Semaglutide gave old mice 92 extra days: what the Nature study shows, and what it does not

This is general educational information, not medical advice and not a recommendation to get hold of anything or to use it. The study behind it was done in mice. You cannot read a human dose off an animal dose, and we are not naming one here.

The short version: on 2 September 2026, Nature published a study in which old mice were given a daily semaglutide injection and lived a median 92 days longer than the control animals. That is roughly 12%. The work comes out of a university lab, was paid for by NIH grants, and had no money from Novo Nordisk behind it. It contains one finding that goes beyond "eat less, weigh less", and several points where the jump to humans breaks down.

What was measured

Danica Chen's group at the University of California, Berkeley started with 20-month-old female mice, an age that puts them somewhere in their early sixties in human terms. From that day on, every animal got an injection under the skin: either semaglutide or saline. The controls were handled and jabbed daily too, so the handling stress was the same in both groups.

The difference is statistically clear (P = 5.7 × 10⁻⁶), which effectively rules out chance as the explanation. No animal was dropped from the analysis. The weight came off mostly as fat, and the lean fraction went up.

If you see 11% in one report and 12% in another, both are right. 92 days out of 742 is 12.4%; 92 out of 834 is 11.0% — same number, different denominator. The press releases say "about 100 days"; the figure is 92, and it is a median, not an average.

One thing almost every write-up drops: there were three separate sets of animals. The 79 mice in the lifespan arm were never tested for motor function, memory or stem cells; those had cohorts of their own. The obvious line, that the longer-lived mice were also the fitter ones, appears nowhere in the paper, because no single animal was ever measured for both.

The real story: semaglutide versus fasting

We have known since the 1930s that less food makes rodents live longer. So the obvious reading of this result is that semaglutide blunts appetite, the mice eat 24% less, and what you are looking at is calorie restriction with a needle.

The authors tested exactly that. In a separate arm, one group had its food cut by exactly what the semaglutide animals had given up of their own accord: the same 24%, delivered through the bowl instead of through the appetite. Both groups and a control were tested repeatedly over five months.

On voluntary activity, balance on the rotating rod, hanging strength and treadmill endurance, semaglutide and fasting came out level, both holding their starting point while the controls declined. On three measures semaglutide beat fasting, and beat the animals' own baseline: exploring a new environment, spatial memory in the maze, and glucose tolerance. Weight loss and fat loss were comparable in both groups, so the calorie saving on its own cannot explain it.

That is the most interesting finding in the paper, and three caveats belong in the same breath:

• The fasting arm was the only unblinded part of the study, because you cannot hide who is being fed less. So the most-quoted headline comes out of the arm most vulnerable to experimenter effects — and from endpoints that rest on watching behaviour.
• With around nine endpoints and ten animals per group, there was no correction for multiple testing across the endpoint family as a whole. Three hits are hypothesis-generating, not proof of superiority.
• For lifespan itself there was no fasting comparison at all. The study cannot say whether semaglutide extends life more than fasting does.

What happened inside the animals

The authors checked whether semaglutide flips the same switches as calorie restriction, and broadly it does: NAD⁺ rose in liver and muscle, several sirtuins were upregulated, blood levels of the growth factor IGF-1 fell, and a FOXO-controlled longevity gene was expressed more strongly. In liver tissue, inflammation and fat-metabolism programmes were turned down, while protein quality control and insulin response were turned up.

Two things belong alongside that. First, it is not a clean sweep. Not all seven sirtuins responded, and the ones that did were not the same in liver as in muscle. Much of this is measured at the level of gene expression, with five to six animals per group, so the mechanism findings stand on far shakier ground than the survival curve. Second, if you follow the longevity debate: mTOR, AMPK and rapamycin do not appear in this paper at all. Anyone reading them into it is making them up.

One detail that keeps getting told backwards: the number of blood stem cells in the bone marrow went down on semaglutide, not up. That counts as a good sign, because these cells get more numerous but worse with age. "More stem cells" is the wrong translation of the result.

The muscle question

For most readers this is the question that matters: what does it do to muscle, especially in old age?

The honest answer is that the study only answers half of it. Performance was better, on the rotating rod, hanging from an upside-down grid, and on the treadmill. For the first two, the advantage survived adjustment for body weight; lighter animals hang more easily, and that was corrected for.

What was not measured is lean mass in grams. Only lean mass as a percentage is reported, and that necessarily rises when total weight falls through fat loss. There is no grip-strength measurement, no individual muscle weights, no tissue sections. This paper can no more show that muscle mass was preserved than it can show the opposite.

There is also a study pointing the other way. Karasawa and colleagues reported in Cell Metabolism in 2025 that mice on semaglutide had reduced force generation in skeletal muscle, and that loss of mass alone could not explain it. Different study design, different question, but the direction is opposite. Letting both stand is more honest than picking a side. We went into the human side of this question in more detail in our fact-check on retatrutide and the liver.

The brain, and the reality check in humans

The finding the authors push hardest themselves is about the brain: in the dentate gyrus, the one region where the adult brain still makes new nerve cells, semaglutide markedly increased how many of them appeared.

That headline already has a reality check in humans, and it came back negative. In the EVOKE and EVOKE+ trials, 3,808 people with early-stage Alzheimer's disease took oral semaglutide daily for two years. The primary endpoint was missed, clearly so in both trials, and the planned extension phase was stopped. A few inflammation and tau markers improved slightly; other markers actually got worse.

That does not refute the mouse finding. The trials tested a different question, whether semaglutide slows Alzheimer's disease that is already under way, not whether it does anything to a healthy ageing brain. But it is the most obvious translation there is, it was tested in the largest trial of its kind so far, and it did not work.

And there is a plainer explanation that should not be waved away. The neuroscientist Tara Spires-Jones, speaking to the UK's Science Media Centre, pointed out that the treated animals simply moved about more, and that exercise is known to do the ageing brain good in much the same way, with no GLP-1 drug involved at all. This study cannot rule that out.

Why the dose does not translate

The mice got 10 nmol per kilogram, every day, under the skin, until the end of life. People inject once a week, usually for months to a few years, and usually when they are a good deal younger than 60.

You can attempt a conversion, but different methods give different answers. Scale it the way the regulators do, by body surface area, and you land in the region of the approved maintenance doses. Scale it bluntly per kilogram of body weight, and the mice were getting eight to twelve times as much. Both figures are correct; they simply answer different questions. A real comparison would have to go by the amount of drug circulating in the blood, and that the study never measured. Semaglutide stays in the human body for roughly a week and in rodents for orders of magnitude less time, which is why animal work has to dose daily. The mice therefore lived through daily peaks and troughs, while a person on a weekly shot sits at a largely flat level. Nobody knows which of the two patterns produces the effect, because the study has no blood levels to show.

Here is the claim that holds up: a comparable effect is defensible, since the mice ate 24% less and lost mostly fat, which is the same order of magnitude as in humans. A comparable dose is not. If you want to know why units and milligrams get muddled so easily, that is in mg, mL and IU; what happens when you get it wrong is in GLP-1 dosing errors.

So is 12% a lot?

The useful yardstick is the best-studied ageing compound there is. Rapamycin, started at the same age as semaglutide was here, extended the median lifespan of female mice by around 15%. The 12% in this study is the same order of magnitude: solid, but no outlier on the high side.

What matters just as much is how rapamycin was tested. It went through the Interventions Testing Program, run by the US ageing-research agency, which puts every substance through three independent sites at once, in genetically mixed mice and in both sexes. That design exists precisely because single-lab results in this field so often fail to replicate. Semaglutide is on none of those lists, for a mundane reason: substances there are given in the food, and a peptide that has to be injected daily does not fit the protocol.

So this is a single-site result, in one mouse strain, in one sex, with just under 40 animals per group. That lifespan went up is statistically clear; that the figure is exactly 12% is a rough estimate. And female animals were chosen deliberately, to avoid dominance fights among males, so for males there is simply no data.

One point that matters especially to this readership and appeared in none of the coverage: the mice were not fat. This was no obesity model. The animals were on ordinary chow and held their weight into old age, so the effect arose without any pre-existing obesity. In humans it is the other way round, because practically the entire body of evidence on semaglutide comes from people carrying excess weight. Whether a lean sixty-year-old would get the same benefit, or whether the weight loss would be more of a problem for them, is wide open.

Who paid for the study

The work was funded by three grants from the National Institutes of Health, not by a manufacturer. Novo Nordisk appears nowhere.

One line of small print belongs here anyway, and it was in none of the coverage we saw: the University of California has a patent application pending on GLP-1 receptor agonists for healthy ageing. The senior author's institution therefore has a commercial interest in exactly the use that was studied. That does not devalue the data, but it is the context in which you read the wording of a press release.

Bottom line

This is a good study with a real finding: a GLP-1 drug, started late in life, made old mice live longer and slowed their decline, and on three of around nine measures it did so more strongly than fasting on the same calorie saving. For ageing research, that is a result worth taking seriously.

For you it changes nothing. There is no human data on semaglutide and lifespan, the most obvious translation to the brain has just failed, the muscle question is open, and the treatment that worked here was a daily injection until the end of life. Semaglutide is prescription-only and approved to treat obesity and type 2 diabetes; which other peptides carry an approval is set out in our list of FDA-approved peptides. "Living longer" is not among them, and nothing in this study is an argument for using a drug to that end.

If you want to know how the drug class works in the first place, start with what peptides are; the drug-specific background is under semaglutide.

Sources

• Feng et al., Nature 2026, Late-life semaglutide treatment slows ageing and extends lifespan in female mice
• Strong et al., Aging Cell 2020, Rapamycin-mediated mouse lifespan extension: Late-life dosage regimes with sex-specific effects
• Karasawa et al., Cell Metabolism 2025, Unexpected effects of semaglutide on skeletal muscle mass and force-generating capacity in mice
• Cummings et al., The Lancet 2026, Efficacy and safety of oral semaglutide in early-stage symptomatic Alzheimer's disease (evoke and evoke+)
• Lincoff et al., New England Journal of Medicine 2023, Semaglutide and cardiovascular outcomes in obesity without diabetes (SELECT)
• Di Francesco et al., Nature 2024, Dietary restriction impacts health and lifespan of genetically diverse mice
• Harrison et al., Nature 2009, Rapamycin fed late in life extends lifespan in genetically heterogeneous mice