HOMA-IR: How to Calculate It From Two Labs, and What Is Optimal

Your annual physical measured fasting glucose. It almost certainly did not measure fasting insulin. That missing number is why most people find out they are insulin resistant years late.

HOMA-IR fixes that with arithmetic. Two fasting labs, one line of math, and you have a single score for how hard your pancreas is working to keep your glucose normal. Here is the formula, the cutoffs that actually come from data, the ways the number can fool you, and what I do when a member's score is high.

The HOMA-IR formula

HOMA-IR stands for Homeostasis Model Assessment of Insulin Resistance. Take both values from the same fasting blood draw.

HOMA-IR = fasting glucose (mg/dL) x fasting insulin (uIU/mL) / 405

If your lab reports glucose in mmol/L, the formula is glucose x insulin / 22.5. Same score either way. That is the whole HOMA-IR calculator; no app required.

Three worked examples:

Look at the bottom two rows. Neither would get much attention at a standard physical. The glucose is under 100 or just over it. The insulin is what gives it away, and the insulin is what nobody ordered.

Where the formula came from

Matthews and colleagues built the model and published it in Diabetologia in 1985. They modeled the feedback loop between fasting glucose and fasting insulin, then checked the model's estimate of insulin resistance against the euglycemic clamp, the research method where insulin is infused and glucose is dripped in to hold blood sugar steady. The correlation with the clamp was 0.88. They also reported the model's weakness in the same paper: a coefficient of variation of about 31% for a single estimate, which is why a lone HOMA-IR should never be read to the decimal.

Bonora and colleagues repeated the comparison in 115 people across the full range of glucose tolerance in a 2000 study in Diabetes Care. The correlation with the clamp was 0.82, and it held in men and women, under and over 50, non-obese and obese, diabetic and not. Wallace and colleagues, with Matthews as coauthor, wrote the definitive guide to using it in Diabetes Care in 2004: HOMA-IR is well suited to cohort and epidemiological studies, needs only one fasting sample, and gives bad answers when the input labs are bad. I use it the same way in one person, as a trend.

So the score is a real measurement, validated against the gold standard, with two known limits: noise on any single reading, and dependence on the quality of the insulin assay.

HOMA-IR normal range: what "normal" and "optimal" mean

No universal cutoff exists, and anyone who quotes one to the decimal is hiding that. The number depends on the population and on the lab. These are the figures I trust.

Isokuortti and colleagues defined the upper limit of normal in two Finnish population cohorts in a 2017 study in Diabetologia. Among people who were not obese, had normal fasting glucose, no known disease, no heavy drinking, and took no medications, the 95th percentile of HOMA-IR was 1.9 to 2.0. The same cutoff, 1.9, identified fatty liver on magnetic resonance spectroscopy with 87% sensitivity and 79% specificity. In that study, a HOMA-IR of 2.0 sat at the upper edge of normal liver fat.

Gayoso-Diz and colleagues studied 2,459 Spanish adults in a 2013 paper in BMC Endocrine Disorders. Using the 90th percentile of the population gave a cutoff of 3.46, but once the authors anchored the threshold to the presence of metabolic syndrome components, it fell to about 2.05, and to 1.85 in non-diabetic men. Those risk-anchored cutoffs sat at only the 70th to 75th percentile of the Spanish population, which tells you how much of the population is already past them.

The thresholds I use in practice:

Above 2 is borderline. In the US survey data Ausk analyzed, the top quartile, where mortality rose, started at 2.8. Below 1 is where I like to see members who are training seriously. These are practice thresholds, not a diagnosis.

Why HOMA-IR predicts disease better than glucose

A fasting glucose of 95 tells you the pancreas is keeping up. It does not tell you what that costs. HOMA-IR does, and the cost predicts outcomes.

Gast and colleagues pooled 65 cohort studies with 516,325 participants in a 2012 meta-analysis in PLoS One. For each one-standard-deviation increase, the relative risk of coronary heart disease rose 21% for fasting glucose, 46% for HOMA-IR, and a statistically insignificant 4% for fasting insulin on its own. The product of the two numbers predicted heart disease better than either number alone.

Ausk and colleagues followed 5,511 US adults without diabetes for up to 12 years in a 2010 analysis of national survey data in Diabetes Care. People in the top quartile of HOMA-IR, above 2.8, had 64% higher all-cause mortality than those in the bottom quartile, at or below 1.4, after adjusting for age, weight, blood pressure, cholesterol, smoking, and activity. The association held only in people with a body mass index under 25.2, the survey median. Insulin resistance in a lean person is not benign.

González-González and colleagues reviewed 38 cohort studies with 215,878 participants in a 2022 meta-analysis in High Blood Pressure and Cardiovascular Prevention. A high HOMA-IR raised the risk of developing type 2 diabetes by 87%, hypertension by 35%, and non-fatal cardiovascular events by 46%. It did not predict death from cardiovascular disease or cancer in that analysis.

Five ways the number can fool you

1. The insulin assay. Marcovina and colleagues tested 12 commercial insulin assays for the American Diabetes Association in a 2007 report in Clinical Chemistry. The between-assay coefficient of variation for the same samples had a median of 24%. Isokuortti's group sent identical samples to seven European laboratories in the 2017 Diabetologia paper, and a HOMA-IR of 2.0 in Helsinki came back as anywhere from 1.3 to 2.1 elsewhere. Use the same lab for every retest, and do not compare your number to a friend's.

2. One draw is noisy. Matthews reported a 31% coefficient of variation for the model's own estimate of insulin resistance, so the score is a rough gauge, not a caliper. Add normal day-to-day swings in fasting insulin and a 2.3 on Tuesday can read closer to 2 or closer to 3 on Friday. Two readings 12 weeks apart tell you more than one reading to two decimals.

3. The fast. HOMA-IR assumes a normal overnight fast of 8 to 12 hours. Fasting much longer, or drawing labs in the middle of an extended fast, makes the number read worse than your baseline, because a long fast produces its own temporary insulin resistance. That is explained in does fasting cause insulin resistance.

4. A failing pancreas. In long-standing type 2 diabetes, the beta cells stop producing enough insulin. Fasting insulin falls, and HOMA-IR understates the insulin resistance that is still there, because the formula reads low insulin as good news. Wallace's 2004 review is explicit that the model should not be used to judge beta-cell function in isolation. A HOMA-IR that drops or stalls in someone with long-standing diabetes is not necessarily an improvement.

5. Insulin injections. If you take insulin, the assay can count the injected drug along with your own, and the score means nothing.

What I do when HOMA-IR is high

A HOMA-IR over 2 in a member with normal glucose is the most useful abnormal result on the panel, because it arrives years before the diagnosis and, in my experience, it is the number that moves first.

First, I look at what is next to it: A1c, triglycerides and HDL, ALT, hs-CRP, waist circumference, and a fasting insulin history if there is one. Then the fix is the same one that lowers fasting insulin, in order: a 12-hour overnight eating window with an early dinner, no liquid sugar or refined flour for 12 weeks, strength training three days a week, seven hours of sleep, 30 grams of protein per meal, and a walk after meals. The full list with the evidence behind each step is in how to lower fasting insulin.

Retest at 12 weeks, same lab, same fast. In our own members who started with a fasting insulin above 15 uIU/mL, the average fell 38%, from 26.7 to 16.7, across 35 members with labs through March 2026, with diet, training, and physician follow-up together. That is a practice audit, not a trial, and individual results vary. The full table with caveats is on our Results page.

When the number will not move after a real 12-week effort, I look for sleep apnea, thyroid disease, polycystic ovary syndrome, and medications that raise glucose. Metformin and GLP-1 medications have their place after that, not before.

Get the two labs

Fasting insulin is one of the cheapest tests a direct-to-consumer lab sells, usually under $50 plus a draw fee, and any lab can run it alongside the glucose you were getting anyway. If your physician will not order it, order it yourself, which most states allow, do the division, and bring the number to the visit.

Rebel Health Alliance members start with an initial panel of more than 100 biomarkers, drawn from more than 3,000 available, that includes fasting insulin, A1c, hs-CRP, and ApoB. A physician you can message any time reads the whole panel, calculates HOMA-IR, and retests on a schedule that shows whether the changes are working. When the number will not move, that physician goes looking for the reason. Rebel Peak members add a dietitian and strength coach who build the protocol around it. See how it works, book a 20-minute call, or read the FAQ.

Frequently asked questions

How do I calculate HOMA-IR?
Multiply fasting glucose in mg/dL by fasting insulin in uIU/mL and divide by 405. If glucose is in mmol/L, divide by 22.5 instead. Both labs must come from the same fasting draw.

What is a normal HOMA-IR?
In healthy, non-obese adults with normal glucose, the 95th percentile is about 1.9 to 2.0. Under 1.0 is very insulin sensitive. Above 2.0 is where fatty liver and metabolic syndrome start to appear. Above 2.8 puts you in the highest-risk quartile of the US survey data.

Can HOMA-IR be high with a normal fasting glucose?
Yes, and that is the most common pattern. A high fasting insulin is what keeps the glucose normal. HOMA-IR catches insulin resistance years before fasting glucose or A1c moves.

Is HOMA-IR accurate?
HOMA-IR correlates 0.82 to 0.88 with the gold-standard clamp, which is good enough for tracking one person over time. A single reading has about 30% noise, and insulin assays differ between labs by about 24%, so retest at the same lab and compare trends, not decimals.

How fast can HOMA-IR improve?
Fasting insulin responds faster than glucose or A1c. With an early eating window, no liquid sugar, strength training, and adequate sleep, I expect movement in 8 to 12 weeks. That is when to retest.

Dr. Alec Weir is the Chief Medical Officer at Rebel Health Alliance (https://rebelhealthalliance.io), a 100% virtual, physician-led longevity practice in all 50 states. This article is for education only and is not medical advice. Talk to your physician before changing medication, diet, or exercise.