← Science

Mechanism overview

Ketosis & short fasts

How a 3–5 day sardine-only intake reaches nutritional ketosis — and what β-hydroxybutyrate looks like across the cycle.

ketosisDossier available

When people talk about "fasting working," they're usually talking about ketosis — though they don't always realize it. The metabolic switch from glucose-and-glycogen fuel to fat-and-ketone fuel is the engine driving most of what people feel and measure during a short fast: the appetite suppression on day two, the mental clarity some people report on day three, the visceral-fat changes over weeks of cycles, the insulin-sensitivity shifts that show up in clamp studies. It is also one of the best-characterized parts of human metabolism — the foundational paper, Cahill 1970, maps the substrate transitions across multi-week starvation in humans with a precision the modern fasting literature still relies on.

This page summarizes what nutritional ketosis is, how short (3–5 day) fasts engage it, and what is distinctive about reaching ketosis on a sardine-only protocol versus a water fast or a long-term ketogenic diet. The applied translation — what ketone numbers to track, when to expect them, what to do if they're not where they should be — is in the ketosis Mechanism Dossier.

What this mechanism is

Ketosis is the metabolic state in which the liver produces ketone bodies — primarily β-hydroxybutyrate (βHB) and acetoacetate — at rates high enough that they become a meaningful systemic fuel. Under normal fed conditions, ketones are present in trace amounts (under 0.2 mmol/L of βHB in blood). "Nutritional ketosis" is conventionally the range from about 0.5 to 3.0 mmol/L of βHB, the level at which brain, muscle, and other tissues begin oxidizing ketones in significant proportion. Diabetic ketoacidosis is a separate, pathological condition reaching 10–25 mmol/L with severe acidosis — it is not what happens during a healthy adult's short fast, and conflating the two is one of the most persistent confusions in popular nutrition writing.

The substrate-level picture is straightforward. The liver stores about 80–120 g of glycogen — perhaps 18–24 hours of fuel at typical resting metabolism, less with exertion. Once liver glycogen is depleted, the liver shifts to gluconeogenesis (producing glucose from amino acids and glycerol) and ketogenesis (producing βHB and acetoacetate from fatty acids delivered by adipose lipolysis). The brain, which cannot oxidize fatty acids, can oxidize ketones — and over the first one to two weeks of fasting, brain ketone uptake can supply 60–70% of cerebral energy demand, sparing the relatively limited gluconeogenic capacity for glucose-obligate tissues. This is the picture mapped meticulously in Cahill's classic paper and confirmed in subsequent isotope-tracer work.

What makes ketones interesting biologically isn't only that they are an emergency fuel. β-hydroxybutyrate also acts as a signaling molecule — it is an HDAC (histone deacetylase) inhibitor, modulates inflammasome activation, and binds receptors that regulate inflammation and oxidative stress. Some of the proposed long-term benefits of ketosis (anti-inflammatory effects, oxidative stress reduction, neuroprotection in animal models) are likely mediated by βHB-as-signal rather than only βHB-as-fuel. Veech 2004 is the canonical review of this signaling biology.

How short fasts engage it

Most healthy adults reach nutritional ketosis (≥ 0.5 mmol/L βHB) somewhere between hour 24 and hour 72 of a fast, depending on:

  • Glycogen status going in. A person who has been low-carb for the prior week enters ketosis within 24 hours of starting a fast. Someone coming off a high-carbohydrate diet may take 48–72 hours.
  • Activity level during the fast. Light exercise depletes glycogen faster and accelerates ketone production. Sedentary fasting prolongs the transition.
  • Body composition. Higher lean mass means more glycogen to burn through. People with very low body fat have less substrate for sustained ketogenesis and may not reach the upper range of nutritional ketosis even on extended fasts.
  • Individual variation. βHB at a given fasting hour varies substantially between people — what's "normal" at 60 hours can be 0.6 mmol/L in one healthy adult and 2.4 mmol/L in another.

A 3-day fast typically peaks somewhere in the 1.5–3.5 mmol/L range by the end of day three. A 5-day fast adds another half-day to a day at the upper end of that range, with βHB often plateauing or modestly declining as adipose lipolysis stabilizes. The de Cabo & Mattson 2019 NEJM review frames this as the "metabolic switch" — once you've crossed it repeatedly across cycles, the switch becomes faster and the metabolic adaptations more pronounced. The Mattson 2017 review argues this switching itself, not just the absolute ketone level, is what produces neurological adaptation in animal models.

The most relevant short-fast human data comes from intermittent fasting protocols rather than continuous multi-day fasts. Halberg 2005 ran 20-hour fasts every other day for 15 days in healthy men and saw improved insulin sensitivity (the ketosis-and-insulin connection). Brandhorst & Longo's 2015 fasting-mimicking diet protocol hits ketosis with very-low-calorie eating for 5 days each month — a closer cousin to the Sardine Protocol's structure than a true water fast.

How sardine fasting specifically engages this mechanism

Sardine fasting is, mechanistically, a protein-sparing modified fast (PSMF) with omega-3 enrichment. Unlike a water fast — where the body is sourcing protein from muscle and viscera to supply gluconeogenic substrate — a sardine fast supplies dietary protein at maintenance-adjacent levels (typically 80–120 g/day across 4–6 cans of sardines). This has two consequences for ketosis:

It slows entry into ketosis modestly. The dietary protein triggers a small insulin response and supplies amino acids that fuel hepatic gluconeogenesis directly, reducing the demand for fatty-acid-derived acetyl-CoA flowing into ketone production. In practice, this means an experienced sardine faster typically reaches βHB ≥ 0.5 by the end of day two rather than mid-day two — roughly half a day later than a water fast.

It produces a more sustainable ketosis. Because muscle is being protein-spared, the body is not burning lean mass to fund gluconeogenesis. Adipose lipolysis becomes the dominant fuel source, and the βHB plateau on day 3–5 of a sardine fast often runs in the 1.5–2.5 mmol/L range — enough for metabolic effect, low enough to feel functional. This is the pattern Bistrian and Blackburn aimed for when they originally designed the PSMF protocol in the 1970s, and it's largely the pattern that holds today, though with sardines as the protein vehicle rather than the original liquid formulas.

The omega-3 angle is mechanistically separate from ketosis but co-occurs with it. EPA and DHA reduce inflammatory markers and may modestly improve mitochondrial efficiency at the cellular level. During a fast, when ketones are signaling anti-inflammatory effects through their HDAC inhibition, the sardine-derived omega-3 layer adds a second, additive anti-inflammatory input. The Mechanism Dossier covers this co-engagement in detail.

Where the evidence is strongest, weakest

Strong: Cahill's substrate-level mapping of human starvation physiology is decades-old, well-replicated, and not seriously contested. The signaling biology of βHB (Newman & Verdin 2014, Veech 2004) is well-characterized in vitro and in animals. Short-term human ketogenic-diet outcomes — weight loss, glycemic improvement in T2D — are supported by clinical work like Westman 2008, Volek 2009, and the Virta Health 2-year cohort data in Hallberg 2018.

Moderate: Long-term cardiovascular outcomes on ketogenic diets remain debated. Most data are 12-week to 2-year studies; mortality endpoints don't exist. Athletic-performance effects of ketogenic adaptation are mixed — Phinney's 1983 endurance cyclists study suggested adaptation preserved sub-maximal endurance, but the Burke 2017 race walkers study showed clear performance impairment at race-pace efforts. Whoever cites only one of these studies is making a partial argument.

Weak: Human RCT data specifically on 3–7 day sardine-only protocols is essentially nonexistent. The case for the Sardine Protocol's specific cycle structure leans on PSMF research, fasting-mimicking-diet research, and short-term ketogenic-diet research — adjacent literatures that are individually well-developed but don't combine into a single trial of "the protocol as such."

What this means for your cycle

A short summary lives here on the public page; the detailed application — βHB targets per cycle day, what to do if you're not reaching ketosis, how to interpret numbers across multiple cycles, and how the Sardine Protocol's quarterly 21-day Transformation handles sustained ketosis differently than a monthly 5-day cycle — is in the Inner Circle Mechanism Dossier.

In broad strokes:

  • Expect βHB ≥ 0.5 mmol/L by end of day 2 if your prior week was moderately low-carb; later if you came off a high-carb week.
  • Expect a peak in the 1.5–2.5 mmol/L range by day 3–5.
  • If you don't reach 0.5 mmol/L by mid-day three, something is wrong — see the dossier troubleshooting guide.
  • Track βHB at the same time of day each cycle for trend-line comparison; absolute numbers vary across labs and meters.

Open questions

  • The human evidence for βHB as a signaling molecule (anti-inflammatory, neuroprotective) is largely extrapolated from animal models and in vitro work. Human dose-response for the signaling effects is not characterized.
  • We don't know with confidence how much of the ketosis-attributable benefit in short fasts is the βHB level itself versus the calorie deficit versus the metabolic-switching frequency. Studies that isolate these factors are hard to design and rare.
  • Whether repeated short ketogenic exposures produce durable insulin-sensitivity or mitochondrial-efficiency gains beyond the fasting window is genuinely open. The closest evidence — Halberg 2005, Sutton 2018 — suggests yes, but these are small studies.
  • Individual variation in ketogenic response is large and not well predicted by baseline characteristics. Some healthy adults plateau at 1.0 mmol/L on day 4 of a fast; others reach 3.5. We don't know why with confidence.

Top sources for this mechanism

The strongest evidence in our library for ketosis & short fasts, by tier and recency. Browse the full library for the long tail.

Tier 2 · Peer-reviewed secondaryreviewstrong

Cahill GF · 1970 · New England Journal of Medicine

George Cahill's 1970 NEJM review remains the single most important paper ever written on human starvation metabolism. Drawing on his lab's careful in-patient studies of obese volunteers undergoing therapeutic fasts (then a common obesity treatment), Cahill mapped the day-by-day fuel transitions that allow humans to survive weeks-to-months of food deprivation: the shift from glucose to fatty acid oxidation in muscle within hours of the last meal, the rise of hepatic ketogenesis over the first few days, and — most consequentially — the progressive switch by the brain from preferring glucose to preferring β-hydroxybutyrate and acetoacetate as primary fuels. This brain-ketone adaptation is what protects body protein. Without it, prolonged fasting would deplete muscle within days through gluconeogenesis demand; with it, daily protein loss falls to a trickle, fat becomes the dominant fuel, and survival extends to the limits of fat reserves. The paper identifies insulin as the principal regulatory hormone of the transitions and remains the foundational citation for almost every modern paper on fasting physiology.

Richard Veech's 2004 review is the most-cited mechanistic argument that ketone bodies — specifically D-β-hydroxybutyrate — are not just an alternative fuel but a more efficient one in metabolic terms. Veech's central claim is that the enthalpy of D-β-hydroxybutyrate combustion is higher per unit oxygen consumed than glucose, meaning more ATP per oxygen molecule. He uses this thermodynamic observation to argue that mild ketosis may be therapeutically useful in conditions where mitochondrial efficiency is compromised: insulin resistance, neurodegeneration, ischemia, and certain rare metabolic disorders. The review covers redox state changes during ketosis (favorable shifts in NAD+/NADH), the role of free fatty acid elevation alongside ketones in ketogenic-diet states, and the activation of PPAR signaling. Veech's framing seeded the modern field of "exogenous ketones as therapy" and is widely cited in research on ketogenic diets for epilepsy, Alzheimer's disease, and traumatic brain injury. The therapeutic claims are speculative for many of the listed conditions; the underlying biochemistry is rigorous.

Tier 2 · Peer-reviewed secondaryreviewstrong

Newman JC & Verdin E · 2014 · Trends in Endocrinology and Metabolism

This Trends in Endocrinology and Metabolism review reframes how the body uses ketone bodies — particularly β-hydroxybutyrate (βOHB) — beyond their traditional role as fuel. Newman and Verdin synthesize evidence that βOHB acts as a signaling molecule through at least two mechanisms. First, βOHB binds at least two cell-surface G-protein-coupled receptors (HCAR2/GPR109A and FFAR3/GPR41), modulating lipolysis, sympathetic tone, and metabolic rate. Second, βOHB directly inhibits class I histone deacetylases (HDACs), which means circulating ketones during fasting or ketogenic diets alter gene expression by changing how DNA is packaged. The review traces implications for caloric restriction, longevity, and aging-related diseases. The paper is a key citation for any claim that ketogenic diets and fasting do work beyond "running on fat instead of carbs" — they trigger gene-expression changes via epigenetic mechanisms with downstream effects on stress resistance, inflammation, and metabolic flexibility. The review is highly cited and has shaped how mechanistic ketosis research is framed.

Five well-trained cyclists ate their usual mixed diet for one week, then switched to a ketogenic diet — under 20 grams of carbohydrate per day — for four weeks. Calories and protein were matched between both diets; only the fuel source changed. After four weeks of ketosis, the cyclists could ride to exhaustion just as long as before (about 150 minutes), and their peak aerobic capacity (VO2max) was unchanged. What did change was where the energy came from. At the same exercise intensity, the body burned roughly three times less glucose and four times less muscle glycogen. The respiratory quotient — the ratio that tells you whether you're burning carbs or fat — dropped from 0.83 (mostly carbs) to 0.72 (almost entirely fat). The study was an early demonstration that humans can stay in ketosis for weeks and still perform endurance work, drawing energy almost entirely from fat and ketones.

This is the largest published study of sustained nutritional ketosis as a T2D management strategy. The Virta Health study enrolled 349 adults with type-2 diabetes — 262 in the continuous care intervention (CCI, an app-mediated remote-care program with macronutrient guidance toward sustained nutritional ketosis) and 87 in usual care. The design was open-label and non-randomized (participants self-selected into the intervention), so it sits below DiRECT's RCT evidence in the hierarchy — but the sample is larger and the duration is longer. At one year, the intervention group's HbA1c fell from 7.6 to 6.3 percent (the threshold for diabetes remission), mean weight loss was 13.8 kg, and 94 percent of insulin users reduced or eliminated insulin therapy. Sulfonylureas were discontinued completely in the CCI group. Secondary markers improved across the board: HOMA-IR dropped 55 percent, hsCRP dropped 39 percent, triglycerides dropped 24 percent, HDL-C rose 18 percent. The usual-care arm showed no meaningful change on any of these endpoints.

Tier 1 · Peer-reviewed primaryrctmoderate

Westman EC et al. · 2008 · Nutrition & Metabolism

This 24-week randomized controlled trial enrolled 84 adults with obesity and type-2 diabetes, randomly assigning them to either a low-carbohydrate ketogenic diet (under 20 g of carbs per day, ad-libitum protein and fat) or a low-glycemic-index reduced-calorie diet (a 500 kcal/day deficit, ordinary macronutrient distribution). Of 84 enrolled, 49 completed the protocol — typical attrition for an outpatient diet trial. The headline results favored ketogenic restriction. HbA1c dropped 1.5 percentage points on the ketogenic diet versus 0.5 points on the low-GI diet (p=0.03). Weight loss was 11.1 kg on the ketogenic arm versus 6.9 kg on the low-GI arm (p=0.008). The most striking endpoint was medication change: 95 percent of ketogenic-arm participants either reduced or eliminated their diabetes medications, compared to 62 percent on the low-GI arm (p less than 0.01). HDL cholesterol improved on the ketogenic diet (+5.6 mg/dL) and was unchanged on low-GI. The trial is one of the foundational small RCTs that established sustained nutritional ketosis as a viable T2D management strategy.

Tier 1 · Peer-reviewed primaryreviewstrong

de Cabo R & Mattson MP · 2019 · New England Journal of Medicine

This NEJM review summarizes evidence that intermittent fasting regimens — alternate-day fasting, time-restricted eating, and periodic multi-day fasts — engage a "metabolic switch" from glucose-derived energy to fat- and ketone-derived energy after hepatic glycogen is depleted, typically within 12–36 hours of fasting depending on the individual and the protocol. The authors argue that repeated exposure to this switch produces adaptive responses across organ systems, including improved insulin sensitivity, reduced inflammation, increased mitochondrial biogenesis, enhanced autophagy, and improved stress resistance in cells. The review compiles findings from animal models alongside the available human trials at the time of publication. The review notes that, despite preclinical signals being strong and consistent, the human evidence base is more heterogeneous: the largest gains in metabolic markers (fasting insulin, HOMA-IR, lipid profile, inflammatory markers) appear in adults with obesity or metabolic syndrome, while effects in lean, metabolically healthy individuals are smaller. The authors flag practical issues — adherence over months, the early-fast hunger and irritability phase, and the lack of long-term outcome data — as the main barriers to clinical adoption rather than safety in healthy adults.

Tier 1 · Peer-reviewed primaryrctmoderate

Brandhorst S et al. · 2015 · Cell Metabolism

This Cell Metabolism paper from Valter Longo's USC group introduced the fasting-mimicking diet (FMD) — a 5-day periodic dietary protocol designed to deliver fasting's molecular benefits while keeping participants able to consume modest amounts of plant-based food. The paper has two parts. In aged mice, monthly FMD cycles for several months produced multi-system regeneration: hippocampal neurogenesis rose, IGF-1 dropped, PKA activity decreased, NeuroD1 expression increased, and cognitive performance improved on standard mouse cognition tests. In a 38-participant pilot human RCT, three monthly FMD cycles (each 5 days) produced reductions in body weight, body fat, blood pressure, fasting glucose, and IGF-1 without significant adverse events. The paper is foundational because it bridged rodent CR research and practical human protocol design — providing a structured, safe framework for delivering fasting benefits without continuous calorie restriction. Longo subsequently commercialized the protocol as ProLon, a packaged 5-day FMD product. The paper's data quality is solid but the commercial development complicates how it should be cited.

Tier 2 · Peer-reviewed secondaryreviewstrong

Mattson MP et al. · 2018 · Nature Reviews Neuroscience

This Nature Reviews Neuroscience paper from Mark Mattson — the most cited researcher on fasting and brain health — synthesizes the case that periodic shifts between fed and fasted metabolic states are essential for optimal brain function. Mattson coined the term "intermittent metabolic switching" (IMS) for the pattern: eating depletes liver glycogen, fasting forces ketone production, and the cycle repeats. The review argues this oscillation is what humans evolved with, and that modern continuous-feeding patterns disrupt it with cognitive and neurological consequences. The mechanistic story focuses on β-hydroxybutyrate (BHB), which is transported into neuronal mitochondria as fuel but also acts as a signaling molecule. BHB induces brain-derived neurotrophic factor (BDNF), which promotes synaptic plasticity, neurogenesis in the hippocampus, and resistance to neuronal injury. Mattson reviews evidence connecting IMS to improved cognition, mood regulation, motor performance, autonomic-nervous-system function, and resistance to neurodegenerative disease. The framework has shaped subsequent fasting-and-brain-health research and is heavily cited in popular literature on fasting's cognitive benefits.

Tier 1 · Peer-reviewed primaryrctmoderate

Volek JS et al. · 2009 · Lipids

This 12-week randomized trial compared a carbohydrate-restricted diet (12 percent carb / 59 percent fat / 28 percent protein) with a low-fat diet (56 percent carb / 24 percent fat / 20 percent protein) in 40 adults with atherogenic dyslipidemia — the metabolic-syndrome phenotype defined by high triglycerides, low HDL, central adiposity, and insulin resistance. Both diets were calorie-restricted to similar levels. Both produced improvements, but the carbohydrate-restricted arm consistently outperformed the low-fat arm across nearly every endpoint that defines metabolic syndrome. Glucose dropped 12 percent in the carb-restricted group; insulin fell 50 percent; insulin sensitivity improved 55 percent; body weight dropped 10 percent; adiposity dropped 14 percent. The lipid panel was the most striking divergence: triglycerides fell 51 percent on carb restriction (versus a smaller drop on low-fat), HDL rose 13 percent (versus no change), and the total-cholesterol-to-HDL ratio improved 14 percent more on carb restriction. The paper's interpretation is that the metabolic syndrome is fundamentally a carbohydrate-intolerance phenotype, and that restricting carbs addresses the upstream driver more directly than restricting fat does.

The Sardine Reset cheat sheet

Free 6-page printable PDF — protocol overview, brand shopping list, day-by-day, refeed rules, and warning signs. Plus the weekly research roundup.

One email when it’s ready. Unsubscribe anytime. We never sell your data.