Multivitamins moved some tests and aging clocks. What does that mean?

The bottom line

Across three related COSMOS cognitive substudies, daily multivitamin-mineral use was associated with small improvements in some cognitive-test scores. The program-level meta-analysis estimated a difference of about 0.07 standard units for global cognition and 0.06 for episodic memory. (COSMOS-Clinic and meta-analysis)

A 2026 substudy then found modestly slower change in two of five DNA-methylation clocks over two years. Those clocks are biomarkers. They do not tell us that participants avoided dementia, stayed independent, developed fewer diseases, or lived longer. (COSMOS clock analysis)

Why this matters

Small standardized effects can be turned into intuitive headlines such as “two years younger” or “four months less aging.” Those translations compare a statistical difference with an age-related slope; researchers did not observe anyone reverse calendar time. A reader deciding what the study means needs the direct outcome first: test scores or methylation patterns.

What researchers did

COSMOS-Mind followed 2,262 adults aged 65 and older with annual telephone cognitive testing for three years. COSMOS-Web used online assessments in 3,562 older adults. COSMOS-Clinic conducted in-person testing in 573 participants and combined the cognitive results across the three substudies.

The 2026 analysis examined serial blood samples from 958 COSMOS participants and tested five DNA-methylation clocks. Because all these reports sit within one parent program, their results are correlated. Agreement across them is useful, but it is not the same as replication by unrelated trials.

What they found

COSMOS-Mind reported standardized differences of roughly 0.07 for global cognition and 0.06 for episodic memory and executive function. COSMOS-Web found a small immediate-recall difference: the multivitamin group gained about 0.71 recalled words compared with 0.44 for placebo. Other tested domains did not all improve. (COSMOS-Mind; COSMOS-Web)

COSMOS-Clinic's own global-cognition estimate was modest and imprecise; the combined COSMOS meta-analysis was statistically clearer. In 2026, two second-generation methylation clocks changed more slowly with the multivitamin, while five clocks had been assessed.

How strong is the evidence?

Randomization strengthens causal interpretation for the measured outcomes. Multiple assessment methods also reduce the chance that one testing format explains everything. But the small effect sizes, correlated parent program, selective differences across cognitive domains, and lack of a direct clinical endpoint limit the conclusion.

The clock analysis is especially easy to overread. Epigenetic clocks are built from methylation patterns associated with age-related outcomes; moving a score does not guarantee moving the outcome itself. Clinical validation requires follow-up showing fewer diagnoses, preserved function, or better survival—not a more favorable surrogate alone.

What this evidence does not show

The COSMOS reports do not demonstrate that a multivitamin prevents dementia, Alzheimer's disease, disability, loss of independence, hospitalization, or death. They do not show that every formulation produces the same result.

They also do not establish human healthspan or lifespan extension. “Slower change in a clock” and “slower aging in daily life” are not interchangeable statements.

Safety, regulation, and funding

Dietary supplements are not preapproved by FDA for safety or effectiveness before sale. A multi-ingredient product can create dose overlap, interactions, or formulation differences that a broad “multivitamin” label conceals. (FDA supplement context)

COSMOS received public and nonprofit support. Pfizer Consumer Healthcare, now Haleon, supplied some study pills and packaging. Commercial summaries emphasizing “years of cognitive aging” are not independent endpoints and do not replace the peer-reviewed effect sizes.

This article is not a recommendation to start, stop, or choose a product. Medication use, diet, diagnosed deficiencies, kidney or liver disease, and other individual factors belong in a clinician or pharmacist discussion.

What happens next

The most useful next evidence would come from independent trials with prespecified clinical outcomes, transparent correction and analysis histories, longer follow-up, and direct measures of dementia, function, and health events. Biomarker findings should be followed to see whether they predict something people can feel or do.