Skip to content
Extended24 studies in people cited

Supervised Multi-Day Fasting

7 days without food · One episode of a week or longer, residential, with a refeeding period built in

Written by Reviewed Sep 2026

Also called: Therapeutic fasting, Water-only fasting, Buchinger fasting, 7 day fast

A week or more without food, in a clinic. The two biggest groups studied report no deaths across 1,422 people and 768 visits. They disagree with each other, because one is water only and the other gives 200 to 250 calories a day. In the older case reports, deaths cluster in the refeeding period, after the fast ends.

Overview

Past about a week, fasting stops being something people do on their own and becomes something clinics run. The two biggest published safety records both come from live-in clinics, and neither reports a death. One is a review of 768 supervised water-only fasting visits by 652 patients, with a typical fast of 7 days. The other follows 1,422 people fasting 4 to 21 days at a specialist clinic.

Those two records are not measuring the same thing, and that matters more than almost anything else on this page. The 1,422-person group is not water fasting. It gives 200 to 250 calories a day, including fruit juice and honey, which adds up to 25 to 35 g of carbs, and people are told to drink 3 L of water. The water-only group gets nothing.

The water-only group had severe (grade 3) problems at 212 of 768 visits, and one grade 4 problem, a step above severe: dangerously low blood salt on fasting day 9. The low-calorie group had side effects in 0.7% of people, with nothing at grade 4 or 5. A small daily dose of salt and carbs is the design difference between them.

The most important thing to take from this section is not about the fast itself. In the older case reports, deaths cluster in the refeeding period. And blood tests do not show the shortfall that makes refeeding dangerous.

How it works

You live at the clinic, see clinical staff every day, and plan the refeeding period before the fast starts.

The water-only clinic with published safety data sets the refeeding period at no less than half the length of the fast. In practice, a later group from the same clinic paired a typical 11-day fast with a typical 5-day refeed. The refeed is low-fat, low-salt, whole plant food.

The low-calorie clinic runs things differently, and it has published the details. A 600-calorie vegetarian day comes before the fast. The gut is emptied with a laxative: 20 to 40 g of sodium sulphate in 500 mL of water. During the fast, people drink 3 L a day of water or herbal tea with no calories, with optional 20 g of honey, 250 mL of fresh juice at noon, and 250 mL of vegetable soup in the evening. That adds up to 200 to 250 calories and 25 to 35 g of carbs. There is an enema or mild laxative every second day, and a nurse and doctor visit every day. Food comes back over 4 days on average, on a vegetarian diet with eggs and dairy that goes from 800 to 1,600 calories a day. Blood is drawn at the start and at the end.

That clinic turns people away outright if they have severe wasting or anorexia, advanced kidney, liver or brain blood-flow problems, dementia or other severe memory and thinking problems, or are pregnant or breastfeeding. That is the clearest published list of who a large supervised fasting program will not take.

Here is exactly what supervision buys you. Your drugs get adjusted before the fast. Your potassium, phosphate and magnesium get checked at the start if you have a low BMI or have lost weight lately. You get thiamine (vitamin B1) before carbs when you start eating again. If you take an SGLT2 inhibitor or have type 1 diabetes, you get blood ketone tests, not just blood sugar tests. You see clinical staff daily. You get screened for eating disorders. And you are refed step by step with daily salt and mineral checks.

What supervision does not buy: a safe long fast on lithium while cutting fluids, protection from gallstones, or protection from ketoacidosis if you keep taking an SGLT2 inhibitor.

What 7 days actually reaches

In 12 people on a 7-day water-only fast, researchers tracked about 3,000 blood proteins every day. They lost 5.7 kg on average. Body-wide protein changes only showed up after 3 days with no calories. More than 1,000 proteins changed, in nine different patterns, many of them proteins from the tissue that holds cells together. The paper's estimates of health effects for 212 proteins, across about 500 outcomes, are predictions from genetic data, not measured health results. Blood sugar keeps falling, and the fall is silent. Glucose monitors on 12 healthy adults through a 7-day water fast showed time spent below 70 mg/dL rising from 3.0% to 66.0% by day 5. The daily low fell from 76 to 50 mg/dL. No one reported symptoms of low blood sugar. Blood salt does worse on a true water fast than in the low-calorie group. In 13 men on a 10-day water fast, blood salt fell 2.88% from day 6 and dropped below the normal floor of 137 mmol/L on day 9. It recovered by day 5 of eating again, and chloride followed the same path. The authors concluded that vitamins were not needed, but a salt supplement should be considered. In the 1,422-person low-calorie group, blood salt went from 140.1 to 138.7 mmol/L and stayed in the normal range. Uric acid roughly doubles by day 7: from 5.9 to 12.5 mg/dL in 15 patients with obesity. The kidneys' clearance of uric acid fell from 5.5 to 1.8 mL/min. Then uric acid fell back to 7.7 mg/dL by day 28, even though ketosis kept going. The rise tracks how deep into ketosis you are, not how dehydrated you are. Inflammation goes the opposite way from the usual story. In the 1,422-person group, CRP rose from 2.85 to 4.30 mg/L. That has been repeated. A study of 20 people fasting 9.8 days on average found higher CRP and other inflammation markers (hepcidin, midkine and IL-8), more active clotting cells, and more activity in the immune and clotting systems. Its look back at the same large group confirmed the CRP rise. A long fast does not lower inflammation while it is happening. On autophagy (the cell's clean-up process), the one measurement in people of how fast clean-up runs at this length found nothing. In a study that ran human and mouse blood side by side, there was no clean-up boost in human white blood cells inside the body after up to 4 days of fasting. It only showed up when the cells were grown in a dish with a drug (leupeptin), and among human cell types only in one kind (neutrophils). So this page does not claim a multi-day fast reaches autophagy. It does not claim mitophagy (clean-up of the cell's power plants) at any length, because no study in people has measured it on any fast.

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

    Still digesting

    Burning
    Sugar from the meal you just ate. Insulin is high enough to keep fat locked away.
    • Insulin is at its after-meal peak. While it is up there, the enzyme that releases fat from fat cells stays switched off, so fat is being stored, not released.
    • The hour count on most fasting charts starts at the last bite. But the body is still digesting a mixed meal here, so the clock and what the body is doing are not yet the same thing.

    Read the biologyInsulin and insulin sensitivity during a fast

    Measured in people
    The hour-by-hour fasting studies in people start later than this. Klein 1993 took its first sample at 12 hours, in 6 healthy men. Rothman 1991 gives its first figure for new sugar making as an average across the first 22 hours. This row is the fed state those measurements are compared against. No measurement from a long fast falls inside it.

  2. 4h to 12h

    Meal digested, overnight fast

    Burning
    A mix of leftover sugar and fat. The liver keeps the brain supplied with sugar.
    Blood sugar
    5.58 mmol/L (100.5 mg/dL) at the 12 hour mark, in 6 healthy men
    Ketones
    Total ketones around 0.20 mmol/L after an overnight fast. That figure is acetoacetate and BHB added together, not BHB on its own.
    Insulin
    64.6 pmol/L (9.3 microunits/mL) at 12 hours, in the same 6 men
    • At 12 hours, glycerol is released at 2.08 and palmitic acid at 1.63 micromol/kg/min. These are two markers of fat being released. They are the starting numbers everything later in the fast is compared against.
    • Making new sugar from scratch already supplies most of the body's sugar at this point. It is not a backup waiting to be called on. This is the first place the popular "switch" story breaks down.

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

    Measured in people
    Klein 1993 measured blood sugar, insulin and fat release with tracers at 12 hours, in 6 healthy men. Féry 1983 measured total ketones in people after an overnight fast and put the starting level at 0.20 mmol/L. Rothman 1991 measured liver glycogen again and again with a special MRI scan through a 68 hour fast in healthy adults. It worked out that new sugar making supplied 64% of all the body's sugar across the first 22 hours.

  3. 12h to 18h

    Ketones have not moved yet

    Burning
    Fat use rising. More and more of the body's sugar is being made fresh, not pulled from storage.
    Ketones
    A middle change of 0% between 12 and 18 hours, measured every 6 hours in 34 adults
    • BHB, the main ketone, does not move in this window. In 34 adults sampled every 6 hours through the standard 72 hour fast doctors use for testing, the middle change from 12 to 18 hours was zero.
    • Making new sugar supplies 64% of the body's sugar across the first 22 hours. A second, unrelated method put it at about 47% by 14 hours. The body does not switch from stored sugar to fat at a set hour.
    • In 11 overweight adults, activity of the autophagy gene LC3A in blood cells was 22% higher after an 18 hour fast than after a 12 hour fast. In the same people, activity of the MTOR gene also rose, by 9%.

    Read the biologyGlycogen depletion and gluconeogenesisKetogenesis and beta-hydroxybutyrateAutophagy

    Measured in people
    Service 2005 measured BHB every 6 hours in 34 adults whose 72 hour test fasts came back clear for a rare insulin-making tumour. Rothman 1991 measured the share of new sugar making with a special MRI scan. Jamshed 2019 measured gene activity in blood cells in an 11-person, 4 day study where each person tried two eating windows: 8am to 2pm, and 8am to 8pm. That is one gene's activity in blood cells. It is not the real speed of autophagy, and it is not muscle. No study in any species has tested a 16 hour time point for autophagy.

  4. 18h to 1d

    The biggest hours for fat release

    Burning
    Fat and glycerol, with ketone making starting its climb
    Ketones
    Climbing. The middle rise from 18 to 36 hours was +333%, the steepest stretch of a 72 hour fast.
    Insulin
    70% of the whole drop seen across a 72 hour fast has already happened by 24 hours
    • The single biggest jump in fat release across a whole 72 hour fast comes between 18 and 24 hours. 60% of the total rise in fat release happens between 12 and 24 hours. The authors put that down to falling insulin, not to any change in blood sugar.
    • In a study of 30 healthy adults, growth hormone was clearly higher after a 24 hour water-only fast than after a normal eating day. Each person did both. The size of that rise is not in the published summary, so it is not given here.
    • In the same study, haemoglobin, red blood cell count and haematocrit all rose together at 24 hours. That pattern means the fluid part of the blood shrank. It does not mean new red blood cells were made.
    • In 20 volunteers, energy burned over 24 hours, measured in a sealed room, fell during a 24 hour fast. How much it fell varied from person to person.

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

    Measured in people
    Horne 2013 had 30 seemingly healthy adults do both a 24 hour water-only fast and a normal eating day, in random order. It measured growth hormone, blood counts, cholesterol, triglycerides, bicarbonate and weight. Klein 1993 measured fat release with tracers. Hollstein 2020 measured energy burned over 24 hours in a sealed room in 20 volunteers. The exact percentage drop is not in the published summary, so it is not given here. On autophagy, a 24 hour fast raised p62 (also called SQSTM1) in the thigh muscle of 50 women. That points toward less recycling in muscle, not more.

  5. 1d to 36h

    Thyroid drops before metabolic rate does

    Burning
    Fat. The remaining sugar is built from glycerol, lactate and protein building blocks.
    Ketones
    1.308 mmol/L on average after day 1, in 13 men starting a 10 day water fast, up from 0.177 at the start. People varied a lot: the spread around that average (1.053) was nearly as big as the average itself.
    • Making new sugar supplies 82% of the body's sugar between 22 and 36 hours.
    • In 8 healthy men, the thyroid hormone T3 fell from 2.30 to 1.84 nmol/L by 30 hours. TSH, the hormone that tells the thyroid to work, fell below 1 mU/L, and its usual night-time peak disappeared. Blood cortisol did not change.
    • So the thyroid slows down within the first 30 hours, well before anyone can measure a drop in resting metabolic rate.
    • Salt loss has not turned yet. In 9 women with obesity who kept eating their usual amount of salt while fasting, salt in the urine started to rise after about 48 hours and then went above what they took in.

    Read the biologyGlycogen depletion and gluconeogenesisKetogenesis and beta-hydroxybutyrateAutophagy

    Measured in people
    Rothman 1991 measured the share of new sugar making with a special MRI scan in healthy adults. Hugues 1984 fasted 8 healthy men for 30 hours and tested their thyroid hormones again and again. Dai 2022 measured blood BHB daily in 13 men fasting in a supervised facility. Sigler 1975 collected urine in 3 hour blocks from 9 women with obesity on a hospital ward. On autophagy, the only study in people with several time points in one fast took thigh muscle samples at 2, 12, 24 and 36 hours. The markers LC3I, LC3II and p62 all fell, but only in untrained people. The authors sum it up as muscle recycling being only mildly affected by 36 hours of fasting.

  6. 36h to 2d

    Growth hormone jumps, IGF-1 does not follow

    Burning
    Fat, with ketones now carrying a real share of the load. Nearly all the body's sugar is now made fresh.
    Ketones
    A middle rise of +210% between 36 and 54 hours, in 34 adults
    • In 9 healthy men, daily growth hormone output rose from 78 to 371 micrograms per litre on the second day of a fast. That is a five-fold rise. The number of bursts went from 14 to 32 a day, and the amount per burst went from 6.3 to 11 micrograms.
    • How long growth hormone lasted in the blood did not change, at 18 against 20 minutes. So the body was truly making more, not just clearing it more slowly.
    • Blood IGF-1 had not changed at 56 hours. It does not fall until around day 5, when it dropped from 1.31 to 0.77 U/mL in 6 men.
    • From 36 hours on, making new sugar supplies 96% of the body's sugar.
    • At 48 hours, in a clamp test where insulin is dripped in, the body's sugar use fell from 39.8 to 24.1 micromol/kg/min. Sugar burned for energy crashed from 21.8 to 3.9. Sugar stored did not change. That is a change in which fuel the body burns. It is not the kind of sugar problem seen in diabetes.
    • Energy burned over 24 hours, measured in a sealed room, was 10.9% lower on day 2 of a fast than on a day of normal eating (9.8 against 11.0 MJ).

    Read the biologyThe growth hormone responseGlycogen depletion and gluconeogenesisInsulin and insulin sensitivity during a fast

    Measured in people
    Hartman 1992 took blood every 5 minutes for 24 hours in 9 healthy men, on a normal day and on the second day of a fast, and worked out the true output with maths. IGF-1 was measured at 56 hours. Ho 1988 measured IGF-1 (as somatomedin C) through a 5 day fast in 6 men. Mansell 1990 did clamp tests on 6 men without diabetes at 48 hours without food. Andriessen 2023 put 12 healthy lean men in a sealed room for 60 hours straight, once fasting and once eating. No study in people has measured autophagy in any tissue between 36 and 72 hours. That is why autophagy is not linked from this row.

  7. 2d to 3d

    Ketones reach the brain

    Burning
    Fat and ketones. The brain now gets a measurable share of its fuel from BHB.
    Blood sugar
    4.14 mmol/L (74.6 mg/dL) at 72 hours in 6 healthy men, about 25% below the 12 hour value
    Ketones
    74% of people whose test fast came back clear went above 2.7 mmol/L before the 72 hour mark. The middle rise from 54 to 72 hours was +167%.
    Insulin
    30.1 pmol/L (4.3 microunits/mL) at 72 hours, about half the 12 hour value
    • Brain BHB was measured directly with a strong MRI scanner aimed at the back of the brain. It rose from 0.05 mmol/L before fasting to 0.60 after the second day and 0.98 after the third. Brain lactate rose from 0.69 to 1.47 mmol/L.
    • Blood and brain BHB rose closely together. The brain level rose about 0.26 times as much as the blood level did.
    • Over a 48 hour fast, fat built up in the livers of men and in the muscles of women. Sugar handling and fuel burning, in the whole body and the liver, were the same in both sexes.
    • The two ways of measuring energy burned split apart here, and both are real. Resting energy burned, measured with a hood, rose 14.1% by day 3 in 11 lean people. Energy burned over the whole day, measured in a sealed room, was still 5.5% below a normal eating day on day 3 (10.3 against 10.9 MJ). The resting part can rise while the whole day falls.
    • The 72 hour supervised fast is a routine hospital test that doctors have used for decades to check hormone problems. That is the strongest evidence there is that healthy people handle this length of fast when they are being watched.

    Read the biologyKetogenesis and beta-hydroxybutyrateInsulin and insulin sensitivity during a fast

    Measured in people
    Klein 1993 sampled 6 healthy men at 12, 18, 24, 30, 42, 54 and 72 hours, using tracers and breath gas tests. Service 2005 measured BHB every 6 hours in 34 adults. Pan 2000 measured brain BHB and lactate with MRI scans in healthy adults. Browning 2012 measured liver and muscle fat with scans across 48 hours in healthy men and women. Zauner 2000 measured resting energy burned with a hood in 11 lean people. Andriessen 2023 measured energy burned over 24 hours in a sealed room in 12 lean men. The two are measuring different things. They do not contradict each other.

  8. 3d onward

    Adapted, and the risk moves to eating again

    Burning
    Fat and ketones. As the fast gets longer, sugar making is split between the liver and the kidneys.
    Blood sugar
    3.5 mmol/L (63.1 mg/dL) on day 4 in 11 lean adults, and 3.2 mmol/L (57.7 mg/dL) on day 5 in 6 men
    Ketones
    Ketone making starts to level off after about 5 days. Across a 21 day fast in 13 volunteers, blood ketones went from 0.1 to 6.61 mmol/L.
    Insulin
    Levels off by day 3 in people with obesity fasting for weeks. Glucagon, which raises blood sugar, about doubles and peaks on the same day.
    • In 8 healthy men at 72 hours, muscle samples showed mTOR activity down about 50% and the recycling marker LC3B-II up about 30%. p62 rose about 10% in the same samples. The authors say that makes the real speed of recycling impossible to read from these markers alone.
    • 12 healthy adults wore sugar monitors through a 7 day water-only fast. By day 5, their time spent below 70 mg/dL rose from 3% to 66%, and their lowest daily sugar fell from 76 to 50 mg/dL. None of them reported symptoms of low blood sugar. So checking how you feel stops working as a safety check at this point.
    • In 15 patients with obesity, blood uric acid about doubled by day 7, from 5.9 to 12.5 mg/dL. It then fell back to 7.7 by day 28, while ketosis carried on. The kidneys cleared uric acid less well, more so than they cleared other waste. So this is a specific change in how the kidneys handle uric acid.
    • In 13 volunteers, resting energy burned fell 20.3% across a 21 day complete fast. This is where the slowdown in metabolism finally shows up.
    • Nitrogen in the urine, a sign of protein loss, falls steadily for about 4 weeks and then holds at about 3 to 6 g a day. How much protein is lost per kilogram of weight lost depends on body fat. It is roughly 20 g of nitrogen per kg in people without obesity, and roughly 10 g per kg in people carrying 50 kg or more of fat.
    • The body runs hundreds of milligrams a day short on minerals, while blood phosphorus and magnesium stay the same. That gap is why the deaths in the medical case reports cluster in the period of eating again, not during the fast.

    Read the biologyKetogenesis and beta-hydroxybutyrateAutophagyRefeeding, the leucine threshold and refeeding syndrome

    Measured in people
    Vendelbo 2014 took muscle samples from 8 healthy men after 72 hours, with tracer tests in the forearm. Kolnes 2026 monitored 12 healthy adults around the clock through a 7 day water-only fast. Fox 1976 measured blood uric acid and how well the kidneys cleared it through 28 days in 15 patients with obesity. Dai 2024 measured resting energy burned and blood ketones across a 21 day fast in 13 volunteers. Owen 1969 measured nitrogen in the urine in 11 people with obesity through 5 to 6 weeks. Forbes 1979 worked out how nitrogen loss scales with body fat. Licata 1981 measured mineral balance directly over 40 days on a hospital ward. On average, people ran 104 mg of calcium, 48 mg of magnesium and 363 mg of phosphorus a day short, while blood phosphorus and magnesium did not change at all.

What studies in people found

There are two large supervised safety records that disagree because they are different kinds of fasts. There is a 7-day blood protein study, two 10-day body make-up studies, a 21-day body study, glucose monitoring through a 7-day fast, and the older case reports that show when the deaths happened.

  • Wilhelmi de Toledo 2019, 1,422 people fasting 4 to 21 days on 200 to 250 calories a day, 8.2 days on average, with no comparison group. The top blood pressure number fell from 131.6 to 120.7 mmHg and the bottom number from 83.7 to 77.9. Blood sugar settled at about 4.7 mmol/L. Uric acid rose from 338.1 to 495.2 micromol/L. Blood salt went from 140.1 to 138.7 mmol/L. CRP, an inflammation marker, rose from 2.85 to 4.30 mg/L. 0.7% of people had side effects. There were no deaths and no lasting harm, and only two people had to stop the fast. 93.2% said they had no hunger.
  • Finnell 2018, a review of supervised water-only fasting at a live-in clinic: 768 visits by 652 patients, 6,265 fasting days in all. The typical fast was 7 days, the shortest 2 and the longest 41. Side effects were graded on a standard scale. The worst problem per visit was mild or moderate (grade 2 or lower) in 555 visits, severe (grade 3) in 212, grade 4 in 1, and fatal in none. Of 5,961 side effects, 75.3% were mild (grade 1). The two serious events were dehydration on fasting day 3 in a 73-year-old man, and dangerously low blood salt on fasting day 9 in a 70-year-old man. Both fully recovered. Longer fasts went with worse side effects.
  • Goldhamer 2001, 174 inpatients with high blood pressure, taken one after another. They ate fruit and vegetables for about 2 to 3 days, did supervised water-only fasting for about 10 to 11 days, then refed for about 6 to 7 days on a low-fat, low-salt vegan diet. Nearly 90% got below 140/90 mmHg. Blood pressure fell 37/13 mmHg on average, and 60/17 mmHg in the most severe high blood pressure. Everyone who came in on blood pressure pills, 6.3% of the group, came off them. There was no comparison group. And the fast sat between two diets that lower blood pressure on their own, so the effect of the fast alone cannot be pulled out.
  • Scharf 2022, worth knowing for its drop-out rate: 48 people started and 26 finished, so 46% dropped out. The typical fast was 17 days, with an 8-day refeed. At the end of the fast, the top blood pressure number, belly size, LDL cholesterol and CRP were down, while blood fats and insulin resistance were up.
  • Laurens 2021, 16 men, age 44 on average, with a BMI of 26.2 on average. They did a 10-day low-calorie fast with up to 3 hours a day of light activity. Weight fell 7% (5.9 kg) and resting energy burn fell 12%. Fat fell 2.3 kg, about 40% of the weight lost. Scan lean mass fell 3.53 kg, about 60%. Of that lean loss, 44% was water outside the cells, 14% was stored sugar with its water, and 42% was working lean tissue. That came to 1.5 kg, or 25% of total weight lost. Protein burning fell 41% by day 5 and then held steady. Daily steps rose 60%, and strength in weight-bearing muscles rose 33%.
  • Dai 2022, 13 men on a 10-day water-only fast in a controlled facility, scanned at the start, on fasting day 6, and on day 5 of eating again. Weight fell 7.28 kg (9.8%). At day 6, fat was down 10.7% and lean mass down 9.2%. By day 5 of eating again, fat was down 17.2% and lean mass was back to normal. Blood sugar hit its low of 3.508 mmol/L on day 4. Blood salt fell below 137 mmol/L on day 9.
  • Dai 2024, 13 people on a 34-day plan: 3 days of normal eating, 21 days of complete fasting, 5 days of light eating and 5 days of full recovery. Body weight fell 14.96%. Blood ketones rose from 0.1 to 6.61 mmol/L. Blood uric acid rose from 385.38 to 866.31 micromol/L. Resting energy burn fell 20.3%. Blood counts and liver tests mostly stayed steady and in the normal range. The heart showed signs of adapting in how it worked, with no damage to its structure.
  • Pietzner 2024, 12 people on a 7-day water-only fast, with about 3,000 blood proteins measured before, every day during, and after. They lost 5.7 kg on average. Body-wide protein changes only showed up after 3 days. More than 1,000 proteins changed, in nine patterns, many of them from the tissue that holds cells together. The health effect estimates for 212 proteins are predictions from genetic data, not measured results.

What this does not tell you: Neither large safety record assigned people to groups, and neither has a comparison group. The 1,422-person group comes from one specialist clinic. It is a 200 to 250 calorie fast, not a water fast. It measured ketones with a urine strip that looks for a different ketone than the blood test does. It never measured blood phosphate, and it did not measure body make-up at all. Its published record also does not add up in places. Low blood salt shows up as 3 cases in the side effects table and 6 in the discussion. The 5-day group is listed as 695 people in the methods and 659 in a table. The water-only review prints its serious event rate as 0.002%, but 2 of 768 visits is 0.26%. That looks like a typo, so the raw counts are used here. The body make-up numbers come from studies of 13 and 16 men. In one long water-only study, 46% dropped out.

Reading the research record

The single most useful comparison here is between the two large groups. Reading either one as evidence about the other is the most common mistake in this research.

The 1,422-person group gets 200 to 250 calories a day, including 25 to 35 g of carbs from fruit juice and honey, and is told to drink 3 L of water. That is why its blood sugar settles around 4.7 mmol/L, while day 5 of a true water fast reaches 3.2 mmol/L. It is also a likely part of why its blood salt stayed in range, while a 10-day water fast dropped below the normal floor on day 9.

Its ketone data come from a rough urine strip that reacts to one ketone (acetoacetic acid), which people tested themselves in their first morning urine. No blood ketone (beta-hydroxybutyrate) was measured in that group at all. So anyone citing it for blood ketone levels has misread it. It also never measured blood phosphate, the mineral that matters most when you start eating again, and it measured no body make-up.

Two more things in that paper go against how it is usually summed up. CRP, an inflammation marker, went up, not down, from 2.85 to 4.30 mg/L, and that has been repeated in a separate water-only study. And the paper does not add up in places: it gives low blood salt as 3 cases in its side effects table and 6 in its discussion, and its 5-day group as 695 people in the methods and 659 in a table. Those are noted here, not resolved.

One more correction belongs here, about resting energy burn, because this is the length where it truly falls. The popular claim that fasting does not slow metabolism rests on resting energy burn, measured under a hood, rising 14.1% by day 3 in 11 lean people. But total energy burned across the day, measured in a sealed room, fell 10.9% on day 2 and 5.5% on day 3. Across a 21-day fast, resting energy burn fell 20.3%. And in a separate group, resting energy burn was down 12% by day 10. The early rise is real, and it does not last the week.

What happens to muscle

This is the length where body make-up has actually been measured. The two studies that measured it best agree on what matters: most of the early "lean" loss is not real tissue. Lean mass means the non-fat part of you, mostly muscle and water.

In 16 men on a 10-day low-calorie fast, scan lean mass fell 3.53 kg, about 60% of total weight lost. That is the number usually quoted as proof that long fasts eat muscle. But the researchers broke it down, using 10 days of urine collections, body water and stored sugar. 44% was water outside the cells, 14% was stored sugar in muscle and liver with its water, and 42% was working lean tissue. That came to 1.5 kg, or 25% of total weight lost.

Function did not follow the scan. Daily steps rose 60%. Strength held steady in muscles that do not carry body weight, and rose 33% in muscles that do. At 3 months, body weight and scan lean mass were still 2.3% and 3.2% below where they started.

In 13 men on a 10-day water fast, scanned three times, lean mass was down 9.2% at fasting day 6 and back to normal by day 5 of eating again. Fat ended 17.2% down. That is the strongest evidence we have that most early lean loss on a fast is water and stored sugar, not muscle protein. In 20 people fasting 9.8 days on average, blood protein tests pointed to muscle and bone being kept.

And the cleanest strength test of all, because it scanned and strength-tested the same people. 13 healthy young adults, 6 women and 7 men, finished a seven-day water-only fast. Scan lean mass fell 4.6 kg, an 8.0% drop, with 1.7 kg of it from the arms and legs. Fat fell 1.4 kg. Stored sugar in muscle halved. Top strength did not change. Top aerobic fitness fell 13% and stamina dropped, so the real cost falls on stamina, not strength. An energy sensor in muscle (AMPK) did not change, which the authors say goes against the idea that it drives muscle breakdown. They say plainly that nobody knows whether the protein broken down includes the muscle's working parts.

The protein picture underneath: protein burned (measured as nitrogen in the urine) dropped 41% by day 5 and then held steady through day 10. Urea in the blood fell 41% alongside it. A marker of muscle breakdown (3-methylhistidine) rose until day 5, then fell. At 3 weeks, 10 women with moderate obesity were losing 5.9 g of nitrogen a day. Their overall protein turnover and that muscle breakdown marker were both down 31% from their normal-diet levels. By 5 to 6 weeks, total nitrogen in the urine was 4.66 g a day, with ammonia at 1.95 and urea at 1.55. That is the reverse of the normal pattern when eating, and it is the sign of full adaptation.

You will often see figures credited to a 1970 paper by Cahill: roughly 12 g of nitrogen a day early on, falling to 3 to 4 g later. We could not confirm them against that paper, so they are not used here.

Who this applies to is the warning that changes everything. Protein lost per kg of weight lost is lower the more body fat you carry. It is roughly 20 g of nitrogen per kg in people without obesity, and roughly 10 g per kg in people carrying 50 kg or more of fat. So a lean person loses about twice as much protein per kg lost. And nearly all the classic research on long starvation was done in people with obesity.

Doing it

Water-only and modified fasting are different interventions
200 to 250 calories a day with 25 to 35 g of carbs gives different blood sugar, different ketone levels and different salt handling from a water fast. The low-calorie group kept blood salt in range. A 10-day water fast dropped below the normal floor on day 9, and its authors concluded a salt supplement should be considered.
Phosphate is the measurement nobody took
The 1,422-person group measured only sodium, potassium, calcium and magnesium. Phosphate is the mineral that matters most when you start eating again, and the biggest fasting group in the research never measured it. Every published refeeding plan says to check it before calories come back, and again and again afterwards.
The blood test will not show the deficit
Over 40 days on a hospital ward, people lost 104 mg of calcium, 48 mg of magnesium and 363 mg of phosphorus a day on average, while blood phosphorus and magnesium did not change at all. Losses in the urine were 58% more than intake for calcium, 75% more for phosphorus and 500% more for magnesium. A normal blood test during a fast does not mean your stores are full.
Thiamine before carbohydrate, not after
Wernicke encephalopathy, a brain emergency caused by lack of thiamine (vitamin B1), has been recorded after a 40-day water fast and on day 53 of a 60-day fast. The second person arrived in a coma and unable to breathe on their own. The body's need for thiamine shoots up the moment carbs come back. The agreed advice is 100 mg before feeding, or before any drip containing sugar, in anyone at risk. Then 100 mg a day for 5 to 7 days, or longer in severe starvation.
Uric acid and gout
Uric acid in the blood roughly doubles by day 7 and peaks somewhere around day 7 to 15 before partly falling back. The rise tracks ketosis, not dehydration, because ketones and uric acid compete for the same exit route in the kidneys. So drinking more does not fix it. In 1,422 supervised people there was one gout attack, in a 72-year-old man already on allopurinol who had frequent attacks before.
Gallstones
On a 500-calorie-a-day diet, the one-year risk of gallstones bad enough to need hospital care was about 1.5%, against 0.44% on 1,200 to 1,500 calories. That is 3.4 times the rate. For every 92 people on the very-low-calorie diet, one extra person needed hospital care. Most of that came from how fast people lost weight. The one treatment tested head to head is a drug called ursodeoxycholic acid. Eating some fat is the other lever, because fat makes the gallbladder squeeze. A water fast has no fat, so that lever is gone, and no study that good exists for water fasting to measure the risk.
What supervision changes, specifically
Drugs adjusted before the fast. Potassium, phosphate and magnesium checked at the start. Thiamine before carbs. Blood ketone tests, not just blood sugar, for anyone on an SGLT2 inhibitor or with type 1 diabetes. Seeing clinical staff every day past about a week. Eating disorder screening. And step-by-step refeeding with daily salt and mineral checks. The serious events in the published groups were dehydration on day 3, and dangerously low blood salt and a heart attack both on day 9. They were caught because someone was watching.

Breaking the fast

This is the dangerous part. It is where a week-long fast is a different kind of thing from a two-day one, not just a longer one. The pattern in the case reports is clear. A death from acid build-up during refeeding after a 3-week fast. A young woman who seemed well, fasted 210 days, and died while eating again. After death, her heart muscle fibres were badly broken up and her liver weighed 600 g. In a hospital series of 18 long fasts, two of three crises involved heavy fluid build-up on refeeding, one needing water pills. In one 382-day fast, the lowest potassium of the whole time, 2.9 mEq/L, came on day 6 of eating again. Magnesium also hit its lowest, around 1.1 mEq/L, in the first few days of eating again. Both of that man's lowest readings came after the fast ended. Here is why blood tests miss it. During a fast, the body runs low on phosphate, potassium and magnesium while blood levels stay normal, because the cells shrink their stores along with the whole body. Over 40 days on a hospital ward, people lost 104 mg of calcium, 48 mg of magnesium and 363 mg of phosphorus a day on average, while blood phosphorus and magnesium stayed the same the whole time. When you eat carbs again, insulin rises. That pushes sugar, phosphate, potassium and magnesium into the cells all at once. Burning that sugar and making energy uses up even more phosphate. Very low phosphate can cause breathing muscle failure, heart failure, muscle breakdown, bursting red blood cells, seizures and confusion. Low magnesium makes low potassium hard to fix. And the carbs use up thiamine (vitamin B1). That is how giving sugar to someone low on thiamine can tip them into Wernicke encephalopathy, a brain emergency. Here are the published numbers for how often this happens. Of 62 intensive care patients fed after 48 hours or more without food, 21 developed low phosphate, below about 2.0 mg/dL. It happened 1.9 days after feeding started, on average. Of 69 teenagers started at 1,200 to 1,400 calories a day, 19 needed extra phosphorus, and 81% hit their lowest phosphorus in the first week. Of 243 adults started on tube or drip feeding, 3 developed the full syndrome, and low magnesium at the start predicted it. And the one study in this area that assigned patients to groups found that cutting calories in people who had already developed low phosphate on refeeding saved lives. At day 60, 149 of 164 were alive against 128 of 163. What the guidelines say to do. For someone at high risk, start at no more than 10 calories per kg of body weight a day. In extreme cases, such as a BMI under 14 or almost no food for more than 15 days, start at 5 calories per kg a day, with constant heart rhythm monitoring. Build up to full needs over 4 to 7 days. Give 100 mg of thiamine before feeding, and 200 to 300 mg a day through the first 10 days for high-risk people, with a B complex and a multivitamin. Add potassium at about 2 to 4 mmol per kg a day, phosphate at 0.3 to 0.6 mmol per kg a day, and magnesium at 0.2 mmol per kg a day by drip or 0.4 by mouth. Check potassium, magnesium and phosphorus before starting, and every 12 hours for the first 3 days in high-risk patients. If they become hard to correct or drop fast, cut the calories or sugar by half, then build back up by about a third of the goal every 1 to 2 days. What the fasting clinics themselves do: the water-only clinic requires a refeeding period of at least half the fast's length. The low-calorie clinic brings food back over 4 days on average, on a vegetarian diet with eggs and dairy that goes from 800 to 1,600 calories a day. Community practice past five days is a ramp measured in days. Broth first, then avocado and fermented food, then eggs, white fish or chicken. Normal portions do not come back until day four or five, and salts and minerals continue through the whole refeed instead of stopping at the first meal. That last part is habit, not evidence. The guideline numbers above are not habit.

Sources

  1. Study in people2025
    Effects of seven days' fasting on physical performance and metabolic adaptation during exercise in humans

    Nature Communications

    13 healthy young adults finished a seven-day water-only fast. They lost 4.6 kg of lean mass (8.0%) and 1.4 kg of fat, and muscle sugar stores halved, yet top strength did not change. Peak fitness fell 13%. An energy sensor called AMPK did not change, which the authors say questions its supposed role in muscle breakdown.

    13 healthy young adults completed a seven day water only fast. DXA lean mass fell 4.6 +/- 0.3 kg (8.0%), fat mass 1.4 +/- 0.1 kg, and muscle glycogen halved, yet maximal isometric and isokinetic strength were unchanged. Peak oxygen uptake fell 13%. AMPK activity was unaffected, which the authors say challenges its proposed role in muscle protein degradation.

  2. Study in people2018
    Is fasting safe? A chart review of adverse events during medically supervised, water-only fasting

    BMC Complementary and Alternative Medicine

    A review of medical charts from 768 stays by 652 patients doing supervised water-only fasting, most often for 7 days. The worst problem was mild to moderate in 555 stays, severe in 212, life-threatening in 1, and nobody died. The 2 serious events were dehydration on day 3 in a 73-year-old man and very low blood sodium on day 9 in a 70-year-old man, and both fully recovered. The paper prints a serious event rate of 0.002%, but 2 of 768 is 0.26%.

    768 visits from 652 patients across 6,265 fasting days of medically supervised water-only fasting, median fast 7 days (IQR 6), shortest 2, longest 41, graded with CTCAE v4.03. Highest-grade adverse event per visit: grade 2 or lower in 555 visits, grade 3 in 212, grade 4 in 1, grade 5 in none. Of 5,961 adverse events, 75.3% were grade 1. The two serious events were a grade 3 dehydration on fasting day 3 in a 73-year-old man and a grade 4 hyponatraemia on fasting day 9 in a 70-year-old man, both fully recovered. The facility requires a refeeding period equal to half the fast length. The paper prints its serious adverse event rate as 0.002% where 2 of 768 is 0.26%, so the raw counts are used here.

  3. Study in people2019
    Safety, health improvement and well-being during a 4 to 21-day fasting period in an observational study including 1422 subjects

    PLoS ONE

    1,422 people fasted 4 to 21 days at one clinic on 200 to 250 calories a day, 8.2 days on average, with no comparison group. Blood pressure fell from 131.6/83.7 to 120.7/77.9 mmHg, while uric acid and an inflammation marker (CRP) rose. Side effects hit 0.7%, with no deaths and two hospital stays, and 93.2% said they felt no hunger. Ketones were checked only with urine strips, and phosphate and body make-up were not measured.

    1,422 subjects fasting 4 to 21 days at a single specialist clinic on 200 to 250 kcal a day including 25 to 35 g of carbohydrate, mean fast 8.2 ± 0.1 days, no control group. Systolic 131.6 ± 0.7 to 120.7 ± 0.4 and diastolic 83.7 ± 0.4 to 77.9 ± 0.3 mmHg; uric acid 338.1 ± 2.3 to 495.2 ± 4.4 micromol/L; sodium 140.1 ± 0.1 to 138.7 ± 0.1 mmol/L; CRP 2.85 ± 0.14 to 4.30 ± 0.20 mg/L (p = 0.001), a rise. Adverse effects in 0.7%, no fatalities, two hospitalisations, only two subjects interrupting the fast, and 93.2% reporting an absence of hunger. Ketones were measured by urine dipstick reacting to acetoacetic acid, not blood beta-hydroxybutyrate; serum phosphate and body composition were not measured at all. The paper reports hyponatraemia as 3 cases in its table and 6 in its discussion, and its 5-day group as 695 in the methods and 659 in a table.

  4. Study in people2024
    Systemic proteome adaptions to 7-day complete caloric restriction in humans

    Nature Metabolism

    12 volunteers, 5 women and 7 men, did a 7-day water-only fast with about 3,000 blood proteins checked daily. They lost 5.7 kg on average. The protein changes showed up only after 3 days, when more than 1,000 proteins shifted, many from the tissue that holds cells together. The health effects the paper links to these proteins are estimates from genetics, not measured outcomes.

    12 volunteers, 5 women and 7 men, through a 7 day water-only fast with plasma proteomics across roughly 3,000 proteins before, daily during and after. Mean weight loss 5.7 ± 0.8 kg. Systemic proteomic changes became evident only after 3 days of complete caloric restriction. More than 1,000 proteins responded significantly across nine distinct response profiles, strongly enriched for extracellular matrix proteins including large changes in the brain-specific protein tenascin-R. The health-consequence estimates for 212 proteins across around 500 outcomes are proteogenomic inference rather than measured clinical outcomes.

  5. Study in people2021
    Is muscle and protein loss relevant in long-term fasting in healthy men? A prospective trial on physiological adaptations

    Journal of Cachexia, Sarcopenia and Muscle

    16 men, average age 44, did a 10-day fast on 200 to 250 calories a day with up to 3 hours of light activity daily. They lost 5.9 kg (7%), and resting calorie burn fell 12%. Only 1.5 kg of the lean tissue lost was working tissue, 25% of total weight lost, and the rest was mostly water and stored sugar. Steps rose 60% and strength 33%, but at 3 months weight and lean tissue were still 2.3% and 3.2% below the start.

    16 men, age 44 ± 14, BMI 26.2 ± 0.9, through a 10 day modified fast at 200 to 250 kcal a day with up to 3 hours a day of low-intensity activity, followed through refeeding and a 3 month follow-up. Weight -7% (-5.9 ± 0.2 kg, p < 0.001), basal metabolic rate -12% (p < 0.01). Fat mass -2.3 ± 0.18 kg and lean soft tissue -3.53 ± 0.13 kg, the latter decomposing into 44% extracellular water, 14% glycogen with bound water and 42% metabolically active lean tissue, which is 1.5 ± 0.1 kg or 25% of total weight loss. Protein oxidation fell 41 ± 7% by day 5 and then held; plasma 3-methylhistidine rose to day 5 then fell. Step count +60% and weight-bearing strength +33% (both p < 0.001). At 3 months, weight and lean soft tissue were still 2.3% and 3.2% below baseline.

  6. Study in people2022
    Effects of 10-Day Complete Fasting on Physiological Homeostasis, Nutrition and Health Markers in Male Adults

    Nutrients

    13 men in a controlled facility did a 10-day water-only fast. They lost 7.28 kg (9.8%), and by day 6 fat was down 10.7% and lean mass 9.2%. By recovery day 5, fat was down 17.2% and lean mass was back where it started. Blood sodium dropped below normal on day 9 and recovered after eating again, and the authors suggest a salt supplement.

    13 male adults in a controlled facility through a 10 day complete water-only fast with DXA at baseline, fasting day 6 and recovery day 5. Weight -7.28 kg (-9.8%). At day 6, fat mass -10.7% and lean mass -9.2%; by recovery day 5, fat mass -17.2% and lean mass back to baseline. Glucose nadir 3.508 ± 0.397 mmol/L on day 4; beta-hydroxybutyrate 0.177 ± 0.044 to 1.308 ± 1.053 mmol/L by day 1, peaking day 5. Serum sodium fell 2.88% from day 6 and dropped below the 137 mmol/L lower limit on day 9, recovering by refeeding day 5. The authors conclude that a sodium chloride supplement should be considered.

  7. Study in people2024
    Analysis of physiological and biochemical changes and metabolic shifts during 21-Day fasting hypometabolism

    Scientific Reports

    13 volunteers did 21 days of complete fasting, then 5 days of light eating and 5 of full recovery. They lost 14.96% of body weight, blood ketones rose from 0.1 to 6.61 mmol/L, and uric acid rose from 385.38 to 866.31 micromol/L. Resting calorie burn fell 20.3%, the clearest measure in people of how metabolism eventually slows. Blood counts and liver tests mostly stayed normal, and the heart adapted without damage.

    13 volunteers through a 34 day protocol: 3 day baseline, 21 days complete fasting, 5 days calorie restriction, 5 days full recovery. Body weight -14.96 ± 1.55%, blood glucose -21.63 ± 0.058%, blood ketones 0.1 ± 0.04 to 6.61 ± 1.25 mmol/L, blood uric acid 385.38 ± 57.78 to 866.31 ± 172.01 micromol/L, resting energy expenditure -20.3 ± 11.13%. Respiratory quotient shifted toward fat. Complete blood count and liver indicators were mostly stable and within normal range, and cardiac function showed functional adaptive changes without structural damage. This is the clearest human measurement of the eventual fall in resting metabolic rate.

  8. Study in people2025
    Prolonged fasting promotes systemic inflammation and platelet activation in humans: A medically supervised, water-only fasting and refeeding study

    Molecular Metabolism

    20 middle-aged volunteers fasted on water for 9.8 days on average, then refed for about 5.3 days, losing 7.7% of their weight. Blood proteins suggested muscle and bone were kept. But fasting also set off inflammation and made platelets, the blood's clotting cells, more active. A look back at the 1,422-person clinic group confirmed the inflammation marker CRP rose from 2.8 to 4.3 mg/L.

    20 middle-aged volunteers, mean fast 9.8 ± 3.1 days plus 5.3 ± 2.4 days of refeeding, 7.7% mean weight loss. Proteomics indicated preservation of skeletal muscle and bone. It also found an acute inflammatory response: elevated CRP, hepcidin, midkine and IL-8, increased platelet degranulation, and upregulated complement and coagulation pathways. A retrospective check of the 1,422-subject cohort confirmed CRP rising 2.8 ± 0.1 to 4.3 ± 0.2 mg/L. Amyloid beta 40 and 42 fell.

  9. Study in people2026
    Marked Increases in Continuous Glucose Monitor-Detected Hypoglycemia During a Seven-Day Water-Only Fast in Healthy Men and Women

    Journal of Diabetes Science and Technology

    12 healthy adults, 7 of them men, average age 29.7, wore glucose monitors through a 7-day water-only fast. Time with blood sugar under 70 mg/dL rose from 3.0% to 66.0% by day 5, and the daily low fell from 76 to 50 mg/dL. Nobody felt any low blood sugar symptoms. So how you feel gives no warning at this length, and none at all for anyone on a drug that lowers blood sugar.

    12 healthy adults, 7 men, mean age 29.7, mean BMI 25.0, wearing continuous glucose monitors through a 7 day water-only fast. Time below 70 mg/dL rose from 3.0% ± 7.1% at baseline to 66.0% ± 25.7% by day 5 (P < .001) and minimum daily glucose fell from 76 ± 14 to 50 ± 7 mg/dL (P < .001). No symptoms of hypoglycaemia were reported, which makes symptom-based self-monitoring unreliable at this duration and removes the warning system entirely for anyone on a glucose-lowering drug.

  10. Study in people1976
    Renal excretion of uric acid during prolonged fasting

    Metabolism

    15 patients with obesity on long medical fasts. Blood uric acid rose from 5.9 mg/dL to 12.5 at day 7, then fell back to 7.7 by days 25 to 28, even though ketones stayed high. The kidneys cleared less uric acid, a specific change in how they handle it, not a general kidney problem.

    15 obese patients on prolonged therapeutic fasting. Serum uric acid 5.9 ± 0.4 mg/dL at day 0, 12.5 ± 1.0 at day 7 and 7.7 ± 1.3 at days 25 to 28, falling back despite sustained ketonaemia. Uric acid clearance 5.5 ± 0.9 to 1.8 ± 0.2 to 4.4 ± 1.5 mL/min against creatinine clearance 114 ± 11 to 80 ± 6 to 64 ± 6.3, so urate clearance fell proportionally more than filtration, making this a specific tubular handling change. There was no glycosuria, bicarbonaturia or increased phosphaturia to suggest general tubular dysfunction.

  11. Study in people1981
    Adverse effects of liquid protein fast on the handling of magnesium, calcium and phosphorus

    American Journal of Medicine

    6 people with obesity spent 40 days on a liquid protein fast on a hospital research ward. They lost more minerals in urine than they took in: 58% more calcium, 75% more phosphorus and 500% more magnesium, while blood phosphorus and magnesium did not change. This was not a total fast, since protein was given. The lesson still holds: normal blood tests during a fast do not mean your stores are fine, which is why eating again is the dangerous part.

    6 obese subjects, 40 days on a metabolic ward. Cumulative urinary losses exceeded cumulative intake by 58% for calcium, 75% for phosphorus and 500% for magnesium, with mean daily balances of -104 mg calcium, -48 mg magnesium and -363 mg phosphorus, while serum phosphorus and magnesium did not change and serum calcium fell 0.5 mg/dL (p < 0.05). Serum bicarbonate fell 20% over the first 8 days. This is a protein-supplemented rather than a total fast, and the principle carries: a normal serum magnesium or phosphate during a fast does not mean the stores are intact, which is why refeeding is the dangerous phase.

  12. Review of studies2020
    ASPEN Consensus Recommendations for Refeeding Syndrome

    Nutrition in Clinical Practice

    Expert guidance from ASPEN, a nutrition society, on refeeding syndrome: a drop in blood phosphorus, potassium or magnesium within 5 days of eating again. It lists long fasting as higher risk. It says to start with 100 to 150 g of sugar or 10 to 20 calories per kg on the first day, build up slowly, check those minerals often, and give thiamine (vitamin B1) 100 mg before feeding and daily for 5 to 7 days. It also says nobody knows how common the syndrome is.

    Defines refeeding syndrome as a fall in serum phosphorus, potassium and/or magnesium of 10% to 20% (mild), 20% to 30% (moderate) or more than 30% or with organ dysfunction (severe), occurring within 5 days of reintroducing calories. Lists prolonged fasting among the higher-risk conditions. Recommends initiating with 100 to 150 g of dextrose or 10 to 20 kcal/kg for the first 24 hours and advancing by a third of goal every 1 to 2 days, checking potassium, magnesium and phosphorus before starting and every 12 hours for the first 3 days in high-risk patients, and thiamine 100 mg before feeding then 100 mg a day for 5 to 7 days or longer in severe starvation. ASPEN states that the incidence of the syndrome is unknown and that quantification of risk is not currently possible.

  13. Study in people2015
    Restricted versus continued standard caloric intake during the management of refeeding syndrome in critically ill adults: a randomised, parallel-group, multicentre, single-blind controlled trial

    Lancet Respiratory Medicine

    339 very sick adults in intensive care developed refeeding syndrome within 72 hours of starting feeding, and got either fewer calories or standard feeding. Days alive after leaving intensive care did not clearly differ, 44.8 against 39.9. But more were alive at day 60 on fewer calories, 91% against 78%. It is the only trial showing that how fast you feed someone matters.

    339 critically ill adults who developed refeeding syndrome within 72 hours of starting nutrition, randomised to caloric restriction or standard feeding. Days alive after ICU discharge were 44.8 against 39.9, not significant (p = 0.19), but more patients were alive at day 60: 149 of 164 (91%) against 128 of 163 (78%), p = 0.002. The only randomised trial establishing that the rate of refeeding matters.

  14. Study in people1996
    Refeeding hypophosphatemia in critically ill patients in an intensive care unit. A prospective study

    Archives of Surgery

    62 intensive care patients were fed after 48 hours or more without food. Blood phosphorus dropped to low levels in 21 of them (34%), and very low in 6. The lowest point came about 1.9 days after feeding started, on average.

    62 ICU patients refed after 48 hours or more of starvation. Phosphorus fell by more than 0.16 mmol/L to below 0.65 mmol/L, about 2.0 mg/dL, in 21 of 62 (34%), and below about 1.0 mg/dL in 6 patients. The nadir came at a mean of 1.9 ± 1.1 days after feeding started.

  15. Study in people2015
    Malnutrition-induced Wernicke's encephalopathy following a water-only fasting diet

    Nutrition in Clinical Practice

    One woman developed Wernicke's encephalopathy, a brain disorder, after 40 days of water-only fasting. She needed thiamine (vitamin B1) through a drip, then 100 to 800 mg a day by mouth for 6 months. The body's need for thiamine jumps when carbs come back, which is why experts say to give it before feeding, not after.

    An adult woman developed Wernicke's encephalopathy after a 40 day water-only fasting diet, requiring parenteral thiamine and then oral thiamine at 100 to 800 mg a day for 6 months. Thiamine demand rises sharply at the moment carbohydrate is reintroduced, which is why the consensus recommendation is to give it before feeding rather than after.

  16. Study in people2015
    A Therapeutic Fast for Lymphoma Resulting in Wernicke Encephalopathy

    Journal of Emergency Medicine

    A 42-year-old man with non-Hodgkin lymphoma tried a 60-day water-only fast to treat it. He reached the emergency room on day 53 in a coma, with breathing failure and Wernicke encephalopathy, a brain disorder.

    A 42-year-old man with non-Hodgkin lymphoma undertook a 60 day therapeutic water-only fast and arrived at the emergency department on day 53 in coma and respiratory failure, with Wernicke encephalopathy.

  17. Study in people2024
    Prolonged Water-Only Fasting Followed by a Whole-Plant-Food Diet Is a Potential Long-Term Management Strategy for Hypertension and Obesity

    Nutrients

    29 people with high blood pressure fasted on water for a typical 11 days (7 to 40), then ate again for a typical 5 days. Everyone stayed in to the end of refeeding, and 93% were still in at 6 weeks. All side effects were mild and short-lived, with nothing serious.

    29 participants with stage 1 to 2 hypertension, median fast 11 days (range 7 to 40) and median refeed 5 days (range 3 to 17). Retention was 100% at the end of refeeding and 93% at 6 week follow-up, with 90% completing at least 7 days. All adverse events were mild grade 1 and transient, with no higher-grade or serious events.

  18. Study in people2022
    The Effects of Prolonged Water-Only Fasting and Refeeding on Markers of Cardiometabolic Risk

    Nutrients

    48 people started a long water-only fast and 26 finished, so 46% dropped out. The typical fast was 17 days, then 8 days of eating again. By the end of the fast, blood pressure, waist size, LDL cholesterol and an inflammation marker were down, but blood fats and insulin resistance were up.

    48 enrolled and 26 completed, a 46% attrition rate, with a median fast of 17 days and an 8 day refeed. At the end of the fast, systolic blood pressure, abdominal circumference, LDL and high-sensitivity CRP were down while triglycerides and HOMA-IR were up.

  19. Study in people2001
    Medically supervised water-only fasting in the treatment of hypertension

    Journal of Manipulative and Physiological Therapeutics

    174 hospital patients with high blood pressure ate fruit and vegetables for about 2 to 3 days, fasted on water for about 10 to 11 days, then ate a low-fat, low-salt vegan diet for about 6 to 7 days. Nearly 90% got below 140/90 mmHg, with an average drop of 37/13 mmHg. Everyone on blood pressure drugs at the start, 6.3% of the group, came off them. There was no comparison group, and the diets around the fast also lower blood pressure, so the fast's own effect is unknown.

    174 consecutive hypertensive inpatients treated with roughly 2 to 3 days of fruit and vegetables, roughly 10 to 11 days of medically supervised water-only fasting, and roughly 6 to 7 days of refeeding on a low-fat, low-sodium vegan diet. Nearly 90% reached below 140/90 mmHg, mean reduction 37/13 mmHg and 60/17 mmHg in stage 3 hypertension, and all subjects on antihypertensive medication at entry, 6.3% of the sample, discontinued it. Uncontrolled case series, and the fast is bracketed by two dietary interventions that independently lower blood pressure, so the fast itself cannot be isolated.

  20. Study in people1970
    Prolonged starvation, a dangerous procedure?

    British Medical Journal

    18 patients with obesity fasted from 62 to 249 days, losing 16.2% to 40.2% of their weight. One middle-aged man got belly pain on day 11 and died on day 13 from a twisted bowel that cut off its blood supply. Two others went into shock from low sodium or leaked sodium and potassium through the kidneys, and both swelled with fluid when they ate again. The authors concluded long fasts are safe only in hospital with full monitoring.

    18 obese patients fasting 60 days or more, range 62 to 249 days, losing 16.2% to 40.2% of initial body weight. One middle-aged man developed colicky umbilical pain on day 11 and died on day 13, with necropsy showing infarction of the small bowel due to an acute volvulus. Two other patients developed hyponatraemic shock or a renal sodium and potassium leak after 55 and 98 uneventful days, and both then developed pronounced fluid retention during refeeding. The authors' own conclusion was that prolonged therapeutic starvation is safe and effective only in hospital with full metabolic monitoring.

  21. Study in people1973
    Features of a successful therapeutic fast of 382 days' duration

    Postgraduate Medical Journal

    One 27-year-old man weighing 207 kg fasted 382 days on calorie-free fluids and daily vitamins, with potassium and salt only at set times, and lost about 125 kg. His blood sugar sat around 30 mg/100 mL from month 4, sometimes lower, yet he felt fine and kept walking about. His lowest potassium, 2.9 mEq/L, and his lowest magnesium came in the first days after he started eating again. Both of his worst mineral readings came after the fast ended.

    A 27-year-old man, admission weight 207 kg, fasted 382 days on non-caloric fluids with daily vitamin supplements, potassium only on days 93 to 162 and sodium chloride only on days 345 to 355, losing about 125 kg. Glucose sat around 30 mg/100 mL from month 4 and occasionally below 20 while he remained symptom-free and ambulant. His lowest potassium of the whole episode, 2.9 mEq/L, occurred on day 6 of refeeding, and magnesium fell further, to about 1.1 mEq/L, in the first few days of refeeding. Both of his lowest electrolyte values came after the fast ended.

  22. Review of studies1982
    Fasting: the history, pathophysiology and complications

    Western Journal of Medicine

    The standard review of problems during medical fasting. Weight loss averaged 0.9 kg a day in week one, slowing to 0.3 kg by week three, and the early loss was mostly sodium, not tissue. Of six deaths reviewed, two were in people with severe heart failure before the fast, and one had a narrowed heart artery. Other problems included dizziness on standing, gout, kidney stones, swelling, heart rhythm trouble, anaemia, lost periods and hair loss, and the results may not fit lean people.

    The standing review of therapeutic fasting complications. Weight loss averaged 0.9 kg/day in the first week, slowing to 0.3 kg/day by the third, with the early rapid loss attributed primarily to negative sodium balance rather than tissue loss. Of the six fatal cases reviewed, two had severe pre-fast congestive heart failure that had initially appeared to improve with fasting and one had focal coronary artery stenosis at autopsy. The catalogue of non-fatal complications includes orthostatic hypotension, acute gout, urate nephrolithiasis, renal insufficiency, oedema, atrial flutter, anaemia, amenorrhoea, alopecia and vitamin deficiency. It also notes that most studies of fasting used obese persons and that the results may not always apply to lean persons.

  23. Study in people1990
    Protein metabolic effects of a prolonged fast and hypocaloric refeeding

    American Journal of Physiology

    10 women with moderate obesity fasted 3 weeks. By the end, the body was losing 5.9 g of nitrogen a day, a sign of protein loss, and protein turnover had slowed a lot, with markers down 31%. A week of eating 80 g a day of protein turned that into a gain of 1.7 g a day. The same amount of carbs only cut the loss to 3.1 g a day.

    10 moderately obese women fasted 3 weeks. Nitrogen balance at the end was -5.9 g of nitrogen a day, with plasma leucine flux and urinary 3-methylhistidine excretion both down 31% from control-diet values (p < 0.01), meaning whole-body protein turnover had slowed substantially. On refeeding with 80 g a day for one week, protein refeeding made nitrogen balance positive at +1.7 g/day (P < 0.05) while carbohydrate refeeding improved it only to -3.1 g/day.

  24. Study in people1969
    Liver and kidney metabolism during prolonged starvation

    Journal of Clinical Investigation

    11 people with obesity starved for 5 to 6 weeks. Nitrogen loss in urine, a sign of protein loss, fell for the first 4 weeks, then held at about 3 to 6 g a day. Fat and ketone levels did not level off until after day 17, and the body's own sugar making fell to about 86 g a day, roughly half from the liver and half from the kidneys. This is the source for late-starvation protein loss figures.

    11 obese subjects followed through 5 to 6 weeks of starvation. Daily nitrogen excretion decreased progressively for the first 4 weeks and then held at about 3 to 6 g a day. At 5 to 6 weeks in 5 subjects, total urinary nitrogen was 4.66 ± 0.62 g per 24 hours with ammonia at 1.95 ± 0.22 and urea at 1.55 ± 0.60, an inversion of the fed-state pattern. Free fatty acids, beta-hydroxybutyrate and acetoacetate did not plateau until after day 17, and total glucose production fell to approximately 86 g per 24 hours, roughly half hepatic and half renal. This is the verifiable source for late-starvation nitrogen figures.

  25. Study in people1979
    Loss of body nitrogen on fasting

    American Journal of Clinical Nutrition

    During a long fast, the body's nitrogen, a stand-in for protein, drops fast at first and then very slowly over months. Lean people lose about 20 g of nitrogen per kg of weight lost, while people carrying 50 kg or more of fat lose about 10 g. Since the classic starvation studies used people with obesity, they may not apply to lean people.

    Total body nitrogen declines biexponentially during an extended fast, a small fraction with a half-life of a few days and the remainder with a half-life of many months. The ratio of nitrogen loss to weight loss is inversely related to body fat: approximately 20 g of nitrogen per kg of weight lost in non-obese people against approximately 10 g per kg in those with 50 kg or more of body fat. This is the caveat on generalising classic starvation physiology, all of which was measured in obese subjects, to lean people.

  26. Study in people2017
    Metabolic effects of fasting on human and mouse blood in vivo

    Autophagy

    Blood from 4 to 5 people per group was studied after up to 4 to 5 days of fasting. White blood cells showed no rise in cell clean-up (autophagy) in the body, only after they were grown in a dish with a blocking drug, and then only in one cell type. In mice, a 48-hour fast raised clean-up in all white blood cell types and cost about 20% of body weight. People lose less than 2% over 4 days.

    Human and mouse blood studied in the same experiment, 4 to 5 subjects per condition. No enhanced autophagy-associated lipidation was detectable in vivo in human leukocytes after up to 4 to 5 days of fasting; flux appeared only after ex vivo culture with leupeptin, and among human cell types only in neutrophils. A 48 hour mouse fast, by contrast, produced LC3B lipidation across all leukocyte subpopulations. The paper also quantifies the dose mismatch: 48 hours of fasting causes approximately 20% weight loss in mice and less than 2% in humans over 4 days.

What people report

These are things people say about their own fasting. Nobody checked them, so they are not evidence. They are here because they tell you what to expect and what to watch for, which the studies do not. They cannot tell you whether anything works.

  • Days three and four described as the low point for energy, followed by a lift.
  • A high on days five and six, described as lots of mental energy and strong focus. Some accounts put that phase much later, around days twelve to fifteen.
  • Feeling cold all the time, like wearing a coat indoors and never getting warm. Almost everyone at this length reports it.
  • Weakness that feels different from tiredness: needing to plan movements, taking stairs slowly, and stopping driving.
  • Dizziness on standing that is bad enough to limit daily life.
  • Trouble sleeping and restless nights, along with daytime tiredness.
  • Hunger staying gone. The community itself treats appetite coming back as one of the signs to end the fast.
  • From clinic patient material rather than forums: runny nose, bloodshot eyes, skin rashes, sore throat, low back pain, stomach upset, emotional upset, irregular or pounding heartbeat, nausea and vomiting, hair loss, changes in periods, and passing kidney stones or gallstones.

Sources: These are self-reports from public fasting communities and from clinic patient material, with no comparison group. Clinic material is a somewhat more reliable kind of report, because a clinic describing its own patients knows how many there were. Reports at this length are the least representative on the site, since they come mostly from clinic patients and people who keep long diaries. The community itself strongly agrees that fasting beyond about five to seven days belongs under supervision with blood tests, not self-managed. Several items have measured matches: 93.2% of 1,422 supervised people said they had no hunger, and trouble sleeping was the most common symptom people reported in that group, at 14.94%, ahead of tiredness at 13.70%.

Who this is wrong for

  • Anyone doing this without supervision. Both large safety records come from live-in clinics with daily clinical contact, and their serious events happened on fasting days 3 and 9. No published study shows what salt and mineral plan, or how often you need checking, would make a long fast safe on your own. The supervised plans work because a clinician is there every day, not because of a supplement schedule.
  • Anyone taking an SGLT2 inhibitor, unless a prescriber has stopped it at least 3 days ahead, or 4 for ertugliflozin. Supervision does not change this risk much. If you keep taking the drug, ketoacidosis (a dangerous acid build-up) stays likely, your blood sugar keeps looking normal, and a urine strip is the wrong test because it looks for a different ketone.
  • Anyone on insulin, a sulfonylurea or a meglitinide without a prescriber adjusting the dose. That is the single most useful step available, and there is no other way to do it.
  • Anyone with type 1 diabetes. Injected insulin cannot be dialled down the way the body's own insulin is. Keep the dose and you risk low blood sugar. Cut it too far and you risk ketoacidosis.
  • Anyone with a BMI under 16, or under 18.5 plus one other risk factor, or who has lost more than 15% of their body weight in the last 3 to 6 months, or who has low potassium, phosphate or magnesium before eating again. Those are the UK NICE high-risk refeeding signs, and they flag risk going in, not only coming out.
  • Anyone with severe wasting or anorexia, which a large supervised fasting clinic turns away outright. The same clinic also turns away people with advanced kidney, liver or brain blood-flow problems, dementia or other severe memory and thinking problems, and people who are pregnant or breastfeeding.
  • Anyone with a past eating disorder, or who now has episodes of eating they cannot control. Screening takes a few minutes and picks out a very different risk group ahead of time, not after the fact.
  • Children and teens.
  • Anyone with known heart failure or heart artery disease outside a hospital. Of the six deaths in the older review of medical fasting, two were in people with severe heart failure before the fast, which at first seemed to get better with fasting. One had a narrowed heart artery found after death. In the 1,422-person group, a 75-year-old man with known heart artery disease had a heart attack on fasting day 9.
  • Anyone with chronic kidney disease, and anyone on an ACE inhibitor, ARB or water pill without those being reviewed first. Stage 4 to 5 kidney disease, dialysis and kidney transplant are in the "advise against" group of the published guide to diabetes and fasting risk.
  • Anyone with primary adrenal insufficiency, gout, or sickle cell trait, or anyone taking lithium while cutting back on fluids.
  • Anyone who cannot commit to the refeeding period. The published water-only plan sets it at no less than half the length of the fast. And the case reports record deaths while people were starting to eat again, not during the fast.

Questions

How safe is a fast of a week or more?
In supervised settings, safer than the older case reports suggest, but not without problems. Across 768 supervised water-only fasting visits, typically 7 days long, the worst problem was mild or moderate (grade 2 or lower) in 555 visits and severe (grade 3) in 212. There was one grade 4 problem, a step above severe, and no deaths. Across 1,422 people on a 200 to 250 calorie fast, 0.7% had side effects, with no deaths and two hospital stays. Both records come from live-in clinics with daily clinical contact, and their serious events happened on fasting days 3 and 9. Neither tells you anything about doing this alone.
Is the Buchinger data evidence about water fasting?
No, and this is the most common misreading in this research. That group gets 200 to 250 calories a day, including 25 to 35 g of carbs from juice and honey, plus 3 L of water. Its blood sugar settles around 4.7 mmol/L, while day 5 of a true water fast reaches 3.2. Its ketone data come from a urine strip that looks for one ketone (acetoacetic acid), not a blood ketone test. It never measured blood phosphate and never measured body make-up. Its published record also does not add up in two places, its count of low blood salt cases and the size of one group. Those are noted here, not resolved.
Does fasting reduce inflammation?
Not while the fast is going on, going by the biggest measurements we have. CRP, an inflammation marker, rose from 2.85 to 4.30 mg/L across the 1,422-person group. A separate water-only study of 20 people found the same pattern: higher CRP and other inflammation markers (hepcidin, midkine and IL-8), and more active clotting cells. Another long water-only study did find CRP down at the end of the fast, while blood fats and insulin resistance went up. The steady finding is that inflammation during a long fast does not look like the calming picture usually described.
How much muscle does a week of fasting cost?
Less than the scan says. In 16 men on a 10-day low-calorie fast, scan lean mass fell 3.53 kg. But 44% of that was water outside the cells and 14% was stored sugar with its water. That left 1.5 kg of working lean tissue, 25% of total weight lost. In 13 men on a 10-day water fast, lean mass was down 9.2% at day 6 and back to normal after eating again. In the first study, function went the other way: steps rose 60% and weight-bearing strength rose 33%. A lean person loses about twice as much protein per kg lost as a person with obesity, so these numbers do not carry over across body types.
Why is refeeding more dangerous than the fast?
Because the shortfall that matters does not show on a blood test. Over 40 days on a hospital ward, people lost 104 mg of calcium, 48 mg of magnesium and 363 mg of phosphorus a day on average, while blood phosphorus and magnesium did not change at all. Eating carbs again raises insulin, which pushes sugar, phosphate, potassium and magnesium into the cells all at once, and uses up thiamine at the same moment. The case reports follow that pattern: a death during refeeding after a 3-week fast, a death on day 7 or 8 of eating again after a 210-day fast, and in the 382-day fast, the lowest potassium and magnesium of the whole time both came in the first days of eating again.
Does a long fast raise autophagy?
The one measurement of clean-up speed at this length found nothing. In a study that ran human and mouse blood side by side, there was no clean-up boost in human white blood cells inside the body after up to 4 to 5 days of fasting. It only showed up when cells were grown in a dish with a drug (leupeptin), and only in one type of cell (neutrophils). In mice, a 48-hour fast boosted clean-up in all types of white blood cells. That mouse fast also makes mice lose about 20% of their weight, while 4 days of fasting makes people lose less than 2%. So this page does not claim a multi-day fast reaches autophagy.