Skip to content
The Longevity Archive
browse 192

The biology of fasting

Refeeding, the leucine threshold and refeeding syndrome

Also called: Breaking a fast, Hypophosphataemia

Written by Reviewed Sep 2026

Coming out of a fast is its own physiological event with its own rules and, in a genuinely malnourished person, its own danger, and the danger is not that phosphate is lost but that insulin drives it inside cells all at once.

What it is

Refeeding runs the fasted signalling state backwards, and it does so quickly. Amino acids arrive and are sensed directly by the mTORC1 machinery. Insulin arrives in parallel and acts on a second arm. Within minutes to tens of minutes, protein synthesis initiation rises, autophagy initiation is suppressed, and the transcriptional programme that was scaling up lysosomal capacity shuts off.

A protein-containing meal is the most decisive way to end the fasted state, more decisive than the same energy as carbohydrate or fat, because protein engages both the amino acid sensor arm and the insulin arm while fat engages neither strongly.

Two separate questions live on this page. How much protein, and with how much leucine, does a meal need to maximise the muscle response, which is a question for anyone eating in a compressed window. And what makes reintroducing food dangerous to a severely malnourished person, which is a question about electrolytes and one specific ion.

The machinery

The signalling reversal. Leucine binds Sestrin2 and arginine binds CASTOR1, both releasing GATOR2. Methionine raises S-adenosylmethionine, which binds SAMTOR. GATOR2 inhibits GATOR1, the Rag GTPases adopt their active configuration, and mTORC1 is recruited to the lysosomal surface where Rheb is waiting. Glucose and amino acids both stimulate insulin, insulin activates PI3K and AKT, AKT inhibits TSC2, Rheb-GTP rises. Downstream, S6K1 and 4E-BP1 are phosphorylated and cap dependent translation initiation increases; ULK1 serine 757 is phosphorylated and autophagy initiation is suppressed; TFEB is phosphorylated and held in the cytosol.

Insulin's role in muscle protein synthesis is best described as permissive rather than driving. At physiological concentrations it is required for the full anabolic response to amino acids, and it strongly inhibits protein breakdown, so net balance improves substantially. Raising insulin further with adequate amino acids present does not raise synthesis proportionately. Amino acid availability is the limiting input.

The dangerous part is the same insulin doing its other jobs. Insulin drives glucose into cells, and phosphate, potassium and magnesium follow. Phosphate moves for an additional and more powerful reason: glucose entering a cell is phosphorylated by hexokinase and phosphorylated again by phosphofructokinase, each step consuming a phosphate from ATP and trapping the product inside. At the same time the resumption of oxidative metabolism drives a surge in ATP synthesis, and red cells resume producing 2,3-bisphosphoglycerate. All of that consumes inorganic phosphate. Serum phosphate can fall precipitously within 24 to 72 hours of refeeding, because it is being consumed and sequestered intracellularly at high rate.

The consequences follow from what phosphate is for. ATP cannot be regenerated adequately, and falling 2,3-BPG shifts the oxyhaemoglobin dissociation curve leftward and impairs tissue oxygen delivery. Clinically that appears as cardiac arrhythmia and failure, respiratory muscle weakness, rhabdomyolysis, haemolysis, seizures, confusion and paraesthesia. Concurrent low potassium and magnesium add arrhythmia risk, and low magnesium makes low potassium refractory to correction. Two further insults: insulin causes sodium and water retention, expanding volume in someone whose myocardium has atrophied; and sudden carbohydrate oxidation in a thiamine-depleted person can precipitate Wernicke's encephalopathy, because pyruvate dehydrogenase requires thiamine pyrophosphate and demand for it rises abruptly.

What switches it on

Amino acids, leucine above all, and the insulin surge that carbohydrate produces. The greater the glucose flux on refeeding, the greater the phosphate consumption, which is exactly why carbohydrate load is the variable clinical protocols restrict first.

On the timing
The signalling reversal is fast: the amino acid sensing steps operate in minutes in cell systems. Whether human autophagic flux is suppressed within 60 minutes of a whey protein drink is the exact question a registered trial was designed to answer; the publication available is a protocol paper and contains no results. The clinical timing is better established. In 62 intensive care patients refed after 48 hours or more of starvation, serum phosphorus fell below 0.65 mmol/L in 21 of them, a mean of 1.9 plus or minus 1.1 days after feeding started. In 69 adolescents with anorexia nervosa, 81% reached their phosphorus nadir within the first week.

What has been measured

In people

Both halves of this page are human. The protein dose-response work is stable isotope tracer infusion with serial muscle biopsies in living people. The refeeding syndrome work is clinical, in intensive care patients, in adolescents with anorexia nervosa, and in one randomised controlled trial of how fast to feed.

  • Whole protein dose response, 6 healthy young men, five trials, leg resistance exercise then 0, 5, 10, 20 or 40 g of whole egg protein, synthesis measured over 4 hours by primed constant leucine infusion. Muscle protein synthesis showed a dose response and was maximally stimulated at 20 g. Albumin synthesis also plateaued at 20 g. Leucine oxidation increased significantly at 20 and 40 g, meaning the protein above the plateau was being burned. It has six participants (American Journal of Clinical Nutrition, 2009).
  • Whey dose response, larger and better powered. 48 resistance-trained young men of approximately 80 kg, given a standardised high protein breakfast then unilateral leg exercise then 0, 10, 20 or 40 g of whey protein isolate, myofibrillar synthesis over 4 hours by phenylalanine infusion. The 0 g rate was 0.041 plus or minus 0.015% per hour. Synthesis rose 49% with 20 g and 56% with 40 g. 10 g produced no significant stimulation. 40 g produced no additional synthesis over 20 g but did increase phenylalanine oxidation and urea production (American Journal of Clinical Nutrition, 2014).
  • Isolating leucine from total protein, which is the direct test of the threshold idea. 40 men, unilateral resistance exercise, then one of five drinks: 25 g whey containing 3.0 g leucine; 6.25 g whey containing 0.75 g leucine; 6.25 g whey topped up to 3.0 g total leucine; 6.25 g whey topped up to 5.0 g total leucine; or 6.25 g whey topped up to 5.0 g leucine plus isoleucine and valine. Over 1.5 to 4.5 hours the greatest responses were 25 g whey at approximately 267% and 6.25 g whey topped to 5.0 g leucine at approximately 220%, P = 0.002. Topping only to 3.0 g did not match. A small protein dose can be rescued toward a much larger one by adding leucine, but the rescue needed the higher leucine target, not merely matching the leucine content of the 25 g dose (American Journal of Clinical Nutrition, 2014).
  • The age comparison, and it needs its caveat attached. The widely quoted figures are that the plateau is reached at 0.40 g/kg body mass in older men versus 0.24 in younger. That comes from a retrospective pooled reanalysis of the group's own prior datasets, not a new prospective trial. The total participant count is not stated in the abstract. And the headline comparison was not statistically significant: 0.40 plus or minus 0.19 versus 0.24 plus or minus 0.06 g/kg body mass, p = 0.055. Only the lean-body-mass comparison cleared significance, 0.60 plus or minus 0.29 versus 0.25 plus or minus 0.13 g/kg lean mass, p < 0.01. Basal fractional synthetic rate did not differ by age, 0.027 versus 0.028% per hour, p = 0.53 (Journal of Gerontology Series A, 2015).
  • The direct challenge to the per-meal ceiling. 36 recreationally active young men, 12 per group, given 0, 25 or 100 g of intrinsically labelled milk protein after 60 minutes of whole-body resistance exercise. 100 g produced a greater and more prolonged response than 25 g, lasting beyond 12 hours, with myofibrillar synthesis roughly 20% higher over 0 to 4 hours and roughly 40% higher over 4 to 12 hours. Protein ingestion had negligible effect on whole-body protein breakdown or amino acid oxidation rates (Cell Reports Medicine, 2023).
  • How well the threshold idea holds up overall. A systematic review of 29 eligible studies that measured both blood leucine and muscle protein synthesis found 16 supporting the leucine trigger hypothesis and 13 not supporting it. Of the 16 supportive studies, 13 were in older adults and 14 used isolated proteins. The concept applies best to older adults eating isolated protein and much less well to protein-rich whole foods (Frontiers in Nutrition, 2021).
  • Human autophagy on refeeding: fibre-type-specific autophagy responses were measured in human skeletal muscle after an overnight fast and after mixed meal ingestion (American Journal of Physiology: Endocrinology and Metabolism, 2022).
  • The mechanistic bridge to refeeding syndrome, measured in healthy people. During insulin infusion in normal subjects before and after a 48 hour fast, serum phosphate fell in every subject (P < 0.001) and the magnitude of the fall correlated with glucose disposal rate at r = 0.76, P < 0.005. The greater the glucose flux, the greater the phosphate consumption. That is the mechanism of refeeding hypophosphataemia measured directly in healthy humans, in the same paper that established selective insulin resistance (Metabolism, 1983).
  • Why serum testing misses the risk. Six obese subjects spent 40 days on a metabolic ward at 300 kcal per day supplying 406 mg phosphorus, 7 mg magnesium and 75 mg calcium daily. Cumulative urinary losses exceeded cumulative intake by 58% for calcium, 75% for phosphorus and 500% for magnesium. Mean daily balances were minus 104 mg calcium, minus 48 mg magnesium and minus 363 mg phosphorus. Serum phosphorus and magnesium did not change. Serum calcium fell 0.5 mg/dL, p < 0.05. This was a protein-supplemented fast rather than a total fast, but the principle carries: a normal serum magnesium or phosphate during a fast does not mean stores are intact (American Journal of Medicine, 1981).
  • Incidence, intensive care. 62 patients refed after 48 hours or more of starvation. Phosphorus fell more than 0.16 mmol/L to below 0.65 mmol/L in 21 of 62, 34%, and below 0.32 mmol/L in six. The mean time to the fall was 1.9 plus or minus 1.1 days after feeding started (Archives of Surgery, 1996).
  • Incidence, adolescents with anorexia nervosa. 69 patients started at 1,200 to 1,400 kcal per day advancing 200 kcal every 24 to 48 hours. Four (5.8%) had moderate hypophosphataemia, 15 (21.7%) mild, and 19 (27.5%) required phosphorus supplementation. 81% reached their phosphorus nadir within the first week. The patient with the lowest phosphorus had short runs of ventricular tachycardia (Journal of Adolescent Health, 2003).
  • The one randomised trial on how fast to feed. 339 intensive care adults who developed refeeding syndrome within 72 hours of starting nutrition, randomised to caloric restriction or standard feeding. Days alive after intensive care discharge were 44.8 versus 39.9, not significant at p = 0.19, but more patients were alive at day 60: 149 of 164 (91%) versus 128 of 163 (78%), p = 0.002. Restricting calories in people who had already developed refeeding hypophosphataemia improved 60-day survival (Lancet Respiratory Medicine, 2015).
  • One refeed does not undo everything. After a 30 hour fast in 8 healthy men with 20 minute sampling, refeeding 800 kcal raised TSH toward midnight, significantly above the starved state but still significantly below never-starved values at the same time of day. Serum T3 was 1.84 plus or minus 0.03 nmol/L starved and 1.80 plus or minus 0.05 after the refeed, not significantly different. Reverse T3 tended to fall but not significantly (Acta Endocrinologica, 1984).

In other species and in cell culture

The amino acid sensing steps that make refeeding work at the molecular level were characterised in cultured cell lines and with purified proteins, and the step that switches autophagy off on refeeding was established in cell culture.

  • Cultured cells and purified proteins: leucine binding Sestrin2 to release GATOR2 (Science, 2016), arginine binding CASTOR1 to do the same (Cell, 2016), S-adenosylmethionine binding SAMTOR to act on GATOR1 (Science, 2017), and the Rag GTPases binding raptor to relocalise mTORC1 to the lysosome (Science, 2008).
  • Cultured cells: mTORC1 phosphorylating ULK1 at serine 757, which is the step that shuts autophagy initiation off when food returns (Nature Cell Biology, 2011).

Why it matters

Refeeding is the half of fasting that almost nobody writes about, and it contains the only reliably lethal event in the whole subject.

On protein, the practical reading of the human tracer work is that in young adults roughly 20 g of a rapidly digested, leucine-rich protein maximises the acute muscle protein synthetic response after exercise, that leucine content is a major determinant independent of total protein, and that older adults need more. These are 4 hour acute synthesis measurements in small numbers of mostly young men, not chronic hypertrophy outcomes. They are also directly relevant to compressed eating windows, because a person eating in a short window has fewer meals in which to clear whatever the per-meal bar is, and that arithmetic gets harder with age.

On safety, the mechanism is worth understanding precisely because the standard summary gets it backwards. Refeeding syndrome is not phosphate loss. Total body phosphate is already depleted from the fast while serum phosphate reads normal, and then insulin drives what is left into cells all at once. The one randomised trial says the rate of feeding matters: restricting calories in people who had already developed refeeding hypophosphataemia improved 60-day survival.

Refeeding syndrome is a phenomenon of significant, sustained malnutrition. It is not a described consequence of a short fast in a well nourished person, and the incidence literature comes from hospitalised and eating disorder populations. That boundary is part of the fact, not a softening of it.

A claim you will see repeated

Two things get repeated here that this page does not repeat. First, the 0.40 versus 0.24 g/kg protein figures for older versus younger adults are quoted as established. They come from a retrospective pooled reanalysis of one group's own prior datasets, the total participant count is not stated in the abstract, and the comparison itself was p = 0.055. Only the lean-body-mass version cleared significance. The direction is probably right and the specific numbers should not be treated as a threshold. Second, that protein above roughly 25 g per sitting is wasted. In 36 young men given 0, 25 or 100 g of labelled milk protein after whole-body resistance exercise, 100 g produced a larger and longer response than 25 g, extending beyond 12 hours, with negligible effect on whole-body breakdown or oxidation rates. Someone eating 150 g of protein across three meals inside an 8 hour window is not, on the human evidence, discarding the surplus. And on the clinical side: refeeding hypophosphataemia is routinely described as phosphate being lost. It is not lost on refeeding. It is consumed and sequestered inside cells, which is why the serum number can be normal right up until the moment feeding starts.

Citations

  1. Human2009
    Ingested protein dose response of muscle and albumin protein synthesis after resistance exercise in young men

    American Journal of Clinical Nutrition

    6 healthy young men, five trials in randomised order, leg resistance exercise then 0, 5, 10, 20 or 40 g whole egg protein, synthesis over 4 hours by primed constant leucine infusion. Muscle protein synthesis was maximally stimulated at 20 g and albumin synthesis also plateaued there. Leucine oxidation rose significantly at 20 and 40 g. Phosphorylation of p70S6K, rpS6 and eIF2B epsilon was unaffected by protein dose.

  2. Human2014
    Myofibrillar muscle protein synthesis rates subsequent to a meal in response to increasing doses of whey protein at rest and after resistance exercise

    American Journal of Clinical Nutrition

    48 resistance-trained young men of approximately 80 kg, standardised high protein breakfast then unilateral leg exercise then 0, 10, 20 or 40 g whey protein isolate, myofibrillar synthesis over 4 hours by phenylalanine infusion. The 0 g rate was 0.041 plus or minus 0.015% per hour; synthesis rose 49% with 20 g and 56% with 40 g. 10 g produced no significant stimulation. 40 g added no synthesis over 20 g but raised phenylalanine oxidation and urea production.

  3. Human2014
    Leucine supplementation of a low-protein mixed macronutrient beverage enhances myofibrillar protein synthesis in young men: a double-blind, randomized trial

    American Journal of Clinical Nutrition

    40 men, unilateral resistance exercise then one of five drinks varying total protein and total leucine, myofibrillar synthesis by phenylalanine infusion with serial biopsies in a rested and an exercised leg. Over 1.5 to 4.5 hours the greatest responses were 25 g whey at approximately 267% and 6.25 g whey topped to 5.0 g total leucine at approximately 220%, P = 0.002. Topping only to 3.0 g total leucine did not match the 25 g dose.

  4. Human2015
    Protein ingestion to stimulate myofibrillar protein synthesis requires greater relative protein intakes in healthy older versus younger men

    Journal of Gerontology Series A: Biological Sciences and Medical Sciences

    A retrospective pooled reanalysis of the group's own prior datasets, not a new prospective trial; total participant count is not stated in the abstract. The plateau was reached at 0.40 plus or minus 0.19 g/kg body mass in older men versus 0.24 plus or minus 0.06 in younger men, p = 0.055, which is not statistically significant. Only the lean-body-mass comparison cleared significance, 0.60 plus or minus 0.29 versus 0.25 plus or minus 0.13 g/kg lean mass, p < 0.01. Basal fractional synthetic rate did not differ by age, 0.027 versus 0.028% per hour, p = 0.53.

  5. Human2023
    The anabolic response to protein ingestion during recovery from exercise has no upper limit in magnitude and duration in vivo in humans

    Cell Reports Medicine

    36 recreationally active young men, 12 per group, given 0, 25 or 100 g of intrinsically labelled milk protein after 60 minutes of whole-body resistance exercise. 100 g produced a greater and more prolonged response than 25 g, lasting beyond 12 hours, with myofibrillar synthesis roughly 20% higher over 0 to 4 hours and roughly 40% higher over 4 to 12 hours. Protein ingestion had negligible effect on whole-body protein breakdown or amino acid oxidation rates.

  6. Review2021
    Evaluating the Leucine Trigger Hypothesis to Explain the Post-prandial Regulation of Muscle Protein Synthesis in Young and Older Adults: A Systematic Review

    Frontiers in Nutrition

    Of 29 eligible studies measuring both blood leucine and muscle protein synthesis, 16 supported the leucine trigger hypothesis and 13 did not. Of the 16 supportive studies, 13 were in older adults and 14 used isolated proteins, so the concept applies best to older adults eating isolated protein and much less well to protein-rich whole foods.

  7. Human2022
    Muscle fiber type-specific autophagy responses following an overnight fast and mixed meal ingestion in human skeletal muscle

    American Journal of Physiology: Endocrinology and Metabolism

    Autophagy responses measured by muscle fibre type in human skeletal muscle after an overnight fast and after mixed meal ingestion.

  8. Review2022
    The Break-Fast study protocol: a single arm pre-post study to measure the effect of a protein-rich breakfast on autophagic flux in fasting healthy individuals

    BMC Nutrition

    A registered trial protocol, not a results paper. It is designed to test whether a whey protein drink suppresses autophagic flux within 60 minutes in overnight-fasted healthy adults. It is listed here because it is frequently cited as though it contained results. It does not.

  9. Human1983
    Insulin action during acute starvation: evidence for selective insulin resistance in normal man

    Metabolism

    6 lean subjects, sequential euglycaemic clamps at four insulin infusion rates before and after a 48 hour fast. Serum phosphate fell during insulin infusion in all subjects (P < 0.001) and the magnitude correlated with glucose disposal rate at r = 0.76, P < 0.005. This is the mechanism of refeeding hypophosphataemia measured directly in healthy humans: the greater the glucose flux, the greater the phosphate consumption.

  10. Human1981
    Adverse effects of liquid protein fast on the handling of magnesium, calcium and phosphorus

    American Journal of Medicine

    6 obese subjects, 40 days on a metabolic ward at 300 kcal per day supplying 75 mg calcium, 406 mg phosphorus, 7 mg magnesium, 33 mEq potassium and 11.5 g nitrogen daily. Cumulative urinary losses exceeded cumulative intake by 58% for calcium, 75% for phosphorus and 500% for magnesium. Mean daily balances were minus 104 mg calcium, minus 48 mg magnesium and minus 363 mg phosphorus. Serum phosphorus and magnesium did not change; serum calcium fell 0.5 mg/dL, p < 0.05. This is a protein-supplemented fast rather than a total fast.

  11. Human1996
    Refeeding hypophosphatemia in critically ill patients in an intensive care unit. A prospective study

    Archives of Surgery

    62 intensive care patients refed after 48 hours or more of starvation. Serum phosphorus fell more than 0.16 mmol/L to below 0.65 mmol/L in 21 of 62 (34%), and below 0.32 mmol/L in six patients. Mean time to the fall was 1.9 plus or minus 1.1 days after feeding started.

  12. Human2003
    Hypophosphatemia during nutritional rehabilitation in anorexia nervosa: implications for refeeding and monitoring

    Journal of Adolescent Health

    69 adolescents with anorexia nervosa started at 1,200 to 1,400 kcal per day advancing 200 kcal every 24 to 48 hours. Four (5.8%) developed moderate hypophosphataemia, 15 (21.7%) mild, and 19 (27.5%) required phosphorus supplementation. 81% reached their phosphorus nadir within the first week. The patient with the lowest phosphorus had short runs of ventricular tachycardia.

  13. Human2015
    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 intensive care adults who developed refeeding syndrome within 72 hours of starting nutrition, randomised to caloric restriction or standard feeding. Days alive after intensive care discharge 44.8 versus 39.9, not significant at p = 0.19, but more patients alive at day 60: 149 of 164 (91%) versus 128 of 163 (78%), p = 0.002.

  14. Human1984
    Rapid adaptations of serum thyrotrophin, triiodothyronine and reverse triiodothyronine levels to short-term starvation and refeeding

    Acta Endocrinologica

    8 healthy male volunteers fasted 30 hours then refed 800 kcal, with blood sampled at 20 minute intervals. Starvation lowered TSH below 1 mU/L and abolished the nocturnal peak; serum T3 fell from 2.30 plus or minus 0.06 to 1.84 plus or minus 0.03 nmol/L (P < 0.01) and reverse T3 rose. Refeeding raised TSH significantly above the starved state but it remained significantly below never-starved values, and 800 kcal did not measurably change T3 (1.80 plus or minus 0.05 nmol/L, not significant). Serum cortisol did not differ across conditions.

  15. Review2008
    Refeeding syndrome: what it is, and how to prevent and treat it

    BMJ

    The standard clinical account of refeeding syndrome: the intracellular shift of phosphate, potassium and magnesium on insulin release, the risk groups, and the prevention and treatment framework.

  16. Review2020
    ASPEN Consensus Recommendations for Refeeding Syndrome

    Nutrition in Clinical Practice

    Current consensus diagnostic criteria, based on a decrease in serum phosphorus, potassium and/or magnesium, together with risk stratification and a management framework including checking those electrolytes before initiating nutrition and delaying calories in patients with severely low values until corrected.

  17. In vitro2016
    Sestrin2 is a leucine sensor for the mTORC1 pathway

    Science

    Cultured cells and purified proteins. Leucine binds Sestrin2 directly and releases GATOR2, which is the first step by which a protein-containing meal reactivates mTORC1 and therefore ends the fasted signalling state.

  18. In vitro2011
    AMPK and mTOR regulate autophagy through direct phosphorylation of Ulk1

    Nature Cell Biology

    Cultured cells. mTORC1 phosphorylates ULK1 at serine 757 to suppress autophagy initiation, which is the step that switches recycling off when food returns.