HbA1c (Hemoglobin A1c)

What HbA1c (Hemoglobin A1c) actually measures — the analyte, and the assay

The analyte is HbA1c: hemoglobin A whose beta-chain N-terminal valine carries a glucose molecule attached without an enzyme, rearranged into a stable ketoamine. It is a ratio, not a concentration — glycated hemoglobin over total hemoglobin, reported either as a percentage on the NGSP scale or in mmol/mol on the IFCC scale Sacks 2012. Both halves of that ratio can move, and that is the whole difficulty with the marker.

Five different physical properties are used to measure it, and a hemoglobin variant hits each one differently. Cation-exchange HPLC separates by charge. Boronate affinity binds the cis-diol of the sugar itself. Immunoassay uses an antibody raised against the glycated N-terminal hexapeptide. Enzymatic methods digest the chain with a protease and count fructosyl valine with fructosyl peptide oxidase. Capillary electrophoresis separates by charge in free solution. A variant that changes the charge of the beta chain wrecks the charge-based methods and leaves the affinity method alone; a variant that changes the amino acids near the N-terminus wrecks the antibody and the protease and leaves the charge methods alone Chen 2022.

The size of that effect has been measured properly. In an evaluation of hemoglobin C, D, E and S traits across fifteen HbA1c methods, every one of the fifteen showed a statistically significant effect from at least one trait, and clinically significant interference was found on the Tosoh G11, the Roche b 101 and the Siemens DCA Vantage — a point-of-care analyzer Rohlfing 2021. A multicenter study of 104 samples carrying 33 different variants, comparing HPLC against a whole-blood enzymatic assay, found clinically relevant discordance for Hb Okayama, HbAE and Hb Lepore Zechmeister 2022. And in a comparison of six commonly used methods against the IFCC reference measurement procedure on 63 variant samples, raised fetal hemoglobin produced a significant negative bias on all six — borate affinity, immunoassay and enzymatic included Li 2024.

So the practical rule: ask whether the result is NGSP-certified and which chemistry produced it. A point-of-care fingerstick and a laboratory HPLC are not interchangeable, and if you carry a hemoglobin variant — which most carriers do not know — the two can disagree by enough to move you across a diagnostic line.

HbA1c (Hemoglobin A1c): what changes the blood, and what only changes the reading

What changes the glycated hemoglobin in your blood — ranked:

1. Mean glucose over roughly the last 8 to 12 weeks, weighted heavily toward the most recent month. Around half the signal comes from the last 30 days, which is why HbA1c is not the flat 3-month average it is sold as.
2. Red cell lifespan — the denominator, and the thing almost nobody adjusts for. A red cell lives about 120 days and accumulates glucose the whole time, so a population of younger cells is a less glycated population at identical glucose. Anything that shortens survival lowers HbA1c: hemolysis, recent blood loss, blood donation, erythropoietin, hypersplenism. Anything that lengthens it raises HbA1c: splenectomy, and above all iron deficiency.
3. Iron deficiency specifically, because it is common and it moves the number the wrong way. In 324 non-diabetic patients across four anemia types, iron-deficiency anemia and sickle cell disease both raised HbA1c, and treating the iron deficiency significantly lowered it Alzahrani 2023. In 59 non-diabetic women, HbA1c tracked hemoglobin concentration, ferritin and red cell count Bindayel 2021. A dieting female lifter with a ferritin of 12 can be handed a 5.8% and told she is prediabetic on the strength of her iron status.
4. Pregnancy, which shortens red cell survival and expands plasma volume, lowering HbA1c by a few tenths.
5. Chronic kidney disease, which does both things at once: it shortens red cell survival through anemia and it carbamylates hemoglobin with urea-derived cyanate. In 1,516 people with diabetes and CKD the association between HbA1c and renal progression was modified by both carbamylation and anemia Tang 2023 Hassanein 2022.

What changes only the reading — and on this marker the two lists look identical from the outside:

1. A hemoglobin variant, on a method that cannot cope with it. This is a measurement artifact that presents exactly like a glycemic change, and 15 of 15 methods showed some effect Rohlfing 2021.
2. Raised fetal hemoglobin, which biased all six methods downward against the IFCC reference procedure Li 2024.
3. Carbamylated hemoglobin in uremia, which co-elutes with HbA1c on some charge-based systems and is not glycated hemoglobin at all Tang 2023.
4. Which analyzer, and whether it was a point-of-care device Rohlfing 2021 Sacks 2012.

The distinction matters because the responses differ. A genuinely high HbA1c is a reason to change what you eat. A high HbA1c from iron deficiency is a reason to check a ferritin, and a low one from donating blood every eight weeks is a reason to run a fructosamine instead.

Reference interval or decision threshold — which kind of number HbA1c (Hemoglobin A1c) is

These are decision thresholds, not a reference interval, and the difference is the reason the numbers look so tidy. Nobody measured the middle 95% of a healthy population and got 5.7 and 6.5. Those cut-points were chosen against outcomes — principally the prevalence of diabetic retinopathy, which rises steeply above roughly 6.5% — and 5.7 to 6.4% is a risk band, a stretch of the scale where progression to diabetes is more likely, not a disease Selvin 2021.

That has a consequence people find counterintuitive. A threshold is a policy, and it can be moved without anybody's blood changing. It also means the threshold on one test does not have to agree with the threshold on another, and it does not. In 40,667 participants, diabetes prevalence was 5.7% by fasting plasma glucose and 7.5% by HbA1c in the same people; against fasting glucose as the comparator, HbA1c had a sensitivity of 58.20% (95% CI 56.43–59.96) and a specificity of 98.59% Abdul Murad 2021. Two tests, two thresholds, one population, and roughly a third disagreement.

The two scales are a separate trap. NGSP percent and IFCC mmol/mol are different units of the same quantity — 5.7% is about 39 mmol/mol, 6.5% is about 48 — and a report in mmol/mol read as a percentage is a number roughly seven times too large Sacks 2012.

Where the optimal number comes from is different again. A target of under 5.4%, or the 4.8 to 5.2% often quoted, is not a threshold anybody validated against an outcome. It is the observation that risk is continuous below the diagnostic line, applied as a goal. That is a defensible use of epidemiology and it is not the same kind of claim as 6.5% Selvin 2021.

How you would know your HbA1c (Hemoglobin A1c) was wrong — and when to redraw

The retest interval is set by the red cell, and it is 120 days long. A red cell that is 100 days old today will still be in circulation in three weeks carrying every glucose molecule it picked up before you changed anything. Roughly half the HbA1c signal comes from the most recent 30 days and roughly a quarter from the 30 to 60 days before that, so a redraw at three weeks is mostly measuring the period you were trying to leave behind. Three months is the minimum honest interval and it is not a round number, it is the cell's lifespan.

Conditions that must match, or the second value is not comparable: the same laboratory and the same method Rohlfing 2021; no blood donation, transfusion or significant blood loss in the preceding 120 days; the same iron status — check a ferritin alongside Bindayel 2021; not pregnant; and stable kidney function Tang 2023.

What would have to change for the retest to mean something. A real glycemic improvement should carry its neighbors: fasting insulin down, triglycerides down, ALT down, waist down. If HbA1c falls and nothing else moves, the first suspects are the red cells, not the pancreas.

How you would know the value was wrong. Run the marker against something with a different denominator. Fructosamine and glycated albumin depend on albumin rather than hemoglobin, so if the HbA1c says one thing and the fructosamine says another, the disagreement localizes to the red cell Chen 2022 Hassanein 2022. Continuous glucose monitoring is the harder test, and it fails more often than clinicians expect: across 641 encounters comparing laboratory HbA1c with the CGM-derived glucose management indicator, only 11% agreed within 0.1% and half differed by 0.5% or more Perlman 2021. A CBC and a ferritin are the cheapest confirmatory pair, and an unexpectedly low HbA1c in someone with no reason to have one is a hematology question before it is an endocrine one.

What HbA1c (Hemoglobin A1c) cannot tell you

It cannot see the shape of your day. HbA1c is an average, and an average conceals excursions. In 153 people without diabetes wearing continuous monitors, the mean glucose management indicator ran 0.59% higher than the laboratory HbA1c, and discordance of 0.4% or more was present in 71% of them Shah 2023. A normal HbA1c does not exclude repeated post-meal spikes; it says the area under the curve, averaged over a season and filtered through your red cells, is unremarkable.

It cannot tell you whether the number moved because of glucose or because of hemoglobin. That is the single most common wrong inference on this page, and it runs in both directions: the athlete who donates blood quarterly reads reassuringly low, and the iron-deficient dieter reads alarmingly high Alzahrani 2023 Bindayel 2021.

It is not a screening test in chronic kidney disease. Anemia, shortened red cell survival, carbamylation and erythropoietin therapy all act at once, in the same person, in different directions Tang 2023 Hassanein 2022.

A single value does not diagnose anything. Diagnosis needs a second abnormal test, and the two available tests disagree with each other about a third of the time Abdul Murad 2021.

The wrong inference readers actually draw is that a 5.2% closes the metabolic question. It does not. HbA1c is a late marker: fasting insulin can be several times optimal, with visceral and hepatic fat accumulating, for years before this number moves at all. A normal HbA1c with an untested insulin is not a clean metabolic panel, it is an incomplete one.

Sources read for these sections

• Rohlfing C, et al. Evaluation of interference from hemoglobin C, D, E and S traits on measurements of hemoglobin A1c by fifteen methods. Clinica Chimica Acta 2021 · PMID 34352282
• Zechmeister B, et al. Analytical interference of 33 different hemoglobin variants on HbA1c measurements comparing high-performance liquid chromatography with whole blood enzymatic assay: A multi-center study. Clinica Chimica Acta 2022 · PMID 35378091
• Li M, et al. Interference of hemoglobin variants with HbA1c measurements: comparison of 6 commonly used HbA1c methods with the IFCC reference method. Laboratory Medicine 2024 · PMID 38801245
• Sacks DB. Measurement of hemoglobin A(1c): a new twist on the path to harmony. Diabetes Care 2012 · PMID 23173136
• Chen Z, et al. Interpretation of HbA1c lies at the intersection of analytical methodology, clinical biochemistry and hematology (Review). Experimental and Therapeutic Medicine 2022 · PMID 36382101
• Alzahrani BA, et al. The effect of different types of anemia on HbA1c levels in non-diabetics. BMC Endocrine Disorders 2023 · PMID 36709277
• Bindayel IA. Influence of iron deficiency anemia on glycated hemoglobin levels in non-diabetic Saudi women. Journal of International Medical Research 2021 · PMID 33641472
• Tang M, et al. The Impact of Carbamylation and Anemia on HbA1c's Association With Renal Outcomes in Patients With Diabetes and Chronic Kidney Disease. Diabetes Care 2023 · PMID 36399777
• Perlman JE, et al. HbA1c and Glucose Management Indicator Discordance: A Real-World Analysis. Diabetes Technology and Therapeutics 2021 · PMID 33253015
• Shah VN, et al. Discordance Between Glucose Management Indicator and Glycated Hemoglobin in People Without Diabetes. Diabetes Technology and Therapeutics 2023 · PMID 36790875
• Abdul Murad NA, et al. Discordance between Fasting Plasma Glucose (FPG) and HbA1c in Diagnosing Diabetes and Pre-diabetes in The Malaysian Cohort. Journal of the ASEAN Federation of Endocrine Societies 2021 · PMID 34966195
• Selvin E. Hemoglobin A1c-Using Epidemiology to Guide Medical Practice: Kelly West Award Lecture 2020. Diabetes Care 2021 · PMID 34548283
• Hassanein M, et al. Assessment of glycemia in chronic kidney disease. BMC Medicine 2022 · PMID 35414081