If you've been hoping a daily multivitamin or a cocoa flavanol capsule might quietly turn down the dial on Alzheimer's-related changes in your blood, here is the news: in one of the most closely watched supplement trials in older adults, it didn't. Over two years, neither intervention measurably shifted three key neurodegeneration biomarkers. The null result is still worth understanding.
What the COSMOS biomarker analysis tested
The COcoa Supplement and Multivitamin Outcomes Study (COSMOS) is a randomized, placebo-controlled 2 × 2 factorial trial. Older US adults were assigned to a multivitamin-multimineral (MVM), a cocoa extract (CE), both, or neither. The participants were free of major cardiovascular disease and recently diagnosed cancer.
- Multivitamin: Centrum Silver
- Cocoa extract: 500 mg/day, including 80 mg of (-)-epicatechin
This ancillary analysis, in the Journal of Alzheimer's Disease from a Brigham and Women's Hospital and Harvard team, looked at COSMOS-clinic participants. It measured two-year changes in three blood-based biomarkers:
- Neurofilament light chain (NfL), a protein linked to nerve-fiber injury
- N-terminal tau fragment (NT1), a tau-related marker
- Oligomeric amyloid-β (oAβ), the clumping-prone form of amyloid
The analysis included 506 people with NT1 or NfL data and 498 with oAβ data.
The headline finding: nothing moved
The researchers compared yearly percentage change in each biomarker between active supplement and placebo. None of the six comparisons was statistically significant.
Multivitamin vs. placebo
| Biomarker | Mean difference in yearly % change | 95% CI | p-value |
|---|---|---|---|
| NT1 | 3.7% | -3% to 10.8% | 0.29 |
| NfL | 3.1% | -1.9% to 8.3% | 0.23 |
| oAβ | -8.9% | -23.9% to 9% | 0.31 |
Cocoa extract vs. placebo
| Biomarker | Mean difference in yearly % change | 95% CI | p-value |
|---|---|---|---|
| NT1 | -0.7% | -7.1% to 6.1% | 0.83 |
| NfL | 0.5% | -4.4% to 5.6% | 0.85 |
| oAβ | -5.9% | -21.4% to 12.6% | 0.51 |
The authors' conclusion is plain: daily MVM or CE, compared with placebo, did not affect two-year changes in NfL, NT1, or oAβ.
How to read these numbers
The cocoa extract results for NfL and NT1 sit very close to zero, with p-values of 0.83 and 0.85. That is about as flat as a result gets.
The multivitamin estimates for NfL and NT1 were slightly positive (3.7% and 3.1%). Higher values on these markers are generally not what you'd want. But the confidence intervals comfortably include zero, so these are not evidence of harm either. They look like statistical noise.
The oAβ results lean negative for both interventions (-8.9% and -5.9%). That might tempt an optimistic reading, but the intervals are wide, spanning roughly -24% to +13%, and the p-values are 0.31 and 0.51. The trial can't tell a real small effect from chance here. Calling this a signal would overstate what the data show.
Why a null result still matters
MVM and cocoa flavanols have both been floated as possible ways to slow cognitive aging, but the biological pathways are unclear. This analysis tested one plausible route: whether the supplements change blood markers of neuronal injury, tau, or amyloid. Over two years, in this group, they didn't.
That narrows the story. If either supplement benefits cognition, as other COSMOS reports have explored, this study gives no evidence that it works by shifting these three biomarkers. The benefit, if real, may act through other mechanisms. It may also be too small or too slow to appear in a two-year window.
Practical takeaways
Dose used in the trial: The cocoa extract provided 500 mg/day of cocoa flavanols, including 80 mg of (-)-epicatechin. The multivitamin was one Centrum Silver product. The abstract doesn't specify timing, such as morning versus evening or with meals, so there is no trial-backed schedule to cite.
Who the findings apply to: Older US adults without major cardiovascular disease or recently diagnosed cancer. Results may not carry over to younger people, to people with those conditions, or to people with established Alzheimer's disease.
What not to conclude:
- Don't take a multivitamin or cocoa extract specifically to lower NfL, tau, or amyloid. This trial found no such effect.
- Don't treat the result as proof that these supplements do nothing for brain health. The study measured three blood markers over two years, not clinical dementia or the full range of cognitive outcomes.
- Don't expect a different outcome from higher doses. This trial didn't test them.
Limitations and context
- Duration: Two years is short for processes that unfold over decades.
- Precision: The wide confidence intervals, especially for oAβ, leave room for modest effects in either direction.
- Surrogate endpoints: Blood biomarkers are informative but are not the same as preserving memory or preventing dementia.
- Funding and disclosures: The authors report that COSMOS received supplement and packaging donations and infrastructure support from Mars Edge and from Pfizer Consumer Healthcare (now Haleon). Some investigators also disclosed other industry grants and honoraria. The trial's null findings run against any commercial incentive, which, if anything, adds to their credibility.
The bottom line
In a randomized trial of roughly 500 older adults, a daily multivitamin or 500 mg/day of cocoa extract flavanols did not significantly change two-year trajectories of NfL, NT1, or oligomeric amyloid-β. These supplements may have other uses, but changing Alzheimer's-related blood biomarkers is not supported by this trial.