Is HbA1c Always Accurate?

HbA1c is one of the more tightly controlled tests a laboratory runs. It is standardised internationally, laboratories are surveyed against known reference samples, and for most people, most of the time, the number that comes back is a fair reflection of the last three months. So the honest answer to "is it always accurate" is: usually, but not always — and the useful part is knowing what kind of inaccuracy you might be looking at.

Three separate things can push an HbA1c result away from the truth, and they have almost nothing to do with each other. The lab has a normal margin of error. Not every HbA1c machine is equally good. And in some people, the biology the test rests on simply does not hold. Each of those calls for a different response, so it is worth being able to tell them apart.

How much wobble is normal, even in a good lab

Every measurement has a margin, and HbA1c is unusual in that we can put a number on it. The NGSP (opens in a new tab), which standardises HbA1c testing worldwide, publishes results from a survey in which samples of a precisely known value are sent out to laboratories to be run blind. In the first 2025 survey (opens in a new tab), samples with reference values of 5.60%, 6.00%, 7.97%, 8.45% and 9.84% went out, and 96.8% to 98.8% of participating laboratories came back within 6% of the target. Average bias across method groups ran from 0.00 to 0.36 percentage points of HbA1c.

Translated into something you can use: a true HbA1c of 7.0% will quite legitimately be reported as 6.8% or 7.2% by a competent laboratory. That is not an error anyone did anything wrong to cause. It is the ordinary resolution of the test.

Not every HbA1c machine is the same machine

The same survey shows how much the method matters. Between-laboratory variation for individual methods ranged from 0.4% to 4.0% — a tenfold spread. Eighteen methods, used by around 60% of participating laboratories, stayed inside the recommended limit on every sample; four methods exceeded it. Both kinds of machine are in use, and a report does not usually tell you which one ran your sample.

This has one practical consequence worth acting on: when you can, get your repeat HbA1c done at the same laboratory as the last one. Comparing a result from one lab against a result from another adds a difference that has nothing to do with your blood sugar, and it is exactly the kind of difference that gets misread as progress or backsliding.

The finger-prick HbA1c machine in the clinic

Point-of-care HbA1c devices — the small desktop analysers used in clinics, pharmacies and screening camps that give a result from a finger-prick in a few minutes — are genuinely useful. They remove the venous draw, the wait, and the second trip. They are also, as a category, less reliable than a laboratory, and the gap is larger than most people assume.

The European Reference Laboratory for Glycohemoglobin tested 19 of these devices (opens in a new tab) against certified reference methods. Only 5 of the 19 met both the IFCC and NGSP quality criteria. Nine met neither — despite most of them holding a valid certificate at the time of testing.

“The analytical performance of the devices was very poor overall, with 74% not meeting both IFCC and NGSP criteria. These evaluations were performed under ideal conditions; performance may worsen further when used in a clinical setting.”

When it is your blood, not the machine

The third category is different in kind. Here the machine measures correctly — it just measures something that no longer means what we assume it means. HbA1c reads as a three-month average because red blood cells are assumed to live a fairly standard length of time. When they do not, the number shifts without your glucose shifting.

The ADA's Standards of Care in Diabetes—2026 (opens in a new tab) names the conditions that interfere: haemolytic and other anaemias, glucose-6-phosphate dehydrogenase deficiency, recent blood transfusion, drugs that stimulate red cell production, kidney failure, and pregnancy. It also notes that A1C cannot be measured at all in sickle cell disease or other homozygous haemoglobin variants.

Two of those deserve emphasis in Pakistan. Anaemia is common enough here that it is a live possibility for a large share of people getting tested — we covered which direction each type pushes the result in can anaemia affect your HbA1c result](/blog/can-anemia-affect-hba1c-results). And beta-thalassaemia trait, with an estimated [carrier rate of 5–7% of the population (opens in a new tab), is a haemoglobin variant that interferes with some testing methods and not others.

Kidney disease is the one most often missed. The NGSP (opens in a new tab) states that HbA1c underestimates glycaemic control in people with diabetes on dialysis, where glycated albumin is a more robust indicator — which means a comfortable-looking HbA1c in someone on dialysis should not be taken at face value. Late pregnancy runs the other way: the same source notes HbA1c rises in late pregnancy even without diabetes, because of iron deficiency.

Accurate and complete are not the same thing

One last distinction, because it accounts for more confusion than all of the above put together. A perfectly accurate HbA1c is still an average, and averages hide their own extremes. Someone swinging between 60 and 300 mg/dL and someone sitting steadily at 170 can produce the same result. That is not the test being wrong — it is the test answering a narrower question than the one you had. If your readings and your HbA1c seem to disagree, this is usually why, and it is worth reading [HbA1c high but sugar readings normal](/blog/hba1c-high-but-blood-sugar-readings-normal) before suspecting the laboratory.

If you do not trust a result

1. 1Check the size of the discrepancy first. A gap of 0.2–0.3 percentage points from what you expected is normal measurement margin, not a problem to investigate.
2. 2Note whether this test was run at the same laboratory as the last one, and whether either was a finger-prick point-of-care device rather than a venous sample.
3. 3Tell your doctor about anything that affects red blood cells — known anaemia, thalassaemia trait in the family, a recent transfusion, kidney disease, pregnancy, or a recent course of iron.
4. 4Bring your own readings for the same period. A discrepancy you can show is far easier to act on than one you can only describe.
5. 5Ask whether a repeat is worth doing before any treatment decision is made on a number that does not fit.

None of this makes HbA1c a bad test. It remains the most useful single number in diabetes care, and for the large majority of people it is doing exactly what it claims. It is simply not a number to read to two decimal places, or to interpret alone.

Give the lab result something to sit against

An HbA1c is one measurement every few months, with its own margin. A record of your own readings — with dates and times — is what tells you whether the lab number fits the rest of the picture, and what makes a genuine discrepancy visible instead of vague.

Start your sugar diary (opens Diatic on Google Play in a new tab)