Convert
The full table
| A1C | eAG mg/dL | eAG mmol/L |
|---|---|---|
| 5.0% | 97 | 5.4 |
| 5.5% | 111 | 6.2 |
| 6.0% | 126 | 7.0 |
| 6.5% | 140 | 7.8 |
| 7.0% | 154 | 8.6 |
| 7.5% | 169 | 9.4 |
| 8.0% | 183 | 10.2 |
| 8.5% | 197 | 11.0 |
| 9.0% | 212 | 11.8 |
| 9.5% | 226 | 12.6 |
| 10.0% | 240 | 13.4 |
| 10.5% | 255 | 14.1 |
| 11.0% | 269 | 14.9 |
| 11.5% | 283 | 15.7 |
| 12.0% | 298 | 16.5 |
How to read the result
An A1C is a percentage: the share of your hemoglobin that has glucose attached. The ADA's Standards of Care describe it as glucose bound to hemoglobin over the life span of the red cell (about 120 days), a weighted average that is more heavily affected by recent blood glucose, which is why a meaningful change can show up in less than 120 days. The estimated average glucose (eAG) translates that percentage into the unit a glucose meter shows, mg/dL in the US or mmol/L elsewhere, so you can set an A1C beside the fasting and after-meal numbers you already know.
The ADA's diagnostic cutoffs are set on the A1C, not the eAG: below 5.7 percent is normal, 5.7 to 6.4 percent is prediabetes, and 6.5 percent or above is diabetes, on a laboratory test in a nonpregnant adult, confirmed by a second abnormal result unless the diagnosis is clinically clear. An eAG of 140 mg/dL is simply what 6.5 percent converts to; it is not a separate threshold. Those cutoffs are laid out with their source on the hemoglobin A1C reference page, and What your A1C means covers what to do at each level.
How precise the estimate is
The converter returns one number, but the study behind it found a spread. The ADA's own A1C-to-eAG table (Table 6.1 in the 2026 Standards of Care) prints a range beside every value, which the table labels a 95% confidence interval and the ADAG paper's text calls its 95% prediction limits, the range of averages consistent with a given A1C: an A1C of 6 percent sits at 126 mg/dL with a range of 100 to 152, 7 percent at 154 with 123 to 185, and 8 percent at 183 with 147 to 217. At 7 percent, by our arithmetic, that range spans 62 mg/dL, so two people with the same A1C can have quite different averages and both fit the equation.
The ADA names further caveats in the same section. The table comes from a single study published in 2008, whose average glucose was built from an early continuous glucose monitor that needed calibration several times a day plus fingerstick readings, and the ADA says it is unclear how well the estimates carry over to modern, more accurate CGM systems. So use the eAG to get a feel for what an A1C means in meter units, not to second-guess a meter or a sensor that disagrees with it.
Where the equation comes from
The equation is from the A1C-Derived Average Glucose (ADAG) study, Nathan and colleagues, Diabetes Care, 2008. Ten international centers enrolled adults with type 1 diabetes, type 2 diabetes and no diabetes; of 661 who completed screening, 507 were analyzed (268 with type 1, 159 with type 2 and 80 without diabetes). Over three months each person wore a continuous glucose monitor for at least two days on four occasions and did seven-point fingerstick profiles at least three days a week, about 2,700 glucose values per person, and their A1C at the end was measured in a central laboratory.
A straight line fit best: eAG (mg/dL) = 28.7 × A1C − 46.7, with an R2 of 0.84. The paper prints its mmol/L equation two ways: rounded, as 1.59 × A1C − 2.59, in its text, and in full, as 1.5944 × A1C − 2.594, under its Table 2. This converter uses the full form, which reproduces every mmol/L value in the paper's table and in the ADA's Table 6.1. The authors found the line did not differ significantly by age, sex, diabetes type, race or ethnicity, or smoking status, and the ADA and the American Association for Clinical Chemistry judged the correlation strong enough to report eAG next to A1C.
Who was left out matters as much as who was in. The study excluded people with hemoglobin variants, anemia (a hematocrit under 39 percent in men or 36 percent in women), high red-cell turnover, blood loss or transfusion, chronic kidney or liver disease, high-dose vitamin C or erythropoietin treatment, and anyone planning a pregnancy during the study, and it excluded children and pregnant women entirely. The authors wrote that the results apply directly only to people in stable control without red-cell disorders, and that people of Asian and African descent were underrepresented, mainly because a center with a large Asian population withdrew and few participants of African descent enrolled; the ADA describes the cohort as 83 percent non-Hispanic White.
When the A1C misleads, and the eAG with it
Most of the conditions the study excluded are ones where the A1C itself can be wrong, and the eAG inherits the error. The ADA's 2026 diagnosis chapter lists factors that alter red-cell turnover (anemia, iron status, splenectomy, blood loss, transfusion, hemolysis, G6PD deficiency, erythropoietin), plus HIV, cirrhosis, kidney failure, dialysis and pregnancy. In plain terms:
- Anemia. Iron-deficiency anemia raises the A1C, so the result can read falsely high (ADA; NIDDK). The ADA says the A1C may not be a suitable diagnostic test in people with anemia.
- Recent blood loss, transfusion, or erythropoietin treatment. Each changes how long red cells circulate, and the ADA says the A1C may not be a suitable diagnostic test in people treated with erythropoietin. The ADA notes that faster turnover usually lowers the A1C and slower turnover raises it, so a result after a bleed or a transfusion does not describe the months before it.
- Hemoglobin variants. HbS (sickle cell trait), HbC, HbE, HbD and elevated fetal hemoglobin make some A1C methods read falsely high or low; the NGSP keeps a method-by-method list (updated October 2026). The ADA says people with sickle cell trait should be tested on a method without that interference. NIDDK notes that most people with a hemoglobin variant have no symptoms and may not know they carry one. A separate variant, G6PD G202A, carried by 11 percent of Black individuals in the US, lowers the A1C by about 0.8 points in homozygous men and 0.7 in homozygous women, per the ADA.
- Pregnancy. Red-cell turnover rises in pregnancy, so the A1C runs slightly lower in people with and without diabetes. The ADA makes A1C a secondary measure in pregnancy, after glucose monitoring, and sets a pregnancy goal of under 6 percent if that can be reached without significant low blood sugar, relaxed to under 7 percent if needed to prevent it. The ADAG equation was never fitted in pregnancy, so do not convert a pregnancy A1C with it.
- Kidney disease. The ADA says A1C is less reliable at advanced stages of chronic kidney disease and may not be a suitable diagnostic test for people on hemodialysis; fructosamine and glycated albumin, which reflect the past 15 to 30 days, are helpful there. NIDDK adds kidney failure and liver disease to the causes of false results.
The tell is a mismatch. NIDDK says a clinician may suspect interference when your A1C and your blood glucose results do not match, and the ADA recommends checking for a problem with either test when the discordance is consistent and substantial. Where the relationship between A1C and glucose is altered, which the ADA names as some hemoglobin variants, pregnancy, G6PD deficiency, HIV and conditions that alter red-cell turnover, it says plasma glucose criteria, such as the fasting plasma glucose or the oral glucose tolerance test, should be used to diagnose diabetes. If your report flags a hemoglobin variant, treat the converted number as unreliable until a clinician says otherwise.
If you wear a glucose sensor: GMI is not your A1C
CGM apps run the conversion the other way and report a glucose management indicator (GMI), an estimate of A1C from your sensor's mean glucose. Its formula, from Bergenstal and colleagues, Diabetes Care, 2018, is GMI (%) = 3.31 + 0.02392 × mean glucose in mg/dL, fitted on 528 people with diabetes wearing one brand of sensor. In those same people, GMI and the laboratory A1C were identical 19 percent of the time, differed by 0.3 points or more 51 percent of the time, and by 0.5 points or more 28 percent of the time. The ADA lists GMI as a calculated value approximating A1C, "not always equivalent".
So a GMI of 7.0 percent and a lab A1C of 7.4 percent are not a contradiction to panic over; the gap is common, and a steady gap is itself information your clinician can use. Neither the eAG from this page nor a GMI diagnoses diabetes. Diagnosis takes a laboratory A1C on an NGSP-certified method, or a plasma glucose test, confirmed and read by a clinician.