HOMA-IR Calculator

How to use this HOMA-IR calculator with mg/dL or mmol/L reports

1. Take both numbers from the same fasting draw. Eight hours or more without food or caloric drinks, water permitted. Glucose and insulin measured on different days, or after a meal, cannot be combined into a meaningful index.
2. Read the unit printed on your report, not the number. United States, Japanese and most Indian laboratories report glucose in mg/dL; the United Kingdom, Canada, Australia and most of Europe report mmol/L. A fasting glucose of 5.3 is mmol/L; a fasting glucose of 95 is mg/dL. Select the matching unit in the dropdown before you calculate.
3. Check the insulin unit too. Most reports use microunits per millilitre (written µU/mL or µIU/mL) or the numerically identical mIU/L. SI laboratories report pmol/L instead, and 1 µIU/mL equals 6 pmol/L. Choose the right entry in the insulin unit dropdown.
4. Press Calculate. The result panel shows HOMA-IR to two decimal places, the interpretation band, HOMA1-%B, a sensitivity figure relative to the model reference subject, and both unit systems side by side so you can confirm the conversion.
5. Use the what-if slider to explore. Drag or arrow-key the slider to scale fasting insulin between half and double your measured value and watch the index and the marker on the scale move. This is the fastest way to see how strongly HOMA-IR is driven by insulin rather than glucose.
6. Copy the permalink if you want to revisit or share the exact inputs; the link carries the values and units in the query string and reloads them automatically.

If either field is empty, non-numeric, zero, negative or physiologically impossible, the calculator refuses to produce a number and explains why. It will never show a stale figure, a zero or an infinity when the input is unusable.

The HOMA1 formula, its two constants, and where 405 comes from

In the units of the original paper, insulin resistance is the product of fasting plasma insulin and fasting plasma glucose divided by a normalising constant:

Here FPI is fasting plasma insulin in microunits per millilitre (µIU/mL, identical to mIU/L) and FPG is fasting plasma glucose in mmol/L. The 22.5 is not a physical constant; it is the product of the fasting values of an idealised, fully insulin-sensitive young adult, namely 5 µIU/mL of insulin and 4.5 mmol/L of glucose. That is why a perfectly insulin-sensitive reference subject scores exactly 1.0:

Laboratories that report glucose in mg/dL use an algebraically identical form with a different constant. Because 1 mmol/L of glucose equals about 18.0 mg/dL, dividing a mg/dL glucose value by 18.0 first and then by 22.5 is the same as dividing once by their product:

The glucose conversion itself follows from the molar mass of glucose, 180.16 g/mol. The National Institute of Diabetes and Digestive and Kidney Diseases publishes the factor as mg/dL × 0.0555 = mmol/L, whose reciprocal is 18.018. The rounded 18.0 baked into the constant 405 therefore makes the mg/dL route disagree with the mmol/L route by about one part in a thousand — 95 mg/dL with 10 µIU/mL gives 2.35 by the 405 route and 2.34 by converting to 5.27 mmol/L first. That difference is far smaller than the assay noise and can be ignored, but it is the reason two reputable calculators can differ in the second decimal place. This page always applies the constant that matches the unit you selected.

Insulin has its own conversion. NIDDK gives µIU/mL × 6 = pmol/L, so an SI report of 60 pmol/L is 10 µIU/mL. Substituting pmol/L directly into either equation inflates HOMA-IR six-fold.

The companion beta cell index published with HOMA1-IR is:

with FPG again in mmol/L. Note the subtraction: the expression is undefined at a fasting glucose of 3.5 mmol/L (63 mg/dL) and meaningless below it, so this calculator withholds %B rather than printing a negative or infinite value. Finally, because HOMA1-IR is scaled so that the reference subject reads 1.0, a convenient reciprocal expresses the same information as a percentage of that reference:

That reciprocal is a linear convenience, not the %S output of the HOMA2 computer model; Oxford explicitly states that no linear approximations of HOMA2 have been attempted.

Worked example: a 95 mg/dL and 10 µIU/mL fasting panel

Consider an adult whose fasting panel reads glucose 95 mg/dL and insulin 10 µIU/mL — values that would be signed off as unremarkable on most laboratory reports.

1. Confirm the units. Glucose is in mg/dL, so the constant is 405, not 22.5.
2. Multiply: 95 × 10 = 950.
3. Divide: 950 ÷ 405 = 2.3457, reported as 2.35.
4. Convert glucose for the beta cell index: 95 × 0.0555 = 5.27 mmol/L.
5. Beta cell function: (20 × 10) ÷ (5.27 − 3.5) = 200 ÷ 1.77 = 113%.
6. Sensitivity relative to the reference subject: 100 ÷ 2.35 = 43%.

The reading is instructive. Glucose alone looks fine, yet the index sits at roughly 2.4 times the reference subject, and the beta cells are running at about 113% of reference output to keep it there. That is the classic picture of compensated insulin resistance: normal glucose bought with elevated insulin. Had the same person been reported in SI units — 5.3 mmol/L and 60 pmol/L — the correct arithmetic is (60 ÷ 6) × 5.3 ÷ 22.5 = 2.36, reassuringly the same answer. Dividing 5.3 × 10 by 405 instead would have returned 0.13 and suggested exceptional insulin sensitivity, which is exactly the failure mode this calculator is built to prevent.

Interpreting the number: bands, reference values and why cut-offs travel badly

The Oxford Diabetes Trials Unit is blunt on this point: there is no absolute value for HOMA indices, because they depend on the specific assays used for glucose and insulin, and consequently there are no defined thresholds for normal versus abnormal. Insulin immunoassays are not standardised against a common reference material in the way glucose assays are, so two laboratories can return fasting insulin values differing by a third on the same serum. Any threshold you read is therefore a property of a particular cohort measured with a particular assay.

With that stated plainly, the bands below are the orientation this calculator uses. They are anchored on the one value that is defined by construction — the reference subject at 1.0 — and on the 2.0 to 3.0 range in which adult research cohorts most commonly place their working threshold.

Two further reading habits are worth adopting. First, HOMA-IR is strongly right-skewed, so the logarithm behaves far better statistically; Endotext notes that log-transformed HOMA correlates much more linearly with clamp estimates across wide ranges of sensitivity. That is why the scale drawn under the calculator is logarithmic rather than linear. Second, read HOMA-IR next to HOMA1-%B. A rising index with preserved or elevated %B is compensated resistance; a rising index with falling %B suggests the compensation is failing, which is a materially different situation.

HOMA-IR compared with other insulin resistance measures

HOMA-IR sits at the cheap, approximate end of a spectrum whose other end is a research-ward infusion study. The comparison below shows what each method actually measures and what it costs to obtain.

In practice HOMA-IR earns its place in epidemiology and in longitudinal follow-up of an individual measured repeatedly in the same laboratory. Wallace, Levy and Matthews, reviewing how the model had been used in the two decades after publication, endorsed it for cohort and epidemiological work while warning against reading beta cell function in isolation and against comparing HOMA1 linear approximations with other models where HOMA2 should have been used.

Limitations and assumptions built into the homeostatic model

Every number this page produces inherits the assumptions of a 1985 steady-state model. The important ones:

• It is not a diagnostic test. No professional body defines diabetes, prediabetes or metabolic syndrome by a HOMA-IR threshold. Diagnosis rests on standardised glucose and HbA1c criteria applied by a clinician.
• It assumes a basal steady state. Hepatic glucose output and insulin secretion must be in equilibrium. Recent food, acute illness, vigorous exercise, glucocorticoids or an intercurrent infection break that equilibrium.
• Insulin assays are not standardised. This is the dominant source of disagreement between studies and the reason Oxford refuses to publish universal thresholds. Serial measurements are only comparable within one laboratory and one assay.
• Precision is modest. The original paper reported a coefficient of variation of 31% for the insulin resistance estimate and 32% for the beta cell deficit. A change from 2.1 to 2.4 on a single pair of measurements is noise.
• It degrades in established diabetes. Once beta cell failure sets in, a person can be severely insulin resistant yet unable to mount a high fasting insulin, producing a falsely reassuring index. HOMA2 with C-peptide handles this better.
• It is invalid on insulin therapy. Exogenous insulin is indistinguishable from endogenous insulin in the assay, so the model's inference collapses. The Oxford calculator is explicitly for people not on insulin.
• Cut-offs are population-specific. Age, sex, ethnicity, pubertal stage and body composition all shift the distribution. Paediatric and pregnancy thresholds differ substantially from adult ones and are not represented here.
• The linear equations diverge at the extremes. Oxford warns that HOMA1 results may differ considerably from the HOMA2 model solutions, particularly at high glucose or high insulin. This calculator flags inputs outside the ranges the HOMA2 calculator accepts as clinically realistic — glucose 3.5 to 25.0 mmol/L (63 to 450 mg/dL) and insulin 20 to 400 pmol/L (about 3.3 to 66.7 µIU/mL).
• Plasma versus serum. The model is specified for plasma. Substituting serum insulin is common practice but should be recorded as a limitation.

Used within those constraints — one laboratory, fasting samples, repeated over time, read alongside the rest of the metabolic picture — HOMA-IR is a genuinely informative index. Used as a standalone verdict on a single blood test, it is not.

Frequently asked questions about HOMA-IR

Sources, provenance and medical disclaimer

Formula and constants verified against primary and institutional sources: the original Matthews model, the Oxford Diabetes Trials Unit HOMA Calculator documentation (which publishes HOMA1_IR = (FPI × FPG)/22.5 and HOMA1_%B = (20 × FPI)/(FPG − 3.5) together with the accepted input ranges), the Wallace, Levy and Matthews review of HOMA use, the NIH Endotext chapter on assessing insulin sensitivity, and the NIDDK unit conversion appendix used for the glucose and insulin factors.

• Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985;28(7):412–419. PubMed 3899825
• Wallace TM, Levy JC, Matthews DR. Use and abuse of HOMA modeling. Diabetes Care 2004;27(6):1487–1495. American Diabetes Association
• University of Oxford, Radcliffe Department of Medicine, Diabetes Trials Unit. HOMA Calculator: frequently asked questions. rdm.ox.ac.uk
• Muniyappa R, Madan R, Varghese RT. Assessing Insulin Sensitivity and Resistance in Humans. Endotext, NCBI Bookshelf NBK278954, updated 2024. NCBI Bookshelf
• National Institute of Diabetes and Digestive and Kidney Diseases. Diabetes in America, 3rd Edition, Appendix 1: Conversions. NIDDK via NCBI Bookshelf

This calculator is an educational implementation of a published research index. It does not provide medical advice, cannot diagnose or exclude any condition, and is no substitute for a consultation with a licensed clinician. Discuss any laboratory result with the professional who ordered it.