What happens during a multi-day water fast, and is it safe?

The first days. The body first uses up stored carbohydrate. As insulin falls, fat is broken down and the liver makes ketones, which the brain can use as fuel¹. In a clinic study of 1,422 people fasting 4 to 21 days on about 200 to 250 kcal a day (a modified fast, not water alone), blood sugar fell to the low-normal range and ketones rose in every group, and 93% reported no hunger³.

Weight and lean mass. Reviews of human trials report 2% to 10% weight loss over 5 to 20 days, but about two-thirds is lean mass, which can include water as well as muscle, and the authors flag possible muscle breakdown as a concern¹.

Blood pressure. Pressure consistently falls¹. In 1,610 people fasting about 10 days, average blood pressure fell from 126/81 to 120/78 mm Hg and by about 17/9 mm Hg in people with untreated hypertension. Among those taking blood-pressure drugs, medication was stopped in 24% and reduced in 44%. That is why supervision matters⁴.

Blood sugar and diabetes. Fasting lowered glucose and insulin in adults with normal blood sugar, but not in people with diabetes¹. In a pilot trial in type 2 diabetes (32 completers), a one-week 300 kcal-a-day fast brought more weight loss (3.5 vs 2.0 kg) and lower blood pressure at 4 months than usual care, with no significant change in HbA1c⁵.

It does not last. Three to four months after fasting, the metabolic gains had disappeared even when weight loss was kept¹. A 5-day modified fast in 64 healthy adults improved some markers briefly, but gut-microbiome and blood-metabolome differences were gone a month later⁶. By contrast, two years of moderate calorie restriction in the CALERIE trial improved LDL cholesterol, blood pressure and insulin sensitivity throughout⁷.

Safety. In a review of 768 supervised water-only fasts, most adverse events were mild and expected, though 212 fasts included at least one severe-graded event and 2 had a serious adverse event². Headaches, insomnia, hunger and metabolic acidosis show up in some studies¹.

Refeeding. Restarting food after prolonged starvation can cause dangerous shifts in phosphate, potassium, magnesium and fluids. It is best documented in malnourished patients, who restart with few calories and thiamine⁸.

Dry fasting. Going without fluids as well as food is a different and far riskier thing. Going more than a few days without water is fatal⁹, and our PubMed search found no controlled trials of multi-day dry fasts.

For gentler options see Does limiting your eating window help?, Who should avoid intermittent fasting and Electrolytes beyond water.

Narrative review of human trials. The 2024 review summarized prolonged fasting of 5 to 20 days: ketones rose sharply, weight fell 2% to 10% (about two-thirds lean mass, one-third fat), and blood pressure consistently decreased. Lipid effects were unclear. Fasting glucose, insulin, insulin resistance and HbA1c improved in normoglycemic adults but stayed unchanged in type 1 or type 2 diabetes. Three to four months after the fast, all metabolic benefits were no longer seen, even where weight loss was maintained. Adverse events included metabolic acidosis, headaches, insomnia and hunger; the authors call prolonged fasting "moderately safe"¹.

Chart review of supervised water-only fasts. From 2,539 charts at a residential medical facility (2006 to 2011), 768 visits involved adults fasting at least two days on water alone and refeeding for half as long. The highest-grade event was grade 2 or lower in 555 visits, grade 3 in 212 (28%), and grade 4 in 1; there were no grade 5 events and 2 visits had a serious adverse event. Of all adverse events, 75% (4,490) were mild and known reactions to fasting². This is a retrospective review of people who chose fasting at one center.

Clinic cohorts. The 1,422-person study ran for one year at a specialized clinic using Buchinger fasting (200 to 250 kcal a day) for 4 to 21 days with a lifestyle program. Weight, abdominal circumference and blood pressure fell; glucose dropped to the low-normal range; ketones rose; 93.2% reported no hunger; of 404 people with existing complaints, 341 (84.4%) reported improvement; adverse effects occurred in under 1%³. In 1,610 fasters (4 to 41 days, mean 10, about 250 kcal a day), 920 were normotensive, 313 hypertensive without medication and 377 hypertensive on medication. Pressure fell by 3.0/1.9 mm Hg in normotensive people, 16.7/8.8 in untreated hypertensive people and 7.3/4.7 on average in the medicated group, while 69 women with low blood pressure (under 100/60) saw it rise by 6.3/2.2⁴. These are uncontrolled before-and-after designs.

Randomized trials. In the type 2 diabetes pilot, 46 people on oral drugs and/or insulin were randomized to a 7-day Buchinger fast (300 kcal a day, then gradual refeeding) or usual care; 32 finished. There were no serious adverse events. At 4 months weight fell 3.5 kg versus 2.0 kg (P = 0.03), and systolic and diastolic blood pressure fell significantly, while HbA1c, insulin and HOMA did not improve significantly⁵. In 64 healthy adults randomized to a 5-day ketogenic modified fast (about 600 kcal a day) or usual eating, weight, glucose and well-being improved, blood pressure fell significantly only after food was reintroduced, and microbiome and metabolome differences were no longer significant one month later⁶.

The calorie-restriction benchmark. CALERIE randomized 218 healthy, non-obese adults to 25% calorie restriction or usual eating. They achieved about 12% restriction and sustained a 7.5 kg loss (71% fat), with lasting improvements in LDL cholesterol, total-to-HDL ratio, blood pressure, C-reactive protein and insulin sensitivity over two years⁷. Fasts, in contrast, bring faster changes that tend to fade.

Harms and gaps. Refeeding syndrome features low phosphate, magnesium and potassium and vitamin deficiency; most data come from case reports and observational studies of malnourished patients, with low-energy restarts and thiamine as the standard prevention⁸. In type 2 diabetes treated with insulin or sulfonylureas, fasting regimens carry an increased risk of low blood sugar¹⁰. Intermittent fasting was prospectively associated with higher eating-disorder psychopathology in 491 adults, though selection may play a role¹¹. For pregnancy, only daylight Ramadan fasting has been studied: birth weight was unaffected across 22 studies, but data on other outcomes were insufficient¹². Water deprivation beyond a few days is fatal, and modest deficits harm vulnerable groups⁹.

Conflicting and missing evidence. The clinic cohorts report big improvements, but there is no control group and participants chose to fast. The small randomized trials used modified fasts rather than water alone and measured outcomes within months. None of the trials above measured heart attacks, new diabetes or survival.

Bottom line (confidence 2 of 5). A multi-day water fast may lower weight and blood pressure within days, but much of the weight lost is lean tissue, benefits fade within months, and risks (refeeding, medication problems, dehydration, eating-disorder relapse) are real. If you are curious, choose gentler patterns first and involve a clinician for anything beyond a day or two.