Your lab A1c, your CGM's GMI and your app's estimated A1C look like versions of one number. Each is built differently, and knowing how helps you read a mismatch calmly and usefully.
Your lab A1c came back at 7.6%. The same week, your CGM report showed a glucose management indicator (GMI) of 7.0%, and your app’s estimated A1C said 7.1%. Which one is right?
Each number answers a slightly different question. A1c is measured in your blood and reflects how much glucose has attached to hemoglobin over roughly three months. GMI and other estimates are calculated from your glucose readings with formulas built on other people’s data. Often they land close together. This article explains how each is made, why they can drift apart, and how to use them together.
Hemoglobin is the oxygen-carrying protein inside red blood cells. Glucose in the blood slowly and permanently attaches to it, a process called glycation. A1c is the percentage of hemoglobin carrying glucose this way, so the higher your glucose runs, the higher A1c climbs. In Canada and the US it is usually reported as a percentage, while much of the world uses mmol/mol. For example, 7.0% equals 53 mmol/mol and 8.0% equals 64 mmol/mol.
Red blood cells circulate for a few months before the body clears them, so A1c reflects a weighted average of glucose over that time, with recent weeks counting most. When glucose was brought down quickly in nine people newly diagnosed with type 2 diabetes, A1c fell halfway toward its new level in about 35 days, and the analysis suggested that A1c reflects glucose going back roughly 100 days 1. A change you made five or six weeks ago will therefore be only partly reflected in your next A1c.
How long red cells survive also differs between people, even when blood counts are normal. A small study that labelled red cells in 12 people with normal blood counts, 6 of them with diabetes, found that the average age of circulating red cells ranged from about 38 to 60 days. That spread was large enough to produce clinically important differences in A1c at the same average glucose 2. Longer-lived cells have more time to collect glucose, which pushes A1c up. Shorter-lived cells pull it down.
A1c earned its place because the trials that shaped modern diabetes care measured it and tied it to complications.
The Diabetes Control and Complications Trial (DCCT) followed 1,441 people with type 1 diabetes for an average of 6.5 years 3, 4. Intensive treatment, which lowered A1c, cut the risk of developing retinopathy by 76% in people who had none at the start and slowed its progression by 54% in those with early changes. It also reduced microalbuminuria by 39%, albuminuria by 54% and clinical neuropathy by 60%. The trade-off was a two- to threefold rise in severe hypoglycemia 3.
Long-term follow-up of the same participants, in the Epidemiology of Diabetes Interventions and Complications (EDIC) study, found that the benefits of the earlier intensive period lasted for years, including on severe complications and cardiovascular disease, even after A1c levels in the two original groups had converged 4.
In type 2 diabetes, the UK Prospective Diabetes Study enrolled 3,867 people with newly diagnosed diabetes. Over 10 years, A1c was 7.0% with intensive treatment and 7.9% with conventional treatment. Intensive treatment lowered the risk of any diabetes-related endpoint by 12% and of microvascular endpoints by 25%, including fewer people needing laser treatment of the retina. Hypoglycemia was more common with intensive treatment, and the reduction in heart attacks didn’t reach statistical significance 5.
Because these trials measured A1c, guidelines express goals and testing schedules in A1c terms. Diabetes Canada recommends measuring A1c at least every 3 months when targets aren’t being met or treatment is being adjusted, and says testing every 6 months may be considered when targets are consistently met 6.
The A1c-Derived Average Glucose (ADAG) study, led by researchers at Massachusetts General Hospital and Harvard Medical School, set out to translate A1c into glucose units. At 10 centres in several countries, 507 people (268 with type 1 diabetes, 159 with type 2 diabetes and 80 without diabetes) wore a CGM on several occasions and did frequent fingerstick checks over three months, giving about 2,700 glucose values per person 7. The relationship was estimated average glucose (eAG) in mg/dL = 28.7 × A1c − 46.7, which converts to eAG in mmol/L ≈ 1.59 × A1c − 2.59. A1c explained about 84% of the variation in average glucose (R² = 0.84) 7.
| A1c | eAG (mmol/L) | eAG (mg/dL) |
|---|---|---|
| 6.0% (42 mmol/mol) | 7.0 | 126 |
| 7.0% (53 mmol/mol) | 8.6 | 154 |
| 8.0% (64 mmol/mol) | 10.2 | 183 |
| 9.0% (75 mmol/mol) | 11.8 | 212 |
| 10.0% (86 mmol/mol) | 13.4 | 240 |
The values in the table are calculated from the ADAG equation 7. They describe the average relationship, and any one person can sit well above or below the line. The GMI authors gave an example in which a lab A1c of 8.0% (64 mmol/mol) could belong to someone whose CGM average is as low as 8.6 mmol/L (155 mg/dL) or as high as 12.1 mmol/L (218 mg/dL) 8.
ADAG also excluded people with conditions known to distort A1c, including anemia, hemoglobin disorders and chronic kidney or liver disease 7. The formula therefore describes people without those conditions best. The study reported that its equation held across the age, sex, diabetes type and ethnic subgroups it included 7. Later research designed to compare groups found differences, described below.
As CGM use grew, CGM software often reported an estimated A1c calculated from the sensor average, which caused confusion when it disagreed with the lab result. In 2018 an expert group proposed a new name, the glucose management indicator (GMI), with formulas derived from CGM data 8.
The formula came from 528 people with type 1 or type 2 diabetes in four randomized trials, with an average of about 48 days of CGM data each 8.
| Mean glucose (mmol/L) | Mean glucose (mg/dL) | GMI (%) | GMI (mmol/mol) |
|---|---|---|---|
| 6.0 | 108 | 5.9 | 40.9 |
| 7.0 | 126 | 6.3 | 45.7 |
| 8.0 | 144 | 6.8 | 50.4 |
| 9.0 | 162 | 7.2 | 55.1 |
| 10.0 | 180 | 7.6 | 59.8 |
| 12.0 | 216 | 8.5 | 69.2 |
The mmol/mol values are as published, and the percentages are calculated from the GMI formula 8. GMI tells you the A1c that people with your CGM average typically have. The authors explain that your lab A1c may be similar, higher or lower, because red cell lifespan and the way glucose attaches to hemoglobin differ between people, and they suggest basing GMI on at least 14 days of CGM data 8.
Differences are common in everyday care. At the University of Washington’s diabetes clinic, researchers compared lab A1c with GMI across 641 clinic visits, mostly by people with type 1 diabetes (91%), using CGM data from within four weeks of the blood test (about 24 days of wear on average). The two agreed within 0.1 points only 11% of the time. They differed by 0.5 points or more in 50% of comparisons and by 1 point or more in 22%, and differences were larger in people with reduced kidney function (an eGFR below 60) 9.
In plain terms: GMI estimates what your A1c would be if your blood behaved like the average blood in the studies behind the formula. If your lab A1c sits a similar distance from your GMI test after test, that steady gap is worth knowing about, because it helps you and your team read both numbers.
Share of 641 clinic visits at the University of Washington, using CGM data from the four weeks before each test. The two larger groups overlap: the 22% are part of the 50%.
GMI (%) = 3.31 + 0.02392 × average glucose in mg/dL. Estimated average glucose (mg/dL) = 28.7 × A1c − 46.7. These are population formulas, for learning: they can’t diagnose diabetes, and your lab A1c and your GMI can differ.
Because A1c depends on red cell lifespan and on how readily glucose attaches to hemoglobin, people produce different A1c values at the same average glucose 2, 8. Researchers call the difference between measured A1c and the A1c predicted from another glucose marker the glycation gap. In 2,263 people with diabetes in Wolverhampton, UK, the gap between measured A1c and the A1c predicted from fructosamine tended to stay the same for each person across tests about 10 months apart 10. A steady gap looks more like a personal trait than a sign that something has changed.
A 2015 systematic review found that iron deficiency anemia tends to push A1c falsely high, while some other anemias may push it lower 11. Diabetes Canada lists iron or vitamin B12 deficiency among causes of higher A1c readings, and treatment with iron, vitamin B12 or erythropoietin among causes of lower ones 6.
In a study of 104 people with type 2 diabetes, 80 of whom had reduced kidney function (eGFR below 60), A1c tracked CGM average glucose about as well as glycated albumin and fructosamine did, though no marker fully captured each person’s average. A1c read low relative to CGM glucose in people with albuminuria (protein in the urine). The authors concluded that A1c remains useful for following trends at that level of kidney function, and that direct glucose measurement is needed to see short-term swings 12. The clinic data above also showed larger A1c and GMI differences when kidney function was reduced 9.
In a Danish study of healthy women, A1c was significantly lower in both early and late pregnancy than in women who weren’t pregnant 13. Diabetes Canada advises against using A1c to screen for gestational diabetes 14.
Inherited hemoglobin variants can shift A1c up or down 6. In 4,620 African American adults from two US cohort studies, 7.9% of whom had sickle cell trait, the trait was linked with A1c values about 0.3 points lower at the same fasting glucose 15.
Because A1c moves only halfway in about five weeks 1, it lags behind recent change. If your glucose has improved over the past month, a 14-day GMI will show it sooner than your A1c. If glucose has crept up, the reverse is true.
In a study of people with type 1 diabetes aged 8 and older, 104 Black and 104 white participants wore CGM for 12 weeks. At the same average glucose, A1c was 0.4 percentage points higher in Black participants. For glycated albumin and fructosamine, which don’t involve red cells, the relationship with average glucose showed no significant racial difference 16. Diabetes Canada notes that A1c runs higher in some ethnic groups (African, Asian and Hispanic) than in white people with similar glucose, by 0.13 to 0.47 points, and that these differences haven’t been shown to change the relationship between A1c and complications 6.
The table summarizes the usual direction of each effect 1, 6, 9, 11–13, 15.
| Situation | Usual effect on A1c |
|---|---|
| Iron or vitamin B12 deficiency | Reads higher |
| Some other anemias; erythropoietin | Reads lower |
| Reduced kidney function | Less reliable |
| Pregnancy | Reads lower |
| Sickle cell trait | Reads lower |
| Recent change in glucose | Lags behind |
When A1c and CGM disagree persistently, or A1c is known to be unreliable, other blood tests can offer a second view. A review by researchers at Johns Hopkins noted that these markers may help where A1c falls short, while long-term studies linking them to clinical outcomes are still lacking 17.
Fructosamine and glycated albumin measure glucose attached to proteins in the liquid part of the blood. Those proteins turn over faster than red cells, so the markers cover a shorter window, about 30 and 40 days respectively in one kinetic analysis 1. They don’t involve red cells, and in the type 1 study above their link with average glucose didn’t differ significantly by race 16. They have their own blind spots. In the kidney study, neither was less variable than A1c at a given CGM average, and both were skewed by age, body mass index, iron levels and albuminuria 12.
1,5-anhydroglucitol (1,5-AG) works differently. The kidneys normally reabsorb it. When glucose rises above the kidney’s threshold and spills into the urine, it competes with 1,5-AG for reabsorption, more 1,5-AG is lost, and blood levels fall. A low value therefore flags recent high peaks over roughly the past one to two weeks. SGLT2 inhibitor medicines, which deliberately send glucose into the urine, lower it, and kidney function affects it too 18.
Like GMI, an app’s estimated A1C is a calculation from the glucose readings the app holds. It is useful for watching your direction of travel between lab tests, provided the data behind it are reasonably complete and cover a similar span each time. With CGM data, gaps from a sensor change, a lost signal or readings that sync late mean fewer readings behind the number. With fingerstick data alone, the readings reflect the moments you chose to test, so an average built from them can sit well away from your true 24-hour average.
An estimated A1C can’t be used to diagnose diabetes or prediabetes. Under Diabetes Canada’s criteria, diagnosis by A1c needs a lab result of 6.5% (48 mmol/mol) or higher, measured with a standardized, validated assay, and, without clear symptoms, a confirmatory lab test on another day. The guideline advises against using A1c alone for diagnosis in some groups, including children and adolescents and people with suspected type 1 diabetes, and notes that A1c may mislead in hemoglobin disorders, some anemias, iron deficiency and severe liver or kidney disease. Prediabetes is defined as an A1c of 6.0% to 6.4% (42-46 mmol/mol) 14.
No. A1c is measured in your blood and reflects how much glucose has attached to hemoglobin over roughly three months. GMI, the glucose management indicator, is calculated from your CGM average and tells you the A1c that people with your CGM average typically have 8.
In one clinic's real-world data covering 641 visits, lab A1c and GMI agreed within 0.1 points only 11% of the time. They differed by 0.5 points or more in 50% of comparisons and by 1 point or more in 22%, with larger differences in people with reduced kidney function 9.
Red cells that live longer have more time to collect glucose, which pushes A1c up. Iron or vitamin B12 deficiency can also make A1c read higher 6. If your glucose has improved over the past month, a 14-day GMI will show it sooner than your A1c.
A1c reflects roughly three months, weighted toward recent weeks. In nine people newly diagnosed with type 2 diabetes, A1c fell halfway to its new level about 35 days after glucose was brought down quickly 1. Changes from five or six weeks ago are only partly reflected in your next A1c.
Using the A1c-Derived Average Glucose (ADAG) formula, an A1c of 7.0% (53 mmol/mol) corresponds to an estimated average glucose of 8.6 mmol/L (154 mg/dL) 7. That is the average relationship, and any one person can sit well above or below the line.
Glyve’s history shows an estimated A1C alongside your averages. Before reading it, check that the period you’re viewing has steady glucose data. Readings that come through Apple Health or the Dexcom and FreeStyle Libre sharing services can arrive later than in the CGM’s own app, so the latest days may still be filling in. Note the date of each lab A1c and compare it with the estimate for the weeks just before that test. If the two differ by a similar amount each time, that pattern is worth showing your care team. An AGP report covering the same weeks gives them the glucose profile behind the numbers, and CSV export lets them look at the readings in detail. Glyve shows these numbers together; it doesn’t diagnose or explain the cause of a difference.
How we write about diabetes. Spotted a mistake? Email info@glyve.ca and we’ll check it against the source.
Glyve keeps glucose, food, insulin, exercise, sleep and stress on one timeline.
For iPhone, iPad and Apple Watch.