The study the geroscience field has been promising for ten years quietly appeared as a preprint in July, was last revised on 19 October, and has still not been peer-reviewed. Nobody put out a press release. Given what is in it, that is a little odd — and given how it will eventually be summarised, it is worth reading before the headline settles.
The trial randomised 145 older adults — average age 72, 48% women, all with glucose intolerance but none with diabetes — to metformin titrated to 1,000 mg twice daily, or to placebo, for two full years.[1] It ran at the University of Texas Health Science Center at San Antonio and finished collecting data in February 2024.[2] The author list is not a minor one: Steve Horvath, who built the original epigenetic clocks, and James Kirkland, who coined the term senolytics, are both on it.
And the headline result is that metformin did not slow frailty. Except that it did. Which one you believe depends entirely on which ruler you pick up — and that turns out to be the real story.
The endpoint that failed because the drug worked
The trial’s pre-registered primary outcome was the Fried frailty phenotype, the standard geriatric instrument. It scores five things: unintentional weight loss, exhaustion, low physical activity, slow gait speed and weak grip.[1]
In year one, the metformin group’s Fried score moved in the wrong direction — 0.72 points per year worse than placebo, with a p-value of 0.0011. On the face of it, a clean negative result. In year two it reversed and stabilised (−0.33 per year, p=0.056).
Then the authors did the obvious check. Metformin is well known to cause weight loss, and it did here: −5.7 kg on drug versus −2.3 kg on placebo (p=0.0002). So they recalculated the Fried score with the weight-loss criterion removed. The difference between groups vanished entirely.
Read that again, because it is a genuinely strange thing to happen in a drug trial. The instrument was built to catch frail older people wasting away. It cannot tell the difference between that and a prediabetic 72-year-old losing five kilos on a glucose-lowering drug. The endpoint did not detect harm; it detected the pharmacology and scored it as decline. Anyone who has watched the muscle-loss debate around GLP-1 drugs will recognise the shape of the problem: weight coming off is not automatically good or bad, and a score that treats it as one thing will mislead you.
The measure that did move
The trial also tracked frailty a second way: a deficit-accumulation index built from 95 separate age-related problems — symptoms, diagnoses, lab values, functional limitations. Rather than five gates, it asks what fraction of 95 things have gone wrong.
On that measure metformin produced a steady improvement of −0.006 per year versus placebo (p=0.0222), and it held up after adjusting for covariates including body mass index.[1] So it is not a weight-loss artefact in the other direction either.
Honesty about the size: 0.006 per year on a 0–1 scale is roughly half a deficit out of 95, per year. Nobody feels that. What matters is the direction and the fact that the placebo group drifted upward while the drug group did not — over two years, in people who were not frail to begin with. A brake, not a reversal.
What the clocks said
This is the part that will travel. Blood DNA methylation was run through principal-component versions of two epigenetic age clocks. PC-Horvath2 slowed by 0.40 biological years per calendar year (p=0.014); PC-Hannum by 0.33 (p=0.047).[1] Over two years, that is roughly eight to ten months of biological age not accrued.
Three caveats, in order of weight. First, a p-value of 0.047 on a secondary outcome in a 145-person trial is a result that wants replication, not a conclusion. Second, these are two clocks out of many available, and clock-to-clock disagreement is routine. Third and most importantly: moving a clock is not the same as benefiting a patient. Regulators draw a hard line between a marker that predicts risk and a marker proven to carry treatment benefit, and essentially no aging clock has cleared the second bar — a distinction that sits underneath the entire question of whether aging can ever be an approved indication, and under most age-reversal claims sold to consumers.
Why anyone expected this to work
The case for metformin was never built on human trials, because until now there weren’t any. It rested on two things.
One was a 2013 study from the National Institute on Aging: mice given 0.1% metformin in their diet from middle age lived longer and stayed healthier, with better physical performance and less chronic inflammation. The same paper reported that a ten-times-higher dose was outright toxic.[3] One species, male mice, one dose that worked and one that poisoned them.
The other was a 2014 analysis of UK primary-care records covering 78,241 people on metformin, 12,222 on a sulphonylurea and 90,463 matched people without diabetes. Adjusted median survival in the non-diabetic controls came out 15% lower than in the diabetics taking metformin (survival time ratio 0.85, 95% CI 0.81–0.90).[4] That single counter-intuitive line — a drug that apparently outran not having the disease — is most of why metformin became a longevity story at all.
It is also a result nobody should take at face value. Metformin is given to healthier diabetics; sicker ones get other drugs or insulin. The comparison is doing work that random assignment is supposed to do. That was the explicit motivation for the TAME proposal in 2016 — a large randomised trial of metformin against a composite of age-related disease, designed by Nir Barzilai and colleagues precisely to settle this.[5] TAME has still not been funded and run. A 145-person trial is what the field has instead.
The catch nobody in longevity circles likes
If you are healthy, exercising, and considering metformin as an insurance policy, there is one finding you should weigh above all of the above. In a double-blind trial, 53 adults averaging 62 did twelve weeks of aerobic training on either metformin or placebo. Metformin blunted the gain in VO2 max and in whole-body insulin sensitivity, and abolished the training-induced improvement in skeletal muscle mitochondrial respiration.[6]
That is a direct collision. VO2 max is one of the strongest predictors of mortality we have, and training response is the intervention with the best evidence behind it in all of longevity. A pill that costs pennies and shaves a bit off your adaptation to the thing that actually works is a bad trade for someone whose metabolism is fine. The response in that study was highly variable — some people were unaffected — but there is no test to tell you in advance which one you are.
What this result is worth
Metformin remains an excellent drug for the people it is licensed for. What this trial adds is narrower and more interesting than “metformin slows aging”: in older adults with prediabetes, two years of it stopped a deficit index from drifting upward and nudged two methylation clocks, without obvious harm, while failing the endpoint it was registered against for a reason that says more about the endpoint than the drug.
That is a real contribution to a field that mostly trades in mouse curves and repurposed-drug wish lists. It is not evidence that a healthy 45-year-old should be taking it, and the authors do not claim it is. Compare it honestly with rapamycin, where the animal data are far stronger and the human data are similarly thin, and the picture is the same one the whole field keeps arriving at: the biology is plausible, the trials are small, and the gap between them is where the marketing lives.
Common questions
Does metformin slow aging in humans?
One randomised trial of 145 older adults with prediabetes found that two years of metformin improved a 95-item frailty deficit index and slowed two epigenetic clocks by roughly 0.3–0.4 biological years per year, while missing its primary frailty endpoint. That is a signal, in one population, in a preprint that has not been peer-reviewed. It is not an established effect, and no drug is approved anywhere for an aging indication.
Should a healthy person take metformin for longevity?
The evidence does not support it, and there is a specific reason for caution: in a randomised trial, metformin blunted the improvement in VO2 max and muscle mitochondrial adaptation that older adults otherwise got from twelve weeks of aerobic training. Metformin is a prescription medicine with real side effects, including long-term vitamin B12 depletion. This is a conversation for a doctor who knows your metabolic numbers, not a supplement decision.
Why did the metformin trial fail its main endpoint?
Because the Fried frailty phenotype counts unintentional weight loss as a sign of frailty, and metformin caused 5.7 kg of weight loss versus 2.3 kg on placebo. When the researchers recalculated the score without the weight criterion, the difference between drug and placebo disappeared — meaning the apparent worsening in year one was the weight change, not a decline in function.