Can insulin resistance be reversed? HOMA-IR and the trials

Key takeaways

• In one trial, losing about 5 percent of body weight improved insulin sensitivity in the liver, muscle and fat tissue.¹¹
• In the DiRECT trial, 46 percent of people with type 2 diabetes were in remission after a year of supervised weight loss, and 36 percent were still in remission at 24 months.¹⁵¹⁶
• There is no agreed normal range for HOMA-IR. Published cutoffs range from about 2 to about 6, depending on the population and method.³⁶
• Glucose and A1C can stay normal for years while insulin rises to hold blood sugar there.⁹¹⁰

Can insulin resistance be reversed? The short answer

Insulin resistance can improve a lot, and weight loss is the best-studied way to improve it. In one trial, losing about 5 percent of body weight improved insulin sensitivity in the liver, muscle and fat tissue. In the DiRECT trial, 46 percent of people with type 2 diabetes were in remission after a year of supervised weight loss. There is no agreed cutoff for when it counts as reversed, though, because there is no agreed normal range for HOMA-IR, the usual score for insulin resistance.¹¹¹⁵³

What is insulin resistance?

Insulin moves glucose out of the blood and into cells. With insulin resistance, the body's tissues respond less to insulin, so the pancreas makes more of it to keep blood sugar normal.⁹

That extra output can hold glucose steady for years. In a long-term study of Pima Indian adults, people whose glucose stayed normal made more insulin as their sensitivity fell. People who went on to develop diabetes could not keep raising their insulin output, and their glucose climbed.⁹

What is HOMA-IR?

HOMA-IR, the Homeostasis Model Assessment of Insulin Resistance, estimates insulin resistance from one fasting blood draw. It was published by Matthews and colleagues in 1985. In their study, it tracked closely with the clamp test, the lab method used as the reference, with a correlation of 0.88. They also noted it was imprecise for a single person, with a coefficient of variation around 31 percent.¹

The model's authors say HOMA works best for comparing groups in research, such as cohort studies, and its numbers need careful reading for any one person.²

• The common formula is HOMA-IR equals fasting insulin in micro units per milliliter times fasting glucose in millimoles per liter, divided by 22.5.³
• If your glucose is in milligrams per deciliter, divide by 405 instead. That's the same formula, since 1 millimole per liter of glucose is about 18 milligrams per deciliter and 22.5 times 18 equals 405.³

What is a normal HOMA-IR range?

There isn't one agreed number. Each study set its own cutoff, based on its own population and method.³

The NHANES authors say directly that there is no consensus on HOMA-IR cutoffs across groups, and that cutoffs specific to sex and ethnicity are still needed. The Spanish study found cutoffs in women also shifted with age.³⁴

You may see very low HOMA-IR or fasting insulin targets online that are called optimal. Those are not guideline numbers. Treat them as one opinion, and ask your clinician how your own lab reads the result.³

The insulin test itself also varies. When an American Diabetes Association work group compared 10 commercial insulin tests, 7 of them were off from the reference method by more than 15.5 percent in 36 percent to 100 percent of samples. The same blood can give a different insulin number at a different lab.⁷

Fasting insulin vs A1C, which shows the problem first?

A1C shows your average blood sugar over the past 2 to 3 months. The standard tests for prediabetes and diabetes all measure glucose.⁸

Because the pancreas makes up for insulin resistance by producing more insulin, glucose and A1C can read normal while insulin is already high. In the Whitehall II study of British civil servants, insulin sensitivity fell steeply over the 5 years before a diabetes diagnosis. Fasting glucose rose slowly at first and only climbed sharply in the last 3 years. Insulin output rose and then fell during the years before diagnosis.¹⁰

This is common. Among US adults without diabetes, the share with high fasting insulin, 10 micro units per milliliter or more, rose from 28.2 percent in 1999 to 2000 to 41.4 percent in 2017 to 2018. The share with a HOMA-IR of 2.6 or more rose from 24.8 percent to 38.4 percent.³

So the two tests answer different questions. A1C tells you whether your blood sugar has risen. Fasting insulin and HOMA-IR can show how hard the pancreas is working to keep it normal.⁸¹⁰

Can insulin resistance be reversed with weight loss?

The trials point the same way. Insulin sensitivity improves as weight comes down.¹¹¹³

• Moderate weight loss. In a small randomized trial of adults with obesity, losing about 5 percent of body weight improved insulin sensitivity in fat tissue, liver and muscle, and improved beta-cell function. Losing more, around 11 percent to 16 percent, improved muscle insulin sensitivity and beta-cell function further.¹¹
• Diabetes Prevention Program, or DPP. In 3,234 adults with prediabetes, a lifestyle program aiming for at least 7 percent weight loss and 150 minutes of activity a week cut new diabetes cases by 58 percent compared with placebo. The lifestyle group had the biggest gain in insulin sensitivity at one year.¹²¹³
• Look AHEAD. In adults with type 2 diabetes, an intensive lifestyle program produced partial or full remission in 11.5 percent at year 1 and 7.3 percent at year 4, compared with 2.0 percent in the control group. The authors called the absolute rates modest.¹⁴

Can the pancreas recover?

The DiRECT trial tested this in people who had type 2 diabetes for up to 6 years. The program used a formula diet of 825 to 853 calories a day for 3 to 5 months, run through primary care, with clinicians stopping diabetes and blood pressure drugs at the start. It was medically supervised and is not something to copy on your own.¹⁵

In DiRECT, the people who recovered had had diabetes for a shorter time. These trials enrolled people who already had diabetes, which is one reason the years when insulin is high but glucose still reads normal get attention.¹⁷

• At 12 months, 46 percent of the intervention group were in remission, compared with 4 percent of controls. Remission tracked weight loss, and 86 percent of those who lost 15 kilograms, about 33 pounds, or more were in remission.¹⁵
• At 24 months, 36 percent were still in remission. Remission was linked to how much weight people kept off.¹⁶
• Liver fat and pancreas fat fell with weight loss. People who returned to normal glucose were the ones whose early insulin response recovered, and they had had diabetes for less time. The authors wrote that this changes the old view that beta-cell loss in type 2 diabetes cannot be undone.¹⁷
• In people who stayed in remission, maximum insulin output rose until it matched a non-diabetic comparison group by 12 months and stayed there at 24 months. The early insulin response improved but stayed below that group.¹⁸

How to track it

Lab numbers move over months, and food and weight are the parts you can see every day.⁸

• Log what you eat. A complete food log shows your real average intake, so you and your clinician can tell whether a plan is happening as written. TrueCal lets you type a meal in plain words, which makes daily logging quicker.
• Watch the weight trend, not single days. Weigh at the same time of day and look at the weekly average.
• Note each lab date next to your log. When new results arrive, you can line them up with what changed in your eating and weight.
• Repeat labs the same way. Same lab if possible, and the same fasting conditions, because insulin tests differ between labs.⁷
• Read insulin and glucose as a pair. Insulin falling while glucose stays flat fits easing resistance. Insulin falling while glucose rises is a different pattern that your clinician should look at.¹⁰⁹

Common myths

• Myth, a normal A1C means no insulin problem. Not necessarily. Glucose can stay normal for years while insulin rises to hold it there.⁹¹⁰
• Myth, healthy HOMA-IR is under 1. No guideline sets that. Published cutoffs range from about 2 to about 6, depending on the population and method.³⁶
• Myth, once you have type 2 diabetes, the damage is permanent. DiRECT showed beta-cell function can recover after diagnosis in many people, especially with shorter duration.¹⁷
• Myth, lower insulin is always good news. Only when glucose holds steady or falls too. Falling insulin output with rising glucose is the pattern seen before diabetes.¹⁰

A note on medical advice

This article is for general information and isn't medical advice. Talk with your clinician about your own situation.

Worked example

Take an invented lab result with a fasting insulin of 10 micro units per milliliter, the level one large US study used to define high fasting insulin, and a fasting glucose of 99 milligrams per deciliter, the top of the normal range. With glucose in milligrams per deciliter, the formula divides by 405. So 10 times 99 is 990, and 990 divided by 405 is about 2.44. That is below the 2.6 cutoff used in the NHANES study but above the cutoff of about 2.05 tied to metabolic syndrome in the Spanish study, so the same result reads differently depending on which cutoff is used.³⁸⁴

A closer look

The insulin test itself varies. When an American Diabetes Association work group compared 10 commercial insulin tests, 7 of them were off from the reference method by more than 15.5 percent in 36 percent to 100 percent of samples, so the same blood can give a different insulin number at a different lab.⁷

Questions and answers

Related guides

• TDEE calculator
• Calorie deficit calculator
• Weight loss plateaus and metabolic adaptation

Sources

1. Diabetologia. Matthews. 1985.
2. Diabetes Care. Wallace. 2004.
3. J Clin Med. Wu. 2025.
4. BMC Endocr Disord. Gayoso-Diz. 2013.
5. Arq Bras Endocrinol Metabol. Geloneze. 2009.
6. Diabetes Care. Tam. 2012.
7. Clin Chem. Miller. 2009.
8. Diabetes Testing. CDC. 2026.
9. J Clin Invest. Weyer. 1999.
10. Lancet. Tabák. 2009.
11. Cell Metab. Magkos. 2016.
12. N Engl J Med. Knowler. 2002.
13. Diabetes. Kitabchi. 2005.
14. JAMA. Gregg. 2012.
15. Lancet. Lean. 2018.
16. Lancet Diabetes Endocrinol. Lean. 2019.
17. Cell Metab. Taylor. 2018.
18. Diabetes Care. Zhyzhneuskaya. 2020.