HbA1c: What the Number Actually Means

HbA1c is the most widely ordered metabolic test in medicine and one of the most widely over-read. It estimates average glucose across roughly three months, which makes it excellent for diagnosis and trend-following. It is also an average, and averages hide exactly the behaviour that does the damage. Knowing what the number is built from tells you when to trust it and when to look elsewhere.

What the number actually is

Haemoglobin A1c is haemoglobin — the oxygen-carrying protein inside red blood cells — with glucose stuck to it. The attachment is non-enzymatic: nothing catalyses or regulates it, and it happens simply because glucose and protein are in the same place at the same time.

That mechanism is the reason HbA1c means anything. It is not a hormone level the body adjusts; it is a physical record of exposure, closer to soot accumulating in a chimney than to a regulated signal. The same reaction is happening to every other long-lived protein in your body — in vessel walls, in the lens of the eye, in kidney tissue. HbA1c is measurable because red cells are easy to sample. The process it reports on is happening everywhere.

Why three months, and why not equally

Red blood cells live about 120 days, and at any moment your circulation holds cells of every age. Older cells have been exposed longer and carry more glycation; younger cells carry less. HbA1c averages across that whole pool.

The weighting is the part people miss. Because older cells are continuously removed, recent glucose exposure contributes disproportionately: the value leans toward the most recent weeks rather than being a flat three-month mean. So a genuine change in control starts showing up well before three months have elapsed — and someone who tightens up for a fortnight before a blood test will move the number somewhat. Not enough to fake a normal result, but enough to matter when interpreting a small change.

Conversion to estimated average glucose (Nathan et al., Diabetes Care 2008):

• HbA1c 5.0% ≈ eAG 97 mg/dL
• HbA1c 6.0% ≈ eAG 126 mg/dL
• HbA1c 7.0% ≈ eAG 154 mg/dL
• HbA1c 8.0% ≈ eAG 183 mg/dL

Roughly, each percentage point corresponds to a substantial shift in average glucose — which is also why very small movements between two tests are frequently assay and biological noise rather than a real change.

Diagnostic ranges

• Normal: <5.7%
• Prediabetes: 5.7-6.4%
• Diabetes: 6.5% or above, confirmed on a second test (American Diabetes Association, Diabetes Care)
• Typical treatment target on therapy: 6.5-7.0%, varying by individual

These are thresholds drawn across a continuum, not biological cliff edges. Nothing changes in a person's physiology between 5.6% and 5.7%; what changes is the category they are recorded in. Risk runs smoothly through the whole range, which is the argument for paying attention well before a threshold is crossed.

Where the number lies

Because HbA1c depends on red cell lifespan, anything that changes red cell turnover changes the result independently of glucose (Radin, J Gen Intern Med 2014). The direction is predictable once you have the mechanism:

• Shortened red cell survival lowers HbA1c — haemolytic anaemia, recent significant blood loss, and the period after a transfusion. Cells are removed before they can accumulate glycation, so the result understates true glucose
• Prolonged red cell survival raises it — iron deficiency anaemia is the common and frequently missed example, where older cells persist and HbA1c reads higher than the glucose justifies
• Haemoglobinopathies — sickle cell trait and disease, thalassaemia. Some assays are affected more than others, and the laboratory method matters
• Pregnancy — red cell turnover increases, and the interpretive framework is different throughout
• Chronic kidney disease — shortened red cell survival plus, in some patients, erythropoietin treatment, both pulling the number down
• Recent iron, B12 or erythropoietin treatment — a burst of new cells dilutes the glycated fraction

There is also the glycation gap: some people consistently glycate more or less than their glucose would predict, for reasons that appear partly genetic. It is why an HbA1c and a set of glucose readings can persistently disagree without either being wrong. When that happens, believe the direct measurements. Fructosamine, which reflects glycated serum proteins over a two-to-three week window, and continuous glucose monitoring are the alternatives.

What an average cannot show

Two people with identical HbA1c can have entirely different glucose days. One holds steady near their average. The other spends the morning low and the afternoon high, spiking hard after meals and dipping in between. The average is the same; the exposure to high glucose, and probably the vascular consequence, is not.

HbA1c is blind to:

• Glucose variability — the amplitude of the swings
• Time in range, and how much of the day is spent outside it
• Post-meal excursions, which is where early metabolic dysfunction usually shows first
• Overnight patterns, including the dawn rise

Continuous glucose monitoring supplies exactly this resolution, and consensus targets exist for interpreting it (Battelino et al., Diabetes Care 2019). For someone in the prediabetes band, two weeks of CGM data usually teaches more than another HbA1c (fasting glucose versus CGM, variability and spikes).

What to run alongside it

HbA1c is a late marker. Insulin resistance typically develops years before glucose rises, because the pancreas compensates by producing more insulin and holds glucose in range as long as it can. Through that entire compensated phase HbA1c can look untroubled while the underlying problem progresses.

Fasting insulin sees that phase, and it is the single most useful companion test — a normal glucose with a high insulin describes a system working hard to stay normal (covered here). Triglycerides and HDL give another cheap read on the same physiology, since insulin resistance raises one and lowers the other. Waist circumference adds more than BMI. Together these describe a trajectory rather than a snapshot.

Targets, and why they differ

• Healthy adults: under 5.4% is generally regarded as metabolically optimal, and the risk gradient continues below the diagnostic thresholds
• Established diabetes: 6.5-7.0% for most, balancing long-term complication risk against hypoglycaemia
• Older patients with frailty or hypoglycaemia risk: 7.5-8.0% may be the appropriate target, because a severe hypoglycaemic event carries more immediate danger than a marginally higher average

That last point surprises people. Tighter is not universally better; the target is set by which risk dominates for that individual, which is a clinical judgement rather than a number.

If yours came back in the prediabetes band

This range is where intervention works best and where it is most often deferred. A reasonable sequence: repeat the test rather than acting on one result, since assay variation and the conditions above move it. Add fasting insulin, triglycerides and HDL to see whether the picture is compensated insulin resistance. Consider two weeks of CGM to find out which meals and which parts of the day are responsible, which converts an abstract number into specific decisions. Then act on the largest levers — resistance training, which increases the muscle glucose disposal capacity most people are short of; total energy intake; and sleep, which changes insulin sensitivity within days. Re-test at three months, not sooner. The insulin sensitivity guide ranks the interventions by effect size.

Bottom line

HbA1c measures glucose physically attached to haemoglobin, which is why it reports roughly three months of exposure weighted toward the recent weeks — a strong diagnostic and trend-tracking tool. It is unreliable wherever red cell turnover is abnormal: anaemia of any kind, haemoglobinopathy, kidney disease, pregnancy, recent transfusion or iron treatment. And it is silent on variability, post-meal spikes and time in range. Read it alongside fasting insulin, and alongside CGM where the question is which behaviours are responsible. Treat the thresholds as administrative lines across a continuous risk gradient rather than the point where anything biological changes.

Educational content, not medical advice. Laboratory interpretation and any treatment decision are made by a licensed physician after individual evaluation. Individual results vary.