HOMA-IR calculator
Enter your fasting glucose and fasting insulin, get HOMA-IR, HOMA-%B and QUICKI.
The values shown are examples, not typical results. Use a fasting sample (the lab’s fasting rules apply). Insulin in pmol/L is not converted here; ask the lab for µU/mL (the same as mIU/L).
HOMA-IR is a quick way to put the fasting glucose and fasting insulin from one blood test into a single index of insulin resistance. It is used in studies of large groups and sometimes in the clinic. It is not a diagnosis, and this page will not tell you that a result is normal or abnormal: no single cut-off is agreed on, as the studies below show.
The formulas and every variable
HOMA-IR = fasting insulin (I) × fasting glucose (G) ÷ 22.5 when G is in mmol/L, or ÷ 405 when G is in mg/dL. I is the plasma insulin in µU/mL, the same as mIU/L. HOMA-%B = 20 × I ÷ (G − 3.5), with G in mmol/L. QUICKI = 1 ÷ (log10 I + log10 G), with I in µU/mL and G in mg/dL. All three come from the same two numbers, which is why they track each other. The constant 22.5 is the product of 4.5 mmol/L and 5 µU/mL, so a pair of that size gives HOMA-IR 1.00 and HOMA-%B 100%. The step between the two glucose units is 18.0156 mg/dL for each mmol/L, from the molar mass of glucose, 180.156 g/mol.
Worked example
A fasting sample shows glucose 90 mg/dL and insulin 10 µU/mL (example values, not typical ones). HOMA-IR = 90 × 10 ÷ 405 = 2.22. In SI units 90 mg/dL is 5.00 mmol/L, and 5.0 × 10 ÷ 22.5 = 2.22. HOMA-%B = 20 × 10 ÷ (5.00 − 3.5) = 134%. QUICKI = 1 ÷ (log 10 + log 90) = 1 ÷ (1 + 1.954) = 0.338. Doubling the insulin to 20 µU/mL at the same glucose doubles HOMA-IR to 4.44 and lowers QUICKI to 0.307.
Why there is no normal range here
Studies set a threshold by choosing a percentile of their own sample, or by finding the value that best separates people with metabolic syndrome from those without. The answers differ with who was studied and how.
| Study and population | HOMA-IR |
|---|---|
| EPIRCE, Spain: 2,459 adults aged 20–92: the 90th percentile of HOMA-IR in the sample (Gayoso-Diz et al., BMC Endocr Disord 2013) | 3.46 |
| EPIRCE, Spain: the same adults: the best cut-off to flag metabolic syndrome (1.85 in men without diabetes) | 2.05 |
| SuRFNCD-2007, Iran: 3,071 adults aged 25–64 without diabetes: optimal cut-off for metabolic syndrome, sensitivity 57.3% and specificity 65.3% by ATP III (Esteghamati et al., Nutr Metab 2010) | 1.775 |
| Urban India: 691 adolescents: the cut-off with sensitivity above 70% and specificity above 60% for metabolic syndrome (Singh et al., J Clin Res Pediatr Endocrinol 2013) | 2.5 |
The table lists four values from four selections, spanning almost a factor of two: 1.775 to 3.46. The Iranian value came with a sensitivity near 57% and a specificity near 65%, which means it misclassified many people. A 2026 narrative review in Frontiers in Pediatrics adds that different insulin immunoassays can give values 20% to 40% apart for the same sample, so a threshold from one laboratory does not transfer to another. Use the reference range your own laboratory prints.
Limits of this method
Precision. In the original 1985 paper, the coefficients of variation of the estimates were 31% for insulin resistance and 32% for beta-cell function, and the authors said this limits the model's use. The correlation with the euglycaemic clamp was 0.88 for insulin resistance.
An approximation. The simple formulas stand in for a computer model. The 2004 review in Diabetes Care states that the model has been used less widely than its approximation formulae, and it lists measuring beta-cell function in isolation as inappropriate.
A single sample. Both numbers depend on a proper fast and on the assay that measured insulin, and one blood draw is one snapshot. For research, HOMA is best at comparing groups.
Not a replacement for other tests. HOMA-IR is not a test for diagnosing diabetes. For blood sugar, see the A1C calculator; for lipids, the LDL calculator; for body size, the BMI calculator.
Common mistakes
Dividing a glucose in mg/dL by 22.5 makes the answer about 18 times too large: for the example above, 90 × 10 ÷ 22.5 = 40, not 2.22. Entering insulin in pmol/L as if it were µU/mL overstates it by a factor of about 6 to 7, and sources even disagree on the exact factor. Using a non-fasting sample, or comparing a result from one laboratory with a cut-off from another, are the other usual errors. Reading QUICKI as a percent is a third: 0.338 is not 33.8%.
What is HOMA-IR?
HOMA-IR is the homeostatic model assessment of insulin resistance: a number worked out from one fasting blood sample, the glucose and the insulin in it. It comes from a model published by Matthews and colleagues in 1985, in which the fasting glucose and insulin concentrations are set by their interaction in a feedback loop. In that model, a higher insulin at the same glucose points to more resistance to insulin. The 1985 paper compared the estimates with clamp studies and found correlations of about 0.88 for insulin resistance, but also a low precision, with a coefficient of variation near 31%, and the authors said that limits its use.
How do I calculate HOMA-IR?
Multiply the fasting insulin in µU/mL by the fasting glucose and divide by a constant that depends on the glucose unit: ÷ 22.5 for glucose in mmol/L and ÷ 405 for glucose in mg/dL. The 405 is 22.5 times about 18, the mg/dL per mmol/L of glucose (18.0156 exactly, from glucose's molar mass of 180.156 g/mol), so the two forms agree to within 0.1%. For glucose 90 mg/dL and insulin 10 µU/mL: 90 × 10 ÷ 405 = 2.22. The same blood in SI units is glucose 5.0 mmol/L: 5.0 × 10 ÷ 22.5 = 2.22.
What is a normal HOMA-IR?
This calculator does not call a result normal or abnormal, because no single cut-off is accepted. Published thresholds depend on the population, the age, the sex, the insulin assay and how the line was drawn. In a Spanish sample of 2,459 adults the 90th percentile was 3.46, while the value that best flagged metabolic syndrome was 2.05. A national Iranian survey of 3,071 adults found 1.775, and a study of urban Indian adolescents 2.5. Ask the doctor who ordered the test, and use the reference range printed on the lab report.
What is QUICKI?
QUICKI is the quantitative insulin sensitivity check index of Katz and colleagues (2000): 1 ÷ (log10 of fasting insulin in µU/mL + log10 of fasting glucose in mg/dL). Unlike HOMA-IR it rises with sensitivity instead of resistance, and it is shown as a decimal such as 0.338, not a percent. It uses the same two numbers as HOMA-IR, combined through logarithms. In the paper's 56 people the index tracked clamp-measured sensitivity with a correlation of 0.78, against 0.57 for a minimal-model estimate, and was suggested for clinical research.
What does HOMA-%B measure?
HOMA-%B estimates how much insulin the pancreas's beta cells release, as a percent of a reference: 20 × insulin ÷ (glucose in mmol/L − 3.5). A pair of glucose 4.5 mmol/L and insulin 5 µU/mL scores exactly 100%. It is only defined above a glucose of 3.5 mmol/L (about 63 mg/dL). The 2004 review "Use and abuse of HOMA modeling" lists measuring beta-cell function in isolation among inappropriate uses, so read it alongside the insulin-resistance number, not alone.
What is HOMA2, and is it used here?
No. HOMA2 is the updated computer model published in 1998 by Levy, Matthews and Hermans, which takes account of variations in hepatic and peripheral glucose resistance. It is solved by computer and has no short formula. The University of Oxford's Diabetes Trials Unit distributes the HOMA2 Calculator, which is now licensed through Oxford University Innovation. This page uses the original 1985 formulas, which the 2004 review calls an approximation of the computer model. Results from the two methods differ and should not be mixed in one comparison.
Can I use HOMA-IR if I take insulin or have diabetes?
Take it with great care and ask your clinician. The model describes the balance between your own insulin production and glucose, so injected insulin breaks its assumption and the number loses its meaning. The 2004 review by Wallace, Levy and Matthews names cohort and epidemiological studies as the appropriate use of HOMA, and in 1985 the coefficient of variation of a single estimate was around 31%, so one result in one person is a rough guide at best. For diabetes monitoring, see the A1C calculator and your care team's advice.
Related: the A1C calculator, the LDL cholesterol calculator, the BMI calculator and the calorie calculator.
For general information only — not medical advice. Talk with your doctor or another health professional about your own health. Sources, read October 27, 2026: Matthews DR et al., Diabetologia 1985;28:412–419; Katz A et al., J Clin Endocrinol Metab 2000;85:2402–2410; Wallace TM, Levy JC, Matthews DR, Diabetes Care 2004;27:1487–1495; Levy JC, Matthews DR, Hermans MP, Diabetes Care 1998;21:2191–2192; Gayoso-Diz P et al., BMC Endocr Disord 2013;13:47; Esteghamati A et al., Nutr Metab 2010;7:26; Singh Y et al., J Clin Res Pediatr Endocrinol 2013;5:245–251; Gutch M et al., Indian J Endocrinol Metab 2015;19:160–164. How we check our calculators.