HbA1c: three months of glucose in one number

Glucose in the blood sticks to haemoglobin inside red cells, slowly and without enzymes. Red cells live for around three months, so the share of haemoglobin that has picked up glucose, HbA1c, reflects average glucose over roughly the previous two to three months, weighted towards the most recent weeks. It is reported as a percentage (the US unit) or in mmol/mol (the IFCC unit used in the UK and much of Europe).

The guideline numbers

For people already diagnosed with diabetes, the ADA's usual treatment goal for many non-pregnant adults is under 7%, adjusted to the person. Online "optimal" targets, often 5.0 or 5.3% and under, are not ADA categories. They lean on cohort data like the study below, which shows risk rising gradually rather than at a cliff, but no guideline sets them.

What raises and lowers it

• Real changes: higher average glucose from diet, inactivity, weight gain, illness, some drugs such as steroids; lower with weight loss, exercise and glucose-lowering drugs.
• False changes: anything that alters red cell lifespan. Iron deficiency can push HbA1c up. Recent blood loss, haemolysis, transfusion, pregnancy and treatment of anaemia can pull it down. Some haemoglobin variants interfere with some assays. Advanced kidney disease muddies it. In those cases fasting glucose or other measures are used alongside.

What it predicts

The clearest data come from the ARIC study in the New England Journal of Medicine (2010), which measured HbA1c in 11,092 adults without diabetes or heart disease and followed them for about 15 years. Against a reference of 5.0 to 5.5%:

Two things in that study matter for how people use the test. HbA1c predicted heart disease even after adjusting for fasting glucose, and fasting glucose stopped predicting it once HbA1c was in the model. And death from any cause had a J-shaped curve: the lowest category was not the safest. This is association in an observational cohort, not proof that lowering HbA1c within the normal range changes outcomes.

On GLP-1 drugs and weight loss

In people without diabetes, HbA1c drifts down a little. In STEP 1, semaglutide 2.4 mg lowered it by 0.4 points against 0.2 on placebo from a baseline of 5.7% over 68 weeks. In SURMOUNT-1, tirzepatide 15 mg lowered it 0.4 points against 0.1 from a baseline of 5.6% over 72 weeks. In people with type 2 diabetes the effect is much larger: in SURMOUNT-2 the Zepbound label reports a fall of 2.1 points on tirzepatide against 0.5 on placebo from 8.0%. The GLP-1 blood work page has the rest of those trial numbers.

How to test

• No fasting needed. HbA1c is not affected by the last meal.
• Timing. Because it reflects about three months, retesting much sooner than that mostly measures the time before a change. The ADA suggests checking at least twice a year in diabetes and about every three months after a treatment change.
• Biological variation is small, which is why HbA1c works for trends. Changes of a tenth or two can still be noise, and a change of method or lab can shift results.
• Read it with a blood count. Anaemia or a recent blood loss changes how much the number can be trusted. Fasting glucose, on the fasting glucose page, is the usual cross-check.

Tracking it in Peptide AI

In Peptide AI, HbA1c sits in the metabolic panel. The app reads it from a lab PDF or manual entry, scores the panel from 0 to 100 alongside an overall grade, flags values that warrant prompt physician consultation, and plots each result against your dose timeline. Log the draw date, whether you fasted, and anything unusual in the days before, such as illness, a hard session or a dose step, so the next result is compared like with like. Bloodwork tracking shows the series next to the dose log, weight and side effects, and the panels page lists every marker the app groups. More markers are in the bloodwork guides.