In the 1970s, pathologist Joseph Kraft began running extended oral glucose tolerance tests that measured insulin alongside glucose. Over more than two decades he collected roughly 14,000 of them. His central finding was striking: a large share of people with completely normal glucose curves had abnormal insulin responses. Kraft called this diabetes in situ. It was a metabolic disease already in progress, invisible to the tests most clinicians relied on.

Standard metabolic screening still largely rests on fasting glucose and HbA1c. Both are lagging indicators. They move only after the body's compensatory machinery begins to fail, often years into the underlying process. Fasting insulin captures that compensation directly. It measures how hard the pancreas is working to keep glucose looking normal, not just whether glucose is normal.

For anyone practising precision prevention, this distinction matters. Insulin resistance is upstream of type 2 diabetes, and it is closely linked to atherosclerotic cardiovascular disease, non-alcoholic fatty liver disease, polycystic ovary syndrome and several cancers, and it has been associated with neurodegeneration. Detecting it early opens a long window for intervention. What follows examines three things: the physiology of compensatory hyperinsulinemia, the evidence on where fasting insulin should sit, and the interventions that reliably move it.

Insulin resistance begins when peripheral tissues, mainly skeletal muscle, liver and adipose tissue, respond less effectively to insulin signalling. Intracellular lipid accumulation, particularly diacylglycerols and ceramides, interferes with insulin receptor substrate phosphorylation. That blunts downstream PI3K–Akt signalling and reduces GLUT4 translocation to the cell membrane. The same amount of insulin now moves less glucose into cells.

The pancreatic beta cell responds elegantly. It increases insulin output, through both greater secretion per cell and expansion of beta-cell mass, until glucose disposal is restored. Fasting glucose stays within range. HbA1c stays unremarkable. By conventional metrics the patient is metabolically healthy, while circulating insulin may be two or three times higher than it was a decade earlier.

The Whitehall II cohort made this trajectory unusually visible. Tabák and colleagues tracked civil servants for up to 13 years before diabetes diagnosis. Insulin sensitivity declined steadily across that period, and beta-cell function rose to compensate. Fasting glucose stayed relatively flat until roughly three to six years before diagnosis. Then glucose rose steeply as beta-cell compensation faltered. The disease had been active for years. Glucose merely announced its late stage.

There is a further complication. A growing body of work, associated notably with Barbara Corkey, argues that hyperinsulinemia is not purely a consequence of resistance. It may also be a driver. Chronically elevated insulin downregulates insulin receptors, promotes lipogenesis and suppresses lipolysis, which can deepen the very resistance it was compensating for. Whether the primary lesion sits in the tissue or in the beta cell probably varies between individuals. Either way, the result is a self-reinforcing loop.

Clinically, this reframes what a normal glucose result actually tells us. It confirms that compensation is succeeding. It says nothing about the cost of that compensation. Elevated insulin carries consequences of its own, including sodium retention, sympathetic activation, increased hepatic de novo lipogenesis and mitogenic signalling through IGF-1 pathways. These effects accumulate long before any diagnostic threshold is crossed.

Takeaway

A normal glucose reading can mean the system is healthy, or it can mean the pancreas is working overtime to hide a problem. Only insulin tells you which.

Laboratory reference ranges for fasting insulin commonly extend to around 20–25 µIU/mL. These ranges are derived statistically from the tested population, and in many developed countries a substantial proportion of that population is already insulin resistant. In NHANES data, median fasting insulin in US adults sits around 8–10 µIU/mL and has trended upward over recent decades. A value inside the reference range simply means you resemble the average. It does not mean you are optimal.

Epidemiological evidence points considerably lower. In the San Antonio Heart Study, elevated baseline fasting insulin independently predicted later type 2 diabetes in normoglycaemic participants. Several cohorts link higher fasting insulin to incident hypertension, coronary disease and all-cause mortality. Metabolically healthy, lean, active adults typically show fasting values in the low single digits. Many longevity-focused clinicians target below roughly 5–7 µIU/mL. They tend to view sustained values above 10 as meaningful early resistance, though no consensus guideline formally defines these cut-offs.

Context sharpens interpretation. HOMA-IR, calculated as fasting glucose (mg/dL) multiplied by fasting insulin (µIU/mL) divided by 405, integrates both variables. Values near 1 suggest good sensitivity. Values above roughly 2 to 2.5 often indicate developing resistance, though population-specific thresholds vary. A fasting glucose of 88 mg/dL paired with an insulin of 4 tells a very different story from the same glucose paired with an insulin of 14.

There are important technical caveats. Insulin assays are not well harmonised. Different immunoassays can return values differing by up to twofold on the same sample, so absolute cut-offs translate imperfectly between laboratories. A single fasting measurement also carries meaningful biological variability, influenced by recent sleep, stress, illness, exercise and the previous evening's meal.

The practical implication is to treat fasting insulin as a trend, not a verdict. Use the same laboratory where possible. Standardise preparation with a genuine 10–12 hour fast, avoiding unusual exertion the day before. Repeat the test to establish a baseline. Where results are ambiguous, an oral glucose tolerance test with timed insulin measurements, Kraft-style, reveals far more about postprandial dynamics than any single fasting value can.

Takeaway

Reference ranges describe the population, not the target. In a metabolically unhealthy population, being average is not the same as being well.

Exercise is the most potent non-pharmacological lever, and it works through a mechanism that bypasses the impaired pathway. Muscle contraction activates AMPK and calcium-dependent signalling that drives GLUT4 translocation independently of insulin. A single bout improves insulin sensitivity for roughly 24–48 hours. Chronic training adds lasting adaptations: greater mitochondrial density from sustained aerobic work and expanded glucose-disposal capacity from resistance training. Skeletal muscle is the largest glucose sink in the body. Building it is metabolic infrastructure.

Reducing visceral and ectopic fat is the second major lever. Fat stored in the liver and pancreas is disproportionately implicated in resistance and beta-cell dysfunction. The DiRECT trial showed that substantial weight loss, around 15 kg, put nearly half of participants with established type 2 diabetes into remission at one year. Mechanistic imaging work by Roy Taylor's group linked this to reductions in liver and pancreatic fat. Even modest weight loss of 5–10% can meaningfully lower fasting insulin in resistant individuals.

Dietary composition and timing matter alongside total energy. Lowering refined carbohydrate and added sugar intake reduces postprandial insulin demand directly. Prioritising protein and fibre blunts glycaemic excursions. Meal timing offers an additional, weight-independent effect. In Sutton and colleagues' 2018 controlled feeding study, early time-restricted eating, with all meals inside a six-hour window ending mid-afternoon, lowered insulin levels and improved insulin sensitivity in men with prediabetes, despite no weight loss.

Sleep is routinely underweighted. Controlled studies of sleep restriction show that just four nights of shortened sleep can reduce whole-body insulin sensitivity by around 16%, with larger reductions measured in adipocytes. Circadian misalignment, as in shift work, compounds the effect. Chronic psychological stress acts through cortisol and sympathetic tone, raising hepatic glucose output and increasing insulin demand. These factors can quietly undermine an otherwise excellent training and nutrition protocol.

Pharmacological tools such as metformin, GLP-1 receptor agonists and SGLT2 inhibitors have clear roles, particularly where lifestyle change alone proves insufficient. They work best layered on top of these fundamentals, not as replacements for them. Retesting fasting insulin every three to six months during an intervention offers objective feedback, and the marker often responds well before weight or HbA1c shift noticeably.

Takeaway

Muscle contraction opens a side door for glucose that resistance cannot close. That makes movement one of the few interventions that works around the problem while also repairing it.

Fasting insulin reframes metabolic health as a dynamic process, not a binary diagnosis. Glucose tells you whether compensation is still succeeding. Insulin tells you how much that success is costing. In the years between those two signals sits one of the largest preventable windows in modern medicine.

A rational protocol starts with a properly standardised baseline of fasting insulin, glucose and calculated HOMA-IR, ideally drawn from the same laboratory each time. Where results are borderline, add a timed insulin response to a glucose load. Intervene on the high-yield variables: consistent resistance and aerobic training, reduced visceral fat, lower refined carbohydrate intake, earlier eating windows, and protected sleep. Then retest to confirm the trajectory is moving.

Perhaps the deeper lesson concerns what we choose to measure. Many conventional markers detect failure. Precision prevention looks instead for strain, the moment when a system is coping but should not have to. Fasting insulin is one of the clearest windows into that moment.