A daily multivitamin slows biological aging — a major trial confirms it.

What epigenetic clocks are and why they matter

Epigenetic clocks are algorithms that estimate biological age from patterns of DNA methylation — a chemical modification (addition of a methyl group to cytosine residues) that turns genes on or off without changing the underlying DNA sequence. These methylation patterns shift in predictable ways with chronological age across the genome, and researchers have trained machine learning models on large population datasets to estimate biological age from them.

The key distinction is between chronological age (how many years since birth) and biological age (how old your tissues and systems appear to function). Two people of identical chronological age can have meaningfully different biological ages as measured by these clocks, and higher biological age on several of the validated clocks correlates with earlier all-cause mortality, cardiovascular disease risk, and functional decline — independently of chronological age.

The five clocks measured in the COSMOS analysis span two generations of development:

• First-generation clocks — PCHannum and PCHorvath: trained on chronological age using genome-wide methylation data. They measure how closely the methylation pattern resembles the typical pattern for a given age.
• Second-generation clocks — PCPhenoAge and PCGrimAge: trained on health and mortality outcomes, not just chronological age. They incorporate methylation patterns associated with phenotypic markers of aging (inflammation, immune function, metabolic health) and are more predictive of actual mortality risk than first-generation clocks.
• Pace-of-aging clock — DunedinPACE: estimates the current rate of biological aging (years of biological aging per chronological year), rather than a point-in-time estimate. A DunedinPACE score of 1.0 means aging at the typical rate; above 1.0 means aging faster than chronological average.

The COSMOS trial measured all five. Only two reached statistical significance. That detail matters and will be addressed directly.

The COSMOS trial: who was studied and how

The COSMOS (COcoa Supplement and Multivitamin Outcomes Study) trial is a large-scale randomized controlled trial based at Mass General Brigham, led by Howard Sesso, ScD, MPH, associate director of the Division of Preventive Medicine in the Mass General Brigham Department of Medicine. The parent COSMOS trial enrolled 21,442 adults across the United States aged 60 and older.

The epigenetic aging sub-study — the one published in Nature Medicine in March 2026 — analyzed a subset of 958 randomly selected participants with available blood samples. This is a critical design detail: these were not selected based on health outcomes; they were randomly selected from the parent trial population, which means the epigenetic aging findings are not biased toward people who happened to do well.

The intervention was straightforward: daily Centrum Silver (a standard over-the-counter multivitamin-multimineral formulation) versus placebo, for two years. DNA methylation was measured from blood samples at baseline and after two years. The statistical analysis compared the rate of change in all five epigenetic clocks between the two groups.

Participants had a mean chronological age of approximately 70 years. Both men and women were included. The COSMOS design used a 2×2 factorial structure, also testing a cocoa flavanol extract supplement — that analysis was included in the same Nature Medicine paper, but the multivitamin arm is the primary focus here.

This is not an observational study. It is a two-year randomized controlled trial with 958 participants and an active control arm. That design is the reason the finding has attracted serious scientific attention.

What the data actually shows — the two clocks, and the three that didn't move

The headline result: compared to placebo, the multivitamin group showed statistically significant slowing on two of the five epigenetic clocks after two years.

• PCPhenoAge: between-group difference in yearly change of −0.214 years per year. The multivitamin group's PCPhenoAge advanced more slowly by approximately 0.21 biological years per chronological year.
• PCGrimAge: between-group difference of −0.113 years per year. The multivitamin group's PCGrimAge advanced more slowly by approximately 0.11 biological years per chronological year.

These are the two second-generation mortality-predictive clocks. Their movement in the direction of slowed aging is the signal the researchers emphasized.

The three clocks that did not show statistically significant differences:

• PCHannum: no statistically significant between-group difference
• PCHorvath: no statistically significant between-group difference
• DunedinPACE: no statistically significant between-group difference

This is not a minor caveat — it is a substantive part of the result. The two clocks that moved are the ones most associated with mortality and health outcomes, which is a favorable interpretation. But three clocks measuring different dimensions of biological aging did not move. The effect is not global across all aging-clock metrics.

Four months of aging: what that effect size means

The "four months" figure that appeared in press coverage is a derived interpretation of the per-year slowing: approximately 1.5 to 2 months of aging averted per year of supplementation, cumulated over two years. Over the full two-year study period, the cumulative between-group difference on the most responsive clock (PCPhenoAge) translated to roughly 3–4 months of biological aging.

This is a small effect. It is also a meaningful one, depending on what comparison you use. The most relevant comparison is the effect of interventions at similar cost and accessibility. Centrum Silver retails for approximately $15–25 for a 100-count supply — among the least expensive supplements available. No dietary, exercise, or pharmacological intervention has produced a larger, better-powered epigenetic aging result in a comparable RCT in this age group. That does not make four months transformative; it makes it worth knowing about.

The comparison to sleep, exercise, and diet is appropriate for calibration. Aerobic exercise training interventions show DunedinPACE reductions of roughly 0.04 units in 6-month trials — a comparable order of magnitude for a higher-effort intervention. The DO-HEALTH trial (omega-3, vitamin D, and exercise in older adults) showed effects on first-generation clocks. The multivitamin effect is not uniquely large, but it is real and peer-reviewed in the highest-impact medical journal in its category.

Who benefited most: the accelerated-aging subgroup finding

One of the most clinically actionable findings in the COSMOS Nature Medicine paper is a subgroup analysis: participants who were biologically older than their chronological age at baseline — that is, those with accelerated biological aging at the start of the trial — showed a substantially stronger benefit on PCGrimAge than those who were biologically younger than their chronological age.

This pattern is consistent with a nutritional deficiency or suboptimal micronutrient status hypothesis: people who are aging faster may be doing so partly because of inadequate micronutrient status that a daily multivitamin corrects. The signal is weaker in people who are already biologically well-maintained — consistent with the ceiling-effect logic of nutritional repletion. Replenishing a deficiency produces a measurable effect; adding to an adequate level produces less.

This subgroup finding does not justify using multivitamins as a targeted therapeutic for people with documented micronutrient deficiencies — that is a separate, established clinical domain. What it suggests is that the aging-clock benefit in the COSMOS trial may be concentrated in the portion of the older-adult population that is nutritionally suboptimal, which is not a small group: micronutrient inadequacy is common in adults over 65, particularly for B vitamins, vitamin D, and magnesium.

Mechanism: which micronutrients are pulling the signal

Centrum Silver contains 23 vitamins and minerals. The COSMOS trial cannot isolate which specific micronutrients are responsible for the epigenetic aging effect — it was not designed to do so. This is the core mechanistic limitation of a multivitamin trial.

The leading candidates based on known biology:

• B vitamins — particularly folate (B9), B12, and B6. All three are required for the one-carbon metabolism pathway, which is directly upstream of DNA methylation: the pathway produces S-adenosylmethionine (SAM), the universal methyl donor for DNA methyltransferase enzymes. Deficiency in any of these B vitamins reduces SAM availability and disrupts methylation patterning. The direction of the methylation change associated with B-vitamin deficiency is broadly consistent with accelerated epigenetic aging.
• Vitamin D — has regulatory effects on immune function and inflammation, both of which are incorporated into PCGrimAge (which includes methylation proxies of inflammatory proteins). Vitamin D deficiency is highly prevalent in older adults.
• Zinc and selenium — involved in antioxidant defense and DNA repair mechanisms. Inadequate zinc has been associated with DNA methylation disruption in some experimental models.

The honest answer is that the mechanism is not resolved. Future studies using individual nutrient supplementation arms or serum nutrient measurement as a moderator variable will be needed to decompose the Centrum Silver signal. The COSMOS trial shows that the package works; it does not show which ingredients are responsible.

What this does not prove: the outcome gap between clocks and longevity

Epigenetic clocks predict mortality risk in population cohorts. They do not prove, in any individual or intervention, that slowing the clock translates to living longer or healthier. This distinction is not a trivial hedge — it reflects a genuine open question in the field.

The clocks were trained on observational data: people with higher biological age on PCGrimAge die sooner, on average, in the datasets used for training. But an intervention that lowers PCGrimAge score has not been directly validated against actual lifespan or disease incidence in a randomized trial. The COSMOS trial is not long enough and is not powered for mortality as an endpoint. No epigenetic-clock intervention study currently is.

This gap matters because other interventions have been shown to modify epigenetic clock scores without clearly translating to clinical outcomes — and some interventions that strongly predict longevity in other models (caloric restriction in rodents, for example) produce surprisingly modest epigenetic clock effects in humans. The clocks are the best available surrogate, but they are still surrogates.

The most intellectually honest framing: the COSMOS trial shows that daily multivitamin use in older adults produces a measurable change in two molecular markers of biological aging. Whether that change translates to more years of healthy life requires follow-up studies that do not yet exist.

The COSMOS trial is the most rigorous evidence yet that a $20 supplement produces a measurable shift in molecular aging markers. What it cannot tell you is whether that shift translates to a longer or healthier life. Both facts deserve to be in the same sentence.

Context: COSMOS cognitive findings and the broader picture

The COSMOS epigenetic aging paper is not the first notable result from this trial program. Earlier COSMOS analyses reported that daily multivitamin supplementation improved memory and slowed cognitive aging in older adults across three independent studies — the COSMOS-Mind, COSMOS-Web, and original COSMOS cognitive substudy. These findings, published in the American Journal of Clinical Nutrition and Alzheimer's & Dementia, showed preservation of memory function over two to three years in the multivitamin group.

The cognitive and epigenetic findings from COSMOS are converging on a consistent signal: daily multivitamin supplementation in older adults produces small but measurable benefits across multiple dimensions of aging — cognitive function, memory, and now molecular aging markers. No single COSMOS finding is transformative on its own. The pattern across multiple pre-specified outcomes in a well-powered RCT is harder to dismiss than any individual result.

For context on where this fits in the supplement landscape: the COSMOS multivitamin signal is more robust than the evidence base for most individual longevity-targeted supplements in the same demographic. The NMN (nicotinamide mononucleotide) and NR (nicotinamide riboside) trials in older adults show modest improvements in NAD⁺ levels and some functional markers, but the head-to-head comparison on epigenetic aging clocks does not currently favor expensive single-molecule supplements over a standard multivitamin for most people.

A practical framework

We do not write prescriptions. We write frameworks that you take to a clinician — or use to calibrate your own reading of the evidence.