Can a Multivitamin Slow Biological Age? What COSMOS Measured—and What It Did Not

A multivitamin has not been shown to make the body younger. A 2026 COSMOS substudy did find that one specific daily formulation modestly slowed the increase in two of five epigenetic clocks over two years. The difference was statistically significant but small, and it does not mean longer life, dementia prevention or preserved function.

The clinically useful question is not “Can I remove four months from my age?” It is: Do I have a nutritional need that justifies supplementation, and how will I know it has been corrected?

What the COSMOS substudy actually did

COSMOS was a randomized, double-blind, placebo-controlled trial with a factorial design that tested a multivitamin–multimineral and cocoa extract separately. The epigenetic-age analysis was prespecified and measured blood DNA methylation at baseline, one year and two years.

Question	What the study did	What it means
Who was studied?	958 people; women aged ≥65 and men aged ≥60, mean age 70.2	It does not establish the effect in younger adults, frail people or other populations
What did the active group take?	One daily tablet of the US Centrum Silver trial formulation	This is not evidence for every product called a multivitamin
For how long?	Two years with serial samples; 91.9% met the trial's adherence definition	It does not tell us what happens after stopping or after decades
What was measured?	Five blood DNA-methylation algorithms	These are research biomarkers, not clinical outcomes
What was found?	A small difference in PCGrimAge and PCPhenoAge; null results in three clocks	An interesting signal that needs replication and clinical translation
What about cocoa?	500 mg/day cocoa flavanols, including 80 mg epicatechin, did not change the five clocks	A plausible mechanism does not guarantee movement in a biomarker

The molecular cohort was unusually selected: participants remained free of myocardial infarction, stroke, coronary revascularization and invasive cancer during the period analysed. That supports a healthy-aging question but limits extrapolation to people with established disease or new clinical events.

From −0.113 years to “four months younger”

Compared with placebo, the annual difference was −0.113 years for PCGrimAge (95% CI −0.205 to −0.020) and −0.214 for PCPhenoAge (−0.410 to −0.019). Extended over two years, those estimates equal about 2.7 and 5.1 months. The widely repeated “about four months” is a shorthand between them.

No participant literally became four months younger. The number describes separation between trajectories calculated by algorithms. The standardized effects were d=−0.033 and d=−0.038: small even within the biomarker itself.

The PCGrimAge difference was larger among participants with accelerated epigenetic age at baseline (−0.236 years/year) and close to zero among the rest (−0.013). That subgroup result is hypothesis-generating, not a treatment-selection test. The study did not adjust every clock and subgroup comparison for multiplicity, and there is no evidence that prescribing from this result improves clinical health.

Why two clocks changed and three did not

PCHorvath and PCHannum were developed mainly to estimate chronological age. PCPhenoAge and PCGrimAge incorporate relationships with phenotype, protein surrogates, smoking exposure and mortality. DunedinPACE aims to estimate the pace of multisystem change. Different responses are not automatically contradictory because the clocks do not measure the same construct.

That diversity also demands restraint. If an intervention moves two models but not three, we cannot simply pick the favourable clock and call it a person's “true age.” Progevita's guide to choosing a biological age test explains why reproducibility, tissue, algorithm and a decision that can genuinely change matter more than a striking number.

What the clinical COSMOS outcomes add

In the main trial, 21,442 older adults were followed for a median of 3.6 years. The multivitamin did not significantly reduce the composite cardiovascular outcome (HR 0.98; 95% CI 0.86–1.12), total invasive cancer (HR 0.97; 0.86–1.09) or all-cause mortality (HR 0.93; 0.81–1.08). Other COSMOS substudies have reported modest cognitive signals.

Both layers can be true: a biomarker may move before we know whether the change causes or predicts a benefit people can feel. For dementia risk reduction, a multivitamin does not replace physical activity, blood-pressure care, hearing, sleep, social connection or adequate food.

Decision table: biomarker or nutritional indication?

Situation	What adds value	What to avoid
Varied diet, no symptoms or deficiency risk	Review diet, fortified foods and supplements already used; another pill may not be needed	Taking it solely to lower a biological-age score
Confirmed deficiency or a focused clinical suspicion	Identify the cause, nutrient, dose, duration and follow-up test	Hiding a specific problem inside a low-dose mixture
Vegan diet, malabsorption, bariatric surgery or very restricted intake	Individual protocol and monitoring; needs may include B12, iron, folate, vitamin D or others	Assuming any general product covers specific needs
Pregnancy or pregnancy planning	A prenatal plan agreed with the healthcare team	Excess preformed vitamin A or a non-prenatal formula
Warfarin, kidney/liver disease, active cancer treatment or polypharmacy	Professional review of ingredients and interactions before starting	Reading “natural” as “risk free”
High epigenetic age without a demonstrated deficiency	Confirm clinical risks that already change care and review sustainable habits	Using a clock to prescribe a brand or megadose

If supplementation is considered: a safer protocol

1. Define the job: correct a deficiency, cover a life stage or address inadequate intake. “Longevity” is not a specific indication.
2. Inventory everything: medicines, other supplements, drinks and fortified foods. Duplication often matters more than one tablet.
3. Use the smallest sufficient intervention: when one nutrient is low, focused replacement can be clearer than twenty ingredients.
4. Check product and dose: formulation, lot, quality controls, daily dose and national regulatory status. Multivitamins are not standardized across brands or countries.
5. Set a review point: symptoms, tolerance, adherence, diet and laboratory testing only when it will change action. Repeating an epigenetic clock every few weeks mostly measures noise.

Adequate food remains the foundation. Our evidence-based eating guide reviews structure and quality without magic lists; the longevity biomarkers guide prioritizes measurements that can change a decision.

Risks, interactions and warning signs

• Warfarin: vitamin K can alter anticoagulation. The goal is a stable intake and coordinated changes, not removing it without advice.
• Pregnancy: excess preformed vitamin A can cause birth defects; an appropriate prenatal formulation matters.
• Smoking or asbestos exposure: high-dose beta-carotene increased lung-cancer risk in previous trials.
• Kidney, liver and iron issues: accumulation or an unnecessary dose can matter. A “50+” formula is not the same as a prenatal or general formula.
• Duplication: stacking a multivitamin, “immune” product, vitamin D, zinc, calcium and fortified foods can exceed tolerable upper levels.

Seek urgent care after breathing difficulty or swelling of the face or throat, marked confusion, fainting, persistent rhythm disturbance, severe vomiting or a large accidental ingestion—especially in a child. For a new but non-urgent symptom, do not keep escalating the dose; review the exact product with a qualified professional.

What changes today—and what does not

COSMOS strengthens a hypothesis: nutritional status may influence some DNA-methylation clocks. It does not turn a multivitamin into an anti-aging treatment, validate buying a clock to decide on one, or allow a US formulation to represent every product sold elsewhere.

At Progevita, the sequence is to listen, measure only when there is a clinical question, interpret in context, act and reassess. If diet, symptoms, medicines and targeted tests reveal a need, it can be corrected with a goal and follow-up. If they do not, the valuable decision may be not adding another intervention.

Sources

1. Li S et al. Multivitamin, cocoa and epigenetic clocks in COSMOS. Nature Medicine. 2026.
2. Sesso HD et al. Main COSMOS multivitamin outcomes for cancer and cardiovascular disease. American Journal of Clinical Nutrition. 2022.
3. NIH Office of Dietary Supplements. Multivitamin/mineral supplements: evidence, safety and interactions.
4. European Commission. Food supplements: rules, permitted sources and safety framework.
5. USPSTF. Vitamin and multivitamin supplementation for cardiovascular disease and cancer prevention. 2022.
6. Bischoff-Ferrari HA et al. Omega-3, vitamin D, exercise and epigenetic clocks in DO-HEALTH. Nature Aging. 2025.

Method: narrative review of the primary trial, its main clinical-outcomes publication and official guidance. Spanish and English search results were reviewed on 27 September 2026. Population, dose, confidence intervals and effect size were retained; a surrogate marker was not presented as a clinical benefit. Educational content, not a prescription.

Transparency: COSMOS received support from the NIH, Mars Edge and Pfizer Consumer Healthcare (now Haleon); FOXO Technologies supported generation and preprocessing of methylation data, and relevant conflicts were declared. According to the authors, funders did not participate in design, collection, analysis or manuscript preparation. This calls for transparency and replication, not automatic dismissal.