Metformin for healthy aging: what the evidence supports

Evidence date: September 22, 2026.

Three questions with different answers

Does metformin treat diabetes? Can it delay diabetes in people at high risk? Can it slow aging in people who would otherwise have no reason to take it? These questions involve different starting risks and different measures of success.

The US prescribing information for oral metformin lists blood-glucose control in type 2 diabetes alongside diet and exercise. It does not list aging or longevity as an indication. An unsettled longevity hypothesis does not cancel the medicine's established role in diabetes care. Equally, that role does not establish a benefit for everyone with normal blood glucose. [1]

• Shown: the Diabetes Prevention Program found fewer new cases of type 2 diabetes with metformin in adults with elevated glucose and high diabetes risk. Its intensive lifestyle program produced a larger relative reduction. [5]

• Not shown: the studies discussed here have not established a net longevity benefit for otherwise healthy adults. Long-term follow-up did not show lower mortality with original metformin assignment. [4]

• What would change the assessment: suitable randomized trials showing fewer important illnesses or better function, with sustained benefit and acceptable harms. A molecular age estimate alone cannot establish that balance.

The diabetes-prevention result is worth understanding

The Diabetes Prevention Program, or DPP, randomly assigned 3,234 adults with elevated glucose and a high risk of type 2 diabetes to metformin, an intensive lifestyle program or placebo. Over an average 2.8 years, metformin reduced new diabetes cases by 31% relative to placebo; the lifestyle program reduced them by 58%. [5]

The rates were 7.8 new cases per 100 person-years with metformin and 11.0 with placebo. “Person-years” combines the number of people with how long they were followed—for example, 100 people followed for one year contribute 100 person-years. These rates describe new diabetes cases. They also concern a high-risk group, not everyone with normal glucose. [5]

Longer follow-up in the DPP Outcomes Study, or DPPOS, allowed researchers to ask a separate question about death. After a median 21 years, original assignment to metformin was not associated with lower mortality from all causes: the hazard ratio was 0.99, with a 95% confidence interval of 0.79–1.25. A hazard ratio compares event rates over time; 1 means no difference. The interval describes uncertainty around the estimate and here includes 1. Original assignment to lifestyle intervention also did not show a statistically significant mortality reduction in this analysis. [4]

This was not 21 years of unchanged, blinded treatment. Placebo stopped after the original trial; metformin continued openly in the original metformin group; and a lifestyle program was offered to everyone. Those changes complicate the later comparison. [4]

Did metformin prevent several chronic diseases together?

A June 2026 JAMA paper examined 1,173 DPP participants with linked Medicare data. The outcome was multimorbidity: having at least two of 15 chronic conditions. That gets closer to the broad prevention question of aging research than a glucose reading does. [8]

Over approximately 21 years, there was no statistically significant difference between the original metformin and placebo groups. The adjusted hazard ratio was 0.91, with a 95% confidence interval of 0.78–1.07. The original lifestyle group had a lower risk than the placebo group. [8]

This was an observational follow-up of part of a randomized cohort. Later treatment changed, and conditions were identified through health-insurance claims, which have limits as measures of disease. The result does not establish that metformin could never prevent multiple diseases in any group. It does show why successful diabetes prevention cannot be relabeled a demonstrated broad anti-aging effect.

What happened in the monkey study?

Yang and colleagues' 2024 Cell paper followed adult male cynomolgus monkeys for 40 months. It examined physical and molecular features, brain-related measures and aging clocks based on gene activity, DNA methylation, proteins and metabolites. DNA methylation is a pattern of chemical marks on DNA; metabolites are small molecules involved in the body's chemistry. [3]

The authors reported slower changes in several aging indicators, including a roughly six-year difference in a brain-aging estimate. That number prompted the “younger brain” headlines. It does not mean that the monkeys lived six years longer, or that a person taking metformin would gain six healthy years. [3]

An aging clock is a model that estimates age-related patterns from biological measurements. If a treatment changes the measurements, the model may assign a different age. Researchers still need to establish whether that change predicts less disease or better function.

The experiment matters because nonhuman primates provide an important setting for aging research. But the animals were male, observation lasted a finite period, and the brain and molecular findings are not a randomized disease-prevention trial in healthy humans. They generate questions for human research rather than settling a prescribing decision. [3]

A human aging-clock trial found no clear primary benefit

METFORAGING, reported in 2026, randomly assigned 40 adults aged 50 or older who had well-controlled HIV and no diabetes to metformin or placebo. The primary analysis at 96 weeks included 35 participants who continued treatment as planned. [11]

Metformin did not significantly improve the PhenoAge measure of epigenetic age acceleration—a DNA-methylation estimate of aging relative to chronological age. This was the primary outcome chosen in advance. The estimated difference was −1.02 years, with a wide 95% confidence interval from −5.30 to 3.26 years. [11]

This small pilot does not establish rejuvenation, and it cannot rule out a smaller effect. It also studied a particular population and a biological marker. Neither a favorable clock result in monkeys nor a null clock result in this human pilot can be treated as a human survival result.

Why the famous survival comparison remains uncertain

A 2014 UK observational study reported longer survival among people with type 2 diabetes starting metformin alone than among matched people without diabetes. It included 78,241 metformin-treated patients, alongside other treatment groups and matched controls. [6]

The finding was striking: could a medicine more than offset the disadvantage of diabetes? But treatment was not randomly assigned. Matching balances selected recorded characteristics; it does not make the groups identical in every relevant way. Illness severity, treatment selection, use of health care and the point at which follow-up starts or stops can all influence a comparison.

A 2023 study of Welsh medical records examined up to 20 years of observation. Metformin-treated patients with diabetes did not retain a survival advantage over matched people without diabetes across the full period. The short-term and long-term comparisons gave different answers. [7]

This is not proof that metformin shortens life. Diabetes itself and the circumstances leading to treatment remain part of the comparison. Neither observational result tells a healthy person what would happen if they started metformin specifically for longevity. That question needs a suitable randomized trial.

Muscle, walking and frailty need their own evidence

In MASTERS, older adults undertook 14 weeks of progressive resistance training while receiving metformin or placebo. Of 109 randomized participants, 94 completed the study. The placebo group gained more lean body mass and thigh muscle mass. Differences in strength gains were not statistically significant. A 2020 correction to relative-strength data is part of the study record; the finding should not be described as proven loss of strength. [9]

MET-PREVENT tested metformin in 72 older people with probable sarcopenia—a disorder involving muscle strength and mass—and prefrailty or frailty, meaning increased vulnerability to health problems. Their mean age was about 80. After four months, metformin did not improve the primary walking-speed outcome and was poorly tolerated in this population. [10]

A July 2026 preprint reported a different, mixed picture. The trial randomly assigned 145 older adults who were not frail or were prefrail to metformin or placebo for two years. All had glucose intolerance, meaning difficulty regulating blood glucose normally. The main Fried frailty score initially worsened with metformin, largely because weight loss counts toward that score. Removing the weight-loss criterion left no difference between groups. [12]

A separate frailty index combining 95 health deficits favored metformin, as did two DNA-methylation clocks. These positive findings deserve attention. They do not erase the different answer from the main frailty measure. As of September 22, 2026, the report was a preprint, meaning it had not undergone journal peer review. Confirmation requires attention to the choice of score, the number of analyses and how incomplete follow-up is handled. [12]

This study concerned older people with glucose intolerance; its frailty and clock findings cannot be assumed to apply to all healthy adults. Across these trials, muscle size, strength, walking speed and molecular markers must remain distinct. An appealing cellular mechanism does not guarantee better mobility.

These findings also do not give people with diabetes a general reason to stop prescribed metformin. They concern particular populations and goals; decisions about established treatment still depend on the clinical reason for taking it.

Where TAME stands

TAME stands for Targeting Aging with Metformin. The American Federation for Aging Research, or AFAR, describes a proposed multisite program involving more than 3,000 adults aged 65–79, with six years of trials once launched. Its goal is to test whether metformin delays the development or progression of several age-related chronic diseases. [2]

As of September 21, 2026, AFAR's official program page continued to seek donations to support launch. It described the design as prepared but did not provide completed trial results. The description offered a plan to answer the broader prevention question; it did not demonstrate that metformin achieves that goal. [2]

The organizers hope the program could help establish a regulatory pathway for treating aging-related diseases collectively. That is a research goal, not an aging indication already approved by the US Food and Drug Administration. A proposed launch or a result from a different, smaller metformin experiment should not be presented as a successful TAME result. [2]

How Healthy Longevity Clinic experts evaluate the evidence

For someone considering metformin, the goal determines the relevant evidence: glucose control, diabetes prevention or broader protection against aging-related disease. HLC's interpretation starts with the strongest supported comparison. In DPP's high-risk population, metformin reduced new diabetes cases, while the intensive lifestyle program produced a larger relative reduction. [5]

The next step is to match the hoped-for benefit to the outcome actually measured. The monkey clock estimate does not answer the survival question. The 21-year mortality follow-up and the 2026 multimorbidity analysis do address broader health outcomes, but they did not establish a benefit from original metformin assignment. Treatment changes and other limits make better trials valuable; they do not demonstrate a benefit missing from these results. [3] [4] [8]

For a generally healthy person considering an additional medicine, the missing evidence is a net benefit in a comparable population over meaningful follow-up. A trial such as the proposed TAME program could make the question more direct if it reports fewer important diseases alongside harms. Replication of the promising frailty-index finding, with agreement across relevant functional measures, would also matter. Until then, diabetes care, prevention in high-risk adults and experimental healthy-aging use should remain separate clinical conversations.

Three questions for a clinical discussion

1. Are we treating diabetes, addressing a high risk of diabetes, or discussing an experimental healthy-aging use—and which evidence applies to me?

2. Is the expected benefit fewer illnesses or better function, or only a change in a biological-age score? What would count as success or treatment failure?

3. How do kidney function, vitamin B12, other medicines and tolerability affect the likely balance of benefit and harm?

Common questions

Should the longevity uncertainty worry someone already taking metformin for diabetes?

It does not cancel metformin's established diabetes role. The US label supports glucose control alongside diet and exercise. Questions about a broader aging benefit should be discussed separately from decisions about a prescribed diabetes treatment. [1]

Does a “six years younger” brain mean six extra years of healthy life?

No. It describes a brain-aging estimate in adult male monkeys over 40 months. A model expressed in years is not a measurement of extra survival, in monkeys or in people. [3]

Do the newer frailty results show that metformin preserves independence?

They do not establish that. One preprint reported a favorable separate deficit index and clock measures, but its main frailty measure gave a different answer. The findings need confirmation in the relevant population. [12]

Is metformin's long history enough to settle safety for healthy-aging use?

No. US labeling identifies stomach and intestinal side effects, reduced vitamin B12 levels and a serious warning about lactic acidosis, a dangerous buildup of lactic acid in the body. Kidney function, other medicines and certain illnesses or procedures can change risk. Familiarity and low cost do not establish a worthwhile benefit for a new use. [1]