A cheap, widely available multivitamin may nudge the body's biological aging clocks in the right direction. A new prespecified ancillary analysis of the COSMOS randomized trial, published in Nature Medicine in 2026, found that a daily multivitamin-multimineral (MVM) modestly slowed the pace of two second-generation epigenetic clocks in older adults. Cocoa extract, by contrast, did nothing measurable. The effects are small, and nobody yet knows whether they translate into healthier or longer lives.
What the researchers tested
COSMOS (the COcoa Supplement and Multivitamin Outcomes Study) is a large randomized trial that tested two interventions in older adults: a daily multivitamin and a daily cocoa extract. Earlier work from the program suggested that these supplements might benefit several age-related chronic conditions. What remained unclear was whether they act on the biology of aging itself.
This ancillary study took a direct look. It enrolled 958 participants (482 women and 476 men) and measured DNA methylation patterns over two years. The interventions were:
- Multivitamin-multimineral: Centrum Silver, a standard over-the-counter formulation designed for adults over 50, taken daily.
- Cocoa extract: 500 mg of cocoa flavanols per day, including 80 mg of (-)-epicatechin.
- Placebo comparison arms.
Biological aging was assessed with five DNA methylation measures: PCHannum, PCHorvath, PCPhenoAge, PCGrimAge and DunedinPACE. The "PC" versions are principal-component-based clocks, designed to be more reliable and less noisy than earlier iterations.
The headline results
Compared with placebo, the multivitamin modestly reduced the rate of increase in two second-generation clocks, which are trained on health and mortality outcomes rather than just chronological age:
- PCGrimAge: a between-group difference of -0.113 years per year (95% CI -0.205 to -0.020; P = 0.017).
- PCPhenoAge: a difference of -0.214 years per year (95% CI -0.410 to -0.019; P = 0.032).
In plain terms, over the typical course of a year, people taking the multivitamin saw their PCGrimAge advance about 0.11 years less than those on placebo, and their PCPhenoAge about 0.21 years less. Extrapolated naively across the two-year trial, that implies a gap of a few months on the clocks. But that framing deserves caution: the confidence intervals are wide, with the lower bounds sitting very close to zero, and an epigenetic clock difference is not the same as a person being literally months younger.
The other three clocks, PCHannum, PCHorvath and DunedinPACE, did not show a reported significant multivitamin effect in the abstract.
Bigger effect in people aging fast
The most interesting finding may be who benefited most. When the team split participants by baseline biological aging status, the multivitamin's effect on PCGrimAge was stronger among those with accelerated aging at the start (-0.236 years per year; 95% CI -0.380 to -0.091). Among those with normal or decelerated aging, the effect was essentially nil (-0.013; 95% CI -0.130 to 0.104). The interaction was statistically significant (P = 0.018).
This pattern fits a plausible biological story: people whose bodies are already showing signs of faster aging may have more room to improve, potentially because of marginal nutrient inadequacies. It echoes a recurring theme in nutrition research, where supplements tend to help most those who are least replete. It is, however, a subgroup finding, and subgroup analyses should be treated as hypothesis-generating until replicated.
Cocoa flavanols: a letdown
Cocoa extract has attracted attention for cardiovascular and cognitive benefits, and flavanols are often marketed as anti-aging compounds. In this trial, however, 500 mg of daily cocoa flavanols had no effect on any of the five epigenetic clocks. That does not rule out benefits through other mechanisms, such as blood vessel function, but it does suggest cocoa is not slowing aging as captured by methylation clocks, at least at this dose over two years.
How much should we read into this?
The authors themselves call the effects "statistically significant but small" and "encouraging," and they explicitly state that more studies are needed to determine clinical relevance. Several caveats matter:
- Epigenetic clocks are surrogate markers. They correlate with disease and mortality risk across populations, but it is unproven that shifting a clock through an intervention changes an individual's outcomes.
- Effect sizes are modest and the confidence intervals are wide. Two of five clocks moved, and the P values (0.017 and 0.032) are not overwhelming, especially given multiple clocks were tested.
- Two years is short. Whether the gap widens, holds or fades over longer periods is unknown.
- Mechanism is unclear. The study cannot say which vitamins or minerals in the formula are responsible.
The study does offer a possible link to earlier COSMOS findings that multivitamins may benefit certain age-related conditions. If the clock effects reflect real slowing of biological aging, they might help explain those results. That remains a hypothesis, not a conclusion.
Practical takeaways
- What was taken: A once-daily Centrum Silver, an inexpensive, mainstream multivitamin, not a boutique longevity stack.
- Who might benefit most: Older adults who are already aging faster biologically appear to gain more, though most people cannot easily know their epigenetic status, and commercial clock tests vary in quality.
- Cocoa flavanols: Do not expect a methylation-clock benefit from cocoa extract. Enjoy it for other reasons if you like, but this trial gives no support for it as an epigenetic anti-aging tool.
- Keep expectations realistic: A multivitamin is not a substitute for diet quality, exercise, sleep, or medical care, and the evidence for hard clinical benefit from this specific effect is still pending.
Bottom line
The COSMOS epigenetic study is one of the larger randomized tests of whether a common supplement touches the biology of aging, and it produced a small but real signal for the multivitamin. It is a promising piece of a puzzle rather than a verdict. Until longer trials show that slower clocks equal better health outcomes, the finding is best viewed as intriguing, low-cost, and unproven in its clinical significance.
