The problem with metformin as a drug to slow aging is that the evidece to support that use is very poor. In animal studies, the results are very unreliable, and the Interventions Testing Program found no effect in its highly overengineered studies. Further, the existing human data is not supportive, taken as a whole. Even if we did want to cherry pick the better data and be hopeful, the effect size compares unfavorably with that achieved through regular exercise, and further appears to be only achieved in people with the abnormal metabolism associated with obesity and diabetes. All of the work that was done to convince the FDA to endorse the TAME human clinical trial to test the ability of metformin to slow aging is useful, but the resulting agreement on trial structure should be applied to an intervention more likely to produce an outcome that is worth the effort, such as senolytic therapies.
The study that is most often cited as evidence that metformin slows the aging process in humans was released with a press release misleadingly titled “Type 2 diabetics can live longer than people without the disease.” But the underlying study had a design flaw that first unintentionally selected only the healthiest diabetic patients (those on metformin) and compared them to patients with poorer glycemic control (those on other drugs) and a random assortment of the nondiabetic population – and then systematically pushed subjects on metformin “off the books” as soon as their diabetes progressed.
The same problem (or related ones) have plagued most of the observational studies that you may have heard cited as showing that metformin lowers the risk of atherosclerosis, total mortality, and especially cancer. Drawing inferences from such studies about effects on aging in otherwise-healthy people would thus be misguided even if these studies didn’t share this design flaw, since none of these other studies include a separate group of people without diabetes. Rather, such studies have compared metformin-taking diabetic people to other people with diabetes taking other diabetic drugs. But actually, even in such diabetics-only studies, the apparent benefits of metformin vanish when the studies are designed to avoid survivorship bias and selection bias.
When put to the test in human trials, metformin has no effect on blood sugar control in obese women with normal glucose tolerance and only modest effects on fasting glucose in normal-weight, nondiabetic men. Similarly, exercise but not metformin tames glycemic variability (dangerously wide swings in blood sugar over the course of the day) in prediabetic people. And importantly, adding metformin to such lifestyle interventions doesn’t lower the risk of developing diabetes any more than lifestyle all by itself.
You may be surprised to learn that there has already been a trial with followup that gives fairly long-term human data on mortality in a group of people who were not yet diabetic – and again, metformin came up short. This was report from the long-term follow-up of the Diabetes Prevention Program (DPP). The volunteers in the DPP were on average 50 years old, and all had prediabetes. The DPP itself lasted only 2.8 years, but the researchers followed up with the participants at 10, 15, and as much as 20 years later. And to get to the punchline, people who had been taking metformin lived no longer than people in the control group.