⚠️ Auto-extracted by Claude on 2026-05-26 (updated same day) — the intervention-level synthesis here is built on three human studies read from local PDFs: Pietzner 2024 (7-day Olink, verified), Commissati 2025 (~10-day SOMAScan, verified), and Dai 2024 (21-day clinical chemistry, partial). Older historical starvation-physiology and detailed refeeding-syndrome clinical literature are still not systematically ingested. Treat cross-study generalizations as provisional.

Prolonged fasting (multi-day complete caloric restriction)

Prolonged fasting is voluntary complete caloric restriction sustained for ≥48–72 hours, typically with ad libitum water (and sometimes non-caloric electrolytes) only — a “water-only fast.” It is mechanistically and physiologically distinct from the wiki’s other fasting/restriction pages:

It is not sustained moderate caloric restriction (CR), which reduces daily intake ~12–25% indefinitely while preserving micronutrients (CALERIE benchmark).
It is not intermittent fasting (IF), whose fasting windows (16:8, ADF, 5:2) are ≤48 h and recur cyclically.
It crosses into fasting-mimicking-diet (FMD) territory in duration but, unlike FMD, permits zero caloric intake.

The defining new insight (Pietzner 2024) is that the systemic, multi-organ molecular response to fasting does not begin until after ~3 days — past the duration of every commonly practiced IF protocol — which reframes prolonged fasting as a qualitatively different intervention rather than “more IF.” ¹

This page covers the prolonged complete-fast modality. For the chronic-CR evidence base (lifespan across organisms, CALERIE biomarker trials, CR mimetics) see caloric-restriction; for cyclic short fasting see intermittent-fasting.

Physiology of a multi-day fast

The classical fuel-utilization cascade, confirmed in the 7-day water-only fast of Pietzner 2024 (n=12 healthy adults) ¹:

Phase	Approx. timing	Dominant physiology
Post-absorptive	0–24 h	Hepatic glycogenolysis; falling insulin
Gluconeogenic	1–2 days	Glycogen depleted; gluconeogenesis from glycerol, lactate, amino acids; glucose falls
Ketogenic switch	2–3 days	Free fatty acids rise and plateau; 3-hydroxybutyrate rises continuously; ketones become primary brain fuel
Protein-sparing	>3 days	Urinary nitrogen falls (renal urea reabsorption spares protein pools); systemic proteome remodeling begins

In Pietzner 2024, 7 days produced 5.7 ± 0.8 kg weight loss (~1.9 BMI units). Notably:

Acute weight loss was split between lean (−3.6 kg) and fat (−1.6 kg) mass, but the lean-mass loss almost fully reversed within 3 days of refeeding (−0.69 kg residual) while fat loss was sustained (−1.85 kg) — i.e., the acute “muscle loss” of a short prolonged fast is largely fluid/labile-protein, not durable sarcopenic loss. ¹
Subcutaneous fat fell but visceral fat did not (over only 7 days); no bone-mass loss was detectable (p=0.42). ¹

Mechanisms of geroprotection

Prolonged fasting drives the same core longevity-pathway shifts as CR/IF, but to a deeper and more sustained degree because nutrient input is zero:

Pathway	Effect under prolonged fasting	Cross-link
mTORC1	Strongly suppressed (amino acids + insulin near-floor)	shared with caloric-restriction, rapamycin
ampk	Sustained activation (low ATP/AMP)
autophagy	Robust, prolonged induction past the ~16 h IF threshold	required mediator
Ketogenesis	Continuous 3-hydroxybutyrate rise; BHB as neuronal fuel + signaling metabolite	shared with ketogenic-diet
IIS	IGF-signaling pathway enriched among fasting-responsive proteins

Beyond energy metabolism: ECM remodeling

The distinctive finding of Pietzner 2024 is that the prolonged-fast plasma proteome is dominated by non-metabolic, extracellular-matrix (ECM) remodeling rather than a purely energy-centric signature ¹:

The fasting signature was strongly enriched for ECM proteins (p<3.6×10⁻⁷), with late-appearing (day ≥2–3) declines in ECM-receptor interaction, elastic-fiber, and vascular-wall proteins — interpreted as structural ECM degradation, plausibly contributing to the acute lean-mass loss.
The single most fasting-associated protein was tenascin-R, a brain-specific ECM component of perineuronal nets (beta=−0.73, p<2.4×10⁻³⁷). Rising 3-hydroxybutyrate tracked lower plasma brain-ECM proteins (brevican, vitrin) and neurotrophic factors — a candidate molecular basis for the long-known efficacy of ketogenic/fasting therapy in drug-resistant epilepsy. ¹

This positions prolonged fasting as engaging loss-of-proteostasis and ECM/structural biology, not only deregulated-nutrient-sensing.

Muscle- and bone-sparing TGF-β suppression

Both proteomic studies converge on a coordinated down-regulation of TGF-β-superfamily negative regulators of muscle and bone — a plausible adaptive mechanism to protect lean tissue during starvation ²:

INHBA −3.3 fold, myostatin −2 fold, GDF11/8 −1.6 fold (Commissati 2025); inhibiting myostatin/GDF11 is associated with preserved muscle and increased bone density.
Parathyroid hormone fell 2.1-fold (compensatory bone-sparing), consistent with Pietzner’s finding of no measurable bone-mass loss over 7 days.

This molecular signature aligns with the clinical observation that acute lean-mass loss during a short fast is largely reversible on refeeding — the body appears to actively defend muscle/bone rather than catabolize them indiscriminately.

Amyloid-β reduction (neuro signal)

Commissati 2025 was the first to report that prolonged fasting lowers circulating amyloid-β — plasma Aβ40 and Aβ42 both fell (targeted mass spec), returning to baseline after refeeding, while the clinically validated Aβ42/Aβ40 ratio was unchanged ². Reduced synaptogenesis- and amyloid-fibril-formation pathways, plus a non-significant BDNF-increase trend (+1.32 fold), echo Pietzner’s ketone→neural-ECM axis. Whether lower production or accelerated clearance, and whether it has any Alzheimer’s-relevant benefit, is unknown. no-mechanism

Human evidence

The high-resolution human evidence base rests on three medically-supervised, single-arm, uncontrolled cohorts read end-to-end here: Pietzner 2024 (7-day, Olink), Commissati 2025 (~10-day, SOMAScan), and Dai 2024 (21-day, clinical chemistry). They span 7→21 days and converge strongly — but all lack a control group.

Pietzner et al. 2024 — 7-day fasting plasma proteome (Nat Metab)

Daily plasma proteomics (2,923 targets, Olink Explore 3072) across a 7-day water-only fast in 12 healthy adults, with proteogenomic disease mapping ¹. Key results:

One-third of the plasma proteome (1,044 targets, 35.9%) changed significantly; 144 by >2 SD. See pietzner-2024-fasting-proteome for the full account.
Delayed onset: only 6 proteins significant at 24 h and 54 at 48 h — the systemic response emerges only after 3 days, with the count of decreasing proteins growing exponentially thereafter. 66 proteins were still abnormal 3 days after refeeding.
Distinct from weight loss: proteome change correlated only weakly (r=−0.20) with weight; far more proteins associated with the fasting/ketone signal (n=452) than with weight change (n=49). The dissociation is driven by the profound post-3-day response.
Proteogenomic map: 212 proteins putatively causally linked to ~500 diseases. For 52.2% of predicted protein–disease links, fasting moved proteins in a compensatory (risk-lowering) direction (e.g., SWAP70 ↓ → potential rheumatoid-arthritis relief; HYOU1 ↓ → potential coronary-disease compensation); a minority moved toward higher risk (e.g., coagulation factor XI ↑ → thrombotic risk).

Commissati et al. 2025 — ~10-day water-only fast + refeeding (Mol Metab)

An independent SOMAScan (1,317-protein) study of n=20 older, overweight adults (mean age 52, BMI 28.8) across a mean 9.8-day water-only fast + 5.3-day guided refeeding (Fontana group) ². It is the key replication + counterpoint to Pietzner:

Replicates the signature with “no discrepancies”: 44 decreased + 5 increased proteins overlapped at the 7-day endpoint despite a different platform; FGF19 and soluble leptin receptor increases reproduced. The water-only-PF proteome signature is highly conserved. Only 6.6% of proteins changed, and <1% remained altered after refeeding — confirming the response is largely transient.
But the primary finding is harm-flavored: PF is acutely pro-inflammatory (hsCRP +129%; ↑ hepcidin, ferritin, IL-8, MMP9, midkine), raises liver transaminases (AST +65%, ALT +64%), and activates platelets (urinary 11-dehydro-TXB2 +21%/+36%). See § Safety below — this materially reshapes the risk picture.
Lipids rose during the fast (total/LDL/non-HDL cholesterol, then triglycerides +32% post-refeeding); PCSK9 decreased (−1.49 fold). Oxidative-stress markers did not improve.
Added the amyloid-β reduction and reinforced the TGF-β muscle/bone-sparing signature (above).

Dai et al. 2024 — 21-day fast hypometabolism (Sci Rep)

A pre-registered Chinese space-medicine study (n=13) of a 21-day complete water-only fast — the upper-duration anchor ³. Profound hypometabolic adaptation (resting energy expenditure −20%), deep ketosis (BHB 0.1 → 6.6 mmol/L), ~15% weight loss, and — notably — uric acid roughly doubled (385 → 866 µmol/L, a concrete adverse effect). Blood counts, liver, and cardiac function showed adaptive but not structurally damaging changes: under supervision, 21 days was organ-safe in healthy adults. Framed for survival/spaceflight, not aging — no aging-biomarker or proteomic endpoints.

Relationship to the IF/CR trial evidence

Calorie-matched RCTs of IF and TRE (Trepanowski 2017; Liu 2022) show no benefit beyond caloric reduction when calories are equalized — see intermittent-fasting. Pietzner 2024 does not overturn this for short fasts; rather, it argues the interesting biology lives past the 3-day mark, which calorie-matched IF trials never reach. Whether the >3-day proteome remodeling translates into durable health or biological-age benefit is untested. needs-human-replication

Cross-organism extrapolation

Dimension	Status	Notes
Pathway conserved in humans?	yes	mTOR/AMPK/autophagy/IIS + ketogenesis all operate in humans; this study is human
Phenotype conserved in humans?	partial	Proteome + metabolic adaptations robustly demonstrated; durable health/lifespan effect not measured
Replicated in humans?	partial	The proteome signature replicated across two independent cohorts + platforms (Pietzner Olink, Commissati SOMAScan) with “no discrepancies”; but no controlled trial, and no durable health/lifespan endpoint

Safety and contraindications

Prolonged complete fasting is investigational and not a casual practice — it warrants medical supervision, especially beyond 3–5 days. The 2025 proteomic data have strengthened the safety concerns, showing PF is not a clean anti-inflammatory intervention:

Acute systemic inflammation — contrary to the chronic-CR anti-inflammatory picture (CALERIE; see caloric-restriction), PF raises inflammation: hsCRP +129% in n=20, and a +66.6% rate of significant CRP increase across a 1,422-person validation cohort, regardless of fast duration (5–20 days) ². CRP normalizes on refeeding. Whether this is adaptive “trained immunity” or harmful is unresolved. contradictory-evidence (vs the assumption that fasting is anti-inflammatory)
Platelet activation / thrombotic risk — urinary 11-dehydro-TXB2 (a gold-standard in-vivo platelet-activation + cardiovascular-risk biomarker) rose +21% (fasting) / +36% (refeeding) with no change in platelet count (degranulation, not production); vWF + soluble GP1Bα rose ². This substantiates the procoagulant signal Pietzner inferred genetically (coagulation factor XI rose, aligning with predicted thrombotic risk ¹). Concerning for anyone with existing cardiovascular disease or unstable atherosclerotic plaque.
Hepatic stress — liver transaminases rose (AST +65%, ALT +64%), persisting into refeeding ².
Hyperuricemia — serum uric acid roughly doubled over a 21-day fast (385 → 866 µmol/L) ³; relevant to gout and renal risk.
Transient post-refeeding insulin resistance — glucose and HOMA-IR rose significantly on reintroduction of food ².
Refeeding syndrome — potentially fatal electrolyte shifts (hypophosphatemia, hypokalemia, hypomagnesemia, thiamine depletion) on resuming food after prolonged fasting; requires monitored, gradual refeeding. (Detailed clinical literature not yet ingested — needs-replication for primary citation.) Hypokalemia and arrhythmia were among the severe adverse events reported ².
Lean-mass loss is acute but largely reversible over a short fast, and the TGF-β/PTH signature suggests active muscle/bone defense ¹ ²; risk of durable sarcopenic loss still rises with fast duration and in older adults.
Common mild AEs: headache, weakness, fatigue, insomnia, dry mouth, orthostatic hypotension (prompting transition to broth/juice in ~30% of one cohort) ².
Contraindicated in: type 1 diabetes / insulin or sulfonylurea use (hypoglycemia), pregnancy/lactation, history of eating disorders, underweight/malnourished individuals, frail elderly (sarcopenia risk), and — given the platelet/inflammation/lipid data — caution in established cardiovascular disease.
The high-resolution evidence base is single-arm, uncontrolled, small (n=12–20), and self-selected — though it now spans both lean-young (Pietzner) and older-overweight (Commissati) cohorts, generalizability to multimorbid populations is unestablished.

Open questions

Benefit vs harm, on net — PF simultaneously moves putative-beneficial markers (amyloid-β ↓, muscle/bone-sparing, autophagy ↑) and harm markers (hsCRP ↑, platelet activation ↑, liver enzymes ↑, uric acid ↑, lipids ↑). Does it net-benefit or net-harm cardiometabolic risk, especially in people with subclinical atherosclerosis? contradictory-evidence
Is the acute inflammation adaptive or pathological — “trained immunity” / hormetic vs a genuine pro-atherogenic insult? Resolved only by hard-outcome follow-up. no-mechanism
Is the molecular signature durable, or does it fully reverse on refeeding (Pietzner: 66/1,044 proteins still abnormal at +3 d; Commissati: <1% persist)? long-term-unknown
Are the proteogenomic ‘compensation’ predictions clinically real — do repeated prolonged fasts measurably lower RA/CAD risk, or is this purely cross-sectional inference? no-mechanism
Optimal duration/frequency for any benefit vs the refeeding-syndrome, thrombotic, and lean-mass risks is undefined. dose-response-unclear
Biological-age endpoints (DunedinPACE etc.) have never been measured across a prolonged fast.

Cross-references

pietzner-2024-fasting-proteome — anchor study, 7-day Olink (verified)
commissati-2025-prolonged-fasting-inflammation — ~10-day SOMAScan replication + inflammation/platelet counterpoint (verified)
dai-2024-21day-fasting-hypometabolism — 21-day upper-duration / hypometabolism / hyperuricemia anchor
caloric-restriction — sustained moderate CR; distinct modality, deeper shared mechanism base. Key contrast: chronic CR is anti-inflammatory (CALERIE); prolonged fasting is acutely pro-inflammatory
intermittent-fasting — cyclic ≤48 h fasting; systemic signature emerges past the IF window
time-restricted-eating — circadian-aligned eating window; shallowest fasting depth
ketogenic-diet — shares the sustained-ketosis / 3-hydroxybutyrate axis (incl. the neural-ECM/epilepsy link)
methionine-restriction — amino-acid-specific restriction; CR-signaling overlap
autophagy · mtor · ampk · insulin-igf1 · sirtuin — shared longevity pathways
loss-of-proteostasis · deregulated-nutrient-sensing · chronic-inflammation — hallmark targets
pcsk9 · lpl — fasting-responsive lipid nodes with existing pages

Footnotes

pietzner-2024-fasting-proteome · n=12 (5F/7M, healthy young white-European adults) · in-vivo human, single-arm uncontrolled · 7-day water-only fast; daily plasma proteomics (2,923 targets, Olink Explore 3072) + proteogenomic disease mapping · doi:10.1038/s42255-024-01008-9 · Nat Metab 2024;6(4):764–777 · local PDF verified (PMC7617311) · 1,044 proteins (35.9%) changed (FDR<5%); systemic response only after 3 days; ECM enrichment p<3.6×10⁻⁷; tenascin-R beta=−0.73 p<2.4×10⁻³⁷; proteome–weight correlation r=−0.20; 212 proteins linked to ~500 diseases, 52.2% of links compensatory · model: human ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹
commissati-2025-prolonged-fasting-inflammation · n=20 (11F/9M, mean age 52, BMI 28.8) · in-vivo human, single-arm uncontrolled + two external validation cohorts (incl. n=1,422 Buchinger-Wilhelmi) · mean 9.8-day water-only fast + 5.3-day guided refeeding; SOMAScan 1,317 proteins + targeted MS (amyloid-β) + ELISAs · doi:10.1016/j.molmet.2025.102152 · Mol Metab 2025;102152 · local PDF verified (Fontana group; accepted pre-proof) · 6.6% proteins changed (<1% persist post-refeeding); replicates Pietzner with “no discrepancies”; hsCRP +129% (p=0.0004), 66.6% CRP-rise in validation cohort regardless of duration; urinary 11-dehydro-TXB2 +21%/+36%; AST +65%/ALT +64%; INHBA −3.3/myostatin −2/GDF11/8 −1.6 fold; PCSK9 −1.49 fold; Aβ40+Aβ42 ↓ (ratio unchanged) · model: human ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰
dai-2024-21day-fasting-hypometabolism · n=13 (8M/5F, mean age 40.5, BMI 24.5) · in-vivo human, single-arm, pre-registered (ChiCTR2100048399) · 21-day complete water-only fast (34-day protocol incl. refeeding) · clinical chemistry + indirect calorimetry · doi:10.1038/s41598-024-80049-2 · Sci Rep 2024;14:28550 (PMC11574170) · local PDF, verified:false (abstract+methods read) · weight −14.96%; resting energy expenditure −20.3%; BHB 0.1→6.61 mmol/L; uric acid 385→866 µmol/L; organ-safe to 21 days under supervision · model: human ↩ ↩²

🧬 Aging Wiki

Explorer

prolonged-fasting