How to Interpret a Fasting Insulin and HOMA-IR Result

HOMA-IR is useful. It can be abnormal years before fasting glucose or HbA1c move, which makes it one of the few widely available windows onto early metabolic dysfunction. It is also an index calculated from an unstandardised assay, with no ratified cut-off, that fails in several predictable situations.

Both of those things are true, and using the test well depends on holding them together.

What the number is

HOMA-IR models the steady-state relationship between fasting glucose and fasting insulin. The original 1985 calculation is a single equation. [1]

In conventional units, HOMA-IR is fasting insulin in µU/mL multiplied by fasting glucose in mg/dL, divided by 405. In SI units it is fasting insulin in mIU/L multiplied by fasting glucose in mmol/L, divided by 22.5. The divisors differ because the units do; mixing them is a common and silent error.

What it estimates is basal insulin sensitivity, which is dominated by the liver. It is a surrogate for the hyperinsulinaemic-euglycaemic clamp, correlates with it reasonably well across populations, and is very much cheaper. It tells you nothing about how the patient handles a meal.

Which HOMA are you looking at?

This is the first thing to establish and it is routinely missed.

The original model, now called HOMA1, is the linear equation above, calibrated against the radioimmunoassays of the 1970s. HOMA2 is a non-linear computer model published in 1998 by the same group, recalibrated for modern assays. [2] It accounts for proinsulin cross-reactivity, handles hyperglycaemia without assuming linearity, and can be run from C-peptide instead of insulin. The Oxford Diabetes Trials Unit distributes the calculator; it cannot be done by hand.

The two produce different numbers on the same sample. HOMA2-IR runs systematically lower, such that a HOMA1 threshold around 2.7 corresponds to roughly 1.8 on HOMA2. A clinician who reads a HOMA2 result against HOMA1 cut-offs will under-call insulin resistance consistently.

The assay problem

Fasting glucose assays are standardised. Insulin immunoassays are not.

Different commercial platforms measure insulin differently, partly because they cross-react with proinsulin and its split products to varying degrees. Reported differences between assays on identical samples run to roughly 20 to 40%, and larger discrepancies have been described across the full range of available platforms. [3] Because insulin is one of only two inputs, that variability passes directly into the index.

Three practical consequences follow.

A HOMA-IR from one laboratory is not comparable to one from another. If you are tracking a patient over time, use the same laboratory throughout, and treat a change following a switch as uninterpretable.

Absolute cut-offs imported from a published study apply to that study's assay and population, not necessarily to yours. The reporting laboratory's own reference range is more relevant than any number quoted in a review.

And this is the main reason no professional body has ratified a universal threshold. It is not an oversight; it is a measurement problem.

Why there is no agreed cut-off

Values between 2.0 and 2.5 circulate widely as the threshold for insulin resistance. They come from research literature and are population-specific rather than validated for routine clinical use.

The degree of variation is substantial. A fifteen-year prospective study in southern Chinese adults found optimal cut-offs of 1.37 for identifying dysglycaemia and 1.97 for type 2 diabetes — both well below the figures commonly quoted in Western practice. [4] Thresholds also differ by age, sex, BMI and the outcome being predicted.

This does not make the test useless. It makes it a continuous risk marker rather than a diagnostic threshold, and it means the clinically meaningful signal is usually the trend in one patient at one laboratory, not the position relative to a line.

Diagnosis of diabetes and prediabetes remains based on glucose criteria — HbA1c, fasting plasma glucose and the oral glucose tolerance test. HOMA-IR does not diagnose anything.

When the number misleads

Five situations account for most misinterpretation.

Exogenous insulin. The assay measures administered insulin along with endogenous, so HOMA-IR is invalid in insulin-treated patients.

Failing beta cells. The model assumes a steady state in which insulin rises to compensate for resistance. As secretory capacity declines, fasting insulin falls, and HOMA-IR falls with it. The result improves while the patient deteriorates — precisely the wrong signal at the point it matters most.

A single measurement. Fasting insulin has considerable within-person variability. One value is a noisy estimate, and a change between two isolated draws may be measurement rather than biology.

Pre-analytical conditions. A true 8- to 12-hour fast, no vigorous exercise in the preceding day, and no acute illness. Sleep deprivation and acute stress both raise fasting insulin. Samples need prompt separation.

Postprandial resistance. HOMA-IR is a fasting snapshot. A patient can have a normal fasting index and markedly abnormal insulin response to a glucose load, which only dynamic testing will reveal.

Before ordering: will the number change anything?

This is the question worth asking first, and it is asked least.

For a patient with central adiposity, raised triglycerides, low HDL and a family history of type 2 diabetes, the management is the same whatever the HOMA-IR shows: resistance training, aerobic activity, protein-adequate diet, sleep assessment, alcohol reduction, and attention to visceral fat. If the number will not alter a single recommendation, it is being ordered for reassurance or for the appearance of thoroughness.

There are three situations where it genuinely earns its place.

When glucose and HbA1c are normal but risk seems high, a raised HOMA-IR can identify compensated insulin resistance years before dysglycaemia appears. That is the test's best use.

When a patient needs to see a physiological reason to change behaviour, a concrete number can do work that advice alone does not — provided the follow-up measurement uses the same laboratory.

And when tracking response to an intervention over months, the direction of travel is informative even though the absolute value is not comparable to anyone else's.

A reading protocol

1. Confirm the model. HOMA1 or HOMA2, and read against matching reference values.
2. Check the units. The divisor is 405 for conventional units and 22.5 for SI.
3. Check the fast and the context. True fast, no recent vigorous exercise, no acute illness, reasonable sleep.
4. Read against the reporting laboratory's range, not a threshold from a paper.
5. Interpret alongside the rest of the picture — waist circumference, triglyceride to HDL ratio, HbA1c, liver enzymes and blood pressure. HOMA-IR rarely changes a conclusion the rest of the assessment has not already suggested.

The honest summary for a patient

A raised HOMA-IR means the body is producing more insulin than it should need to keep fasting glucose normal, which is an early and reversible signal. It is not a diagnosis, the exact figure is less reliable than it appears, and the response is the same set of unglamorous interventions that were indicated anyway.

That is a less impressive sentence than a number with a reference range. It has the advantage of matching what the test can actually support.

For why a changed marker is not automatically a changed outcome, see Your biomarkers improved — did your healthspan?. For grading the evidence behind any test or intervention, see Dr. Hassan Bennani's Evidence-based longevity medicine.