A two-year randomized, placebo-controlled trial of metformin in older adults with glucose intolerance has reported that the diabetes drug slowed the rise of one frailty measure, a deficit accumulation frailty index, but not the Fried frailty phenotype, the other main frailty score it tracked. The results were posted on the preprint server medRxiv on 28 July 2026 and have not yet been peer reviewed1.
Metformin is one of the most discussed repurposed drugs in longevity research, in large part because of TAME, a planned large trial of metformin against aging that, as of September 2026, still had not enrolled a participant5. The new study is much smaller, but it is a completed, double-blind trial with two years of follow-up in people aged 65 and over13.
What happened
The trial randomized 145 people, 72 to metformin and 73 to placebo, and analyzed 141 by intention to treat1. Participants had a mean age of 72, 48% were women, 94% were White and 35% were Hispanic or Latinx, with a mean BMI of 30.91. All were non-frail or pre-frail and had glucose intolerance, defined as a two-hour oral glucose tolerance test result of 140 to 199 mg/dL. People who were already frail, had reduced kidney function or had significant cognitive or mood disorders were excluded1.
Immediate-release metformin started at 500 mg a day and was increased about every two weeks to a maximum tolerated dose of up to 2,000 mg a day, for 24 months1. Both groups received diet and exercise counseling3. The doses are given here only to describe the study, not as a recommendation.
The trial was run at the University of Texas Health Science Center at San Antonio with the National Institute on Aging as collaborator. It began in April 2016 and was completed in February 20242. Its registry record lists the Fried frailty phenotype as the primary outcome and the deficit accumulation frailty index as a secondary outcome2. That matters for reading the results: the positive finding is on the registered secondary measure.
The evidence so far
The preprint reports these differences between metformin and placebo1:
| Measure | Metformin vs placebo | p-value |
|---|---|---|
| Fried frailty score, year 1 | +0.72 ± 0.22 per year (worse) | 0.0011 |
| Fried frailty score, year 2 | −0.33 ± 0.17 per year | 0.056 |
| Deficit accumulation frailty index | −0.006 ± 0.0026 per year (slower rise) | 0.0222 |
| Body weight at 24 months | −5.7 kg vs −2.3 kg | 0.0002 |
| HbA1c | −0.04% vs +0.05% | 0.0229 |
| Conversion to diabetes | 19.4% vs 15.1% | 0.50 |
| PC-Horvath2 epigenetic clock | −0.40 ± 0.16 | 0.014 |
| PC-Hannum epigenetic clock | −0.33 ± 0.16 | 0.047 |
Source:1
The worse Fried score in year one was driven by weight loss, which is one of the five Fried criteria. When the weight loss criterion was removed, the groups did not differ1. The frailty index difference held after adjusting for covariates, including body mass index, and after censoring participants who developed diabetes1.
Gait speed, grip strength and the Short Physical Performance Battery did not change significantly in the intention-to-treat analysis1. Of 13 epigenetic clocks examined, only the two above reached statistical significance. GrimAge, PhenoAge, DunedinPACE and PC-DNAmTL showed no significant effect1.
On safety, nearly all participants reported at least one adverse event, mostly mild and gastrointestinal. There were 12 serious adverse events across both groups and one death, which the authors judged unrelated to the study drug1. An earlier conference abstract from the same team reported no differences in side effects between the groups3.
What it does not show yet
- Peer review. The paper is a preprint and has not been checked by independent reviewers14.
- Effect on real frailty. Nobody was frail at the start, so the trial cannot say whether metformin prevents people from becoming frail4. Coverage by Longevity Germany summed up the size of the effects plainly: "The differences were small."4
- Muscle. Metformin reduced both fat and lean mass, although the authors observed no functional impairment1.
- Clock selection. Two of 13 clocks moved, and the newer clocks did not. Epigenetic clocks are estimates, not health outcomes1.
- Who it applies to. Participants were older adults with prediabetes-range glucose, mostly White, from one site12. The results may not carry over to healthy adults or other groups.
- Size. The authors call the sample modest and say the findings should inform the design of larger trials. They also note that frailty progressed slowly in the placebo group, which may reflect the benefits of close monitoring and lifestyle counseling1.
Who is behind it
The corresponding authors are Sara Espinoza and Nicolas Musi, now at the Center for Translational Geroscience and Center for Advanced Gerotherapeutics at Cedars-Sinai in Los Angeles1. The trial itself was sponsored by UT Health San Antonio2.
Funding came mainly from National Institute on Aging grants, with additional support from a Geriatric Research Education Clinical Center, the Connor Fund, the Ryan Foundation and the Noaber Foundation1. Metformin is a generic drug. One co-author, Steve Horvath, declared patents on epigenetic clocks, a founder and consultant role at the Epigenetic Clock Development Foundation, and past work at Altos Labs, where he held equity until March 2026. The other authors declared no conflicts1.
How it fits the field
The best-known metformin aging study, TAME, was designed by Nir Barzilai of Albert Einstein College of Medicine as a six-year, randomized, placebo-controlled trial in more than 3,000 adults aged 65 to 79 without type 2 diabetes, at 14 US institutions. Its primary endpoint is time to the first of cardiovascular disease, cancer, dementia or death5. As of late September 2026, TAME was still awaiting funding, with estimates of $45 to $75 million needed; because metformin is off-patent, no company stands to profit from the result5.
Meanwhile, federal money for repurposed drug trials has gone elsewhere. In March 2026, UT Health San Antonio's Barshop Institute announced VITAL-H, a trial backed by up to $38 million from ARPA-H that will enroll 726 adults in their 60s and compare rapamycin, dapagliflozin and semaglutide with placebo. Enrollment is due to begin in 2027. Metformin is not one of the drugs6.
Against that background, the San Antonio frailty trial offers completed two-year, placebo-controlled data on metformin in older adults who did not have diabetes at entry12. Its mixed result, a slower rise on one frailty scale and no benefit on the other, shows how much the reading of such a trial depends on which measure is chosen.
What to watch next
- A peer-reviewed version of the paper, and how it reports the registered primary outcome (Fried phenotype) alongside the secondary frailty index2.
- Larger trials designed with the frailty index as the main measure, which the authors suggest their findings can inform1.
- Any funding news for TAME5.
- The start of VITAL-H enrollment in 20276.
For now, the trial shows that two years of metformin in older adults with glucose intolerance shifted some aging-related measures by small amounts. Whether that adds up to healthier aging has not been tested.
