Metabolic flexibility
The capacity to switch fuels — burning glucose when you're fed, fat when you're not — is one of the clearest markers of metabolic health, and one of the first things to break under sedentary, overfed living. It is a better gauge than a target: the habits that improve it (regular exercise, keeping fat out of liver and muscle, holding on to muscle) pay off anyway, while the fashionable shortcuts — time-restricted eating, permanent ketosis, mitophagy supplements — test worse than their reputations.
Metabolic flexibility is the body's ability to burn fat and carbohydrate interchangeably, depending on what's available. The formal definition — switching from mostly fat-burning during a fast to suppressed fat-burning with increased glucose uptake and storage once insulin rises — dates to the late 1990s and builds on the Randle cycle, in which rising fat oxidation reciprocally suppresses glucose oxidation.[1][2] A flexible person glides between burning glucose after a meal and burning fat between meals; an inflexible person stays stuck on glucose, fails to fully mobilise fat during fasting, and develops the classic phenotype of hyperinsulinemia, visceral and ectopic fat accumulation, and insulin resistance. Inflexibility is at minimum a robust marker of cardiometabolic ill-health — but whether it is a cause or mostly a consequence of insulin resistance is genuinely unresolved (below), and there is no single standardised definition or measurement protocol.
What it is, and what it isn't
The clinical proxy for fuel-switching is the respiratory exchange ratio (RER) — the volume of carbon dioxide produced divided by the volume of oxygen consumed during indirect calorimetry. An RER near 1.0 indicates near-pure carbohydrate oxidation; an RER near 0.7 indicates near-pure fat oxidation. A metabolically flexible person shows a wide swing in RER between fasted and fed states. An inflexible person shows a narrow swing — stuck on glucose regardless of feeding status.[3]
On average, the blunted RER pattern is the laboratory picture of metabolic inflexibility in people with severe obesity, type 2 diabetes, or significant insulin resistance: they can't fully suppress fat oxidation when glucose is abundant, and they can't fully ramp it up when glucose is scarce.[4] Their muscle over-relies on glucose in the post-absorptive state, when insulin is low and it should be burning fat — and then fails to raise glucose oxidation when insulin does rise. Their adipose tissue keeps releasing fatty acids when insulin should be shutting lipolysis down, and stores less of each successive meal's fat than lean tissue does — so those fatty acids stay in circulation and land ectopically, in liver and muscle.
Why it matters: the unifying-biomarker case
Loss of metabolic flexibility travels with the whole cardiometabolic cluster, not just insulin resistance. No prospective cohort has yet shown that a measured flexibility index predicts incident diabetes, cardiovascular disease or fatty liver, so read the table below as company kept rather than risk forecast.[5]
| Outcome | Link to inflexibility |
|---|---|
| Insulin resistance and type 2 diabetes | Associated, and linked mechanistically via ectopic lipid intermediates (below); causal direction contested — see Cause or consequence? — though the clamp-measured gap between type 2 diabetes and simple overweight is small and looks mostly like a body-mass effect[6] |
| Visceral and ectopic adiposity (liver, pancreas, muscle) | The downstream physical signature of unable-to-oxidise-fat tissue |
| Metabolic dysfunction-associated steatotic liver disease (MASLD, formerly NAFLD) | Hepatic fuel handling is deranged rather than switched off — fasting mitochondrial oxidative flux runs about twice normal in simple steatosis[7], and respiratory capacity falls only once steatohepatitis develops[8] |
| Heart failure with preserved ejection fraction (HFpEF) | HFpEF myocardium is metabolically inflexible, but the lesion is in glucose oxidation, not fat-burning — direct substrate tracing finds fatty-acid oxidation maintained or increased and glycolysis unchanged, and the driver is obesity and diabetes rather than chronological age[9] |
| Sarcopenia | Mitochondrially stressed muscle secretes stress signals, and one of them — GDF15 — prospectively tracks frailty and sarcopenia in older adults[10]. Calling them drivers of aging goes further than the evidence: the tissue of origin is unresolved, and FGF21 extends lifespan when overexpressed in mice[11] |
| Accelerated biological aging | Chronic inflexibility tracks reduced mitochondrial-biogenesis capacity; the specific claim that PGC-1α hypermethylation accelerates epigenetic clocks is not established |
The interventions that improve one — energy deficit, exercise, adequate sleep — tend to improve the rest, which is much of why flexibility reads as a unifying marker. Whether the shared AMPK/SIRT1/PGC-1α machinery is the reason is a separate and unsettled question.
Cause or consequence?
Evidence rating: Caution — the direction of causation is unsettled. One camp (Kelley; Muoio's "metabolic gridlock") reads inflexibility as an early driver: impaired substrate switching appears early in glucose intolerance and in isolated mitochondria, and incomplete fatty-acid oxidation with acylcarnitine accumulation can precede overt insulin-signalling defects.[12] The opposing camp reads it as mostly a consequence: in 59 people with type 2 diabetes and 42 non-diabetic controls matched for obesity, sex and race, the clamp fuel swing was smaller in diabetes (ΔRQ 0.06 versus 0.10) — but the gap disappeared once insulin-stimulated glucose disposal was accounted for, and glucose disposal alone explained 46% of the variance in flexibility.[13] A Shulman-lab study then dissociated the two outright, concluding that altered mitochondrial substrate preference is "not an essential step in the pathogenesis of muscle insulin resistance" and that drugs aimed at it would likely fail.[14] The defensible stance: flexibility is best treated today as a biomarker of metabolic health; its status as an independent therapeutic target — changing fuel selection without raising energy demand the way exercise does — remains unproven.
The cellular mechanism
Three nutrient-sensing kinases sit at the centre of metabolic flexibility: AMP-activated protein kinase (AMPK), the mechanistic target of rapamycin (mTOR), and the SIRT1 deacetylase. They integrate systemic energy status into cellular decisions about whether to grow, repair, store, or break down.[15]
- AMPK is the catabolic master switch. Activated by a rising AMP/ATP ratio (fasting, exercise, energy deficit), it stops anabolic processes, ramps fatty-acid β-oxidation, increases glucose uptake into skeletal muscle, and suppresses mTORC1.[16]
- mTOR (complex 1, mTORC1) is the anabolic master switch. Activated by amino acids and insulin, it drives protein synthesis and lipogenesis. AMPK shuts it off directly, by phosphorylating one of its scaffold proteins, while chronic overfeeding keeps it constitutively on.[17]
- SIRT1 is a nicotinamide adenine dinucleotide (NAD⁺)-dependent deacetylase. AMPK raises the NAD⁺ pool that SIRT1 needs to function, at least in rodent muscle.[18] Active SIRT1 deacetylates PGC-1α — the master transcriptional co-activator of mitochondrial biogenesis — and the FOXO transcription factors that regulate stress resistance and longevity.
When fasting, exercise, or caloric restriction periodically activate AMPK and SIRT1, the system cycles through energy production, cellular cleanup (autophagy), and mitochondrial renewal. Sustained overfeeding and sedentary living keep AMPK quiet and mTOR on, and the oscillation flattens. Mitochondrial biogenesis slows, autophagy is thought to be impaired, and lipid intermediates accumulate in the wrong compartments.[19] That flattened oscillation is the usual molecular picture of inflexibility.
The lipid-intermediate problem
When tissues can't oxidise fatty acids efficiently, free fatty acids accumulate inside cells as diacylglycerols and ceramides. These intermediates aren't inert, but they act through a narrower route than a generic block: the diacylglycerol pool sitting at the cell membrane activates a family of protein kinase C enzymes, which in the liver phosphorylates the insulin receptor itself and blunts its signalling, shown in rodents.[20] In muscle a related enzyme, protein kinase C-theta, is thought to act on insulin receptor substrate 1, though which phosphorylation site matters is uncertain.[21] Ceramides do not share that route, and whether they are causal at all is a live dispute.[22] The pancreas compensates by secreting more insulin, while insulin-resistant fat tissue keeps releasing fatty acids that land in muscle and liver. That vicious cycle is a large part of the mechanism of insulin resistance, but not all of it — how much lipid spills into liver and muscle in the first place is gated by the storage capacity of subcutaneous fat, which is why people with lipodystrophy are severely insulin resistant with almost no fat at all.[23]
Mitochondrial dynamics: fusion, fission, and "megamitochondria"
Mitochondria aren't static bean-shaped organelles. They exist as dynamic networks that fuse (joining together to share contents and dilute damage) and undergo fission (splitting off damaged segments for clearance). Damaged segments get isolated, tagged with ubiquitin by a sensor-and-effector pair that detects failing protein import, and then degraded through mitophagy — a specialised autophagy that selectively recycles broken mitochondria.[24] AMPK's role is upstream: it drives the splitting and starts the recycling machinery, but it is not the tag.
In aged or chronically overfed tissue the balance tips rather than both arms failing: in human skeletal muscle, chronic substrate excess increases splitting and leaves a fragmented, glycolytic-leaning network, while the fusion arm — the one associated with high fat oxidation — is what falls.[25] The "megamitochondria" framing — oversized, dysfunctional mitochondria as the signature of inflexibility — is illustrative at best; the better-supported substrate-switching machinery is the Randle cycle, pyruvate dehydrogenase and the CPT1/carnitine shuttle described below.
The substrate-competition hub
At the biochemical core, fat and glucose compete for oxidation. The Randle cycle and pyruvate dehydrogenase (PDH) form the switch: rising fatty-acid oxidation inhibits PDH and suppresses glucose oxidation. Fatty acids enter the mitochondrion through CPT1 and the carnitine/acylcarnitine shuttle — the rate-limiting gateway. When β-oxidation runs incompletely, partially oxidised acylcarnitines accumulate ("mitochondrial overload")[26] — a rodent-derived model that human muscle-biopsy work has not reproduced, finding no association between muscle acylcarnitine content and insulin resistance.[27] Branched-chain amino acid (BCAA) metabolites are proposed to further congest the pathway — restricting them in obese rats normalised muscle fatty-acyl-CoA accumulation and improved clamp insulin sensitivity.[28] Human genetics points the other way, though: in about 250,000 UK Biobank participants, raised circulating BCAAs looked like a consequence of poor metabolic health rather than a cause.[29]
Upstream, adipose tissue is decisive: when insulin fails to suppress lipolysis, the resulting high free-fatty-acid load forces muscle toward fat oxidation and blunts its switch — adipose insulin responsiveness sets the flexibility of other organs.[30] The liver is its own flexibility site, with fuel handling deranged in fatty liver disease (see the table above) — a state that bile-acid signalling also shapes (see Bile acids and metabolism).
What drives the loss
- Chronic caloric surplus. Sustained excess fills glycogen stores, then forces fat into ectopic deposition. Without periodic energy deficit, AMPK doesn't activate and mTOR doesn't shut off.
- Sedentary behaviour. Mitochondria adapt to demand, and withdrawing it is measurable — though the first week is inconsistent: four days of strict bed rest cut citrate synthase activity, the standard marker of mitochondrial content, by about 18% in young men[31], five days in older men moved neither citrate synthase, the electron-transport complexes, nor the fission and fusion proteins[32], and ten days reduced both mitochondrial respiration and content markers in older adults.[33] The mitochondrial change may be a consequence rather than the trigger: a 2024 human bed-rest study places it downstream of nutrient overload rather than ahead of it.[34] The everyday version is milder and reversible: cutting about 10,000 steps a day for two weeks lowered whole-body and muscle insulin sensitivity, lowered fitness and raised liver fat — and nearly all of it reversed within a fortnight of resuming normal activity — only people with a close relative with type 2 diabetes still had lower muscle insulin sensitivity.[35]
- Constant grazing. Eating across a long window shortens the uninterrupted post-absorptive stretch in which lipolysis and fat oxidation run. Half of free-living adults already eat across a window longer than about fifteen hours[36] — a description of how people eat, not a harm threshold. The popular mechanism, chronically raised insulin, does not hold: when the same food was spread over seventeen snacks a day instead of three meals, mean serum insulin fell by about a quarter, with free fatty acids and ketones unchanged.[37] That was seven men, and it contrasts meal frequency rather than window length, so it refutes the stated mechanism without showing a long window is harmless. See Fasting.
- Refined-carb-heavy ultra-processed diet. Drives overeating and repeated large glucose and insulin excursions, on top of additive harms from emulsifiers and processing — see Ultra-processed food.
- Sleep loss. Short sleep lowers clamp-measured insulin sensitivity: one night of four hours cut it by about 25% in nine healthy adults[38], and a week of five-hour nights cut it by about 11% in twenty healthy men[39] — and catching up afterwards does not reliably undo it (see Sleep and circadian alignment below). The detail belongs to the dedicated Sleep article.
- Aging itself. Independent of behaviour, mitochondrial quality control deteriorates with chronological age — but the trajectory is highly modifiable.[40]
The behavioural inputs and the biological aging trajectory overlap. Chronic overfeeding produces a phenotype that looks like accelerated aging — ectopic lipid, blunted energy-sensing and an inflammatory baseline — even in chronologically young adults. The parallel is looser than it sounds: in the one randomised human overfeeding trial to measure it, five days of high-fat overfeeding lowered muscle PGC-1α expression only in men born at low birth weight, and the promoter methylation it induced was reversible and uncorrelated with expression.[41]
What restores it
Exercise (the best-supported single lever)
- Zone 2 aerobic training is a potent stimulus for mitochondrial biogenesis and fat-oxidation capacity — though not a uniquely potent one: when training weeks and starting fitness are held constant, easy, hard and sprint training raise mitochondrial content by similar amounts (about 23%, 27% and 27%) and capillaries per fibre by 15%, 13% and 10% respectively.[42] The first lactate threshold is properly the first rise above your own resting baseline rather than a fixed number; in studies of people with obesity, measured lactate at the fat-oxidation optimum spans roughly 1.3 to 2.7 mmol/L.[43] Working below it shifts the fuel mix toward fat without ever making it exclusive — in trained cyclists fat supplies about 55% of energy at 40% of maximal workload and only 24% at 75%.[44] A 10-week aerobic intervention in type 2 diabetes, assessed by 24-hour whole-room calorimetry, improved one specific feature of flexibility — the post-breakfast rise in respiratory quotient — though in a single-arm study of 23 adults with no untrained comparator.[45] See Zone 2 training.
- Resistance training matters because muscle is the body's largest insulin-stimulated glucose-disposal tissue. Training it lowers fasting insulin, fasting glucose and HbA1c in adults with type 2 diabetes and overweight, on very-low-to-moderate certainty evidence.[46] The popular mechanism — build more muscle, get a bigger sink — is the part the evidence does not confirm: across 33 training studies, gains in fat-free mass did not consistently predict the improvement in glucose handling.[47] See Resistance training.
- High-intensity interval work may shift fuel switching a little further than easy work does — in a short energy-matched crossover in young men with obesity, interval training produced a larger fuel swing across a glucose load, while insulin sensitivity and glucose tolerance improved equally in both arms.[48] The longer trials do not separate them[49] — in one, interval training did no better than a no-exercise control for clamp-measured flexibility in women with overweight or obesity[50] — and the fusion machinery is not interval-specific: in type 2 diabetes the main fusion protein rose across moderate aerobic, interval and resistance training alike.[51] See VO₂ max.
Combining an easy aerobic base, resistance work and a weekly hard session is a sensible default — but no head-to-head trial has shown that combination beating any single modality for metabolic flexibility. "Polarised" properly names a distribution of endurance intensity, roughly 80% easy and 20% hard, not the addition of strength training; where it has been compared directly the evidence is about athletic performance, and the only edge was a small aerobic-capacity advantage confined to highly trained athletes.[52]
Eating pattern
- Adequate protein preserves muscle and lowers the post-meal glucose excursion when it displaces refined carbohydrate — though protein added to a carbohydrate meal cuts glucose while raising insulin, because several amino acids stimulate the insulin-secreting beta cell directly.[53] See Protein.
- Lower glycaemic load meals flatten the post-meal glucose curve. That this improves insulin resistance is plausible on mechanism but unproven — in a five-week controlled-feeding crossover in 163 overweight adults, lowering glycaemic index at matched carbohydrate content actually worsened insulin sensitivity.[54] See Glycemic index.
When you eat looks more useful than how long you fast. Pooling 41 randomised trials in 2,287 people, eating early rather than late — finishing the last meal by late afternoon or early evening — lowered body weight and fasting insulin at high certainty, while the length of the eating window showed no consistent effect.[55] Time-restricted eating specifically is weaker than its reputation for this endpoint. An early window raises overnight fat oxidation in short chamber studies without weight loss — but in eleven people over four days, and the entire gain came at night, with daytime and post-meal values unchanged.[56] The claim does not survive the one trial that measured flexibility properly, with a clamp: over 12 weeks an 8-hour eating window produced less metabolic flexibility than a 15% calorie-restriction arm (about 9% achieved) and no gain over eating freely — a small difference, but pointing the wrong way.[57] That trial used a self-selected window that averaged 9.8 hours, so it does not test the early protocol directly. A small early-window trial in prediabetic men improved insulin sensitivity and blood pressure without weight loss — but in only eight completers, with glucose unchanged at every point of the tolerance test and fasting triglycerides and total cholesterol rising.[58] It has not generalised: pooling ten randomised trials in 599 adults with prediabetes or type 2 diabetes, the effects of time-restricted eating on insulin-sensitivity markers were inconclusive.[59] The larger early-window trials since, including a twelve-month randomised trial, found metabolic risk factors no better than matched calorie restriction.[60] See Fasting and time-restricted eating.
A note on the ketogenic diet: fat-adaptation enhances fat-burning, but it is unidirectional. Ultra-endurance athletes who had chosen a low-carbohydrate diet for an average of twenty months showed peak fat oxidation roughly 2.3-fold higher than habitually high-carb athletes (1.54 vs 0.67 g/min, at a higher exercise intensity), with fat supplying 88% versus 56% of submaximal fuel.[61] That study is reported to have been supported by Quest Nutrition and the Atkins Foundation, with authors receiving book royalties. The gain is one-way: keto-adaptation reduces the ability to rapidly upregulate carbohydrate oxidation, the other half of true flexibility. In a crossover in seven trained cyclists — after only five days of high-fat eating, not long-term keto-adaptation — a day of carbohydrate refeeding equalised muscle glycogen, yet the cyclists still oxidised about 30% less carbohydrate, with the enzyme that gates glucose oxidation blunted at rest, during exercise and during an all-out sprint.[62]
The cardiovascular concern about permanent ketosis is not an animal-model extrapolation. Pooling 47 randomised trials in 3,450 adults aged 35–75 with at least one cardiovascular risk factor, ketogenic diets raised LDL cholesterol by about 0.43 mmol/L against a low-fat comparator, alongside higher HDL; triglycerides fall in the keto-only pooled analyses.[63] What is missing is hard-outcome data rather than human data, and the pooled certainty for the LDL signal is graded low.[64] The cardiac-fibrosis findings do remain confined to rodents. In fatty liver disease, randomised trials show improved weight and insulin resistance with limited short-term effect on liver fat or enzymes, rather than harm.[65] Cyclical ketosis is often proposed as the safer route, but no trial has compared cyclical with continuous ketosis in people, and the two are not obviously equivalent — twelve weeks of ketogenic eating lowered skeletal-muscle AMPK protein, the very switch fasting is supposed to activate.[66]
Body composition
Visceral and ectopic fat reduction is the most direct lever in adults with excess belly fat, and the benefit outlasts the weight loss: five and ten years after two lifestyle trials, each 10% of visceral fat lost during the intervention was associated with 28% lower incidence of type 2 diabetes despite complete weight regain — while liver and pancreas fat came back in full.[67] The cleanest pathway is calorie balance plus the exercise interventions above — a combination, not either alone.
See also Ozempic-class drugs for the pharmacological route in adults who qualify — with the caveat that rapid weight loss takes muscle with it, whether or not a drug drives it. Across 35 randomised trials, a median of just over a quarter of the weight lost came from muscle-based measures, more than a third of the non-drug comparator arms also exceeded the expected share, and no trial has yet measured whether any of it costs physical function.[68] Measured as a share of total mass, lean tissue actually rises.[69] Resistance training and adequate protein belong alongside the drug.
Sleep and circadian alignment
- Consistent sleep protects, but catching up does not reliably repair. After five short nights, two ten-hour recovery nights restored the fatty-acid response but left insulin secretion, scaled to insulin sensitivity, still suppressed[70], and an ad libitum recovery weekend left insulin sensitivity below baseline into the following week.[71] Only after a very short exposure — four nights of 4.5 hours — did two recovery nights restore it.[72] The same hours every night beats catching up. See Sleep.
- Eating, sleeping, and exercising on a regular schedule supports the same circadian-clock machinery that organises the AMPK/mTOR oscillation. See Circadian rhythms.
Supplements and geroprotector molecules (early signals, not first-line)
The supplement with the best flexibility-specific trial evidence is L-carnitine: 2 g/day for 36 days in a double-blind crossover in eleven adults with impaired glucose tolerance restored clamp and meal-test ΔRQ toward normal and raised muscle acetylcarnitine, though whole-body insulin sensitivity was unchanged[73] — a larger, longer, uncontrolled follow-up in type 2 diabetes at a higher dose did move clamp-measured insulin sensitivity, but with no placebo arm.[74] Oral carnitine also carries a pharmacokinetic caveat: in healthy volunteers given 0.5 or 1.5 g, absorption was low and about 90% of the dose was metabolised to trimethylamine N-oxide (TMAO), reaching 50 µM in plasma — levels the authors note have been associated with adverse health outcomes; an acute single-dose study, not an outcome trial.[75] Small samples and surrogate endpoints — promising, not practice-changing.
Several other molecules trigger parts of the same AMPK/SIRT1/mitophagy machinery pharmacologically. What each one is supposed to do to the fuel-switching apparatus:
- Urolithin A — a gut metabolite of the polyphenols in pomegranate, berries and walnuts, and the clearest dietary activator of mitophagy, the recycling of damaged mitochondria that keeps the fat-oxidising machinery in working order.
- Spermidine — induces autophagy in cells and animals, the cleanup arm of the fasting response. Oral doses, though, do not reliably reach the bloodstream at all.
- NAD⁺ precursors (NMN, nicotinamide riboside) — aim to refill the falling NAD⁺ pool that SIRT1 needs in order to switch on PGC-1α and mitochondrial biogenesis.
- Senolytics (dasatinib + quercetin, fisetin) — clear senescent cells, whose inflammatory secretions are one proposed reason aging tissue handles fuel badly.
None of the four has been tested against a fuel-switching endpoint, and the one pooled analysis against the insulin and glucose markers this page tells you to track — eight randomised trials of NMN, one of the NAD⁺ precursors, in 342 mostly non-diabetic middle-aged and older adults — was null.[76] See Geroprotectors for the trial-by-trial state of the human evidence — the doses, the sample sizes, and which primary endpoints were missed.
How to measure it
The research methods differ in what they capture. Resting/clamp switching and exercise fat-oxidation are different constructs, their summary metrics can be unstable, and even two resting measures need not agree: in eight healthy adults, chamber-measured flexibility to a high-fat meal repeated well across a week yet showed no correlation at all with clamp-measured flexibility in the same people.[77]
| Method | What it captures | Key limitation |
|---|---|---|
| ΔRQ during hyperinsulinemic-euglycemic clamp (fasting → insulin-stimulated RQ) | Fasting→insulin fuel switch; the original gold standard | Invasive; not scalable; ΔRQ largely tracks glucose disposal rate[78] |
| RER response to an oral glucose tolerance test (OGTT) or mixed meal (ΔRER, incremental area under the curve, iAUC) | Postprandial switching | The instantaneous respiratory exchange ratio repeats well day to day; it is the area-under-the-curve summary that doesn't (see below) |
| Exercise FATmax / crossover point / lactate | Fat-oxidation capacity across graded exercise | Different construct from resting flexibility; no agreed protocol yet, though a 2025 standardisation proposal calls for formal consensus[79] |
| 24h whole-room calorimetry | Real-world fuel rhythm across sleep/meals/exercise | Requires a specialised chamber — and the answer depends on the challenge imposed inside it: a high-fat day shifted the day–night swing while acute exercise in energy balance moved no flexibility index at all[80] |
Concretely: measured twice 48h apart, the day-to-day coefficient of variation for the OGTT respiratory-exchange area under the curve was ~22% (energy-expenditure AUC ~56%), meaning intervention studies "would need large sample sizes" to detect change.[81] Exercise protocols are unstandardised too: the choice of stoichiometric equation alone shifts the measured maximal fat-oxidation rate by up to nearly 7%, though FATmax, the intensity at which that peak occurs, is unaffected.[82] And much of the reported day-to-day variation may be nothing more than the measurement error of the gas analysers themselves.[83]
You don't need a metabolic lab to track your own trajectory. The subjective markers below are not validated against any of the lab measures above — treat them as rough self-report, not a substitute for the lab panel:
Markers of good flexibility:
- Comfortable doing zone-2 aerobic work without needing pre-workout carbs.
- Can skip or delay a meal without a mood or cognitive collapse.
- Stable energy through the day, not the post-lunch crash pattern.
- Wake up alert without immediate hunger.
Markers of poor flexibility:
- Energy crashes 2–3 hours after meals.
- Need to eat every 3–4 hours to function.
- Can't exercise without pre-fuelling carbs.
- Wake up shaky if you didn't eat late.
Lab measures worth running
| Test | Useful threshold | Notes |
|---|---|---|
| Fasting insulin | <70 pmol/L (<10 µIU/mL) | Rises before glucose does, which is what makes it useful — but not the sharpest of the surrogates: it tracked the clamp slightly worse than HOMA-IR in the founding study[84], and insulin immunoassays are not standardised between laboratories, so read the number against your own lab's reference range rather than as a universal line[85] |
| HOMA-IR | (fasting insulin × fasting glucose) ÷ 22.5 (mmol/L) or ÷ 405 (mg/dL). There is no universal cut-off — risk-anchored thresholds ran from 1.85 to 3.46 within a single cohort depending on sex and definition[86]; this site uses >2.5 as the working flag | Combines insulin and glucose into one number, and reflects mostly liver insulin sensitivity rather than the muscle fuel-switching this article is about |
| HbA1c | <5.7% normal; 5.7–6.4% prediabetes; ≥6.5% diabetes[87] | Three-month glycation average — but a red-cell measurement as much as a glucose one: Black adults run 0.1–0.5 percentage points higher at identical measured glucose[88] |
| Triglyceride / HDL ratio | <1.5 when both values are in mmol/L — equivalently <3.5 when both are in mg/dL, the cut-point derived against the insulin-suppression test in 449 adults[89] | Accessible surrogate for insulin resistance in white and Hispanic adults — but it transfers poorly: in one study of overweight African Americans it discriminated at close to chance[90], and in African-American women no usable cut-off could be defined at all[91] |
| Fasting glucose | <5.6 mmol/L (100 mg/dL) normal; 5.6–6.9 mmol/L (100–125 mg/dL) prediabetes; ≥7.0 mmol/L (126 mg/dL) diabetes. The lower prediabetes bound is the American Diabetes Association's; the World Health Organization still sets it at 6.1 mmol/L (110 mg/dL)[92] | Late marker — moves after insulin |
| Continuous glucose monitor (CGM) | No validated target — in a community cohort with a normal HbA1c and a normal fasting glucose, time in the 3.9–7.8 mmol/L (70–140 mg/dL) range averaged about 87%, leaving roughly three hours a day above it[93] — a description, not a goal | Most useful for visualising postprandial patterns; see continuous glucose monitors for what the device can and can't tell a healthy person |
None of these measures flexibility directly; together they show whether insulin resistance — its closest clinical companion — is developing. See Glycemic index for the postprandial side of the story.
One caveat about the evidence base: a single-centre clamp study in healthy young adults found women more metabolically flexible — a wider fasting-to-fed fuel swing (clamp ΔRQ ~0.14 vs ~0.09 in men), driven mainly by greater insulin-stimulated glucose oxidation[94] — but that does not generalise: pooling 35 clamp studies in 985 people, sex did not contribute to the variance in the fuel swing (body mass index was the only significant predictor), and the lean groups were split evenly between the sexes, so the clamp literature is not a male-only one.[95]
Practical synthesis
Most metabolically inflexible adults benefit most from these moves, applied together:
- Train most weeks, in more than one way. A reasonable default is 3–4 hours of zone 2, two resistance sessions and one hard session — not because that mix is proven superior, but because each part earns its place separately. See Zone 2 for the dose.
- Lose visceral fat if you carry it. The most direct lever, and the diabetes benefit persists even if the weight comes back.
- Eat earlier in the day rather than late. Finishing dinner by early evening appears to matter more than how short the eating window is — and time-restricted eating is not a better route to flexibility than simply eating less.
- Cut ultra-processed carbohydrate-fat combinations, and put protein and fibre in their place. Displacing refined carbohydrate blunts the post-meal glucose rise, and protein helps preserve muscle.
- Sleep 7–8 hours on a regular schedule. Worth doing on its own merits — but not a quick metabolic fix: adding about an hour of sleep a night for six weeks improved sleep health without improving clamp-measured insulin sensitivity or 24-hour glucose control in short-sleeping adults with overweight and insulin resistance[96] — though that trial ran in only 29 people, and the extra hour a night it actually achieved probably left them still under seven hours. Where sleep can't be fixed, exercise offsets much of the metabolic damage from sleep loss.[97]
- Sequence meals: vegetables and protein first, carbs last. Pooling seventeen studies in 389 people with type 2 diabetes, eating carbohydrate last lowered glucose by about 2.4 mmol/L (43 mg/dL) at one hour and about 0.7 mmol/L (13 mg/dL) at two hours, with a small durable effect on HbA1c of 0.16 percentage points.[98] The much-quoted ~73% reduction comes from a single eleven-person letter and is measured against eating the carbohydrate first, not against a normal mixed meal.[99] In healthy adults the base is thinner still — six acute studies in 107 young participants, consistent in direction but too heterogeneous to pool.[100] See Glycemic index.
Flexibility metrics typically move within weeks, and the biochemistry of mitochondrial capacity moves faster still: citrate synthase activity and mitochondrial DNA were already up by the third session of a seven-session training block.[101] Morphological density is a slower and noisier readout.
What's overrated
- Continuous ketosis as a longevity strategy. Short-term weight loss is real but modest, with little effect on glycaemic markers;[102] the reproducible signal against it is the LDL rise across pooled randomised trials, with no long-term outcome data either way[103] — and the popular "cyclical instead" fix is untested (above).
- NAD⁺ precursors and other geroprotector supplements as substitutes for exercise. The one pooled analysis, of NMN trials, is null on the very markers this page tells you to track (above).[104] See Geroprotectors.
- "Metabolic-type" diet matching from commercial blood-test panels. The Stanford glucotype work is a real finding but a thin one — the taxonomy was formally challenged in the same journal on generalisability and interpretability[105], and it has not been replicated in a non-diabetic population since. The consumer-marketed personalised-nutrition panels have less behind them still (see Glycemic index).
- Endless cardio without resistance training. Builds aerobic capacity but not muscle, and misses the improvement in fasting insulin and glucose that resistance training brings on its own.
Further reading
- Goodpaster BH, Sparks LM. Metabolic Flexibility in Health and Disease. Cell Metab 2017.[106]
- Smith RL et al. Metabolic flexibility as an adaptation to energy resources and requirements in health and disease. Endocr Rev 2018.[107]
- San-Millán I, Brooks GA. Assessment of Metabolic Flexibility — Blood Lactate, Fat, and Carbohydrate Oxidation Responses to Exercise. Sports Med 2018.[108]
- Ang JC, et al. Perspectives on whole body and tissue-specific metabolic flexibility and implications in cardiometabolic diseases. Cell Rep Med 2025.[109]
- Amorim JA, et al. Mitochondrial and metabolic dysfunction in ageing and age-related diseases. Nat Rev Endocrinol 2022.[110]
- Galgani JE, Moro C, Ravussin E. Metabolic flexibility and insulin resistance. Am J Physiol Endocrinol Metab 2008.[111]
- Sparks LM et al. Effect of adipose tissue on the sexual dimorphism in metabolic flexibility. Metabolism 2009.[112]
- Song JD et al. Dissociation of muscle insulin resistance from alterations in mitochondrial substrate preference. Cell Metab 2020.[113]
- Bruls YMH et al. Carnitine supplementation improves metabolic flexibility and skeletal muscle acetylcarnitine formation. EBioMedicine 2019.[114]
- Volek JS et al. Metabolic characteristics of keto-adapted ultra-endurance runners. Metabolism 2016.[115]
- Alcantara JMA et al. Reproducibility of the energy metabolism response to an oral glucose tolerance test — influence of a postcalorimetric correction procedure. Eur J Nutr 2023.[116]
- Kelley DE, Mandarino LJ. Fuel selection in human skeletal muscle in insulin resistance — a reexamination. Diabetes 2000.[117]
- Carnero EA et al. Twenty-four hour assessments of substrate oxidation reveal differences in metabolic flexibility in type 2 diabetes that are improved with aerobic training. Diabetologia 2021.[118]
- Chávez Guevara IA, Amaro-Gahete FJ. Methodological issues related to maximal fat oxidation and FATmax reproducibility — a narrative review. Int J Obes 2025.[119]
- Oldenburg N et al. Effects of time-restricted eating and caloric restriction on metabolic flexibility — a randomized trial. Obesity (Silver Spring) 2025.[120]