HOMA-IR Calculator

Background.

HOMA-IR is a way of turning two ordinary fasting blood results — glucose and insulin — into a single number that describes how hard the pancreas is working to hold that glucose down. It comes from the homeostasis model assessment published by David Matthews and colleagues at Oxford in Diabetologia in 1985. Their insight was that fasting glucose and fasting insulin are not two independent facts about a person; they are the resting equilibrium of a feedback loop, and where that equilibrium sits tells you something about how sensitive the tissues are to insulin. The linear approximation the paper gave is startlingly simple: multiply fasting insulin in µU/mL by fasting glucose in mmol/L and divide by 22.5. Someone with textbook-normal values of 5 µU/mL and 4.5 mmol/L lands on exactly 1.0, which is where the 22.5 comes from.

Before the number is worth anything, one thing has to be said clearly, because it changes how every HOMA-IR result should be read: there is no universal normal value. This is not a hedge invented for a web page. It is the position of the Oxford Diabetes Trials Unit, the group that owns and licenses HOMA, stated on their own site: 'There is no absolute value for HOMA indices. These will depend on the specific assays used for glucose, insulin and C-peptide. Because of this, there are no defined thresholds for normal vs. abnormal values.' Insulin immunoassays are notoriously non-interchangeable — they cross-react differently with proinsulin and its split products and are calibrated against different standards — so the same serum sent to two laboratories can produce HOMA-IR values that differ by more than the gap between the thresholds people argue about. That is why this calculator makes the comparison threshold an input you fill in rather than a constant it decides for you.

The field default of 2.05 is a real, named, dated number and not a consensus. It is the optimal cut-off for non-diabetic men under IDF metabolic-syndrome criteria in the EPIRCE study, a random sample of 2,459 Spanish adults. That same study found 1.85 for men under ATPIII criteria, and for non-diabetic women a cut-off that moved with age — 2.31 at age 30, 2.07 at age 50, 2.47 at age 70 — while a 90th-percentile criterion put the threshold at 3.46. The authors noted that all their values sat between the 70th and 75th percentiles of the Spanish adult distribution, and that cut-offs 'are different according to ethnicity, clinical methods of estimation, and metabolic conditions'. Replace the default with whatever your own laboratory or study protocol uses. If you have no such number, treat the comparison as decorative and use the HOMA-IR value itself for tracking change over time, which is what it is genuinely good for.

Precision is the second thing to know. The original paper reported that its insulin-resistance estimate correlated with the euglycaemic clamp at Rs = 0.88, which is impressive for two fasting tubes of blood, but the same authors wrote in the same abstract that 'the low precision of the estimates from the model (coefficients of variation: 31% for insulin resistance and 32% for beta-cell deficit) limits its use'. A coefficient of variation of 31% means a single HOMA-IR of 2.4 is entirely compatible with a true value near 1.7 or near 3.1. Small differences between two measurements, and small margins to a threshold, are noise. Differences of a factor of two, sustained across repeated tests, are signal.

This page computes HOMA1, the 1985 linear approximation, not HOMA2. Oxford's HOMA2 is an iterative computer model published in 1998 that simulates the underlying physiology rather than approximating it, and it accounts for variation in hepatic and peripheral glucose resistance, renal glucose loss and the proinsulin content of insulin assays. Oxford themselves describe the HOMA1 formulae as providing 'only linear approximations'. HOMA1 and HOMA2 do not produce the same numbers, and a value from one cannot be compared with a threshold from the other. If you are reading a paper, check which it used before comparing your result to it.

There is a small arithmetic point that is worth being explicit about, because it will make this calculator disagree very slightly with most others. The model is defined in mmol/L. The mg/dL version everyone quotes divides by 405, which is exactly 22.5 × 18 — it embeds a glucose conversion factor rounded from 18.0156 down to 18. This calculator converts with the exact factor derived from the molar mass of glucose (C₆H₁₂O₆ = 180.156 g/mol, from IUPAC's 2023 abridged standard atomic weights) and then divides by 22.5. At the values this page loads with, 90 mg/dL and 8 µU/mL, the exact route gives 1.7762 and the 405 route gives 1.7778 — a difference of 0.087%, roughly three hundred times smaller than the model's own 31% coefficient of variation. It is disclosed rather than hidden, but it should never change a decision.

Two things are deliberately missing. First, there is no pmol/L option for insulin. The conversion between µU/mL and pmol/L depends on which insulin standard the assay is calibrated against, and different reference preparations give materially different factors; rather than pick one and bake it in, this page asks for µU/mL (equivalently mIU/L) and tells you to get that figure from your laboratory. Second, HOMA-%B, the beta-cell function companion index, is not computed. Its linear approximation is widely quoted, but I could not retrieve it from a primary source — the Diabetologia and Diabetes Care full texts are both behind access controls and Oxford's public FAQ gives the equation for insulin resistance only. A clinical constant that cannot be verified does not get shipped, so the page is narrower and honest rather than complete and guessed.

Finally, a boundary of applicability that follows from the model itself. Matthews and colleagues opened by saying that 'the steady-state basal plasma glucose and insulin concentrations are determined by their interaction in a feedback loop'. The loop is between glucose and the pancreas's own insulin output. If the fasting insulin in your blood arrived from a syringe rather than from your beta cells, that loop is not the thing being measured and HOMA-IR does not describe what it claims to describe. The same caution applies to anyone whose insulin assay may be cross-reacting with an insulin analogue. HOMA-IR is a research and epidemiology instrument used in enormous numbers of studies; it is not a diagnostic test, and no clinical decision should rest on it alone.

What is homa-ir calculator?

The homeostasis model assessment is a mathematical description of the resting relationship between blood glucose and blood insulin. In a person whose tissues respond normally to insulin, a modest amount of insulin holds fasting glucose in the normal range. As tissues become less responsive, the pancreas compensates by secreting more insulin, so fasting insulin rises while glucose stays normal for a time — and it is that combination, normal glucose bought with high insulin, that HOMA-IR is designed to detect. HOMA1-IR is the linear approximation to the full model: fasting insulin in µU/mL multiplied by fasting glucose in mmol/L, divided by 22.5. The divisor normalises the index so that a healthy reference individual scores 1.

HOMA1-%S is the same information turned inside out. Oxford describe HOMA-IR as the inverse of HOMA-%S, and %S is expressed as a percentage of a normal reference population, so %S is simply 100 divided by HOMA-IR. A HOMA-IR of 2 corresponds to 50% sensitivity; a HOMA-IR of 4 to 25%. Because it is an exact reciprocal it adds no information, and this page shows it only because a large part of the literature reports sensitivity rather than resistance and readers need to be able to move between the two. Neither index is a diagnosis. Both describe a fasting equilibrium at one moment, measured with an assay whose calibration is not standardised between laboratories.

How to use this calculator.

1. 01Use a fasting sample — at least eight hours without caloric intake, and both tubes drawn at the same time. Glucose and insulin from different visits do not describe one equilibrium.
2. 02Enter fasting plasma glucose and choose its unit. mg/dL is usual in the United States and Japan; mmol/L almost everywhere else.
3. 03Enter fasting insulin in µU/mL, which your laboratory may print as mIU/L — the two are the same. If your report shows pmol/L, ask the laboratory for the µU/mL figure rather than converting it yourself: the factor depends on the assay's calibration standard.
4. 04Replace the reference cut-off with the threshold your own laboratory, clinician or study protocol uses. The default of 2.05 is one population's value, not a standard.
5. 05Read HOMA1-IR alongside the margin to your cut-off, and remember the model's 31% coefficient of variation before treating a small margin as meaningful.
6. 06If you are comparing against a published paper, check whether it used HOMA1 or HOMA2, and which insulin assay. Values from different models and different assays are not interchangeable.
7. 07Take the result to a clinician. HOMA-IR is a research and epidemiology instrument, not a diagnostic test.

The formula.

The equation is one multiplication and one division. Fasting insulin in µU/mL is multiplied by fasting glucose in mmol/L, and the product is divided by 22.5. Oxford's Diabetes Trials Unit states it as 'HOMA1_IR = (FPI x FPG)/22.5', and the EPIRCE investigators state it independently as 'fasting serum insulin (µU/ml) × fasting plasma glucose (mmol l⁻¹)/22.5' — two sources, identical equation, identical units. The constant 22.5 is the product of a reference fasting insulin of 5 µU/mL and a reference fasting glucose of 4.5 mmol/L, which is what makes a healthy reference individual score 1.0 rather than some arbitrary number.

When the input is in mg/dL it has to be converted first, and the conversion is where this calculator diverges very slightly from convention. The universally quoted mg/dL form divides by 405, and 405 is exactly 22.5 × 18 — so using it is identical to converting glucose with a factor of 18 rather than 18.0156. This page converts with 18.0156, derived from the molar mass of glucose: IUPAC's 2023 abridged standard atomic weights give C = 12.011, H = 1.0080 and O = 15.999, so C₆H₁₂O₆ = 6(12.011) + 12(1.0080) + 6(15.999) = 180.156 g/mol, and one mmol/L of glucose is 18.0156 mg/dL. Working the shipped defaults both ways: 90 mg/dL ÷ 18.0156 = 4.99567 mmol/L, so HOMA1-IR = 8 × 4.99567 ÷ 22.5 = 1.7762, whereas 8 × 90 ÷ 405 = 1.7778. The gap is 0.087%. It is disclosed because a reader comparing two calculators deserves to know why the last digits differ, not because it matters clinically.

Rounding happens at one place only: the return boundary. The unit conversion, the product, the division by 22.5 and the reciprocal for HOMA1-%S are all carried at full decimal precision, and each output is rounded once, at the end, to ten decimal places. In particular, the comparison against your cut-off is made on the unrounded value, so a HOMA1-IR that lands exactly on the threshold reports 'at or above' with no rounding ambiguity — the shipped test suite asserts this at exactly 2.0, and immediately either side of it.

HOMA1-%S is the exact reciprocal, scaled: 100 ÷ HOMA1-IR. Because both inputs are guarded to be strictly positive, HOMA1-IR is strictly positive and the reciprocal can never divide by zero. The relationship is inverse and non-linear, which is worth internalising: moving from a HOMA1-IR of 1 to 2 halves %S from 100% to 50%, but moving from 4 to 5 takes it only from 25% to 20%. Equal steps in resistance are not equal steps in sensitivity, and papers that report one can look very different from papers that report the other while describing identical data.

A worked example.

Frequently asked questions.

References& sources.

1. [1]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. doi:10.1007/BF00280883. Original model. Source of the feedback-loop premise, the clamp correlation Rs = 0.88, and the authors' own precision caveat (coefficients of variation 31% for insulin resistance and 32% for beta-cell deficit). Abstract open on PubMed (PMID 3899825); full text paywalled. Retrieved 29 July 2026.
2. [2]University of Oxford, Radcliffe Department of Medicine, Diabetes Trials Unit. "HOMA — Frequently Asked Questions." The custodians of HOMA. Source of the equation 'HOMA1_IR = (FPI x FPG)/22.5', of the statement that HOMA1 formulae provide 'only linear approximations of HOMA_%B and HOMA_IR, the inverse of HOMA_%S', and of the key scope limit: 'There is no absolute value for HOMA indices. These will depend on the specific assays used for glucose, insulin and C-peptide. Because of this, there are no defined thresholds for normal vs. abnormal values.' Retrieved 29 July 2026.
3. [3]University of Oxford, Radcliffe Department of Medicine, Diabetes Trials Unit. "HOMA Calculator" overview page: "The Homeostasis Model Assessment (HOMA) estimates steady state beta cell function (%B) and insulin sensitivity (%S), as percentages of a normal reference population." Source of the definition of %S as a percentage of a normal reference population. Retrieved 29 July 2026.
4. [4]Gayoso-Diz P, Otero-González A, Rodriguez-Alvarez MX, Gude F, García F, De Francisco A, Quintela AG. "Insulin resistance (HOMA-IR) cut-off values and the metabolic syndrome in a general adult population: effect of gender and age: EPIRCE cross-sectional study." BMC Endocrine Disorders. 2013;13:47. doi:10.1186/1472-6823-13-47. Independent statement of the equation and units ('fasting serum insulin (μU/ml) × fasting plasma glucose (mmol l⁻¹)/22.5'); source of the population cut-offs quoted on this page (men 1.85 ATPIII / 2.05 IDF; women 2.31 at 30, 2.07 at 50, 2.47 at 70; 3.46 at the 90th percentile) from 2,459 Spanish adults, and of the statement that cut-offs 'are different according to ethnicity, clinical methods of estimation, and metabolic conditions'. Open access; retrieved 29 July 2026.
5. [5]Wallace TM, Levy JC, Matthews DR. "Use and abuse of HOMA modeling." Diabetes Care. 2004;27(6):1487-1495. doi:10.2337/diacare.27.6.1487. The standard review of appropriate and inappropriate uses of HOMA, and of the relationship between the 1985 approximation formulae and the later computer model. Publisher full text paywalled; open bibliographic record with abstract at Oxford University Research Archive, retrieved 29 July 2026. Cited for the HOMA1/HOMA2 distinction only.
6. [6]International Union of Pure and Applied Chemistry, Commission on Isotopic Abundances and Atomic Weights. Standard atomic weights, 2023 abridged table: C = 12.011, H = 1.0080, O = 15.999. Used to derive the glucose molar mass 180.156 g/mol and hence the exact mg/dL-to-mmol/L factor of 18.0156 that this calculator uses in place of the rounded 18 embedded in the conventional 405 divisor. Retrieved 29 July 2026.