMOS study results demand. The cocoa flavanols heart story lives in the gap between the primary and secondary lines.
What Happened Among People Who Took the Pills
There is a third layer, and it pushes the cocoa case the other way. Researchers also looked at participants who stuck with the protocol.
In analyses among adherent participants taking at least 75% of their study pills, the hazard ratio for the primary endpoint of total CVD events was reduced by 15%, and CVD death was reduced by 39%. Those are bigger effects than the full-group numbers.
This is called a per-protocol analysis, and it is seductive. If you only count people who actually took the supplement, of course the effect looks stronger.
The catch is that people who take their pills reliably tend to be different from people who do not. They often exercise more, eat better, and follow medical advice closely. That "healthy adherer" effect can inflate a benefit that the supplement did not fully earn. So the 39% number is encouraging and unreliable at the same time.
Safety, at least, was clean. Overall, the cocoa extract intervention appeared safe and well tolerated.
How the Three Numbers Compare
The same trial produced three different impressions depending on which analysis you read. Here is the full picture in one view.
| Result | What it measured | Effect size | Statistically significant? | How much weight it carries |
|---|---|---|---|---|
| Primary endpoint | Total cardiovascular events, all participants | 10% reduction (HR 0.90) | No (P = 0.11) | Highest. This is the trial's main test. |
| Secondary endpoint | Cardiovascular death, all participants | 27% reduction | Yes, but secondary | Suggestive. Generates new hypotheses. |
| Per-protocol, primary | Total CVD events, adherent users | 15% reduction (HR 0.85) | Yes, in subgroup | Lower. Vulnerable to healthy-adherer bias. |
| Per-protocol, death | CVD death, adherent users | 39% reduction (HR 0.61) | Yes, in subgroup | Lowest. Strongest bias risk. |
Notice how the cocoa supplement cardiovascular story improves as you move down the table, and how the reliability drops at the same time. That trade-off is the whole point.
So Should You Take a Cocoa Extract Supplement?
The honest answer: COSMOS did not give cocoa a clean win, and it did not rule it out either.
The trial's own authors landed in a measured place. Cocoa extract supplementation did not markedly reduce total cardiovascular events among older adults but reduced CVD death by 27%, and potential reductions in total cardiovascular events were supported in per-protocol analyses.
If you want a defensible takeaway, it is this. A cocoa flavanol supplement appears safe, and the signal for cardiovascular death is interesting enough to justify more research. That is different from proof that it extends your life.
For most people, the boring advice still wins. Eat well, move daily, and treat supplements as a small addition rather than the foundation. A cocoa extract supplement is a reasonable bet for some older adults, not a guarantee for anyone.
Conclusion
COSMOS is valuable precisely because it refuses to give a tidy answer. One trial, one population, three layers of results, and a lesson about reading each line for what it is worth.
The primary endpoint missed. The secondary endpoint and the per-protocol numbers were encouraging but came with real limits. Anyone who tells you the study "proved cocoa prevents heart disease" skipped to the part of the table they liked best.
That is the actual skill worth keeping. When a study lands in the news, ask what was the main question, did it hit, and how much of the excitement is riding on a secondary or subgroup result. Most of the time, the truth is sitting one line below the headline.
Frequently Asked Questions
Did the COSMOS trial prove cocoa prevents heart disease?
No. The trial's primary endpoint, total cardiovascular events, showed a 10% reduction that was not statistically significant. The strong-sounding 27% drop in cardiovascular death was a secondary endpoint, which scientists treat as a promising signal rather than proof. COSMOS narrowed the question and supported more research, but it did not confirm that cocoa extract prevents heart disease in the general population.
What is the difference between a primary and secondary endpoint?
A primary endpoint is the main question a trial is statistically designed to answer, and it carries the most weight. Secondary endpoints are additional measurements collected along the way. When a primary endpoint misses but a secondary one hits, as happened in COSMOS, the secondary result is treated as a hypothesis for future studies, not as a confirmed finding. The distinction is the core lesson of the trial.
How much cocoa flavanol did COSMOS participants take?
Participants in the cocoa group took a daily supplement delivering 500 mg of cocoa flavanols, including 80 mg of epicatechin, the bioactive compound researchers focus on for blood vessel function. This is a standardized cocoa extract supplement, not equivalent to eating chocolate. A typical chocolate bar contains far less flavanol and much more sugar and fat, so the trial findings do not translate to dessert.
Why did adherent participants show bigger benefits?
In the per-protocol analysis, people who took at least 75% of their pills showed a 15% reduction in total cardiovascular events and a 39% reduction in cardiovascular death. Those numbers are larger than the full-group results, but they are vulnerable to the healthy-adherer effect. People who take supplements consistently often have healthier habits overall, which can inflate the apparent benefit.
Was the cocoa extract supplement safe?
Yes. The COSMOS investigators reported that the cocoa extract intervention appeared safe and well tolerated over the median 3.6-year follow-up. Safety was not the concern in this trial. The open question was efficacy, specifically whether the cardiovascular benefits were large and reliable enough to recommend cocoa extract broadly for heart protection.
Does COSMOS apply to younger adults?
Not directly. The trial enrolled U.S. women aged 65 and older and men aged 60 and older, none with prior heart attack or stroke at baseline. Results from an older population do not automatically extend to younger adults, who have different baseline cardiovascular risk. Applying the findings to people in their 30s or 40s would stretch the evidence beyond what the study tested.
Reading the Evidence, Not Just the Headline
COSMOS is a clean example of why the gap between a primary and a secondary endpoint matters. The exciting number was real, but it was not the number the trial was built to prove. Knowing the difference is what separates a careful reader from someone repeating a press release.
That is the standard we hold ourselves to at Roon. We make a sublingual cognitive performance pouch built on four ingredients with published evidence behind them: 80 mg caffeine, 60 mg L-theanine, 25 mg methylliberine (Dynamine), and 5 mg theacrine (TeaCrine), designed for a 5 to 10 minute onset and 6 to 8 hours of focus without the jitters, crash, or tolerance creep.
Roon is not a heart medication and not a substitute for the diet and exercise that actually move cardiovascular risk. It is a focus tool, and we write about studies like COSMOS so you can read the next big headline the way a scientist would. If that approach fits how you think, try Roon and see how it reads.
Written by Roon Team
