Fasting and time-restricted eating

Fasting earns its reputation from the aging biology it engages — it lowers the cell's growth signalling, switches on cellular self-cleaning, and triggers a wave of tissue renewal when you eat again — but the human longevity evidence is still a single moved biological-age clock, not a lifespan trial, and the weight loss it is actually marketed for is no better than ordinary calorie restriction. The most useful version is also the easiest: a consistent 12–14 hour overnight fast, anchored to the morning rather than the evening.

The reason fasting belongs in a longevity discussion at all is mechanistic. Going without food for long enough flips a metabolic switch — insulin and the growth pathways fall, the cell's low-fuel sensors come on, and the recycling machinery that clears damaged proteins and organelles is released. In animals this reliably extends lifespan. In humans the strongest signal so far is a small trial in which a fasting-style diet moved a validated biological-age clock; no randomised trial has ever tested whether fasting makes people live longer. Intermittent fasting has also been the most popular dietary "movement" of the last decade, and the weight-loss trials are now substantial enough to settle that question separately: it works, but no better than eating less the ordinary way. This article walks the longevity case first, then the metabolism, the weight-loss reality, the window-length question — how many hours actually matter — and the safety considerations.

Four protocols, distinguished

Protocol	Description	Best evidence
Time-restricted eating (TRE)	Daily eating compressed into a 6–10 hour window (most commonly 16 hours fasting / 8 hours eating, written "16:8")	Best-studied; early-day windows clearly superior to late-day
Alternate-day fasting (ADF)	Days of normal eating alternating with days of ~25% of energy needs	~5–8% weight loss over 8–12 weeks; adherence drops sharply long-term
Periodic fasting	Single multi-day fasts (24–72 hours, occasionally longer)	Stronger autophagy signals; harder to maintain
Fasting-mimicking diet (FMD)	Five-day low-calorie, low-protein, plant-based protocol designed to trigger fasting biology while still allowing some food	Multiple human trials (Wei et al., Sci Transl Med 2017); typically 4 cycles per year

The vast majority of public attention has focused on 16:8 time-restricted eating. The evidence base on TRE has matured the most; the others are less well-studied in humans.

What the evidence actually says

How to weigh it. The evidence is uneven across the claims. Strong that fasting engages the aging-relevant nutrient-sensing and self-cleaning machinery, and strong in animals that this extends lifespan; weak and unproven that it lengthens human life — the human case rests on a single trial that moved a biological-age estimate, not survival. Moderate-to-strong for blood-sugar control, the best-measured human benefit. Strong that intermittent fasting produces weight loss, but no better than ordinary calorie restriction. Moderate that eating earlier beats eating later at matched calories. Weak and fading for the cardiometabolic markers, which ride on continued adherence. The much-publicised eight-hour-window mortality signal is weak and contested.

The longevity case — strong in animals, unproven in humans

This is the honest centre of the topic. No randomised trial has tested whether intermittent fasting extends human lifespan. The animal evidence is strong but far less uniform than it is usually presented. When caloric restriction was tested across 41 genetically different mouse strains, it extended lifespan in only 5% of male and 21% of female strains — and shortened it in others.^[1] The largest such experiment ever run, in 960 genetically diverse mice, confirmed that both restriction and fasting extend life in proportion to their severity, but found that genetics mattered more than diet, that fasting did not help the heaviest animals, and — most awkwardly for the surrogate markers this field relies on — that the metabolic improvements produced by restriction were not associated with living longer.^[2] Its authors' conclusion is the one to carry forward: improving health and extending lifespan are not the same thing. The mechanism is nonetheless coherent — fasting hits the same nutrient-sensing pathways that, when dialled down, lengthen life across species (see Deregulated nutrient sensing). The translation to human lifespan is plausible but unproven, and marketers routinely blur that line. Don't.

The best human evidence is a surrogate, not a lifespan outcome. An exploratory analysis of blood samples from two fasting-mimicking-diet trials reported that three monthly cycles were associated with a biological age about 2.5 years lower — independent of weight loss. The measure is not an epigenetic clock but a composite score built from seven routine blood and blood-pressure readings, and the improvement tracked changes in four of its own seven inputs, so it is substantially a restatement of the short-term drops in inflammation, blood pressure and blood sugar reported alongside it. About a fifth of participants on the diet came out biologically older. The trial also reported lower insulin resistance, less liver fat measured by imaging, and a shift in the immune-cell profile (a higher lymphoid-to-myeloid ratio) toward a younger state. The trial diet was supplied by a company in which two of the authors hold equity.^[3] That is the same intervention, and the same roughly 2.5-year figure, discussed under deregulated nutrient sensing. It is genuinely encouraging — but it is one small trial, measuring a biological-age estimate rather than survival.

The metabolic switch and autophagy

The cellular mechanism is the part that is real and well-characterised. A long enough fast raises the cell's ratio of "spent" to "charged" energy currency, which activates the low-fuel sensor AMP-activated protein kinase (AMPK); at the same time, falling insulin, falling insulin-like growth factor 1 (IGF-1), and dropping amino-acid levels switch off the master growth regulator mTOR. With the growth signal down and the fuel sensor up, the cell shifts from building and storing to repairing and recycling.

That recycling is autophagy — the system that clears damaged organelles and misfolded proteins, and a hallmark of aging in its own right (see Disabled macroautophagy). The first clinical trial to measure it found that a single five-day fasting-mimicking diet cycle moves autophagy markers in human blood cells, with the effect persisting into refeeding.^[4] Ramadan fasting (dawn-to-sunset for a month, an extended form of time-restricted eating) shifts the same markers toward active autophagy. Two caveats keep this honest:

Exercise activates autophagy too. Human studies showing it have used prolonged endurance exercise — around two hours of cycling — rather than a typical session, so the honest version is that you do not need a fast to engage the system, not that any workout will do it.^[5]
Autophagic capacity appears to decline with age, though no human study has given the same fasting stimulus to young and old adults and compared the response directly — the age claim rests on indirect evidence.

Refeeding is half the story

The fast gets the attention, but recent work points to the refeeding transition as where much of the longevity benefit — and a real risk — is actually generated.

On the benefit side, a 2026 study in the roundworm C. elegans found that fasting's lifespan extension depends on how cleanly the animal switches back to a fed state. During the fast, a lipid-metabolism regulator (the receptor NHR-49) keeps fat-burning genes switched on; on refeeding, an enzyme (KIN-19) chemically silences it, and the metabolic gears re-engage. Engineer the worms so that regulator can't be switched off at refeeding, and the lifespan gain disappears entirely — even though the fast itself was unchanged.^[6] The finding is in worms, not people, so it is a mechanistic clue rather than human guidance — but it reframes fasting as a fast-and-refeed cycle, not a fast alone.

On the risk side, the same refeeding window carries a downside. Animal models show that the burst of cell proliferation after a long fast — exactly the tissue renewal that makes refeeding regenerative — comes with a transient elevated proliferation signal that can, in some contexts, accelerate tumour growth if a cancer-driving mutation is already present. In mice, post-fast refeeding drove intestinal stem-cell proliferation through the growth-sensor mTORC1 and polyamine synthesis — and when the tumour-suppressor gene Apc was deleted in refed stem cells, tumour incidence was higher than in either the fasted or the freely-fed state.^[7] Human relevance is unclear, but it argues against habitual very-long fasts for anyone with a cancer history.

Blood-sugar control — the strongest measured human benefit

Where fasting clearly delivers in humans is glucose handling. Intermittent fasting produces modest improvements in fasting glucose, fasting insulin, and HbA1c (the three-month blood-sugar average) in healthy adults, and the signal is biggest in type 2 diabetes and prediabetes. Pooling the four randomised trials that allow it (280 people), fasting lowered HbA1c by about 1.9 percentage points against control care (−1.85%, 95% CI −2.86 to −0.84; the bracketed range is where the true effect most plausibly lies) — though the trials disagree so violently with one another that the average is close to uninterpretable, and the same pooling found no significant change in body weight (−1.45 kg, 95% CI −5.51 to 2.61). The often-quoted figures of roughly 0.5% for people on oral diabetes drugs and 2.8% for those on insulin should be treated with much more caution: they are simple averages of before-and-after changes with no control group, the insulin figure rests on a single 12-week trial of 46 people, and that trial's own reported result works out closer to 0.7 percentage points. Anyone taking insulin who fasts needs close medical supervision because of low-blood-sugar risk — that part is well supported.^[8] The effect is real but modest: in a three-month randomised trial of adults with metabolic syndrome and prediabetes (the TIMET trial), a personalised time-restricted-eating schedule added to standard care improved HbA1c by 0.10 percentage points versus standard care alone, with no major adverse events.^[9] The trial was designed to detect exactly that much, and its authors regard a 0.1-point drop at this baseline as clinically meaningful — it is what the lifestyle arm of the Diabetes Prevention Program achieved.

The mechanism is straightforward: long fasting windows let insulin fall and fat-burning run, which uses up the toxic fat stored inside muscle cells that drives insulin resistance. See Metabolic flexibility for the cellular detail.

Early eating beats late eating at matched calories

The clearest piece of timing evidence: eating earlier in the day produces better metabolic outcomes than eating later, even when total calories are identical. A 2026 network meta-analysis ranked early-day windows ahead of late-day ones for both body weight and fasting insulin, at high certainty — though "early" there means finishing the last meal before 17:00, stricter than most people picture, and the ranking leans on indirect comparison because, as its authors note, the head-to-head trials have been conflicting.^[10] The best head-to-head test since is null: a crossover trial that put the same women through two weeks of eating from 08:00–16:00 and two weeks from 13:00–21:00, matched for food, found no difference in insulin sensitivity or any cardiometabolic marker — even though the late window measurably shifted their internal clocks.^[11] The timing case rests on ranked indirect evidence and mechanism, not on a decisive trial. The landmark mechanistic study — five weeks of early time-restricted feeding (eating between roughly 8:00 and 14:00) in men with prediabetes — improved insulin sensitivity, blood pressure, oxidative stress, and evening appetite without any weight loss.^[12]

The biology underneath this is the same circadian story that shows up everywhere on this site (see Circadian rhythms):

Pancreatic insulin secretion is faster and more efficient in the biological morning — the early insulin response to an identical meal is about 27% lower in the biological evening.^[13]
The same meal eaten in the biological evening produces a glucose rise about 12–17% larger than in the biological morning, measured under laboratory conditions that separate body-clock time from clock time.^[14]
The liver clock anticipates the feeding window, and in animals feeding time resets it independently of the brain's master clock — eating outside it produces metabolic friction with the liver's glucose handling.^[15]
Late eating makes you hungrier the next day — roughly doubling reported hunger and shifting appetite hormones toward eating more — and pushes fat tissue toward storage. It does not appear to disturb sleep: in the best-controlled test, total sleep time, sleep efficiency and every sleep stage were unchanged, and late eating slightly lowered rather than raised both energy expenditure and core body temperature.^[16] That trial's late supper still began two and a half hours before bedtime, so it does not speak to eating immediately before sleep.

Practical implication: a "skip breakfast, big late dinner" version of 16:8 is probably the worst chronotype for fasting — the half of this that cohort data supports most clearly is not skipping breakfast, where a later first meal tracks steadily higher mortality. Pushing dinner very early is less well supported than it sounds: in the largest cohort to examine both ends of the window, finishing before 19:00 was associated with slightly higher mortality than finishing between 19:00 and 20:00, as was eating after midnight.^[17] A "breakfast and lunch as the main meals, lighter dinner in the early evening" pattern is the defensible version.

Weight loss — the marketed benefit, no better than calorie restriction

This is the claim fasting is sold on, and the one the human trials have most decisively cut down to size. The 2025 BMJ network meta-analysis pooled 99 randomised trials covering 6,582 participants. All intermittent fasting strategies produced weight loss versus eating ad libitum. The comparison versus continuous calorie restriction was much closer: alternate-day fasting had a small statistical edge (roughly 1.3 kg better) in short trials, and that difference disappeared entirely in trials longer than 24 weeks.^[18] A 2026 Cochrane review of 22 randomised trials in 1,995 adults with overweight or obesity reached the same conclusion: intermittent fasting makes little to no difference to weight or quality of life versus standard dietary advice, and likely little to no difference versus no advice at all.^[19]

The honest reading: intermittent fasting is one acceptable structure for eating less, not a metabolically privileged route to weight loss. People who find a structured window easier to maintain than counting calories often benefit; people who find skipping breakfast destabilising do not.

Cardiometabolic markers — modest, and they fade without adherence

Blood pressure and inflammatory markers (C-reactive protein, interleukin-6) generally move favourably across fasting protocols, but the effect sizes are smaller and more variable than for glucose, and lipids are not uniformly good news: a 2026 meta-analysis stratified by age found LDL ("bad") cholesterol rose in adults over 45 under a conservative analysis, alongside loss of fat-free mass.^[20] The certainty on that is low, but it is a reason to re-check lipids rather than assume they improved. Most of the favourable marker changes wash out when fasting is compared against an equivalent calorie-restriction control. They are also not durable: the cardiovascular improvements typically appear within a few weeks of starting a structured regimen and dissipate within weeks of returning to unrestricted eating — the benefit rides on sustained adherence, not on a one-time reset.

The 8-hour caution and other concerning signals

A 2024 American Heart Association meeting abstract reported a 91% higher cardiovascular mortality in adults with self-reported eating windows under 8 hours, compared with those eating across 12 to 16 hours, in an analysis of a large US nutrition survey.^[21] That abstract has since been published in full.^[22] The cardiovascular signal not only persisted but grew: more than double the cardiovascular mortality against a 12–14 hour reference window (hazard ratio 2.35), and it survived all fourteen of the authors' sensitivity analyses, including moving the reference window, excluding deaths in the first year, and adjusting for measured blood pressure, blood sugar and cholesterol. Cancer mortality showed nothing (a 27% difference well within the range expected from chance), and all-cause mortality was only borderline — about 40% higher, but with a range that just clears no effect (HR 1.40, 95% CI 1.01–1.94) — and unlike the cardiovascular result it did not hold up under sensitivity analysis. The estimate rests on a small group: 33 cardiovascular deaths among 383 people, roughly 2% of the sample, who were younger, more often smokers, and food-insecure, and were never verified to be deliberate fasters. The risk was concentrated in people who already had cardiovascular disease or diabetes. In absolute terms the paper puts the excess at about 0.7 percentage points of cardiovascular death over fifteen years from age 50 in the sample overall, rising to about 3.8 points in those with existing cardiovascular disease or diabetes. Many fasting researchers — including 34 who wrote jointly to the American Heart Association — argue the signal is most plausibly confounded rather than causal. That letter responded to the 2024 conference abstract; no peer-reviewed rebuttal to the fuller 2025 publication has appeared, and the published version reported a larger cardiovascular hazard than the abstract did, not a smaller one. A separate cohort finds the same U-shape: lowest mortality at an ~11–12 hour window, with both very short (<8 h) and very long (≥15 h) windows linked to higher risk. Its authors tested for reverse causation by excluding early deaths and the short-window signal held, and argue that the crude dietary measure would if anything push their estimates toward zero rather than away from it.^[23] Worth knowing; not a reason to abandon early TRE.
Severe time-restricted eating under 6 hours sustained long-term has not been adequately studied for safety — particularly cardiovascular safety in adults with established disease.
The refeeding proliferation signal (above) argues against habitual very-long fasts in adults with a cancer history.
Rapid weight loss raises gallstone risk. Losing weight faster than about 1.5 kg per week prompts the liver to dump extra cholesterol into bile while the gallbladder empties less often, a well-established trigger for gallstones.^[24] Ordinary intermittent fasting rarely gets near that rate — the trials produce roughly 0.4–0.7 kg per week — so this is a caution for very aggressive protocols and very-low-calorie phases, not for 16:8.

The window-length question: how many hours?

This is the question most people actually arrive with — fast for 8, 10, 12 hours, or more? The trial data sorts daily eating windows into three useful tiers.

Short (eating window under 8 hours — i.e. fasting 16+ hours daily). Produces the fastest metabolic and weight effects, but adherence is poor — hunger, fatigue, irritability, and social friction drive high dropout — and this is the band where the cardiovascular-mortality signal sits. Not a sensible default. The 2026 timing meta-analysis points the same way from trial data rather than cohorts: windows under 8 hours ranked worse than longer ones for fasting glucose, triglycerides and LDL ("bad") cholesterol, even while they lowered fasting insulin.^[25]
Moderate (8–12 hour eating window — i.e. fasting 12–16 hours). The practical default, chosen for liveability rather than for measured superiority. The honest position is that the trials have not shown any window length to be metabolically better than another: the network meta-analysis that compared them concluded that "no particular metabolic advantages of various eating windows were found", and where it did detect differences they modestly favoured shorter windows for fasting insulin and glucose.^[26] A three-month randomised trial of a 10-hour window in middle-aged and older adults at risk of diabetes found no clinically relevant effect on body weight.^[27] What recommends this band is that it fits an ordinary day and an ordinary social life, which is what determines whether anyone is still doing it in a year.
Long (eating window over 12 hours — i.e. fasting under 12 hours). Effortless to sustain, but probably too short a fast to trigger a robust metabolic switch or much autophagy.

The key point is that there is no simple "longer is better" dose-response. Beyond about 16 hours of daily fasting the added benefit is small and the adherence and safety costs rise, and when you eat (early) matters at least as much as how long you fast. For most people the moderate band — a 12–14 hour overnight fast, nudged toward 16:8 if they want more and tolerate it well — captures essentially all of the realistic benefit.

The training interaction

A reasonable question: can you build muscle and intermittently fast? The answer is yes, with caveats.

Resistance training requires adequate total protein and energy. Fasted training is fine; the post-training meal, with protein, is what supports adaptation.
A 16:8 eating window is compatible with hypertrophy. Most adults can hit ~140–160 g of protein in an 8-hour window if they plan it — three meals at ~40–50 g protein each, with the post-training meal at the top end.
The pre-sleep protein finding — a ~40 g pre-sleep casein meal improves overnight muscle protein synthesis — directly conflicts with strict early-TRE protocols. See Protein for the detail.

The reason this matters: fasting-induced weight loss is not all fat. In a four-week alternate-day-fasting trial in young men, roughly a third of the weight lost was lean tissue — and a low dose of protein (25 g) on fasting days did not prevent it, though the authors themselves judged that dose too small to answer the question, and body composition was measured by bioelectrical impedance rather than a scan.^[28] Across the wider TRE literature the same caution holds: fat loss is reliable, but lean mass can fall unless protein and resistance training are deliberately added.^[29] The fix is the one this section already describes — adequate total protein (aim for at least 1.2–1.6 g/kg/day) plus resistance training preserves muscle while the fat comes off; the timing of the window matters far less than hitting that protein target.

The reasonable compromise: most days, an early or moderate TRE window (eating between, say, 8:00 and 18:00). On heavy training days, allow flexibility — including a post-training meal even if it pushes the window slightly later than usual.

Fasting while taking a GLP-1 drug

Glucagon-like peptide-1 (GLP-1) receptor agonists raise a question a lot of people now arrive with, and the honest answer is that nobody has tested it. As of mid-2026 there is no randomised trial of time-restricted eating or intermittent fasting added to semaglutide, tirzepatide or liraglutide — the published human evidence is a three-patient case series. The joint advisory from four US nutrition and obesity societies does not recommend adding a fasting window, and notes that people on these drugs often end up fasting unintentionally because they are not hungry, while warning that fasting raises hypoglycaemia risk in anyone taking glucose-lowering medication.^[30] Given that these drugs already cause meaningful lean-mass loss, the defensible position is that protein intake and resistance training matter more here than the eating window does — which is the same conclusion the fasting-and-training section reaches for everyone else. See GLP-1.

Who should not fast

The contraindication list is short and important.

Pregnancy and lactation. The largest evidence base is Ramadan fasting during pregnancy — dawn-to-sunset, no fluids — rather than the daily windows this page is about. There the newborn outcomes are reassuring, with no association with birth weight or prematurity, but children exposed in the womb score slightly lower on later cognitive tests, with the first trimester the sensitive window.^[31]
Type 1 diabetes, except under explicit medical supervision.
History of disordered eating — fasting can trigger or reinforce restrictive patterns, and the data on safety is not on the side of the people most drawn to the protocol. In a survey of nearly 2,800 Canadian adolescents and young adults, intermittent fasting was common — reported over the past year by about 48% of women, 38% of men, and 52% of transgender and gender non-conforming participants, the highest of the three groups — and in all three it was associated with more eating-disorder symptoms. The link to specific behaviours such as binge eating, vomiting and compulsive exercise was most consistent in women.^[32] That survey measured everyone at a single point in time, so it cannot show which came first; a smaller study that followed 491 adults over eight months found that fasting at the start predicted worse eating-disorder symptoms later, even after accounting for symptoms they already had.^[33] This is the most clinically important safety signal on the page.
Underweight adults, and anyone prone to osteoporosis. Bone is the most consistent human harm signal: bone-formation markers fall within the first days of a fast, and two years of sustained calorie restriction reduced spine and hip bone density.^[34] Those data come from severe interventions — a ten-day zero-calorie fast, two years of 25% calorie restriction, anorexia nervosa — and no evidence suggests daily 16:8 harms bone.
Children and adolescents.
Insulin or sulfonylurea users — extreme low-blood-sugar risk; only under explicit medical supervision.
Frail older adults — protein intake and muscle preservation are more important than fasting at this stage, and the risk of muscle loss dominates. See Protein.

Fasting is often said to disrupt women's hormones, and the trials do not bear this out at the doses people actually use. A twelve-month randomised trial that measured sex hormones separately in men, pre-menopausal women and post-menopausal women on an eight-hour eating window found no change in testosterone, dehydroepiandrosterone (DHEA) or sex-hormone-binding globulin in any group, and a 2026 review of time-restricted eating and female reproductive hormones reports the same across trials.^[35] Reproductive-axis suppression is real, but it belongs to a much deeper energy deficit — multi-day fasting and anorexia-level restriction — and at that dose it lowers reproductive hormones in men too.^[36] Menstrual-cycle changes are still worth treating as a stop signal, as with any diet that produces rapid weight loss, but there is no evidence for a female-specific window limit.

Polycystic ovary syndrome. Fasting is often promoted specifically for PCOS, and the trials support a narrower claim than the marketing. A six-month randomised trial found a six-hour eating window produced about 4% weight loss in women with PCOS — real, but statistically indistinguishable from ordinary calorie restriction, the same verdict as in everyone else.^[37] Pooled trials show insulin resistance improving and the free-androgen index falling, but that reflects more testosterone being bound rather than less being made: total testosterone does not measurably change.^[38] The outcome most women with PCOS actually want — more regular cycles and better fertility — has not been pooled by any of these analyses. So it is a reasonable option for weight and insulin resistance, on the same terms as any other structured way of eating less, and an open question for everything else.

Practical protocols

For most healthy midlife adults, the simplest evidence-based starting point is the easiest one — and the gains from going more aggressive are small.

Daily 12–14 hour overnight fast (the recommended default)

Last bite of food around 19:00; first food at 7:00–9:00 the next morning.
Aligns with circadian biology.
Effortlessly compatible with normal social patterns.
Low downside risk.

Daily 16:8 with an early window (for those who want more)

Eat between 9:00 and 17:00, or 10:00 and 18:00.
Strongest mechanistic signal in the trial data for adults without metabolic disease.
Compatible with morning training; awkward with late-evening social meals.

Periodic deeper fasts (optional)

A 24-hour fast (dinner to dinner) once a week is a moderate intervention with low cost.
One fasting-mimicking diet cycle (five days) per year is a reasonable option for adults drawn to the longevity and autophagy framework; the evidence base is real but small. If you do a multi-day fast, the refeeding matters as much as the fast — reintroduce food gradually rather than breaking a long fast with a large meal.

What to skip. Aggressive alternate-day fasting, multi-day water-only fasts without medical supervision, and one-meal-a-day patterns sustained long-term — the adherence cost is high, the metabolic upside over moderate TRE is small, and the muscle and bone consequences accumulate.

What's overrated

"Fasting reverses aging." The cellular mechanisms are real and the animal lifespan data is strong, but no human trial has shown fasting extends life — and the equivalent claim in humans is doing most of the marketing work behind the protocols people pay for.
OMAD (one meal a day) sustained long-term. Adherence and muscle-mass cost versus the metabolic upside over moderate TRE; not justified by the data.
Multi-day water-only fasts as "cellular cleanse." Autophagy is real; the supplement industry has wrapped legitimate biology in spa-day language that doesn't match the evidence.
Late-day TRE. Skipping breakfast to eat a heavy late dinner is the worst chronotype combination for the same total fasting window.