Skip to content
The Longevity Archive
browse 192
Intermittent8 human studies cited

The 5:2 Diet

1 day without food · Two non-consecutive days a week

Written by Reviewed Sep 2026

Also called: Intermittent energy restriction, Two fast days a week, Modified fasting

Two very-low-calorie days a week, usually 500 to 600 kcal, with normal eating on the other five. Across five randomised trials of 6 to 12 months it produced weight loss and cardiometabolic change similar to daily calorie restriction. The consistent difference is adherence and reported hunger, not physiology.

Overview

5:2 is not a true fast. In every trial that tested it the fast days carried food: 500 to 600 kcal in Carter, 400 kcal for women and 600 for men in Sundfor, roughly 2710 kJ in Harvie, and a 75% energy deficit on those days in the HELENA trial. What varies between trials is how much, not whether.

The question it was built to answer is whether concentrating a weekly energy deficit into two days does something a spread-out deficit does not. Five randomised trials running 6 to 12 months have now compared the two, and on weight, HbA1c, lipids and blood pressure the answer has been consistently that they land in the same place.

The more useful findings are the ones about behaviour. Hunger was rated higher on intermittent restriction in a year-long trial. Real-world adherence fell from 74% at 6 weeks to 22% at one year in a 300-person pragmatic trial. And in the trial that ran longest with the largest sample, 146 of 332 participants completed.

How it works

Two non-consecutive very-low-calorie days a week, five days of normal eating. The specific numbers used in the trials: 500 to 600 kcal/d on fast days in Carter's type 2 diabetes trial against a continuous arm at 1200 to 1500 kcal/d every day; 400 kcal for women and 600 kcal for men in Sundfor; 2100 kJ for women and 2520 kJ for men in Headland; and in HELENA a 75% deficit on the two days, giving a net weekly deficit of about 20% to match the continuous arm.

Note what those designs have in common: the two arms were matched for total weekly energy. 5:2 in these trials is a delivery schedule for a deficit, not an additional deficit.

Hajek 2021 is the only trial that tested it the way a person actually encounters it, as self-help instructions with or without six weekly group support sessions. That is where the adherence curve comes from.

What 1 day actually reaches

A 24-hour period at 500 to 600 kcal is not a water fast and the physiology reflects that. The one place 5:2 separated from continuous restriction in a randomised comparison was insulin: with a matched 25% energy restriction over 6 months in 107 premenopausal women, the intermittent arm showed greater reductions in fasting insulin and insulin resistance, with between-group differences of 1.2 microU/mL and 1.2 microU mmol/L respectively (both P = 0.04). Everything else in that trial, including leptin, free androgen index, high-sensitivity CRP, cholesterol, triglycerides and blood pressure, moved comparably in both arms. On autophagy, the closest human measurement to this protocol is a trial of 50 women, mean age 51, fasting 24 hours on three non-consecutive days a week for 8 weeks, with vastus lateralis biopsies taken after both 12-hour and 24-hour fasts. The 24-hour fast raised SQSTM1, and BECLIN1, SQSTM1 and LAMP2 mRNA fell in the reduced-energy arm after an overnight fast. The authors concluded that autophagy markers in human muscle were reduced, and attributed that to weight loss. In the same paper, intermittent fasting raised autophagy markers in mouse liver. Liver is the tissue where the effect is clearest in animals and it is the tissue nobody biopsies in a healthy volunteer. This page therefore does not claim that 5:2 reaches autophagy: the one human measurement at this protocol shape pointed the other way in the tissue that was sampled.

Fig. 1 · what has been measured, hour by hour
  1. 0h to 4h

    Still absorbing

    Burning
    Glucose from the meal just eaten, with insulin high enough to hold fat release down
    • Insulin is at its post-meal peak, and while it is there, adipose tissue lipase stays suppressed and fat is being stored rather than released.
    • The hour count on most fasting charts starts at the last bite. Absorption of a mixed meal is still running here, so the clock and the physiology are not yet the same thing.

    Read the biologyInsulin and insulin sensitivity during a fast

    Measured in people
    The human hour-by-hour fasting series begin later than this. Klein 1993 took its first sample at 12 hours in 6 healthy men, and Rothman 1991 reports its first gluconeogenesis figure as an average across the first 22 hours. This row is the fed state those measurements are compared against, and no extended-fasting measurement exists inside it.

  2. 4h to 12h

    Post-absorptive

    Burning
    A mix of remaining glucose and fatty acids, with the liver covering the brain's glucose demand
    Blood glucose
    5.58 ± 0.08 mmol/L (100.5 mg/dL) at the 12 hour mark, in 6 healthy men
    Ketones (BHB)
    Total ketone bodies around 0.20 mmol/L in overnight-fasted subjects. That figure is acetoacetate plus beta-hydroxybutyrate, not BHB on its own.
    Insulin
    64.6 ± 12.9 pmol/L (9.3 microunits/mL) at 12 hours, in the same 6 men
    • Glycerol release runs at 2.08 ± 0.22 and palmitic acid release at 1.63 ± 0.20 micromol/kg/min at 12 hours. Those are the numbers everything later in the fast is measured against.
    • Gluconeogenesis is already the majority contributor to glucose production rather than a reserve waiting to be called on, which is the first place the popular switch model breaks.

    Read the biologyInsulin and insulin sensitivity during a fastGlycogen depletion and gluconeogenesis

    Measured in people
    Klein 1993 measured glucose, insulin and stable-isotope lipid kinetics at 12 hours in 6 healthy men. Féry 1983 measured total ketone body turnover in overnight-fasted subjects and put the baseline at 0.20 mmol/L. Rothman 1991 measured hepatic glycogen serially by 13C nuclear magnetic resonance through a 68 hour fast in healthy adults and calculated that gluconeogenesis supplied 64 ± 5% of total glucose production across the first 22 hours.

  3. 12h to 18h

    Ketones have not moved yet

    Burning
    Fatty acids rising, with glucose production increasingly made rather than released from store
    Ketones (BHB)
    Median change of 0% between 12 and 18 hours, sampled every 6 hours in 34 adults
    • Beta-hydroxybutyrate does not move in this window. In 34 adults sampled every 6 hours through the standard diagnostic 72 hour fast, the median change from 12 to 18 hours was zero.
    • Gluconeogenesis is supplying 64 ± 5% of glucose production across the first 22 hours, confirmed by an orthogonal method at about 47% by 14 hours. Glucose production does not switch from glycogen to fat at a threshold hour.
    • Whole blood LC3A messenger RNA was 22 ± 5% higher at 18 hours than in a 12 hour control arm in 11 overweight adults (p = 0.001). In the same subjects MTOR expression also rose, by 9 ± 3%.

    Read the biologyGlycogen depletion and gluconeogenesisKetogenesis and beta-hydroxybutyrateAutophagy

    Measured in people
    Service 2005 measured beta-hydroxybutyrate every 6 hours in 34 adults whose 72 hour fasts were negative for insulinoma. Rothman 1991 measured the gluconeogenic share by 13C NMR. Jamshed 2019 measured whole blood messenger RNA in an 11-person 4 day randomised crossover of an 08:00 to 14:00 eating window against 08:00 to 20:00. That is one gene's expression in blood cells, not autophagic flux and not muscle. No study in any species has tested a 16 hour timepoint for autophagy.

  4. 18h to 1d

    The steepest hours for fat release

    Burning
    Fatty acids and glycerol, with ketone production starting its climb
    Ketones (BHB)
    Climbing. The median rise from 18 to 36 hours was +333%, the steepest proportional stretch of a 72 hour fast.
    Insulin
    70% of the entire decline seen across a 72 hour fast has already happened by 24 hours
    • The largest single interval increase in lipolysis across a whole 72 hour fast falls between 18 and 24 hours, and 60% of the total rise in lipid kinetics happens between 12 and 24 hours. The authors attribute that to the falling insulin rather than to any change in glucose.
    • Growth hormone was significantly higher after a 24 hour water-only fast than after a fed day in a randomised crossover of 30 healthy adults (p = 1.1 x 10^-4). The size of that rise is not in the accessible record and is not published here.
    • Haemoglobin, red cell count and haematocrit all rose together at 24 hours in the same trial. That combination is the signature of plasma volume contraction, not of new red cells.
    • 24 hour energy expenditure measured in a whole-room calorimeter fell during 24 hour fasting in 20 volunteers, with the size of the fall varying between individuals.

    Read the biologyInsulin and insulin sensitivity during a fastThe growth hormone responseKetogenesis and beta-hydroxybutyrateAutophagy

    Measured in people
    Horne 2013 randomised 30 apparently healthy adults to a 24 hour water-only fast or a day of usual eating in crossover and measured growth hormone, blood counts, cholesterol, triglycerides, bicarbonate and weight. Klein 1993 measured the lipid kinetics by stable isotope tracer. Hollstein 2020 measured 24 hour energy expenditure by whole-room calorimetry in 20 volunteers; the exact percentage decrease is not in the accessible record and is not published here. On autophagy, a 24 hour fast raised p62/SQSTM1 in the vastus lateralis of 50 women, which is movement in the direction of less autophagic degradation rather than more.

  5. 1d to 36h

    Thyroid drops before metabolic rate does

    past this window
    Burning
    Fat, with the remaining glucose built from glycerol, lactate and amino acids
    Ketones (BHB)
    1.308 ± 1.053 mmol/L after day 1 in 13 men beginning a 10 day water fast, from a baseline of 0.177 ± 0.044. The spread on that mean is as wide as the mean.
    • Gluconeogenesis supplies 82 ± 5% of glucose production between 22 and 36 hours.
    • Serum T3 fell from 2.30 ± 0.06 to 1.84 ± 0.03 nmol/L by 30 hours in 8 healthy men (p < 0.01), TSH fell below 1 mU/L and the nocturnal TSH peak was abolished. Serum cortisol did not differ across conditions.
    • Thyroid suppression therefore begins inside the first 30 hours, well before any measurable fall in resting metabolic rate.
    • Sodium excretion has not turned yet. In 9 obese women fasting with pre-fast salt intake maintained, urinary sodium began to rise after roughly 48 hours and then exceeded intake.

    Read the biologyGlycogen depletion and gluconeogenesisKetogenesis and beta-hydroxybutyrateAutophagy

    Measured in people
    Rothman 1991 measured the gluconeogenic share by 13C NMR in healthy adults. Hugues 1984 fasted 8 healthy men for 30 hours with serial thyroid sampling. Dai 2022 measured blood beta-hydroxybutyrate daily in 13 men fasting in a controlled facility. Sigler 1975 collected urine in 3 hour blocks from 9 obese women on a metabolic ward. On autophagy, the only human study with several within-fast timepoints took vastus lateralis biopsies at 2, 12, 24 and 36 hours: LC3I, LC3II and p62 all fell, in untrained subjects only, which the authors summarise as skeletal muscle autophagy being only modestly affected by 36 hours of fasting.

Human evidence

Five randomised trials of 6 to 12 months comparing intermittent against continuous energy restriction, including one noninferiority trial in type 2 diabetes and one 300-person pragmatic trial that measured real-world adherence as an outcome.

  • Harvie 2011, 6 months, 107 overweight or obese premenopausal women, matched 25% energy restriction delivered intermittently or continuously. Weight fell 6.4 kg intermittent and 5.6 kg continuous (P for difference = 0.4). Reductions in fasting insulin and insulin resistance were modest in both arms but greater with intermittent restriction (both P = 0.04). Leptin, free androgen index, CRP, cholesterol, triglycerides and blood pressure changed comparably (International Journal of Obesity, 2011).
  • Schubel 2018, the HELENA trial, 150 overweight and obese non-smokers, 12-week intervention plus 12-week maintenance and 26-week follow-up. Over the intervention phase weight fell 7.1% intermittent, 5.2% continuous and 3.3% control (P for intermittent versus continuous = 0.053). By week 50 the arms were level at 5.2% and 4.9% (P = 0.89). No significant differences between the restriction arms in expression of 82 preselected adipose genes or in circulating biomarkers (American Journal of Clinical Nutrition, 2018).
  • Carter 2018, 12-month noninferiority trial in 137 adults with type 2 diabetes, 500 to 600 kcal on two non-consecutive days versus 1200 to 1500 kcal daily. HbA1c fell 0.5% continuous and 0.3% intermittent (P = .65), a difference of 0.2% (90% CI -0.2% to 0.5%) that met the prespecified equivalence margin. Weight fell 5.0 kg and 6.8 kg, a difference of -1.8 kg (90% CI -3.7 to 0.07) that did not meet the 2.5 kg equivalence margin, and neither fat mass nor fat-free mass met their margins. Both statements are true and both belong in the summary (JAMA Network Open, 2018).
  • Sundfor 2018, 112 adults with abdominal obesity over one year. Weight fell 8.0 kg intermittent and 9.0 kg continuous (P = 0.6) with regain of 1.1 kg and 0.4 kg (P = 0.6) and no between-group differences in waist circumference, blood pressure, triglycerides or HDL. Hunger was rated higher on intermittent restriction, 4.7 against 3.6 on a subjective numeric scale (P = 0.002) (Nutrition, Metabolism and Cardiovascular Diseases, 2018).
  • Headland 2019, one year, 332 randomised across continuous restriction, week-on week-off restriction and 5:2. 146 of 332 completed, a 44% completion rate that is the single most useful number in the trial. Among completers weight fell 6.6 kg, 5.1 kg and 5.0 kg respectively (P = 0.2 for time by diet), with HDL up 7% and triglycerides down 13% and no differences between groups (International Journal of Obesity, 2019).
  • Hajek 2021, 300 adults randomised to standard brief advice, 5:2 self-help, or 5:2 self-help plus six weekly group sessions, followed one year. Adherence was 74% at 6 weeks, 31% at 6 months and 22% at one year. Weight at one year fell 1.9 kg on 5:2 against 1.8 kg on standard advice (P = 0.79). 5:2 received higher participant ratings and did not undermine fat and fibre intake or physical activity (PLoS ONE, 2021).

What this does not tell you: Every trial here compared 5:2 against continuous restriction or brief advice, not against unrestricted eating, so the question they answer is about delivery schedule rather than about fasting. Diet was self-reported throughout. Body composition was reported in only some trials and never as a powered primary outcome. Completion was 44% in the largest trial. No randomised trial of this protocol was designed or powered to measure cardiovascular events or mortality.

Reading the research record

The trials in this section agree with each other more than most areas of nutrition research do: matched weekly energy delivered in two blocks or spread across seven days produces about the same weight and about the same cardiometabolic change. Harvie's insulin result is the main exception and it has not been replicated in the later, longer trials, where fasting glucose, fasting insulin and insulin resistance did not separate.

The disagreement worth reading carefully is not between these trials. It is between the trial literature and an observational analysis that reached the public before it reached peer review.

In March 2024 a research abstract presented at an American Heart Association conference reported that people with a derived eating window under 8 hours had a 91% higher risk of cardiovascular death. It was covered heavily. It was a conference abstract, which the AHA itself describes as preliminary and not peer reviewed.

It has since been published in full. The 2025 paper reports 19,831 NHANES adults from 2003 to 2018 with mortality through December 2019 and a median follow-up of 8.1 years, and gives a cardiovascular mortality hazard ratio of 2.35 (95% CI 1.39 to 3.98) for an eating duration under 8 hours against a reference of 12 to 14 hours, which is a larger figure than the 91% the abstract implied, with a wide confidence interval. The association was not significant for all-cause mortality or for cancer mortality, and the all-cause association did not survive many sensitivity analyses. The cardiovascular association was consistent across 8 subgroups and survived 14 sensitivity analyses. The authors' own conclusion is that further research is required to determine whether the risk is attributable to the short eating duration itself or to residual confounding.

A commentary signed by a large group of circadian and nutrition researchers, including principal investigators on several trials cited across this site, was published in the same journal arguing that short-term 24-hour dietary recalls from observational studies cannot support claims on mortality. Several signatories report industry relationships in that paper's competing-interests statement. The exposure in question is a derived eating window from two days of self-reported recall standing in for a habitual pattern across a median of 8 years. The named criticisms at the abstract stage were that the short-window group differed at baseline in ways that predict cardiovascular risk and likely included shift workers rather than deliberate fasters. This page reports the published hazard ratio, the null results alongside it, and the rebuttal, and does not treat any of it as evidence from a randomised trial.

What happens to muscle

5:2 has been measured less carefully for body composition than the time-restricted protocols, and where it was measured the equivalence people assume was not demonstrated.

Carter 2018 is the clearest case. In a 12-month noninferiority design, the between-group fat mass difference was -1.3 kg (90% CI -2.8 to 0.2) and the fat-free mass difference was -0.5 kg (90% CI -1.4 to 0.4), and neither met its prespecified equivalence margin. That is not a finding that 5:2 lost more lean mass. It is a finding that the trial could not demonstrate the two approaches were equivalent on body composition, which is a different and more honest statement than the one usually made.

Headland 2019 reported body fat loss as similarly non-different across its three arms among completers. Harvie and Sundfor reported comparable cardiometabolic change without separating lean mass as a headline figure.

The muscle biopsy trial closest to this protocol shape, 50 women fasting 24 hours on three days a week for 8 weeks, found autophagy markers in human muscle reduced rather than raised, which the authors attributed to weight loss.

What the wider protein literature contributes is the general rule that applies here as much as anywhere: across an energy deficit, the proportion of weight lost as fat-free mass was lower at 1.6 and 2.4 g/kg/day than at 0.8 g/kg/day, resistance training is present in every trial where lean mass was maintained or gained during a deficit, and a slower rate of loss preserved more lean tissue for the same total weight lost. None of the 5:2 trials controlled protein intake, and none included a supervised training programme.

Doing it

What a fast day actually contained
500 to 600 kcal in the type 2 diabetes trial, 400 kcal for women and 600 for men in the one-year Norwegian trial, and a 75% deficit on the two days in HELENA. None of these trials tested a zero-calorie day.
Protein on fast days
No 5:2 trial controlled protein. The nearest randomised evidence on protein in severe restriction is a trial of 17 obese women on two isocaloric 500 kcal diets where 1.5 g protein per kg of ideal body weight held nitrogen balance at zero while 0.8 g/kg gave -2 g N/day. Note the denominator: that prescription is grams per kilogram of ideal body weight, not lean body mass.
Electrolytes and fluid
No published protocol monitors electrolytes below 24 hours, and the natriuresis and diuresis of a fast peak in the 24 to 72 hour window, which is when orthostatic symptoms are most reported. A 500 kcal day that includes salt and fluid is a different exposure from a dry or zero-calorie day.
Space the days out
Every trial used two non-consecutive days. Nobody has tested two consecutive days against two separated ones.
Plan for month six, not week six
Adherence in the pragmatic trial ran 74% at 6 weeks, 31% at 6 months and 22% at one year, and the group-support arm's early advantage was gone by twelve months. The largest trial completed 146 of 332.
Medication timing on fast days
One published guide for non-religious intermittent fasting advises skipping a sulfonylurea on a 24-hour fast day, halving it for a partial day, skipping prandial insulin when no carbohydrate follows, and reducing basal insulin to a third or half of the usual dose. That is a prescriber's decision, not a self-adjustment.

Breaking the fast

The trials broke each fast day by returning to normal eating the next day, with no prescribed refeed. What the metabolic record adds is that after a genuinely prolonged fast, protein refeeding turns nitrogen balance positive while carbohydrate refeeding does not: in 10 moderately obese women after a three-week fast, one week of 80 g of protein moved nitrogen balance from -5.9 g/day to +1.7 g/day (P < 0.05), while 80 g of carbohydrate moved it only to -3.1 g/day. That measurement comes from a three-week fast rather than a 24-hour day, and refeeding syndrome is not a concern at this protocol unless a person met the NICE high-risk criteria before they started.

Citations

  1. Human2011
    The effects of intermittent or continuous energy restriction on weight loss and metabolic disease risk markers: a randomized trial in young overweight women

    International Journal of Obesity

    107 overweight or obese premenopausal women, 6 months, matched 25% energy restriction delivered intermittently (about 2710 kJ/day on 2 days a week) or continuously. Weight -6.4 kg versus -5.6 kg (P = 0.4). Fasting insulin and insulin resistance fell more with intermittent restriction, between-group differences -1.2 microU/mL and -1.2 microU mmol/L (both P = 0.04). Leptin, free androgen index, CRP, cholesterol, triglycerides and blood pressure comparable.

  2. Human2018
    Effects of intermittent and continuous calorie restriction on body weight and metabolism over 50 wk: a randomized controlled trial

    American Journal of Clinical Nutrition

    150 overweight and obese non-smokers randomised to 5:2 (2 days at a 75% deficit, net weekly deficit about 20%), continuous restriction at the same weekly deficit, or control. Intervention phase weight -7.1% versus -5.2% versus -3.3% (intermittent versus continuous P = 0.053). At week 50, -5.2% versus -4.9% (P = 0.89). No significant differences between restriction arms in 82 preselected adipose genes or circulating biomarkers.

  3. Human2018
    Effect of Intermittent Compared With Continuous Energy Restricted Diet on Glycemic Control in Patients With Type 2 Diabetes: A Randomized Noninferiority Trial

    JAMA Network Open

    137 adults with type 2 diabetes, mean baseline HbA1c 7.3%, 12 months. HbA1c -0.5% continuous versus -0.3% intermittent (P = .65), difference 0.2% (90% CI -0.2% to 0.5%), meeting the equivalence margin. Weight -5.0 versus -6.8 kg, difference -1.8 kg (90% CI -3.7 to 0.07), not meeting the 2.5 kg margin; fat mass -1.3 kg and fat-free mass -0.5 kg also did not meet their margins. Hypoglycaemic or hyperglycaemic events in the first 2 weeks were similar between arms.

  4. Human2018
    Effect of intermittent versus continuous energy restriction on weight loss, maintenance and cardiometabolic risk: A randomized 1-year trial

    Nutrition, Metabolism and Cardiovascular Diseases

    112 adults with abdominal obesity, 6-month weight-loss phase then 6-month maintenance. Weight -8.0 kg intermittent versus -9.0 kg continuous (P = 0.6); regain 1.1 kg versus 0.4 kg (P = 0.6). Waist, blood pressure, triglycerides and HDL improved with no between-group difference. Hunger was rated higher on intermittent restriction, 4.7 versus 3.6 (P = 0.002).

  5. Human2019
    Effect of intermittent compared to continuous energy restriction on weight loss and weight maintenance after 12 months in healthy overweight or obese adults

    International Journal of Obesity

    332 randomised to continuous restriction, week-on week-off restriction or 5:2 for one year. 146 of 332 completed, a 44% completion rate, with discontinuation not differing across arms (P = 0.4). Among completers, weight -6.6 kg, -5.1 kg and -5.0 kg (P = 0.2 for time by diet), with body fat loss similarly non-different. HDL rose 7% and triglycerides fell 13% with no group differences. An erratum was published for this paper.

  6. Human2021
    A randomised controlled trial of the 5:2 diet

    PLoS ONE

    300 adults with obesity randomised to standard brief advice, 5:2 self-help instructions, or 5:2 self-help plus six weekly group support sessions, followed one year. Adherence 74% at 6 weeks, 31% at 6 months, 22% at one year. Weight at one year -1.9 kg versus -1.8 kg on standard advice (P = 0.79). Group support gave more loss at 6 weeks (P = 0.02) but the difference was gone by one year (P = 0.37). 5:2 received higher participant ratings.

  7. Human2022
    Intermittent fasting activates markers of autophagy in mouse liver, but not muscle from mouse or humans

    Nutrition

    50 women, mean age 51, randomised to a 24-hour fast on 3 non-consecutive days a week at 70% or 100% of energy requirements for 8 weeks, with vastus lateralis biopsies after 12 and 24 hour fasts. The 24-hour fast raised SQSTM1; BECLIN1, SQSTM1 and LAMP2 mRNA fell in the 70% arm after an overnight fast. The authors concluded autophagy markers in human muscle were reduced and attributed this to weight loss. The same intervention raised autophagy markers in mouse liver.

  8. Human2025
    Association of eating duration less than 8 h with all-cause, cardiovascular, and cancer mortality

    Diabetes and Metabolic Syndrome

    Observational, not a trial. 19,831 NHANES adults from 2003 to 2018 with mortality through December 2019, median follow-up 8.1 years, eating duration derived from two 24-hour dietary recalls. Cardiovascular mortality hazard ratio 2.35 (95% CI 1.39 to 3.98) for under 8 hours versus 12 to 14 hours, consistent across 8 subgroups and 14 sensitivity analyses. Not significant for all-cause or cancer mortality, and the all-cause association did not survive many sensitivity analyses. The authors state further research is needed to separate the exposure from residual confounding.

  9. Review2025
    Short-term 24h dietary recalls from observational studies cannot support claims on mortality

    Diabetes and Metabolic Syndrome

    Published commentary in the same journal, signed by a large group of circadian and nutrition researchers, arguing that an eating window derived from two days of dietary recall cannot support mortality claims. Several signatories are principal investigators on the intervention trials cited across this site, and several report industry relationships in the competing-interests statement.

What people report

These are uncontrolled self-reports, not evidence. They are here because they tell you what to expect and what to watch for, which the trial literature does not. They cannot tell you whether anything works.

  • Higher hunger than on daily restriction, measured at 4.7 against 3.6 on a subjective scale in a one-year trial
  • Preferring two hard days to seven restricted ones, reflected in the higher participant ratings in the pragmatic trial
  • Adherence decaying over months rather than failing at the start
  • Fast days getting easier after the first few weeks
  • Eating more than prescribed on fast days, which was measured in the alternate-day literature and is the same behaviour

Sources: Uncontrolled self-report, except for the hunger rating and the adherence figures, which were prespecified measured outcomes in Sundfor 2018 and Hajek 2021 respectively. A separate review of what fasting communities report is still being compiled and nothing here is drawn from it.

Who this is wrong for

  • Anyone taking an SGLT2 inhibitor without a prescriber stopping it first. A 24-hour fast day is the exposure that has produced euglycaemic ketoacidosis in published cases, and the blood glucose reading looks normal or only mildly raised while it happens.
  • Anyone on insulin, a sulfonylurea or a meglitinide without a prescriber writing a fast-day plan. Symptomatic hypoglycaemia in a 1378-person sulfonylurea cohort ran 25.6% on glibenclamide, 16.8% on glimepiride and 14.0% on gliclazide, with 6.7% severe events overall.
  • Anyone with type 1 diabetes outside a structured programme, where the risk runs in both directions: hypoglycaemia if basal insulin is not reduced, ketoacidosis if it is cut too far.
  • Pregnancy and breastfeeding. Lactation draws roughly 30 to 50 g of glucose a day out of the mother, and dieting or exercise to lose weight was the precipitant in 76% of 18 reported cases of lactation ketoacidosis.
  • Anyone with a history of an eating disorder or current loss-of-control eating. Restriction followed by unrestricted eating is structurally the cycle clinicians screen for.
  • Children and adolescents.
  • Anyone with a BMI under 18.5 or meeting the NICE refeeding risk criteria, which include little or no nutritional intake for more than 5 days as one of the paired criteria.
  • Anyone with chronic kidney disease on an ACE inhibitor or ARB. In 131 patients with CKD, serum creatinine rose during fasting in 60.4% by day 7, and the rise was associated with taking renin-angiotensin-aldosterone system antagonists (relative risk 2, P = 0.002).
  • Anyone with adrenal insufficiency. In a randomised crossover in 30 patients on replacement therapy, complications occurred in 67% during fasting against 23% outside it and 37% broke the fast because of one, even though continuous glucose monitoring showed no significant difference in measured hypoglycaemia.
  • Anyone with gout or on urate-lowering treatment, since ketone bodies compete with urate for tubular secretion so serum urate rises, and drinking more water does not address it.

Questions

Is 5:2 better than eating less every day?
Across five randomised trials of 6 to 12 months, the two produced similar weight loss and similar cardiometabolic change. One trial found greater reductions in fasting insulin and insulin resistance with intermittent restriction (both P = 0.04) and later, longer trials did not reproduce it. The noninferiority trial in type 2 diabetes met its equivalence margin for HbA1c and did not meet it for weight or body composition, in either direction.
What about the study saying short eating windows raise heart death risk?
That began as a conference abstract reporting a 91% higher risk and was published in full in 2025. The paper reports a cardiovascular mortality hazard ratio of 2.35 (95% CI 1.39 to 3.98) for an eating duration under 8 hours against 12 to 14 hours, and null results for all-cause and cancer mortality. The exposure is derived from two days of self-reported dietary recall standing in for a habitual pattern across a median of 8 years. A formal rebuttal signed by a large group of researchers was published in the same journal, and the authors of the paper itself decline to interpret the result causally.
Do fast days have to be zero calories?
No trial in this section tested that. The fast days carried 500 to 600 kcal in the type 2 diabetes trial, 400 kcal for women and 600 for men in the one-year Norwegian trial, and a 75% deficit in HELENA. All the results on this page belong to those numbers.
Will I stick to it?
The pragmatic trial that measured this in 300 people found adherence at 74% at 6 weeks, 31% at 6 months and 22% at one year, and the largest trial completed 146 of 332 participants over twelve months. Participants also rated 5:2 more highly than standard advice, and hunger on fast days was rated significantly higher than on daily restriction in a separate year-long trial.