Metformin's case as an anti-ageing drug now rests on three null results

Metformin has been the most-cited candidate geroprotector for over a decade. It is cheap, generic, taken by tens of millions of people, and it affects several biological systems relevant to ageing [s2]. The geroscience argument has always been that a drug like this could delay not one disease but many at once.

Three long studies have now tested versions of that claim in people. None of them supports it.

Twenty-one years, no reduction in cancer

The Diabetes Prevention Program randomised 3,234 adults at high risk of diabetes to intensive lifestyle intervention, metformin at 850 mg twice daily, or blinded placebo, between 1996 and 2001 [s3]. Its long-term follow-up study kept tracking them.

After a median follow-up of 21 years, 546 participants had a first incident cancer — 173 on metformin, 182 in the lifestyle arm and 191 on placebo [s3]. Incidence rates were 9.8, 10.5 and 10.8 per 1,000 person-years respectively [s3]. The hazard ratio for metformin versus placebo was 0.90, with a 95% confidence interval of 0.73 to 1.10; for lifestyle versus placebo it was 0.96 (0.79 to 1.18) [s3]. There were no differences for obesity-related cancer or in sex-specific analyses [s3].

That matters because observational meta-analyses had reported cancer incidence reductions of roughly 10% to 40% with metformin use among people with diabetes [s3]. Randomised assignment, followed for two decades, did not reproduce it. The authors note the result may be affected by non-study metformin use rising over time as participants developed diabetes, and by the lifestyle intervention becoming less intensive [s3] — real caveats, and not enough to convert a null into a benefit.

Multimorbidity: lifestyle yes, metformin no

The same cohort was analysed for the outcome geroscience actually cares about — accumulating several chronic conditions rather than one.

Medicare morbidity data were available through 2021 for 1,173 consenting participants, with a median age of 74 (IQR 70 to 80) and 795 (68%) women [s1]. By the end of follow-up, 997 of them (85%) had developed at least two of 15 chronic conditions, with a median of five [s1]. Broken down by original randomised assignment: 316 of 385 (82%) in the lifestyle group, 327 of 385 (85%) on metformin and 350 of 403 (87%) on placebo [s1].

After adjustment, the risk of multimorbidity was lower in the lifestyle group than placebo, at a hazard ratio of 0.79 (95% CI 0.68 to 0.93) [s1]. For metformin against placebo there was no difference: 0.91 (0.78 to 1.07) [s1]. Restricting to the costliest pairs of conditions widened the lifestyle advantage to a hazard ratio of 0.57 (0.38 to 0.85) [s1]. The relationships held when diabetes itself was excluded from the definition of multimorbidity [s1].

This is the closest thing the field has to a decades-long test of a geroprotective drug against a multi-disease endpoint, and the drug did not separate from placebo while a behavioural intervention did.

The frailty trial, and a tolerability signal

MET-PREVENT tested metformin directly on a functional ageing outcome. It was a double-blind, randomised, placebo-controlled trial in England enrolling adults aged 65 and over with a 4-metre walk speed below 0.8 m/s and probable sarcopenia, defined by low handgrip strength or a slow five-times sit-to-stand [s2]. Of 268 people screened, 72 were randomised to 500 mg of metformin three times daily or matching placebo for four months [s2]. Mean age was 80.4 years (SD 5.7) and 42 of 72 participants (58%) were women [s2].

Mean 4-metre walk speed at four months was 0.57 m/s (SD 0.19) on metformin and 0.58 m/s (SD 0.24) on placebo, an adjusted treatment effect of 0.001 m/s with a 95% confidence interval of −0.06 to 0.06 and p = 0.96 [s2].

The tolerability data are the part that should give pause. There were 108 adverse events in the 35 participants who received metformin, against 77 events in 33 of 36 (92%) on placebo, and hospital admissions occurred in 12 of 35 (34%) in the metformin group against 3 of 36 (8%) on placebo [s2]. One death occurred, in the metformin group, judged unrelated to study treatment [s2]. The investigators' conclusion was that metformin did not improve walk speed and was poorly tolerated in this population [s2].

What the field still lacks

None of these studies was designed as a lifespan trial, and none of them can rule out that metformin does something useful in a different population, at a different age, or over a different horizon. The trial the geroscience field has long argued for — a randomised study of metformin with a multi-disease endpoint in people without diabetes, powered and funded as such — has not been run.

But the arguments for it now have to be made against, rather than alongside, three long looks at real outcomes. In the largest and longest, what worked was the lifestyle arm [s1].

Metformin is a prescription medicine. This article reports trial findings and is not medical advice; no dose described here should be read as guidance.

Sources

1. [s1] Lifestyle and Metformin Interventions and Risk of Multimorbidity in Adults With Prediabetes. JAMA, 2026. https://doi.org/10.1001/jama.2026.8492
2. [s2] Metformin and physical performance in older people with probable sarcopenia and physical prefrailty or frailty in England (MET-PREVENT): a double-blind, randomised, placebo-controlled trial. The Lancet Healthy Longevity, published online March 24, 2025. https://doi.org/10.1016/j.lanhl.2025.100695
3. [s3] Randomized Study of Metformin and Intensive Lifestyle Intervention on Cancer Incidence over 21 Years of Follow-up in the Diabetes Prevention Program. Cancer Prevention Research, published online April 17, 2025. https://doi.org/10.1158/1940-6207.CAPR-23-0461