Two People With the Same Blood Sugar Can Get A1c Results a Full Point Apart

Hemoglobin A1c may be the most trusted number in metabolic medicine. It anchors the diagnosis of diabetes, it defines prediabetes, and for most people it is the only glucose-related marker their annual physical ever reports. The pitch is simple: red blood cells pick up glucose as they circulate, the glucose sticks to hemoglobin, and the percentage of hemoglobin that ends up glycated works out to a three-month average of your blood sugar.

The pitch is mostly true. What almost nobody explains is the size of the error bars, or where they come from. The short version: A1c is not a measurement of glucose. It is a measurement of glucose exposure filtered through the lifespan of your red blood cells, and that lifespan varies from person to person by enough to move the result a full percentage point in either direction. In A1c terms, a full point is the entire distance between normal and diabetic.

An average of an average

Start with what the assay actually reports. The number on your lab slip is the percentage of your hemoglobin that carries a glucose adduct. Glucose attaches slowly and essentially irreversibly, so an older red blood cell has had more time to accumulate glycation than a young one. Your A1c is therefore a blend: a population of cells of every age from freshly released to nearly recycled, each carrying a glycation load proportional to both the ambient glucose and its own time in circulation.

That second variable is the problem. The standard model assumes everyone's red blood cells live about the same length of time, roughly 120 days. They do not.

In 2008, Cohen and colleagues measured red blood cell survival directly in hematologically normal people, tagging cells with a biotin label and following them in circulation, and published the results in Blood (DOI). The mean age of circulating red cells ranged from 38 to 60 days in the nondiabetic subjects, and the authors concluded that this variation in survival was large enough to cause clinically important differences in A1c at the same mean blood glucose. Nothing was wrong with anyone's glucose metabolism. Nothing was wrong with the assay. The cells simply lived longer in some people, accumulated more glycation, and produced a higher number.

A person whose cells turn over quickly runs an A1c that reads lower than their true glucose exposure. A person whose cells linger runs one that reads higher. Neither of them can see this on a standard panel, because red cell lifespan is not measured in routine care. It is assumed.

The size of the gap, measured properly

The most rigorous accounting of this problem comes from Malka, Nathan, and Higgins at Massachusetts General Hospital, published in Science Translational Medicine (DOI). They built a mechanistic model of hemoglobin glycation and red cell kinetics, then tested it against large sets of continuous glucose monitoring data, where the true average glucose is actually known rather than inferred.

Three of their findings deserve to be quoted almost verbatim. First, patients with identical A1c values had true average glucose concentrations that differed by more than 60 mg/dl. For scale, the true average glucose of a nondiabetic and a poorly controlled diabetic may differ by less than 15 mg/dl. Second, the standard estimated-average-glucose conversion, the one printed on lab reports, produced errors greater than 15 mg/dl in one of every three patients. Third, when the authors derived each patient's red cell age from the data, that single variable explained essentially all of the glucose-independent variation in A1c.

Read that middle number again. The tool used to grade three months of your metabolic life misestimates one in three people by more than the entire gap between healthy and diseased.

This is also why the tidy conversion formula deserves skepticism. The equation everyone uses, average glucose equals 28.7 times A1c minus 46.7, comes from the international ADAG study published in Diabetes Care (DOI). It is a real regression built from roughly 2,700 glucose readings per subject across 507 people, and it fits the population well. But a population regression is not a personal guarantee, and the ADAG authors never claimed otherwise. Beck and colleagues later showed in Diabetes Care (DOI) that for a given A1c of 8 percent, an individual's actual mean glucose could plausibly range from roughly 155 to 218 mg/dl. The average is honest. Applying the average to an individual is where the trouble starts.

The screening consequence: A1c misses people

If A1c only wobbled around a correct center, it would be a precision problem. The screening data suggest something worse: used alone as a diagnostic gate, it systematically under-detects.

Cowie and colleagues ran the comparison on the U.S. population using NHANES data, published in Diabetes Care (DOI). Using the A1c threshold of 6.5 percent, the prevalence of undiagnosed diabetes came out at one-third of what glucose-based criteria found in the same population. For the high-risk category, the A1c definition captured roughly one-tenth as many people as glucose criteria. Same bodies, same blood, different test, and two-thirds or more of the at-risk group walks out with a normal result.

This is the practical reason a fasting glucose and a fasting insulin sit next to A1c on any panel worth drawing. A1c integrates the past. Fasting insulin flags the compensation that precedes glucose elevation, often by years. The failure mode of relying on A1c alone is not a noisy number. It is a reassuring one.

Predictable interference: iron, anemia, variants

Beyond the individual lifespan lottery, there are named conditions that push A1c in known directions, and they are common.

Iron deficiency, with or without anemia, raises A1c without any change in glucose. English and colleagues reviewed the evidence systematically in Diabetologia (DOI): iron-deficient states produced spuriously elevated A1c against matched controls, while several non-iron-deficiency anemias, the hemolytic ones in particular, shortened red cell survival and dragged A1c down. Iron deficiency is among the most common nutritional deficits in the world, and it is disproportionately present in exactly the demographic most likely to be screened on A1c alone: adult women.

Ancestry matters too, and not as a rounding error. Bergenstal and colleagues put continuous glucose monitors on 104 Black and 104 white participants for twelve weeks and compared measured glucose against A1c, published in Annals of Internal Medicine (DOI). At the same measured mean glucose, A1c averaged 0.4 percentage points higher in Black participants. The alternative glycemic markers in the same study, glycated albumin and fructosamine, showed no such difference, which points at hemoglobin-specific glycation biology rather than glucose control. Depending on which side of a diagnostic threshold you live near, that 0.4 can label one person diabetic and clear another with identical physiology.

Hemoglobin variants, chronic kidney disease, recent blood loss or transfusion, and pregnancy each distort the number further, in directions that are documented but rarely flagged on a report.

The cross-checks that exist

None of this makes A1c useless. It is cheap, stable, standardized, and strongly associated with long-term complications at the population level. The correct response to a marker with known blind spots is not to discard it but to refuse to let it testify alone.

Fructosamine and glycated albumin glycate serum proteins instead of hemoglobin, so they reflect the past two to four weeks and owe nothing to red cell lifespan. In the ARIC cohort, Selvin and colleagues followed more than 12,000 adults for two decades and found both markers predicted incident diabetes, retinopathy, and kidney disease with prognostic strength comparable to A1c, published in Lancet Diabetes and Endocrinology (DOI). When A1c and fructosamine disagree, that disagreement is itself information: it usually means the red cell assumption is failing.

Direct glucose measurement is the other check. A fasting glucose is a snapshot, a continuous glucose monitor is the whole film, and neither passes through a red blood cell on the way to the result. When a two-week CGM average and an A1c point in different directions, the CGM is measuring glucose and the A1c is measuring glucose plus cell biology. Trust the film.

Reading your own

Three questions do most of the work when an A1c comes back.

Whether anything on the same panel undermines the red cell assumption: ferritin and iron studies for deficiency, a CBC for anemia or unusual red cell indices, kidney markers for reduced clearance. An A1c read next to a ferritin of 8 is a different document than the same number next to a ferritin of 90.

Whether the A1c agrees with the direct glucose evidence. A normal A1c beside an elevated fasting glucose or a climbing fasting insulin is not reassurance. It is a discrepancy, and discrepancies between a filtered average and a direct measurement usually resolve in favor of the direct measurement.

And whether your own A1c has been stable relative to itself. Because red cell lifespan is a personal constant that changes slowly, your A1c trend against your own baseline carries more meaning than your position against a population range. A 0.4 rise within the normal range is a real signal even though both numbers pass.

What any of it means for an individual case, and whether a discordant result warrants an oral glucose tolerance test, a CGM trial, or nothing at all, is a judgement for a licensed clinician looking at the whole picture. The general point survives every caveat: A1c is a good average built on an assumption that is false for a meaningful fraction of the people it is applied to, and the fix is not a better average. It is more than one line of evidence.

The Peptide Dispatch publishes research summaries for educational purposes. This material describes what the published literature reports about laboratory markers and physiology. It is not medical advice, it is not a diagnosis, and it does not recommend or endorse any treatment, product, or provider. Laboratory results require interpretation by a licensed clinician in the context of your individual history. Always discuss testing and treatment decisions with a qualified healthcare professional.

Research summaries in this article are based on articles retrieved from PubMed.