Intermittent Fasting

What it is

Intermittent fasting is an eating pattern rather than a particular diet. Common versions are time-restricted eating, which limits eating to a daily window of about 8 to 10 hours [1]; the 5:2 diet, with two non-consecutive low-calorie days a week [2]; alternate-day fasting, which alternates a 600 kcal fast day with a normal day [3]; and 4:3 fasting, which cuts energy by 80% on three non-consecutive days a week and allows normal eating on the other four [4].

How it might work

Reviewers propose that weight loss comes from eating less overall, more fat burning and better insulin sensitivity; these mechanisms are still under study [5].

What the research says

Weight: intermittent fasting helps adults with overweight or obesity lose weight. In a meta-analysis of 15 trials with 758 people, body weight fell 3.7 kg more than in control groups [6]. In another meta-analysis of 24 trials with 2,032 people, trials of six months or longer showed 2.8 kg more weight loss than controls, about 3 kg less fat mass, and a waist about 3.9 cm smaller [7]. A 2026 Cochrane review pooled trials against two kinds of control. Compared with no intervention or a waiting list (6 trials, 427 people), fasting led to 3.4 percentage points more weight loss from baseline, with moderate certainty [5]. Compared with regular dietary advice (21 trials, 1,430 people), fasting gave 0.33 percentage points more loss, not significant (95% CI -0.92 to 0.26), with low certainty [5]. Most trials were short: the median follow-up in one umbrella review was three months [8].

Compared with daily calorie cutting, fasting works about as well. A network meta-analysis of 99 trials with 6,582 adults found that fasting and calorie cutting both beat unrestricted eating for weight loss; alternate-day fasting was the only fasting method that beat calorie cutting, by about 1.3 kg [9]. Other reviews found differences under 1 kg: fasting gave 0.9 kg more short-term weight loss in 10 trials, which the authors did not consider clinically significant [10], and a pooled difference of 0.26 kg, not significant, across 24 trials [11]. In a 12-month trial of 165 adults with obesity, 4:3 fasting led to 2.9 kg more weight loss than daily calorie cutting (95% CI 0.1 to 5.7 kg), a result the authors describe as modest [4]. In a 4-week head-to-head trial of 76 adults, alternate-day fasting cut total fat volume by about 700 cm3 more than 16:8 eating, and only alternate-day fasting lowered visceral fat and non-HDL cholesterol; the trial was small and short [12].

Blood sugar: in adults with prediabetes or type 2 diabetes, fasting lowered HbA1c, the standard blood sugar test, by 0.8 percentage points more than control groups across 14 trials of 1,101 adults [13]. In an 8-hour time-restricted eating trial of 75 adults with type 2 diabetes, HbA1c fell 0.91% against controls, similar to the 0.94% fall with daily calorie cutting [14]. A 12-week trial in people on insulin found HbA1c fell 7.3 mmol/mol, against a rise of 0.1 in controls, with no severe low blood sugar [15]. In a 12-month trial of 137 adults with type 2 diabetes, intermittent restriction lowered HbA1c by 0.3%, similar to the 0.5% fall with continuous restriction [16]. In a 16-week trial of 405 Chinese adults with early type 2 diabetes, a 5:2 plan with meal replacements lowered HbA1c by 0.3 percentage points more than metformin; the trial was open-label [2].

Blood pressure: diastolic pressure fell in three meta-analyses, by 2.0 to 3.3 mmHg against controls [17][7][6]. Systolic pressure fell by 4.4 mmHg in one meta-analysis of 15 studies with 929 adults [17], but showed no change in the other two [7][6]. A meta-analysis in prediabetes or type 2 diabetes found no significant blood pressure change [13].

Cholesterol: results are mixed. One meta-analysis found total cholesterol 6.3 mg/dL lower and LDL 5.4 mg/dL lower than controls [6]. Another found total cholesterol 0.31 mmol/L lower in prediabetes or type 2 diabetes but no change in LDL [13], and a third found no change in total cholesterol or LDL [7].

Insulin resistance: results are mixed. Fasting lowered HOMA-IR, a measure of insulin resistance, by 0.14 more than continuous calorie restriction in one meta-analysis [10]. In five small studies of women with polycystic ovary syndrome, insulin resistance fell by 0.94, though the authors call the evidence preliminary [18]. Another meta-analysis found no significant change in insulin [13].

Liver fat: in 840 people with fatty liver disease across 10 studies, liver fat, liver enzymes and waist-to-hip ratio improved [19]. In a 3-month trial of 80 adults with obesity and fatty liver, combining fasting with exercise cut liver fat by 5.48%, against 0.17% in controls; fasting alone was not significantly different from the combination [3].

Hunger: fasting did not curb hunger more than daily calorie cutting. A meta-analysis of 17 trials with 1,111 people found no clear difference in hunger, fullness or desire to eat, though only 13 trials measured hunger [20]. Every trial had some concerns or a high risk of bias, and hunger was measured as an appetite rating [20].

Mood: a narrative review of 39 human studies linked short-term time-restricted eating and supervised prolonged fasting to modest falls in depressive symptoms and stress, in small samples [21].

Longevity and biological aging: no human study has shown that fasting extends life. In the 2-year CALERIE trial, 220 healthy adults were randomized to cut calories by 25% or eat normally; they achieved about 12% and lost about 10% of their body weight [22][23]. The trial slowed the DunedinPACE DNA measure of aging pace by about 2–3% — about nine in ten participants' trajectories still overlapped with controls — while the PhenoAge and GrimAge aging clocks did not change; the authors estimate, based on earlier observational data, that a 2–3% slower pace could correspond to as much as 10–15% lower mortality risk, but deaths were not measured [22]. T3 and TNF-α also fell [23]. Three monthly cycles of a 5-day fasting-mimicking diet were linked to a 2.5-year lower median biological age estimate in a 100-person randomized trial, with nearly identical findings in a second 44-person study; the researchers reported the change was independent of weight loss, but the analyses were exploratory and the company that makes the diet supplied it, with two authors holding equity [24]. Four weeks of alternate-day fasting lowered LDL cholesterol, T3 and an age-related inflammation marker (sICAM-1) [25]. Six months of intermittent fasting plus time-restricted eating was followed by a possible rise in autophagy, a cell-cleaning process tied to aging, in 121 adults with obesity; the authors call the result uncertain [26]. In people with multiple sclerosis, secondary analyses found intermittent calorie restriction lowered a metabolomic aging score while daily calorie restriction did not [27].

Mortality data are observational and mixed. In 33,052 US adults followed for a median of 8 years, the link between eating window and death was U-shaped: the lowest risk was at about 11–12 hours, windows of 8 hours or less were linked to 30% or more higher all-cause mortality (most strongly in older adults), and windows of 15 hours or more to 25% higher mortality [28]. In 10,561 US adults over 60, nightly fasts longer than 12.38 hours were linked to higher cardiovascular mortality, with the lowest risk at 11.49 hours [29]. A meta-analysis of 9 cohorts (242,095 adults) found skipping breakfast was linked to 27% higher all-cause mortality, though certainty was rated very low and some sensitivity analyses disagreed [30]. A 2026 systematic review of adults over 60 found 16:8 eating and Islamic Sunnah fasting were the most effective patterns for weight across 7 pooled RCTs, but noted that very restrictive eating windows (10 hours or less) and prolonged fasting were linked to adverse outcomes including lower cognitive scores [31]. No randomized trial has reported deaths, heart attacks or strokes [32].

Meal timing: when the eating window starts appears to matter somewhat. A 2023 network meta-analysis of 12 trials found early windows improved insulin resistance more than later windows, though the effect was small and its range was wide [33]. A 2024 meta-analysis found fasting glucose fell only in early-window studies [34]. In a 3-month trial, early windows with calorie restriction lowered fasting glucose more than late windows (0.35 mmol/L), with no significant difference in weight loss [35].

Doses used in studies

These are the doses and durations from the trials, not recommendations. The 4:3 trial cut energy by 80% on three days a week for 12 months [4]. The 5:2 arm of a 16-week trial used two fasting days a week with meal replacements [2]. The alternate-day trial used a 600 kcal fast day alternating with a feast day for three months [3]. In a 12-month trial in type 2 diabetes, intermittent restriction used 500 to 600 kcal on two non-consecutive days a week, against 1,200 to 1,500 kcal daily for continuous restriction [16]. Time-restricted eating trials used eating windows of 8 to 10 hours for three months [1], an 8-hour window for six months in type 2 diabetes [14], or a 16:8 pattern for 12 weeks [36].

Safety and interactions

Low blood sugar with diabetes medicines: this is the clearest safety signal. In a 12-week trial of 37 people with type 2 diabetes on glucose-lowering medicines, fasting increased the rate of hypoglycaemia despite medicine reductions (risk ratio 2.05, 95% CI 1.17 to 3.52) [37]. In a 12-month trial, 16 of 46 people on sulfonylureas or insulin had a low or high blood sugar event in the first two weeks [16]. In a separate 12-week trial of insulin users, no severe hypoglycaemia occurred [15]. We found no study data on other medicine interactions.

Common side effects: in a meta-analysis of 15 studies, fasting groups had more vomiting (risk ratio 1.11) and irritability (risk ratio 1.22) than controls; the authors say these were mostly early and resolved on their own [17]. A separate meta-analysis of 15 trials with 1,365 adults found no higher rates of tiredness, headache or dropout than control groups [38]. Some participants reported mild headaches [32]. In a 12-week trial of time-restricted eating, lean mass index fell 0.16 kg/m2 more than in controls [36].

Mood: a narrative review reports possible mood destabilisation, anxiety and, rarely, psychotic or manic episodes in vulnerable people, while randomised trials reported few adverse events [21].

Eating disorders: the same review notes that people with eating-disorder traits showed more preoccupation with food and a higher risk of relapse [21].

Pregnancy and breastfeeding: the only human trial we found tested time-restricted eating plus exercise before and during pregnancy in 167 people at higher risk of gestational diabetes. It did not change blood sugar at week 28, and its abstract reports no safety outcome [39]. We found no human data on breastfeeding.

Kidney disease: our searches found no trials in people with kidney disease.

Older adults: a review of fasting and cardiovascular ageing urged caution before wide use, especially in older people [40]. In an observational US cohort of 33,052 adults followed for about eight years, eating windows of 8 hours or less were linked to 30% or higher all-cause mortality, most strongly in older adults; this cannot show cause and effect [28]. A second analysis of 10,561 US adults over 60 found the longest nightly fasts (over 12.4 hours) were linked to higher cardiovascular mortality, with the lowest risk at about 11.5 hours [29].

Upper limits: no upper limit was reported in these studies. The trials used fast-day energy intakes of 500 to 600 kcal [16].

Caveats and open questions

Heart events are not measured: no trial in the 2021 Cochrane review reported heart attacks, strokes, heart failure or deaths [32]. The 2021 umbrella review called for long-term trials of clinical outcomes [8].

Most trials are short and small. The 2020 review found trials with 10 to 244 participants lasting 2 to 26 weeks [41]. The 2026 Cochrane review, covering trials up to 12 months, found weight loss no different from regular dietary advice (mean difference -0.33 percentage points, 95% CI -0.92 to 0.26; low certainty) [5]. In trials of 24 weeks or more, the network meta-analysis found benefit only against unrestricted eating [9]. Another review found no superiority over daily calorie cutting in long-term outcomes [10].

Time-restricted eating results conflict. A 12-week trial of 116 adults found 16:8 eating no more effective for weight than eating throughout the day (difference -0.26 kg, not significant) [36]. A meta-analysis of 29 trials found time-restricted eating 1.4 kg more effective than control, but 22 of the 29 trials had high concerns about bias, and the authors call the effect small and of uncertain importance [42].

Adherence: compliance above 80% was reported only in trials shorter than three months [11].

Funding: several authors of the network meta-analysis declared industry funding or consulting, including from food and sweetener companies [9]. The 2023 meta-analysis comparing alternate-day fasting, the 5:2 diet and time-restricted eating was funded by Nestle, which employed some of its authors [11]. The 4:3 trial was funded by the US National Institute of Diabetes and Digestive and Kidney Diseases [4].

Open questions: a recruiting trial of 225 adults, due to finish in 2027, is testing a modified intermittent eating plan for weight loss (NCT06302803).

Longevity evidence rests on biomarkers, not outcomes: the 2-year CALERIE trial slowed DunedinPACE but not the PhenoAge or GrimAge clocks, and whether a 2–3% slower pace of aging means fewer deaths is unknown [22]. The fasting-mimicking diet biological-age results are exploratory and the trial had industry ties [24]. Mortality studies cannot show cause and effect — illness itself can change appetite, eating windows and fasting duration [28].

A 177-person randomized trial (ENSATI) is testing a 10-hour eating window against 25% calorie restriction for six months, with molecular markers of aging among its outcomes [43].

Early versus late timing is not settled. An isocaloric 2-week crossover in 31 women found no difference in insulin sensitivity between early and late windows [44], and a 2021 meta-analysis of early eating rated most outcomes very low to low quality [45].