A fasting insulin of 12: what it can, and cannot, tell you.

A fasting insulin of 12 may fall inside a laboratory's reference interval beside a normal glucose. It can add context that glucose and A1c do not give on their own, but it is a proxy, not a diagnosis: there is no universal cut-off, laboratories do not agree, and it is read beside standard testing and repeated before it means anything.

Daniel Tagge, MDJuly 30, 2026Updated September 19, 20266 min read

Reviewed and revised by Daniel Tagge, MD, September 19, 2026

Your fasting insulin is “normal.” Here’s the range that matters. On YouTube

A fasting insulin of 12 µIU/mL may fall inside a laboratory's reference interval, sometimes beside a normal fasting glucose and A1c. None of those results necessarily carries a flag.

Fasting glucose, HbA1c and the two-hour glucose after a sugar drink are the tests the American Diabetes Association uses to screen for and diagnose diabetes and prediabetes, and a normal answer is good news I take at face value. Fasting insulin is not on that list. A normal glucose is not the same as a complete answer about insulin sensitivity, though, and that gap can run for years.

Two people, the same glucose

Two people can have the same fasting glucose and very different fasting insulin levels. That difference may contain useful information, but it does not by itself establish insulin resistance.

What it can point at is compensation: the pancreas holding a normal glucose in place, which is one reason a glucose test can look unremarkable for years. The Whitehall II study followed British civil servants for about thirteen years. In the people who went on to develop diabetes, insulin sensitivity, estimated from fasting insulin and glucose, fell steeply across the five years before diagnosis, while fasting glucose drifted up slowly inside the normal range and rose steeply only from about three years before.

Read that as a population pattern, not a schedule. Later work in the same cohort found that the paths into diabetes differ substantially between people, and type 2 diabetes is a heterogeneous condition. What Whitehall II supports is that insulin sensitivity can deteriorate before the late acceleration in glucose. It does not establish a fixed sequence in any one person, and it says nothing about A1c, which it did not measure.

So glucose answers one question well, and it does not answer every question about insulin sensitivity. That is my reading, not a standard.

What the number is, and is not

Insulin resistance is a state: muscle, liver and fat respond less to insulin's signal. The reference method for measuring it is the hyperinsulinemic-euglycemic clamp, a research procedure that is rarely practical in routine clinical care. Fasting insulin is a stand-in. It rises with resistance, but it also reflects how much the pancreas secretes and how fast the liver clears it. And it has two further problems.

The first is that it wobbles. Fasting insulin varies from day to day in the same person, so one value is a signal to repeat, not a verdict. The second is that laboratories do not agree with each other. Insulin assays remain incompletely standardized. Current commercial methods can give materially different results on the same sample, which is why I prefer to compare repeat values from the same laboratory when I am following a trend.

What HOMA-IR is

HOMA-IR combines fasting glucose and insulin into an estimate of insulin resistance. It is useful in research and can add context clinically, but it is not a diagnostic test with one universal cut-off. Published cut-offs vary with the population studied, and because the estimate is built from an insulin value, it moves with the assay and the way the sample was handled as well.

HOMA was built to approximate the clamp across groups of people, and its authors' own guidance is that it is at its best in cohort and epidemiological work. A 12 is not a catastrophe and it is not a diagnosis. It is a number worth repeating and reading in context.

What it tends to travel with

Insulin resistance commonly clusters with a particular lipid pattern, with higher blood pressure, with central adiposity and with fat in the liver. Across eleven prospective studies, people in the highest HOMA-IR category had about 1.4 times the risk of developing high blood pressure.

Each of those findings can also arise for its own reasons, and type 2 diabetes itself is heterogeneous. These are associations. They tell me where to look, not what to treat.

What clarifies a 12

Most of it you may already have. A repeat fasting insulin and glucose, drawn under comparable conditions and ideally at the same laboratory. The fasting glucose and A1c that remain the standard tests. Where it is clinically indicated, the two-hour glucose after a sugar drink, which also moved early in Whitehall II and is a standard diagnostic test in its own right. And the trajectory: weight and waist over time, and the rest of the cardiometabolic picture.

Lipids, including ApoB, and an assessment of the liver belong to that broader cardiometabolic evaluation. They are worth having, and they are not tests that confirm insulin resistance.

Sometimes two weeks of a continuous glucose monitor, and it is optional. The randomised evidence for glucose monitoring as a behaviour-change tool is dominated by people with diabetes, where the glycaemic benefit is modest; in people without diabetes it is much thinner. The American Diabetes Association also holds that the evidence is insufficient to use continuous glucose monitoring to screen for or diagnose prediabetes or diabetes. I use it as an optional teaching tool, not a treatment and not a test.

What improves insulin sensitivity

Resistance training has direct evidence here: across studies in adults with overweight or obesity without diabetes, it has improved fasting insulin and HOMA-IR. Skeletal muscle is a major site of insulin-stimulated glucose disposal.

Walking after meals has evidence for post-meal glucose rather than for fasting insulin, and the two are worth keeping apart. Ten minutes after each main meal lowered post-meal glucose by about 12 percent against the same walking done at one time of day, in a crossover trial of 41 adults with type 2 diabetes. Sleep matters too: six nights of four hours in bed reduced glucose tolerance in eleven healthy young men, a small study, and I say so. And in the right patient, pharmacotherapy, which is a decision made on the whole picture, never on a fasting insulin alone.

Then re-test the insulin, under comparable conditions and ideally at the same laboratory, beside the glucose and A1c that remain the standard.

The point

Fasting insulin is not a diagnosis and there is no single number that defines healthy metabolism. But in the right context it can add information that glucose and A1c do not provide by themselves. I use it as one piece of a longitudinal metabolic assessment: interpreted beside standard testing, repeated when necessary, and acted on only when the whole picture supports a decision.

References

Sources for this essay.

Creating Health with Dr. Tagge

The science of health, connected with the way we live: sourced and translated, with no hype. A new issue by email every Friday.

Get Creating Health by email

Daniel Tagge, MD

Written by

Daniel Tagge, MD

Board-certified family physician and founder of Tagge Health. Certified in Health Optimization Medicine.

About Dr. Tagge

Lab reference

Fasting insulin: the laboratory range, the evidence, and what moves itNo guideline defines an optimal fasting insulin. In the cohorts that followed healthy adults for years, the lowest-risk groups sat in the single digits, and risk of later diabetes and heart disease rose from about 9 to 12 µIU/mL upward, well inside most laboratory reference ranges.
Hemoglobin A1c: the laboratory range, the evidence, and what moves itIn the long cohorts, adults whose hemoglobin A1c sat between 5.0 and about 5.4 percent had the lowest risk of later diabetes and coronary disease; risk rose step by step from 5.5 upward, well inside the laboratory's normal range, and below 5.0 all-cause mortality was higher rather than lower. No trial has tested a target inside the normal range.
Fasting glucose: the laboratory range, the evidence, and what moves itA fasting glucose below 100 mg/dL is normal by the diagnostic cutoff, but in cohorts that followed healthy adults for a decade or more, the risk of later diabetes began rising from about 87 mg/dL and was about twice as high at 95 to 99 as below 85, while all-cause mortality was lowest at about 80 to 94 and rose again below about 70 to 80.
Triglycerides: the laboratory range, the evidence, and what moves itLaboratories call triglycerides under 150 mg/dL normal, the Copenhagen cohorts found heart attack risk already higher above about 88 than below it, the American Heart Association has called a fasting value under 100 'optimal', and genetic studies show the risk travels with the number of ApoB particles, so a triglyceride number is best read as a signal about remnant particles and insulin resistance rather than a target of its own.
HDL cholesterol: the laboratory range, the evidence, and what moves itIn cohorts, mortality is lowest with HDL cholesterol in the middle, about 55 to 70 mg/dL in pooled data and 73 (men) to 93 (women) in Copenhagen, and rises again at both ends; but Mendelian randomization and four large drug trials show that raising HDL cholesterol does not lower heart risk, so the number is a marker of metabolic health, not a lever.
Uric acid: the laboratory range, the evidence, and what moves itUrate causes gout, and the risk climbs steeply across the laboratory range: about 1 percent over 15 years below 6 mg/dL, 2.7 times that between 6 and 6.9, and about half of people above 10, which is why the gout guideline target is below 6. Higher urate also tracks with heart disease, kidney disease and death in cohorts, but genetic studies and three randomized trials of urate lowering found no causal effect on those outcomes, so outside gout and stones the number is best read as a marker of insulin resistance, fructose, alcohol, body weight, diuretics and kidney function.
HOMA-IR: the laboratory range, the evidence, and what moves itHOMA-IR is a calculation, not a measurement, and no guideline defines a normal or optimal value. In the one US cohort that followed adults without diabetes to death, all-cause mortality was lowest in the quarter at or below about 1.4 and 64 percent higher above about 2.8, and cutoffs that separated healthy from at-risk adults in other populations fell between about 1.8 and 2.6, all well under the 2.0 to 2.9 that some laboratories print.
SHBG: the laboratory range, the evidence, and what moves itSHBG is the liver protein that decides how much testosterone and estradiol is free, and its level reads out the liver's metabolic state: in nested case-control studies of 359 women and 170 men, the highest SHBG quartile had about one tenth the odds of type 2 diabetes of the lowest, genetic variants that raise SHBG lower diabetes risk, and SHBG falls with liver fat and rises with age, thyroid excess and oral estrogen. In 149,436 UK Biobank men and nine pooled cohorts, higher SHBG went with higher, not lower, mortality, so no guideline names an optimal value; what SHBG changes is the reading of a total testosterone, which is why guidelines ask for a free testosterone whenever SHBG is abnormal or the total is borderline.

More on choosing care

Read alongside.

Bring it to the practice

Want this read for your biology?

Reading a pattern in general is not the same as reading it in you. The Precision Health Assessment is one visit that takes your history, your records and your results together, and ends with a written plan.

Explore working with Dr. Tagge