The Daily Multivitamin That Slowed Biological Aging in the COSMOS Trial
COSMOS randomised more than 21,000 older US adults to a standard multivitamin, a cocoa extract, or placebo. In a 1,000-person subset with DNA methylation measured at baseline, one and two years, the multivitamin slowed the rise of the second-generation clocks GrimAge and PhenoAge by roughly 2.7 to 5.1 months over two years. The researchers stress the standardised effect is modest, and that the benefit was clearest in people who already showed accelerated biological aging.
Overview
Dr Howard Sesso, co-principal investigator of COSMOS at Harvard Medical School, and Dr Ce Dong Lee, lead author of the epigenetic clock paper, walk through what is still the largest randomised trial ever to measure changes in epigenetic aging. COSMOS ran as a two-by-two factorial trial testing a common off-the-shelf multivitamin (Centrum Silver) and a cocoa extract supplement against placebo in about 21,500 adults over roughly three and a half years.
A subset of 1,000 participants gave blood at baseline, one year and two years, allowing the team to track three generations of epigenetic clocks. The first-generation clocks, trained to predict chronological age, moved in the expected direction but not significantly. The second-generation clocks GrimAge and PhenoAge, which are trained on mortality and phenotypic aging, showed a statistically significant slowing with the multivitamin.
DunedinPace, a third-generation pace-of-aging clock, trended in the same direction but was developed in a much younger cohort, which may limit how well it applies here. Cocoa extract showed no effect on any clock in this analysis. Subgroup analysis found the clearest benefit in people with accelerated GrimAge at baseline, who also tended to have nutritional deficiencies or insufficiencies. Both researchers repeatedly return to the same framing: this is a low-cost, safe, widely available intervention with a modest measurable signal, not a substitute for a good diet, exercise and regular check-ups.
Key quotes
5These epigenetic clocks were developed to predict different aspects of aging using mathematical algorithms.
When we look at the standardised effect, it is only about 0.03. So it is just a modest effect on delays of biological aging.
In truth, it is probably less about one specific component, but hopefully them working in concert with each other.
No single pill is going to replace a healthy lifestyle.
Prevention can never start early enough.
Key ideas
9COSMOS was built as a two-by-two factorial trial
About 21,500 older US adults were randomised to a standard multivitamin, a cocoa extract, both, or placebo, with roughly three and a half years of follow-up. The design lets each intervention be read independently.
The epigenetic arm is a 1,000-person subset
One thousand participants provided blood at baseline, one year and two years for DNA methylation analysis. It remains the largest trial-based analysis of change in epigenetic clocks.
Multivitamin and cocoa did not interact
The team saw no compelling evidence that the two supplements influenced each other. That absence is a design strength, because it lets each effect be estimated cleanly.
Three generations of clocks measure different things
First-generation clocks predict chronological age, second-generation clocks such as GrimAge and PhenoAge are trained on mortality and phenotypic aging, and third-generation DunedinPace estimates the speed of aging.
Only the second-generation clocks moved significantly
The multivitamin slowed the increase in GrimAge and PhenoAge. First-generation clocks showed direction without significance, which matches earlier work showing they respond poorly to interventions.
DunedinPace may not translate to an older cohort
The pace-of-aging clock was built in a cohort aged roughly 20 to 45, while COSMOS enrolled older adults carrying more accumulated aging burden. The direction was favourable but not statistically significant.
Two to five months over two years, but a small standardised effect
The second-generation clocks showed roughly 2.7 to 5.1 months less biological aging over two years. The standardised effect is about 0.03, and whether it maps onto lifespan is still unanswered.
The product was an ordinary shelf multivitamin
COSMOS deliberately tested Centrum Silver, a wide-spectrum, low-dose formulation, not a mega-dosed or herbal blend. The point was to test what people actually take.
Accelerated agers benefited most
Participants with accelerated GrimAge at baseline showed the strongest response, and they were also more likely to have nutritional deficiencies or insufficiencies. Existing risk factors such as hypertension pointed the same way.
Practical takeaways
6- 1
Simple beats specialised 43:30
If a multivitamin fits your situation, the researchers lean toward a basic wide-spectrum product from a major brand rather than a hyper-specialised formulation.
- 2
Treat it as a floor, not a ceiling 46:00
The multivitamin raised nutritional biomarkers such as vitamin D, B12 and lutein from low toward adequate. It is about closing gaps, not stacking mega doses.
- 3
Know where you start from 49:30
The people who moved most were those already aging faster or carrying risk factors. Baseline matters more than the supplement itself.
- 4
Consumer epigenetic tests are still early 38:00
Dr Sesso considers single-point-in-time consumer clock tests premature. What is missing is data on how methylation naturally evolves within one person over time.
- 5
Start with the whole picture 53:20
Diet quality and diversity, movement, social connection and regular preventive check-ups come first. Only against that background does a supplement question make sense.
- 6
Earlier is likely better 55:00
Both researchers suspect middle-aged adults have more room to respond, which is why the next generation of trials is moving the window forward.
Topics & chapters
15Why this trial matters
The host frames the three headline points: the scale of COSMOS, the epigenetic subset, and the fact that the multivitamin outperformed the cocoa extract.
What COSMOS actually was
A large randomised trial in about 21,500 older US adults testing a multivitamin and a cocoa extract against placebo in a factorial design.
The 1,000-person methylation subset
Blood collected at baseline, one year and two years allowed DNA methylation measurement in a substantial subgroup.
No interaction between the two supplements
The interventions were chosen so their effects could be read separately, and the data supported that assumption.
What a methylation clock measures
Dr Lee describes methylation as a switch on gene expression and explains how algorithms turn those patterns into age estimates.
Why clocks rather than raw biomarkers
Epigenetic clocks are currently the best-validated aging biomarkers, and they capture broad effects across systems.
What a better clock reading does not tell you
Dr Sesso cautions that we do not yet know how a clock improvement translates into sleep, energy, blood pressure or lipids.
Results by clock generation
GrimAge and PhenoAge slowed with the multivitamin, first-generation clocks did not, and cocoa extract showed no clock effect.
The DunedinPace caveat
The pace clock was trained in a younger cohort, so its generalisability to older COSMOS participants remains open.
How big is the effect really
Roughly 2.7 to 5.1 months over two years sounds large, but the standardised effect is about 0.03 and should not be over-interpreted.
Are consumer epigenetic tests worth it
Not yet mainstream, in Dr Sesso's view. Within-person natural history studies are the missing piece.
What is in the tablet
A standard shelf Centrum Silver with essential vitamins and minerals at low doses plus a few carotenoids, no mega dosing, no herbals.
The insurance policy framing
Host and guests discuss the multivitamin as inexpensive nutritional insurance, and the limits of that metaphor.
Who benefits most
Subgroup analysis points to those with accelerated GrimAge at baseline and to participants with existing risk factors.
Funding, GrimAge components and what comes next
How industry, government and academia co-funded the work, what the GDF15 and telomere signals mean, and the Singapore-based Sedera trial in middle-aged adults.