Short answer. For most people the right fast is the shortest one they keep. Daily windows of 14 to 16 hours are where the daily-plan trials sit, and in trials that matched calories no fasting pattern beat the same cut spread over every day. A 24 or 36-hour fast is an occasional tool with thin evidence; from 48 hours we found no trial of people fasting at home, and the main multi-day dataset is a clinic cohort, supervised daily and refed over several days.
How long should a fast be? The shortest one you keep
For most people, a fast should be as long as the plan they will still keep in three months. Daily windows of 14 to 16 hours are where the daily-plan trials sit, with 16:8 fasting as the usual default. In head-to-head trials of 100 to 150 adults over 50 weeks to 12 months, alternate-day fasting, 5:2 and an 8-hour window added to a calorie cut each matched, and did not beat, the same calorie cut spread over every day (Trepanowski et al., JAMA Internal Medicine, 2017; Schübel et al., American Journal of Clinical Nutrition, 2018; Liu et al., New England Journal of Medicine, 2022).
The effect of any of these patterns is modest. A 2022 review (Varady et al., Nature Reviews Endocrinology, 2022), as summarised by the lead author’s university (UIC press release, March 2022), puts weight loss at 3 to 8% from baseline over 8 to 12 weeks, and notes that only a few studies have examined the health benefits in humans. So the hours matter less than whether you keep them.
The length ladder, 12 to 72 hours
In a daily plan, the first number is the hours without calories and the second is the eating window. From 24 hours up, the fast is a whole day or more, done now and then rather than daily. The last column shows which lengths Bare Fast, the app we make, has free and which need Pro.
| Length and use | Evidence | Watch for | In Bare Fast |
|---|---|---|---|
| 12:12, a first step: sleep plus a late breakfast or an early dinner | No trial we found of 12:12 itself; half of adults in one app study ate across more than 14.75 h (Gill and Panda 2015) | Late snacks | 13:11 free; 12 h custom, Pro |
| 14:10, a daily plan that fits most working days | 19 adults, 12 weeks, no control group: lower weight and blood pressure (Wilkinson 2020) | Eating creeping past the window | Free |
| 16:8, the default daily plan | 116 adults, 12 weeks: about 1 kg, no better than three meals a day (Lowe 2020) | Eating more inside the window | Free |
| 18:6, a tighter daily window | About 3% in 8 weeks, the same as 20:4 (Cienfuegos 2020) | Hunger late in the fast | Free |
| 20:4, one main meal and a snack | Same trial: no difference from 18:6 | Fitting enough food into 4 h | Free |
| 23:1 (OMAD), one meal a day | Short crossovers in normal-weight adults; more hunger and higher blood pressure in one (Stote 2007) | Hunger, blood pressure, eating enough | Free |
| 24 h, an occasional full day, usually once a week | 103 adults, 26 weeks: LDL unchanged, insulin-resistance score better (Bartholomew 2021) | Headache, nausea, medicine timing | Pro |
| 36 h, occasional: dinner to breakfast the day after next | Physiology only: ketones above 0.5 mmol/L at about 21 h (Deru 2021) | Dizziness, fluids, salt, the first meal after | Pro |
| 48 h, rare: two full days | No at-home trial we found; lab only, 12 lean men (Andriessen 2023) | Electrolytes, medication, someone around | Pro |
| 72 h, rare: three full days | No at-home trial we found; one supervised clinic cohort (Wilhelmi de Toledo 2019) | Medical supervision, refeeding | Custom length, Pro |
Daily plans: 12:12 to one meal a day
12:12 and 14:10
No trial we found has tested 12:12 on its own, but it is a real change for many people: in a study that tracked meals with a phone app (Gill and Panda, Cell Metabolism, 2015), half of the adults ate across more than 14.75 hours a day. 14:10 has one small study behind it: 19 adults with metabolic syndrome kept a self-chosen 10-hour window for 12 weeks and lost weight, with lower blood pressure and blood fats, but with no control group the study cannot say how much of that the window caused (Wilkinson et al., Cell Metabolism, 2020). Both are where a beginner starts; how to start intermittent fasting covers the first two weeks.
16:8, 18:6 and 20:4
Of the daily plans, 16:8 has the most trial time. In a 12-week randomised trial of 116 adults, eating between noon and 8 pm did no better than three meals a day (Lowe et al., JAMA Internal Medicine, 2020); windows and a first week are in the 16:8 fasting schedule. A tighter window did not do more: in an 8-week trial in adults with obesity, 4-hour and 6-hour windows produced comparable weight loss of about 3% (Cienfuegos et al., Cell Metabolism, 2020). If 18:6 suits you, that trial gives no reason to squeeze down to 20:4.
23:1, one meal a day
The trials of one meal a day we found are short, small crossovers, all in people without overweight. In two 8-week periods in normal-weight, middle-aged adults (Stote et al., American Journal of Clinical Nutrition, 2007), one meal a day with calories held equal to three meals brought more hunger and raised blood pressure and total, LDL and HDL cholesterol. An 11-day crossover in lean adults (Meessen et al., Frontiers in Physiology, 2022) found slightly more weight lost with a single evening meal than with three meals. Neither says what 23:1 does over months, and the 2007 result is a reason to have blood pressure and cholesterol checked if you stay on one meal a day.
Occasional longer fasts: 24, 36, 48 and 72 hours
24 hours
A 24-hour fast has a randomised trial in people living at home, and the result is mixed. The WONDERFUL trial (Bartholomew et al., European Heart Journal Open, 2021) enrolled 103 adults with modestly raised LDL cholesterol and at least one feature of metabolic syndrome or type 2 diabetes, on no statin or diabetes medicine. They fasted on water for 24 hours twice a week for 4 weeks, then once a week for 22 weeks. LDL did not change against control, an insulin-resistance score (HOMA-IR) improved, and 71 of the 103 finished with usable data. The Cleveland Clinic (What is intermittent fasting?, January 2026) warns that side effects of a 24-hour fast can be extreme: fatigue, headaches, irritability, hunger, low energy and nausea.
36 hours
A 36-hour fast, usually from dinner one day to breakfast the day after next, has no outcome trial we found in people doing it at home. The evidence we found is physiology: in a crossover in which 20 adults each completed two 36-hour fasts (Deru et al., Medicine and Science in Sports and Exercise, 2021), blood ketones passed 0.5 mmol/L at about 21 hours. The rest of the picture, phase by phase, is on what happens when you fast, hour by hour. The same phases sit on the ring in Bare Fast, so you can see which one you are in.
48 and 72 hours
No trial we found has tested a 48 or 72-hour fast in people fasting at home, so nobody can tell you what it does for your health. The lab evidence is a 60-hour fast in 12 healthy, lean men in a sealed room that measures energy use; they burned less energy per day during the fast (Andriessen et al., Clinical Nutrition, 2023).
The main safety dataset is a clinic cohort: 1,422 people fasted for 4 to 21 days on 200 to 250 kcal a day under daily supervision of nurses and specialised physicians, and food came back stepwise over an average of four days (Wilhelmi de Toledo et al., PLoS ONE, 2019). The authors say the results may not apply to people fasting without medical supervision, and the study had no control group. It shows that a clinic can run a long fast carefully, not that a 72-hour fast at home is safe.
Alternate-day fasting and 5:2, the cousins
Alternate-day fasting and 5:2 have the largest trials among those above, and in the head-to-head trials neither beat daily calorie restriction.
In the 12-month Trepanowski trial of 100 adults with obesity (Trepanowski et al., JAMA Internal Medicine, 2017), «fast days» allowed 25% of energy needs and «feast days» 125%. Alternate-day fasting lost 6.0% of body weight against 5.3% for daily restriction (both relative to the control group), which the authors called not superior, and the fasting group had the highest dropout, 13 of 34. A strict version of alternate-day fasting (the authors’ term) ran for 4 weeks in healthy, non-obese adults (Stekovic et al., Cell Metabolism, 2019); it cut calories by 37% on average, and the authors reported no adverse effects, even in participants who had followed it for more than six months.
5:2 has at least three randomised trials of six months or longer, and each found it about as effective as daily restriction, not better: 150 adults over 50 weeks (Schübel et al., 2018), 107 overweight or obese premenopausal women over 6 months (Harvie et al., International Journal of Obesity, 2011) and 137 adults with type 2 diabetes over 12 months, measured by HbA1c (Carter et al., JAMA Network Open, 2018).
The risks of long fasts
Side effects depend on the length of the fast. Harvard Health (4 intermittent fasting side effects to watch out for, undated, read 2026-11-07) lists headaches, lethargy, crankiness and constipation. The ones below can do real harm.
Dizziness and dehydration
Feeling dizzy or lightheaded is on the NHS list of dehydration signs, with thirst, tiredness, headache and dark, strong-smelling urine (NHS, Dehydration, reviewed May 2026). Those are easy to mistake for the fast «working». If you are dizzy, sit down and drink; if it does not pass, eat. A fast is not worth a fall.
Electrolytes and medicines
Blood sodium and potassium can shift during a fast, and people who take medications for blood pressure or heart disease may be more prone to imbalances (Harvard Health). Even in the supervised clinic cohort, the authors report six cases of mild hyponatraemia (low blood sodium). At home nobody measures these, which is the plain reason to keep home fasts short.
Glucose-lowering drugs are the clearest case. In a 12-week trial of 41 adults with type 2 diabetes, two low-calorie days a week roughly doubled the rate of hypoglycaemia despite reduced doses (Corley et al., Diabetic Medicine, 2018). Who should not do intermittent fasting has the rules from an NHS hospital trust’s leaflet on fasting with diabetes.
Refeeding after a multi-day fast
Refeeding syndrome is a shift in blood minerals when calories come back after a period of very low intake, and it is the risk specific to multi-day fasts. The ASPEN consensus (da Silva et al., Nutrition in Clinical Practice, 2020) defines it as a fall in blood phosphorus, potassium or magnesium within five days of calories being reintroduced. The Cleveland Clinic (Refeeding syndrome, reviewed June 2022) puts food deprivation for more than seven days with evidence of stress and depletion among the risk factors, and a 2025 review (Borriello et al., Nutrients, 2025) adds a low BMI and unintentional weight loss.
A 36 or 48-hour fast at home is well short of seven days, and none of the sources above says it carries this risk. None of them tests the first meal after a long fast either, so, as a rule of thumb, break a long fast with a normal meal rather than a feast, and do not run long fasts back to back.
Medical supervision for multi-day fasts
The Cleveland Clinic advises doing prolonged fasting under medical supervision, and when it treats refeeding syndrome, the care team measures electrolyte levels daily. Neither it nor the clinic cohort gives a threshold in hours; a 48-hour line is a rule of thumb, not theirs. An app can time 72 hours; it cannot measure your sodium.
Who should not do a long fast
Anyone who should not do intermittent fasting at all should not try a long fast: pregnancy or breastfeeding, children and teenagers, an eating disorder, diabetes, medicine that must be taken with food, being underweight or malnourished. The full lists, with their sources, are in who should not do intermittent fasting. If you take medicine for blood pressure or heart disease, that is not a ban, but ask your doctor first. If you have a long-term condition, ask.
How to choose
Pick by what you will keep, then by hours. In order:
- Start from the gap you already have: last food at 8 pm and breakfast at 8 am is 12:12, and for many people a 13 or 14-hour fast is one skipped evening snack away.
- Move in one or two-hour steps and hold each for two weeks before the next.
- Lengthen only when the current plan is boring, not when it is hard. Hunger that does not fade after two weeks is a reason to go back a step.
- Judge a plan at 12 weeks, the length of the Lowe trial, not at day three.
- Try a 24-hour fast only once a daily plan feels ordinary, and plan the day: fluids, a normal meal to end it, nothing that needs your full attention. Treat anything past 48 hours as a medical decision, not an app setting.
- If you take any regular medicine or have a condition, ask before you start.
In Bare Fast, the long fasts and a custom length from 12 hours to 7 days need Pro, and before your first fast longer than a day, and before every fast longer than three days, the app shows safety notes.
Ending early is fine
A fast that ends at 14 hours instead of 16 is a 14-hour fast, not a failed one. People leave plans that do not fit them; in the Trepanowski trial above, the alternate-day group had the highest dropout. What you keep over months is the whole point, and a shorter day now and then is part of keeping it. Bare Fast records that day as «of 16 h», never «late» or «failed».
Stop early without a second thought if you feel dizzy, lightheaded or nauseous, or if you are unwell. If you have diabetes, stop and treat a glucose reading under 4 mmol/L, as the Hull NHS leaflet on fasting with diabetes advises.
Common questions
Is 12 hours of fasting enough?
Enough to count as a fast, yes. No trial we found has tested 12:12 on its own, so what it does on its own is unknown. For someone who eats across 14 or 15 hours a day, as half the adults in the 2015 phone-app study did, 12:12 is a real change.
Is a 24-hour fast better than 16:8?
None of the trials above compared them directly. The one 24-hour trial (103 adults with raised LDL, 26 weeks) found no change in LDL and a better insulin-resistance score; the main 16:8 trial (116 adults, 12 weeks) found no advantage over three meals a day. Pick the one you would still be doing in three months.
Is a 36-hour fast safe to do every week?
No trial we found has tested a weekly 36-hour fast, so the honest answer is unknown. The nearest pattern, strict alternate-day fasting, ran for four weeks in healthy, non-obese adults with no adverse effects reported. That is a safety signal from healthy adults, not a test of weekly 36-hour fasts, and anyone on glucose-lowering drugs should ask first.
What counts as a long fast?
None of the sources above sets a cut-off. Daily plans run up to 23:1; from 24 hours the fast is a whole day, and that is where the Cleveland Clinic says side effects can be extreme. Past 48 hours, the question is less «how long» than «who is watching».
This post is not medical advice. Talk to your doctor before you change how you eat, especially if you take medication or have a medical condition.