GMI Calculator: GMI vs A1C

GMI vs A1C in one minute

The Glucose Management Indicator (GMI) is a number calculated from your continuous glucose monitor (CGM) data. It takes your average sensor glucose over the last two weeks or more and converts it into a value on the same scale as the A1C (HbA1c) blood test – a percentage or mmol/mol. The A1C is a laboratory measurement of how much haemoglobin in your red blood cells has glucose attached to it, which reflects average glucose over roughly the previous two to three months.

So when people search "GMI vs A1C", the short answer is:

• A1C is measured in a blood sample. GMI is estimated from CGM readings.
• A1C covers about 8–12 weeks, weighted towards the most recent month. GMI usually covers 14 to 90 days, whatever window your CGM software uses.
• They are designed to be comparable but not identical. A difference of a few tenths of a percentage point is common, and larger gaps happen in a meaningful minority of people.
• When they disagree, neither is automatically "wrong": the gap itself can be useful information to discuss with your diabetes team.

The calculator above uses the published GMI equation to turn an average glucose in mg/dL or mmol/L into a GMI in % and mmol/mol. It also shows the A1C that the older ADAG estimated-average-glucose formula would predict for the same average, and – if you enter a lab A1C – the difference between the two. For a full lab A1C chart and conversions, see the A1C calculator and chart.

The GMI formula

GMI was defined in 2018 by Bergenstal and colleagues, who fitted a regression between mean CGM glucose and laboratory A1C in several clinical trial populations. The equations are:

The two forms are the same relationship expressed in different units. To move between the percentage (NGSP/DCCT) scale and the mmol/mol (IFCC) scale, the usual master equation is mmol/mol = (% − 2.15) × 10.929 – the same conversion our HbA1c mmol/mol to % converter uses.

Worked examples

• Mean glucose 154 mg/dL (8.5 mmol/L): 3.31 + 0.02392 × 154 = 7.0%, about 53 mmol/mol.
• Mean glucose 125 mg/dL (6.9 mmol/L): 3.31 + 0.02392 × 125 = 6.3%, about 45 mmol/mol.
• Mean glucose 200 mg/dL (11.1 mmol/L): 3.31 + 0.02392 × 200 = 8.1%, about 65 mmol/mol.
• Mean glucose 7.0 mmol/L entered directly: 12.71 + 4.70587 × 7.0 = 45.7 mmol/mol, about 6.3%.

GMI chart: average glucose to GMI

The table below shows GMI for common average-glucose values, alongside the A1C that the ADAG formula (A1C = (glucose + 46.7) ÷ 28.7) would predict. Notice that the two lines cross at around 154 mg/dL, and diverge above and below that point.

Why are the two columns different? The ADAG formula was derived in 2008 from a mixture of CGM and finger-stick data in about 500 people, many of whom did not have diabetes, and its purpose was to translate an A1C into an average glucose that patients could relate to. GMI was derived a decade later from modern CGM data in people with diabetes, with the purpose of going in the opposite direction – from CGM average to an A1C-like number. Because the regressions were fitted to different populations and in different directions, they do not give identical answers, especially at the extremes.

Three numbers, one average: A1C, eAG and GMI

It helps to keep three related numbers apart:

Because eAG and GMI run in opposite directions and come from different datasets, converting an A1C to eAG and then putting that eAG into the GMI formula does not return the original A1C. That is expected, and it is why this page shows both the GMI and the ADAG-predicted A1C side by side.

A worked case

Suppose a CGM report shows a 14-day mean glucose of 170 mg/dL with the sensor active 92% of the time, and a lab A1C taken the same week is 7.9%. The GMI is 3.31 + 0.02392 × 170 = 7.4%. The difference is +0.5 points (lab higher). On its own that is within the common range. If the next two A1C tests are also about half a point above the GMI, the gap is probably a stable personal feature – useful to know, because it means that for this person a GMI of 7.0% corresponds to a lab A1C of roughly 7.5%, and treatment targets can be discussed with that in mind.

Now suppose a different person has a GMI of 7.4% and a lab A1C of 6.2%, with a recent diagnosis of iron-deficiency anaemia that has just been treated. The newly made red cells have had little time to collect glucose, pulling A1C down. Here the CGM-based numbers are likely to describe recent glucose more faithfully until red-cell turnover settles.

Why it was renamed from "estimated A1C"

Early CGM reports printed an "estimated A1C" (eA1C). That label caused confusion: people saw a sensor-derived "A1C" that differed from their lab result and assumed one of them was faulty. In 2018 an expert group proposed the new name Glucose Management Indicator precisely to make it clear that the value is an indicator derived from CGM glucose, not a measurement of glycated haemoglobin. The 2019 international consensus on time in range adopted GMI as one of the standard metrics in the ambulatory glucose profile (AGP) report that most CGM software now produces.

Why GMI and A1C can disagree

A1C depends on two things: how much glucose the red blood cells are exposed to, and how long those cells live. Anything that changes red-cell lifespan or the glycation process changes A1C without changing average glucose. GMI, by contrast, depends only on the glucose readings – so it is blind to red-cell factors but sensitive to sensor accuracy and wear time.

Reasons A1C may read higher than GMI

• Longer red-cell lifespan, for example in iron-deficiency anaemia before treatment, or after removal of the spleen. Older cells have had more time to collect glucose.
• Individual "glycation gap" – some people consistently glycate haemoglobin more than others at the same average glucose. This tendency is fairly stable over time and appears partly inherited.
• Higher glucose earlier in the A1C window than in the CGM window. If glucose has recently improved, A1C lags behind.
• Ethnic and genetic variation. Studies have reported small average differences in A1C between ethnic groups at similar glucose levels.

Reasons A1C may read lower than GMI

• Shortened red-cell survival: haemolysis, recent blood loss, recent transfusion, treatment of iron or B12 deficiency (a wave of young red cells), pregnancy, and advanced chronic kidney disease, particularly with erythropoietin treatment.
• Haemoglobin variants such as HbS, HbC or HbE can interfere with some A1C methods. The NGSP keeps a list of which assays are affected.
• Recent worsening of glucose – CGM picks it up within days, A1C over weeks.

Reasons that come from the sensor

• Too little data. The consensus recommends at least 14 days of data with the sensor active for at least 70% of the time before metrics such as GMI are used.
• Sensor bias. Every CGM has some measurement error, and a sensor that reads systematically high or low shifts the average, and GMI, accordingly.
• Gaps that are not random – for example, if the sensor tends to fall off during exercise or is removed overnight.

How big a difference matters?

In the original analysis, GMI and laboratory A1C were often within a few tenths of a percentage point of each other, but a meaningful share of people had differences of 0.5 percentage points or more, and a smaller number had differences of 1 point or more. In other words, a gap of 0.3% is ordinary; a persistent gap of 0.5–1% or more is worth noticing.

A practical way to read your numbers:

The key word is repeatedly. A single mismatch after a change in treatment, illness, or a short sensor window is expected; a consistent gap over several A1C tests is more informative.

How to use GMI alongside other CGM metrics

GMI is only one line on a CGM report. The international consensus recommends looking at a set of metrics together:

Targets are individualised – for older or high-risk adults the consensus suggests less tight targets, with more emphasis on avoiding lows, and pregnancy has its own tighter ranges. Your diabetes team sets the targets that apply to you.

Two people can have the same GMI of 7.0% with very different experiences: one with steady readings between 110 and 200 mg/dL, another who swings between frequent lows and long highs. GMI averages those patterns away, which is why time in range and time below range are considered at least as important. A GMI that looks "good" can hide frequent hypoglycaemia.

GMI in people without diabetes

CGMs are increasingly worn by people without diabetes. The GMI equation was developed in people with diabetes, so it should be applied cautiously outside that group. In healthy adults, average sensor glucose is often close to 100 mg/dL (5.6 mmol/L), which corresponds to a GMI of about 5.7%. Lab A1C values in healthy people are frequently a little lower than that, which fits the table above, where GMI sits above the ADAG-predicted A1C at low averages. A GMI in the 5.7–6.4% range in someone without diabetes does not by itself mean prediabetes – the diagnostic thresholds of the American Diabetes Association apply to laboratory A1C, fasting glucose and oral glucose tolerance tests, not to GMI.

Getting the most accurate GMI

• Use at least 14 days of data, ideally the same 14 or 90 days your clinic report uses, and make sure the sensor was active for 70% or more of that time.
• Use the average glucose from your CGM software rather than an average of finger-stick readings, which over-represent fasting and pre-meal times.
• Compare like with like: a 14-day GMI reflects the last two weeks, whereas your A1C reflects the last two to three months. If glucose has changed recently, expect a gap.
• Note the units. The calculator accepts mg/dL or mmol/L; mixing them is the most common cause of implausible results.