Does fasting actually extend human lifespan? An honest look at the evidence

Most people think the fasting question is “does it work?” It isn’t. Fasting demonstrably works — for weight, for some metabolic markers, sometimes for how you feel at 4pm. The actual question, the one the whole longevity pitch rests on, is much narrower: does fasting make humans live longer?

And the honest answer is that nobody has ever measured it.

That sounds like a technicality. It isn’t. It’s the entire gap between what the science shows and what the podcasts claim, and once you see it you can’t unsee it. So let me walk through what we do have, steelman it properly, and then show you the one study that quietly undermined the reasoning everyone was using.

The claim, as everyone states it

The everyday version goes like this. Calorie restriction is the most reproducible lifespan-extending intervention in biology — true, it works in yeast, worms, flies, and mice. Fasting triggers the same cellular machinery (autophagy, lower insulin, lower mTOR signalling) without the misery of eating 25% less every single day. Therefore, if you compress your eating into eight hours, you get the longevity effect on easy mode.

Three steps, and each one sounds reasonable. The trouble is that the argument smuggles in two assumptions: that what extends life in a mouse extends life in a human, and that the biomarkers fasting improves are the same thing as the lifespan it’s meant to extend.

Both assumptions have now been tested. Neither survived intact.

Where it came from

The rodent evidence is genuinely excellent, and it’s worth being fair about how good it is. The best single piece of it is recent: Di Francesco and colleagues, published in Nature in 2024, put 960 genetically diverse female mice through graded calorie restriction (20% and 40%) and intermittent fasting (one or two fasting days per week), then followed them for life. Both interventions extended lifespan, in proportion to how much was restricted. That’s about as clean a demonstration as the field has. (According to PubMed: Di Francesco et al., DOI.)

The monkey evidence is where it starts wobbling. Two long-running rhesus studies began in the late 1980s. Wisconsin reported a survival benefit from 30% restriction. The National Institute on Aging study, published in Nature in 2012, reported no significant survival effect at all — same species, same broad protocol, opposite headline. (According to PubMed: Mattison et al., DOI.)

The two teams then did the admirable thing and pooled their data. Their 2017 joint analysis in Nature Communications found the health benefits were conserved across both studies, and traced the survival divergence to design differences: when restriction started, what the diet was made of, and — crucially — that the NIA’s control monkeys were themselves fed a portion-controlled diet rather than allowed to eat freely. In other words, the NIA study may have been comparing restriction to mild restriction. (According to PubMed: Mattison et al., DOI.)

That’s the honest state of the primate evidence: health effects reproducible, survival effect real but conditional on doing it a particular way, at a particular age, against a control group that isn’t already eating sensibly.

What the strongest human studies actually show

Now the humans, where the ceiling is much lower than people assume.

The flagship is CALERIE — 220 adults without obesity, randomised to a 25% calorie-restriction target or eating freely, for two full years. It is the most serious long-term restriction trial ever run in healthy humans, and it is still only two years long. A 2023 analysis in Nature Aging by Waziry and colleagues found that restriction slowed the pace of aging as measured by the DunedinPACE epigenetic algorithm — but produced no significant change in the biological-age clocks PhenoAge or GrimAge. The authors themselves describe the effect sizes as small, and say plainly that a conclusive test would need long-term follow-up measuring actual disease incidence and mortality. (According to PubMed: Waziry et al., DOI.)

Worth noting the bit that rarely makes it into summaries: CALERIE participants were asked for 25% restriction and, on average, managed roughly half of that over two years. Even in a well-supported trial with dietitians and free food, sustained restriction is hard. That matters for anyone extrapolating from it to their own kitchen.

On fasting specifically, the two best randomised trials are both mildly deflating.

Trepanowski and colleagues, in JAMA Internal Medicine (2017), randomised 100 adults with obesity to alternate-day fasting, daily calorie restriction, or a control group for a year. Weight loss was essentially identical between fasting and daily restriction (-6.0% vs -5.3% at twelve months). There were no significant between-group differences in blood pressure, triglycerides, fasting glucose, insulin resistance, or C-reactive protein. Dropout was highest in the fasting group — 38%, against 29% for daily restriction — and LDL cholesterol was significantly higher in the fasting group by month twelve. (According to PubMed: Trepanowski et al., DOI.)

The TREAT trial, also in JAMA Internal Medicine (2020), tested the internet’s favourite protocol — 16:8, eating noon to 8pm — in 116 adults over twelve weeks. Between-group weight difference: -0.26 kg, not significant. No significant changes in fasting insulin, glucose, or HbA1c. And in the in-person subgroup, a significant drop in appendicular lean mass index, meaning a meaningful share of the little weight that came off was muscle. (According to PubMed: Lowe et al., DOI.)

There is one more positive data point worth giving its due. Brandhorst and colleagues in Nature Communications (2024) reported that three cycles of a fasting-mimicking diet were associated with reduced insulin resistance, lower liver fat, and a decrease of about 2.5 years in a validated blood-based measure of biological age — independent of weight loss. Interesting, genuinely. Also a secondary and exploratory analysis of two modest trials, measuring a proxy. (According to PubMed: Brandhorst et al., DOI.)

That’s the full evidence base. Not a single one of these studies measured how long anyone lived.

The study that broke the logic

Here’s the part that changed my mind, and it comes from inside the strongest pro-fasting dataset we have.

Everyone’s reasoning runs: fasting improves your metabolic markers → metabolic markers predict mortality → therefore fasting extends life. It’s such a natural chain that it barely registers as an argument.

The 2024 Nature mouse study tested that chain directly, because it had both the biomarkers and the lifespans for all 960 animals. And the metabolic responses to restriction — reduced adiposity, lower fasting glucose, the exact numbers a human would celebrate on a blood panel — were not associated with longer life. The mice that lived longest were not the ones whose metabolic markers improved most. The traits that actually tracked with lifespan were stranger and less flattering: retaining body weight through periods of handling stress, immune-cell profiles, and higher adiposity in late life.

The authors put it about as bluntly as a Nature paper ever does: improving health and extending lifespan are not synonymous, and this raises questions about which endpoints are meaningful for evaluating aging interventions. They also found that genetics influenced lifespan more than the diet did, that intermittent fasting failed to extend life at all in mice that started heavy, and that the most aggressive restriction — 40%, the biggest lifespan winner — cost lean mass and shifted the immune repertoire in ways that could increase infection risk.

So the single best fasting-lifespan dataset in existence says: yes it extends lifespan in mice, and no — at least for the metabolic markers it tested — you can’t read the lifespan effect off the blood panel, and the protocol that extends life the most also does the most collateral damage.

Be careful about how far that travels, though, because the study only tested some of the markers we care about. Adiposity and fasting glucose are exactly the two that the human fasting trials lean on hardest — and they’re the two that came apart from lifespan in the mice. Other things the human trials measured, like blood pressure, lipids, CRP, lean mass and the epigenetic clocks, weren’t part of that mouse analysis at all, so they’re untested rather than disproved. And several of them, especially blood pressure and lipids, have their own large body of evidence linking them to mortality in humans that has nothing to do with fasting. Nobody should read this and stop caring about their blood pressure.

The narrower and more honest conclusion is the one that still hurts: for the specific markers fasting is most often sold on, we now have direct evidence — in the one species where we’ve actually measured lifespan — that improving them didn’t translate into living longer. That’s not a reason to dismiss fasting. It’s a reason to stop treating a better fasting glucose as a receipt for extra years.

The honest verdict: real for metabolism, unproven for lifespan

Partial credit, then, and it splits cleanly.

Supported: fasting and restriction improve several cardiometabolic markers; a compressed eating window helps many people eat less without counting anything; CALERIE showed a small but real slowing on one measure of the pace of aging. These are worth having.

Not supported: that any of it makes you live longer. There is no human trial, there is no ongoing human trial that will report in the next decade, and the mouse data actively warns that at least two of the markers we’re using as stand-ins — body fat and fasting glucose — may not carry the signal. Meanwhile the evidence in the other direction is thin too — a 2024 American Heart Association conference abstract linking eating windows under eight hours to higher cardiovascular death got enormous coverage, but it was an unreviewed abstract based on two days of self-reported diet recall. It shouldn’t change your behaviour either.

The fair summary is that fasting is an eating schedule with decent short-term metabolic evidence and no longevity evidence, sold as a longevity intervention. That’s a smaller claim than the one on the podcast. It’s also not nothing.

What to do instead

If you already fast and enjoy it, carry on — this isn’t a case against it. Just spend your attention where the evidence is denser.

• Guard your muscle while you do it. Both the TREAT trial and the mouse data flag lean-mass loss as the standing cost of restriction, and muscle mass is one of the better predictors of how well you age. If you fast, lift, and hit your protein inside the window — don’t let the schedule eat your legs.
• Front-load the window rather than shortening it. The metabolic evidence for when you eat is better than the evidence for how long you go without, which is the actual finding buried in the time-restricted eating research. An earlier eight-hour window beats a later one at the same length.
• Spend the effort where the returns are measured. Nothing in the fasting literature comes close to the mortality evidence behind cardiorespiratory fitness or strength — the longevity habits ranked by healthspan per hour exist precisely because attention is the scarce resource, and fasting is not near the top of that list.
• Judge it on what it actually does for you. Track the things that respond in months, not decades: fasting glucose, resting heart rate, how you sleep, whether you’re still doing it in March. That’s the real test, and it’s the only one you’ll get an answer to.

The thing I keep coming back to is that fasting got adopted as a longevity intervention largely because it was easy to believe and cheap to try. Both of those are still true. The evidence for the actual longevity part just never arrived — and the best study in the field spent 960 mice telling us why we shouldn’t assume it will.