Estimated Average Glucose (eAG): The Conversion and the Range It Hides

Labs print only the middle number. The study also reported the range around it, and at an A1c of 7 percent that range runs from 123 to 185 mg/dL. The spread reflects real differences between people: red cells survive for different lengths of time and take up glucose at different rates, which a formula fitted to a group cannot see. Two people with the same 7 percent A1c can genuinely be averaging 130 and 180 mg/dL. That is one reason a series of results, read as a trend, says more than any single eAG.

People who check with a meter often find their average is below their eAG. That is expected rather than a lab error. Fingerstick checks are mostly taken before meals and during the day, when glucose is lowest, while the A1c behind the eAG also records the hours after meals and overnight. A single glucose reading is one moment; the eAG and A1c are one result in two units and cannot disagree with each other, as the HbA1c vs eAG comparison explains.

The formula itself is approximate. The ADAG researchers got versions about 6 mg/dL apart at an A1c of 7 percent, depending on whether they used sensor data or fingerstick readings from the same people.

People using a continuous glucose monitor (CGM) see a calculation that runs the other way. The glucose management indicator (GMI) estimates an A1c from the sensor's average glucose: GMI (%) = 3.31 + 0.02392 × mean glucose in mg/dL. It is a separate formula fitted to sensor data, not the eAG formula run backwards. The two agree near an average of 154 mg/dL, about 7 percent, and drift apart further out: an average of 200 mg/dL gives a GMI of 8.1 percent, while reversing the eAG formula would give 8.6.

In the 2018 paper that introduced GMI, among 528 people with both measured over the same period, GMI matched the lab A1c exactly in 19 percent, differed by 0.3 points or more in 51 percent and by 0.5 or more in 28 percent. A GMI of 6.8 percent next to a lab A1c of 7.3 percent is ordinary. Because a sensor records day and night, the gap is less about timing than about the person: red cell lifespan and how readily glucose attaches to hemoglobin both vary, which makes the gap useful information about that person.

Because eAG is simple arithmetic on the A1c, anything that distorts the A1c distorts the eAG equally. The ADA's Standards of Care advise using plasma glucose for diagnosis when a condition changes the relationship between A1c and blood sugar, and looking for interference when glucose readings and A1c disagree consistently. Common causes:

• hemoglobin variants such as sickle cell trait, which affect some A1c methods;
• G6PD deficiency, an inherited enzyme shortfall; one variant, carried by about 11 percent of Black Americans, lowers A1c by roughly 0.8 points in affected men and 0.7 in women with two copies;
• anything that changes how long red cells survive: recent blood loss, transfusion, hemolysis, erythropoietin treatment, removal of the spleen, or iron deficiency;
• kidney failure and dialysis, cirrhosis, and HIV;
• pregnancy, in which A1c is not used to diagnose gestational diabetes.

Iron deficiency is the most common. The NIDDK describes falsely high A1c results in people very low in iron, so an eAG alongside a low ferritin can read high while blood sugar is unremarkable. When the A1c is in doubt, a fasting glucose does not depend on red cells, and fructosamine and glycated albumin measure a two-to-three-week average from blood proteins instead.

Every diagnostic cutoff is written in A1c units, so a doctor reading a high eAG is really reading the A1c beside it. Without clear symptoms of high blood sugar, a diabetes diagnosis needs two abnormal results, either two different tests from the same sample or the same test on two occasions; the HbA1c vs fasting glucose comparison covers the cutoffs for each. For people with diabetes, the ADA advises an A1c at least twice a year, more often when treatment changes. A repeat much sooner mostly measures the same red cells again.