Metformin vs GLP-1 for Longevity: The TAME Trial Has Not Started

Neither drug is proven to extend a healthy person’s life. Only the GLP‑1 class has large modern randomized trials showing fewer deaths or major events, and only in specific high-risk groups.^[16][17] Metformin’s longevity case rests on one small diabetes trial from 1998, a 2014 observational study that later, better-designed cohorts did not reproduce, short biomarker trials, and TAME, a trial proposed in 2015 that has no registry record and was still raising money at its last public update.^[2][3] Metformin also blunted the gains older adults made from exercise in two randomized trials.^[14][15] For someone without diabetes, there is no trial showing metformin adds a single year.

Metformin vs GLP-1 drugs: what each has actually shown

The TAME trial: what it was meant to be and where it stands

TAME, short for Targeting Aging with Metformin, was laid out in 2016 by Nir Barzilai and colleagues as a test of a new idea: that a drug could be judged on whether it delays a cluster of age-related diseases, not one disease at a time.^[1] The plan was about 3,000 people aged 65 to 79, half on metformin and half on placebo, followed for six years at 14 medical centers, with heart disease, cancer and dementia among the outcomes.^[2][3] Metformin was picked because it is old, cheap and well understood, not because it was the most promising drug on the list.

That cheapness is also why the trial stalled. Metformin is generic, so no company stands to profit from a positive result. In April 2024 NPR reported that the National Institute on Aging had set aside about $5 million, against an estimated cost of $45 million to $70 million.^[2] The American Federation for Aging Research, which leads the effort, said at the time that it was raising money to launch the trial.^[3]

As of October 2026, TAME has no record on ClinicalTrials.gov, the registry where US trials are listed before they enroll, and the federation’s old TAME page no longer loads. No participants have been reported as enrolled and no results exist. Any website that cites TAME as evidence that metformin slows aging is citing a trial that has not happened.

The study that started it, and why it did not hold

The longevity claim took off in 2014. Researchers at Cardiff used UK primary-care records to compare 78,241 people with type 2 diabetes who started metformin, 12,222 who started a sulfonylurea, and 90,463 matched people without diabetes. After adjustment, the people without diabetes had a median survival 15% shorter than the metformin group.^[4] If a diabetes drug left its users outliving healthy peers, the reasoning went, it might do the same for anyone. A 2017 meta-analysis of 53 studies found the same direction, a hazard ratio of 0.93 against people without diabetes, while noting that baseline differences between the groups could have biased it.^[5]

Observational studies of a drug like metformin carry a known trap. Doctors do not give metformin to people with poor kidney function, advanced illness or frailty, so the group taking it starts out healthier than the group that is not. Some designs also count time before a person starts the drug as if they were already on it, which flatters the drug. When a Danish team rebuilt the comparison with national registers and added twins, the advantage reversed.

The Danish authors found the excess mortality at every level of cumulative use and across several study designs, and concluded that metformin was not associated with survival equal to or better than the general population’s.^[6] Comparing twins removes much of the shared genetic and family background that a matched stranger cannot. In a 2025 review, the same group and two senior drug-safety epidemiologists described the methodological limits of the early findings and summed up the field as one of “emerging uncertainty” about metformin as an anti-aging drug.^[7]

The animal data are thinner than the marketing suggests. In the National Institute on Aging’s Interventions Testing Program, which runs each compound at three sites in genetically mixed mice, metformin on its own did not significantly extend lifespan.^[8]

What randomized trials of metformin show on deaths and events

The one randomized result most often quoted is from UKPDS 34. Among 753 overweight people with newly treated type 2 diabetes, those assigned to metformin had 36% lower all-cause mortality over a median 10.7 years than those on diet alone.^[9] It is a real result and it is small: 342 people on metformin. The same paper reported a 96% higher risk of diabetes-related death when metformin was added early to a sulfonylurea, which no one has explained.^[9]

Pool every placebo- or lifestyle-controlled trial of metformin in diabetes and the mortality signal fades. A 2017 meta-analysis of 13 trials found a risk ratio for death from any cause of 0.96, with a confidence interval running from 0.84 to 1.09, and no outcome reached statistical significance. UKPDS supplied most of the data.^[10]

The longest trial in people without diabetes is the most relevant to a longevity buyer, and it is negative. In the Diabetes Prevention Program and its follow-on study, 3,234 adults with impaired glucose tolerance were randomized to metformin 850 mg twice daily, lifestyle change or placebo. After a median of 21 years, metformin did not reduce heart attacks, strokes or cardiovascular deaths: hazard ratio 1.03.^[11] The authors note that later cross-over use of metformin and wide use of statins may have diluted any effect. Even so, two decades of metformin in people at risk of diabetes did not move the events that decide how long people live.

MILES and the biomarker trials: surrogate signals, small numbers

The trials designed around aging biology are short and small. MILES, the Metformin in Longevity Study, gave 14 people around age 70 six weeks of metformin and six of placebo in a crossover design, then biopsied muscle and fat.^[12][13] Metformin changed the activity of 647 genes in muscle and 146 in fat, including pathways tied to metabolism and DNA repair.^[12] That is evidence metformin does something in aging tissue. It is not evidence the something is good for a person, and gene-expression shifts are a surrogate, the same caution that applies to the epigenetic-clock findings in does semaglutide slow aging.

Metformin blunted the benefit of exercise in older adults

Two randomized, double-blind trials asked whether metformin adds to exercise in older people. Both found it subtracted.

• Aerobic training. 53 adults around age 62 did 12 weeks of aerobic exercise on metformin or placebo. Metformin attenuated the training gain in whole-body insulin sensitivity and in VO2 max, the standard measure of fitness, and wiped out the rise in muscle mitochondrial respiration that exercise produced on placebo.^[14]
• Strength training (MASTERS). 94 adults aged 65 and over took metformin 1,700 mg a day or placebo through 14 weeks of supervised resistance training. The placebo group gained more lean mass and more thigh muscle; strength gains trended lower on metformin without reaching significance.^[15]

Fitness and muscle mass in later life are among the strongest predictors of how long people live, so a drug that taxes both needs a hard-outcome reason to take it. The aerobic trial’s authors put it directly: before metformin is prescribed to slow aging, more work is needed on why some people respond well and others badly.^[14] Exercise and GLP-1 drugs interact differently: in a randomized trial the drug alone did not improve fitness, and exercise added on top did, as covered in exercise plus a GLP-1. Muscle on a GLP-1 after 65 has its own page: strength and frailty after 65.

The GLP-1 side: outcome trials, in specific people

The GLP‑1 case is different in kind, not just degree. SELECT randomized 17,604 adults with established heart disease and overweight or obesity, but no diabetes, to weekly semaglutide 2.4 mg or placebo. Over a mean 39.8 months, major cardiovascular events fell from 8.0% to 6.5%, a hazard ratio of 0.80.^[16] In FLOW, 3,533 people whose type 2 diabetes had already damaged their kidneys were randomized, and those on semaglutide 1 mg had 20% lower death from any cause, a prespecified confirmatory outcome.^[17]

Those are hard outcomes, which is more than metformin has outside diabetes. They are also limited: every one of these trials enrolled people who were already sick or at high risk, and none tested a lean, healthy adult taking the drug to age more slowly. Whether that extends to a healthy person is unknown. The mortality data across the class are collected in GLP-1s as longevity drugs, and the tirzepatide record in tirzepatide and longevity.

Metformin or a GLP-1 for longevity: the verdict

• If you have type 2 diabetes, both are standard treatments and the choice is a clinical one about your diabetes, heart and kidneys. Many people take both; the practical overlap is in metformin and a GLP-1 together.
• If you have obesity with heart disease, semaglutide has a randomized reduction in major cardiovascular events. Metformin has none in that group.^[16]
• If you are healthy and want to live longer, neither drug has been shown to do that. Metformin’s best observational evidence did not replicate,^[6] its 21-year prevention trial showed no effect on cardiovascular events,^[11] and it blunted exercise gains in older adults.^[15]

Older, cheaper drugs sometimes beat newer ones once cost or tolerability enters the comparison, as in the pediatric obesity analysis where an older drug came out ahead. Metformin for aging is not that story yet. It is a hypothesis with a trial design attached and no trial. Other molecules sold on the same promise fare no better; see what the NAD trials show.