A1c Calculator

What A1c actually measures

Glucose in the bloodstream attaches itself to haemoglobin inside red blood cells. The reaction is slow, non-enzymatic and essentially irreversible: once a haemoglobin molecule is glycated, it stays glycated for the life of the cell. Because red cells circulate for roughly 90 to 120 days before being cleared, the proportion of haemoglobin carrying that sugar is a running record of how much glucose has been around.

That record is not evenly weighted, which is the first thing most explanations skip. The most recent month contributes roughly half of the value, the month before that around a quarter, and the oldest month the remainder. A person who spent two difficult months and then changed something four weeks before the blood draw will see a result closer to the recent weeks than a simple three-month average would suggest. It is a weighted memory, not a photograph.

This is also why an A1c drawn two weeks after a change in treatment says almost nothing about that change. The red cells carrying the old record are still in circulation. Most laboratories and guidelines settle on a repeat interval of about three months for exactly this reason.

Four scales, and the error that costs an order of magnitude

A single glycated haemoglobin result can be written four different ways, and a person moving between countries, between a lab report and an app, or between a clinician and a forum post will meet all four.

The two A1c scales are related linearly, but not by a bare multiplication:

IFCC (mmol/mol) = (NGSP % − 2.15) × 10.929

That intercept of 2.15 is not decoration. The NGSP scale is anchored to the assay used in the DCCT trial, which carried along a fraction of glycated haemoglobins that were not strictly HbA1c; the IFCC reference method measures only the specific molecule. Drop the intercept and multiply 7 by 10.929 and you get 76.5 instead of 53 — a result that is not just wrong but wrong in a direction that looks plausible on a lab report.

The genuinely dangerous confusion is between the two A1c scales themselves. Reading 53 mmol/mol as 53% would imply an average glucose above 1,400 mg/dL, a value incompatible with life. This calculator refuses to compute anything that resolves to an A1c below 3% or above 20% and says why, rather than printing a four-digit number with the confident typography of a real result.

On the glucose side, mg/dL and mmol/L are related by the molar mass of glucose: mg/dL ÷ 18.0182 = mmol/L. An "average of 8" is unremarkable in mmol/L and impossible in mg/dL.

Where the eAG formula comes from

The conversion between A1c and estimated average glucose used here is:

eAG (mg/dL) = 28.7 × A1c(%) − 46.7

It comes from the A1c-Derived Average Glucose study — Nathan and colleagues, published in Diabetes Care in 2008. The design matters for judging how much weight the equation deserves: 507 participants, including people with type 1 diabetes, people with type 2 diabetes and people without diabetes, each contributing roughly 2,700 glucose measurements over twelve weeks from a combination of continuous monitoring and fingerstick testing. The linear regression across those pairs produced the coefficients above, and the American Diabetes Association adopted them for its own A1c-to-eAG conversion.

Provenance is worth stating because older conversions still circulate. The DCCT-era regression was eAG = 35.6 × A1c − 77.3, and it gives noticeably different answers: at 7% it returns 172 mg/dL rather than 154. A health calculator that does not say which equation it used is asking to be trusted on nothing.

The spread is the finding, not the footnote

A regression line describes a cloud of points, and the cloud in the ADAG data is wide. The study published a 95% interval alongside each A1c value, and those intervals are what this page shows next to the point estimate:

Read the 7% row carefully. Two people with an identical laboratory result can have three-month averages that differ by more than 60 mg/dL. That is not measurement noise in the assay; it is real biological variation in how quickly haemoglobin glycates and how long red cells survive. Some people are consistent "high glycators" and some are consistent "low glycators", and the difference persists across repeated tests.

The practical consequence is that comparing your own A1c to your own previous A1c is far more informative than comparing it to someone else's, and that a conversion to mg/dL should be read as a neighbourhood rather than an address.

How A1c is used in diagnosis, and what the ADA targets are

Since 2010 the American Diabetes Association has accepted A1c as a diagnostic test in its own right, using a threshold of 6.5% (48 mmol/mol) on a properly standardised assay, with a range of 5.7% to 6.4% (39–47 mmol/mol) described as increased risk. A diagnosis normally requires two abnormal results, either two A1c values or an A1c plus a fasting or post-load glucose, unless symptoms are unmistakable.

For people already living with diabetes, the ADA Standards of Care describe a general target of below 7% (53 mmol/mol) for many non-pregnant adults, a tighter target of below 6.5% where it can be reached without significant hypoglycaemia, and a looser target of below 8% where the risks of intensive treatment outweigh the benefits — limited life expectancy, hypoglycaemia unawareness, advanced complications, extensive comorbidity.

Those are guideline thresholds, not a verdict on any individual, and this page deliberately does not compare your number to them or colour it in. Which target applies to a given person depends on age, duration of diabetes, hypoglycaemia risk, kidney and heart status, pregnancy, and what that person wants from treatment. That conversation belongs to the clinician who follows the case.

Time in range: what the average cannot see

Continuous glucose monitoring has changed what "good" data looks like in American practice, and it has exposed the central weakness of A1c: an average conceals its own composition. Two people can share an A1c of 7% while one spends the day gently between 120 and 160 mg/dL and the other alternates between 50 and 280. The second person is having a much harder time, and the A1c cannot tell them apart.

The 2019 international consensus on CGM metrics set out targets that describe the shape of the day rather than its mean: for most non-pregnant adults with type 1 or type 2 diabetes, more than 70% of readings between 70 and 180 mg/dL (3.9–10.0 mmol/L), less than 4% below 70 mg/dL, less than 1% below 54 mg/dL, and less than 25% above 180 mg/dL. Roughly, each 10 percentage points of time in range corresponds to about 0.5 to 0.8 percentage points of A1c — a useful rule of thumb and nothing more.

CGM reports also carry a Glucose Management Indicator, which is a mean glucose from sensor data converted onto the A1c scale. GMI and laboratory A1c routinely disagree by a few tenths of a point in the same person, and that disagreement is information rather than error: it is the individual glycation difference showing up in the open. None of this makes A1c obsolete — it remains the outcome measure that decades of trials were built on, and it is cheap, standardised and available where sensors are not.

When the number is measuring something else

Because A1c depends on red cells staying alive for a predictable length of time, anything that changes their lifespan changes the result without any change in glucose:

• Shortened red-cell survival — haemolytic anaemia, recent significant blood loss, recent transfusion, splenomegaly, erythropoietin therapy. Younger cells have had less time to glycate, so A1c reads low.
• Lengthened red-cell survival — iron-deficiency anaemia, vitamin B12 deficiency, splenectomy. Older cells have had more time, so A1c reads high. Treating the iron deficiency can drop an A1c by half a point with no change in glucose at all.
• Haemoglobin variants — HbS, HbC, HbE and HbD interfere with some assay methods and not others. The NGSP publishes which assays are affected, and a laboratory can usually switch method.
• Chronic kidney disease and dialysis — shortened red-cell survival plus carbamylation of haemoglobin makes A1c unreliable in both directions.
• Pregnancy — red-cell turnover accelerates and plasma volume expands, so A1c runs lower than the same glucose would produce outside pregnancy, and gestational targets are different in kind rather than degree.

When A1c and day-to-day glucose readings tell different stories, that mismatch is a reason to investigate rather than to pick the more comfortable number. Fructosamine, glycated albumin and CGM all give a shorter-window view that does not depend on red-cell lifespan.

How this calculator works

All four fields are live and linked in both directions. Type an A1c in percent and the IFCC value, the mg/dL average and the mmol/L average follow; type an average glucose from a meter or a CGM report and the equivalent A1c appears. Every conversion routes through a single pivot value in percent, so there is no alternate path that could produce a slightly different answer.

Going backwards — from a measured average to an A1c — carries extra uncertainty, and the page says so when you do it. The ADAG regression was fitted to predict glucose from A1c; reading it in reverse inherits the whole spread without the benefit of a blood test.

Nothing here is sent anywhere. The arithmetic runs in your browser, there is no account, no upload and no log, and the page keeps working offline once loaded. That matters because the value you type is health information about you.