GMI vs A1c: Why They Disagree

Two numbers, two different measurements

A1c measures glycation — the proportion of your haemoglobin that has sugar attached to it. Because red blood cells live around three months, it reflects roughly two to three months of average glucose, weighted towards the most recent weeks.

GMI measures nothing about haemoglobin at all. It takes your CGM’s mean glucose over a period, usually 14 days, and converts it into an A1c-like percentage using a published regression equation. It is an estimate of what an A1c would read if glucose continued at that average.

So the two disagree for a straightforward reason: one is a chemical measurement of your blood cells over three months, the other is arithmetic applied to a fortnight of sensor readings. Expecting them to match is the actual mistake. GMI was introduced in 2018 specifically to stop clinicians calling the CGM figure an “estimated A1c”, because that name implied an equivalence that does not exist.

Mean glucose to GMI

These come from the published GMI equation, GMI% = 3.31 + 0.02392 × mean glucose (mg/dL). Note that it is not the same relationship as the older eAG conversion — at a mean of 154 mg/dL the eAG table gives an A1c of 7.0%, and GMI happens to agree there, but the two equations diverge either side of it. Our GMI calculator works this out from your own mean glucose in either unit.

How big a difference is normal

Differences of a few tenths of a percentage point are routine, and differences approaching a full point are not unusual. What matters clinically is not the size of a single gap but whether it is consistent.

A stable, repeatable gap in the same direction tells your care team something real about your physiology, and it changes how both numbers should be read. If your A1c habitually runs half a point above your GMI, then an A1c at target may correspond to a CGM average that is not, and treatment decisions based on the A1c alone would be systematically off.

A gap that appears suddenly is a different matter and usually has a mundane explanation: insufficient CGM wear time, a recent change in control that the three-month A1c has not caught up with, or a sensor running consistently high or low.

What widens the gap

Red cell lifespan. This is the big one. A1c assumes red cells survive about 120 days. Anything that shortens that — haemolytic anaemia, recent blood loss, chronic kidney disease, pregnancy — means less time for glycation and an A1c that reads lower than the glucose data justifies. Conditions that lengthen it, such as iron-deficiency anaemia, push A1c the other way.

Haemoglobin variants. Sickle cell trait, thalassaemia and other variants interfere with some A1c assays directly, producing results that do not reflect glucose at all.

Individual glycation rate. People genuinely differ in how readily glucose attaches to haemoglobin at the same average glucose. This is why two people with identical CGM traces can return different A1c values, and it is the main reason the two measures cannot be reconciled by arithmetic.

Insufficient CGM data. GMI needs adequate sensor coverage to mean anything — conventionally at least 14 days with 70% or more wear. Calculated from patchy data it describes the hours you happened to be wearing the sensor, which may not be representative.

Which should you act on

They answer different questions, so the useful approach is to use both rather than pick one.

GMI updates within days, so it is the better feedback signal when you are actively changing something. A1c is what long-term outcome data was built on, so it remains the reference for clinical targets. And neither describes variability: two people can share a GMI of 7.0% while one runs steadily and the other swings between hypo and hyper all day. That is what time in range captures, and it is why modern CGM reporting shows GMI and time in range together rather than either alone.

Limitations & Safety Notes

• GMI is an estimate derived from CGM mean glucose, not a haemoglobin measurement. It is not expected to equal lab A1c and a difference is not an error.
• It requires adequate CGM data — conventionally at least 14 days with 70% or more sensor wear — to be meaningful.
• Anything affecting red blood cell lifespan (anaemia, recent transfusion, pregnancy, chronic kidney disease, haemoglobin variants) changes A1c independently of glucose and widens the gap.
• Neither GMI nor A1c describes glucose variability. Two people with the same value can have very different day-to-day patterns.
• Sensor accuracy limits GMI. A sensor reading consistently high or low shifts the mean and therefore the GMI.

Frequently Asked Questions