What Counts as a Normal A1C Level?

What this actually measures

A1C (glycated haemoglobin, also written HbA1c) measures the fraction of haemoglobin in red blood cells that has glucose chemically attached to it. Because red blood cells circulate for roughly three months before being replaced, the percentage reflects an average of blood glucose exposure over that stretch rather than the level at any single moment.

An A1C between 5.7% and 6.4% is not a diagnosis. It is also not nothing. That band — “prediabetes” on the American Diabetes Association’s chart — is where most people who get an A1C back actually land, and it is the range the popular explainers tend to gloss over fastest, right when the reader most needs the nuance.

The three numbers, and what they’re actually cut points for

The bands are simpler than the anxiety around them suggests. Below 5.7% is normal. From 5.7% to 6.4% is prediabetes. 6.5% and above, confirmed on a second test, is diabetes. That’s the ADA’s line, and the WHO and NHS land on the same diabetes threshold — they mostly just report it differently, in mmol/mol rather than percent, with 48 mmol/mol as the equivalent of 6.5%.

A cut point is a decision, not a discovery. Somewhere a body of clinicians looked at outcome data — who went on to develop diabetes complications, at what average glucose exposure — and drew a line where the risk curve started climbing meaningfully. That’s a genuinely well-supported line, not an arbitrary one. But it is still a line drawn across a continuum, which means a 5.6% and a 5.8% describe two people who are more alike than the different labels on their lab reports suggest.

What the test is actually measuring

A1C, also written HbA1c, measures the percentage of haemoglobin — the oxygen-carrying protein in red blood cells — that has glucose attached to it. Glucose in the bloodstream sticks to haemoglobin gradually and semi-permanently over the life of the red blood cell, which normally circulates for around three months before the body replaces it.

That’s the entire mechanism, and it’s why the number behaves differently from a fasting glucose reading. A fasting test is a photograph: it shows you the blood sugar at 8am on the day you had it drawn, and a bad night’s sleep or an unusually late dinner the evening before can move it. A1C is closer to a long exposure — it can’t tell you what happened yesterday, but it’s much harder to spike or dip with one unusual day.

That averaging property is also its main limitation. Two people can have identical average glucose over three months and very different day-to-day patterns — one steady, one swinging widely between highs and lows — and get the same A1C. The number tells you the average was elevated or wasn’t; it does not tell you whether the average was steady or the product of large swings that cancelled out.

Where the number can mislead

This is the part worth being direct about, because it’s the part most explainers skip. A1C is a proxy — glucose attached to haemoglobin — and the proxy depends on an assumption: that red blood cells are turning over at the normal rate.

When that assumption breaks, so does the accuracy of the result, in either direction:

• Conditions that shorten red blood cell lifespan — some anaemias, recent significant blood loss, pregnancy — give glucose less time to attach, which can make A1C read lower than actual average glucose.
• Conditions that lengthen it, or certain haemoglobin variants that interfere with some assay methods, can push the number the other way, or make it unreliable regardless of direction.
• Recent blood transfusion effectively dilutes the red blood cell population being measured with cells of a different age, muddying the three-month window the test relies on.

None of this makes A1C an unreliable test in general — for someone without one of these conditions, it’s one of the better-validated tools in routine bloodwork, which is exactly why it’s used for diagnosis rather than dropped for something else. It does mean that a result that doesn’t match how someone feels, or doesn’t match home glucose meter readings, is a reason to ask a clinician whether one of these situations applies — not a reason to distrust lab testing generally.

Why one result gets a second test

The ADA’s diagnostic criteria call for confirmation: a second A1C, on a different day, or a second test of a different type — fasting glucose or an oral glucose tolerance test — before diabetes is formally diagnosed on the basis of A1C alone, unless someone already has clear symptoms alongside a clearly elevated result.

That’s not bureaucratic caution. Lab assays carry a small amount of run-to-run variability even in someone whose biology hasn’t changed at all, and a single result sitting right at a cut point — 6.4% versus 6.6%, say — is exactly where that variability matters most. Confirming on a second sample, sometimes with a second method, is how the number earns the diagnosis rather than the diagnosis riding on one measurement.

Prediabetes: the band that gets flattened in headlines

“Prediabetes” sounds like a countdown, and coverage of it often reads that way. It is more accurately a risk category: a documented, elevated-versus-normal range that is associated with higher future risk of developing diabetes, which is different from a guarantee of progression.

The honest, uncomfortable part is that we should not overstate what a single A1C in that band tells any one individual about their own trajectory — population risk curves describe groups, and where any one person’s number goes next depends on factors an A1C alone doesn’t capture. What the band reliably does is flag a result worth discussing with a clinician and worth rechecking rather than filing away, particularly alongside other risk factors such as excess abdominal weight, physical inactivity, or a family history of type 2 diabetes.

A1C across age and life stage

The ADA and WHO cut points for diagnosis are not published as age-adjusted; the 5.7% and 6.5% lines are the same lines regardless of the age on the lab slip. Where age does enter the picture is downstream of diagnosis, in how tightly a diagnosed person’s blood sugar gets managed — some clinical guidance for older adults, once diabetes is already established, allows a higher individualised treatment target, weighing the risks of overly tight control against the benefits. That’s a different decision from where “normal” sits on a first reading.

One genuine technical limitation: A1C is not considered reliable in infants under roughly six months, because fetal haemoglobin — still present at that age — interferes with some assay methods used to run the test. Pregnancy is its own separate case, with its own glucose testing protocols, precisely because red blood cell turnover and blood volume both shift during pregnancy in ways that can affect the reading.

Reading your own result

The table above gives you the bands. What it can’t give you is context a clinician has and a static reference range doesn’t: your other bloodwork, your symptoms if any, your family history, and whether anything in your recent health — a blood donation, a new diagnosis, pregnancy — falls into the list of things that can shift the number independent of glucose. A result close to a boundary is a reason to ask a specific question — “should we recheck this, and with what test” — rather than a reason to self-diagnose off a chart, however good the chart is.