🩸Reading an A1C result

What is being measured

A1C is the proportion of your haemoglobin that has become glycated - that is, has had glucose attach to it non-enzymatically. Glucose binds to haemoglobin at a rate that depends on how much glucose is around, and haemoglobin lives inside red blood cells, which circulate for roughly 120 days before being replaced. Measure the glycated fraction and you have a physical record of glucose exposure over the life of that population of cells.

That is why an A1C is described as reflecting the previous two to three months. The weighting inside that window is not even: the most recent weeks count for more, because the cells carrying the oldest glycation are steadily being retired. It is a weighted look backwards, not an average of equal days.

It is also a property of your blood, not an arithmetic function of readings. This distinction matters when you compare a laboratory A1C with a figure calculated from meter readings: the first is measured, the second is inferred.

Two unit systems, one measurement

A result reported as 6.5% and one reported as 48 mmol/mol are the same measurement in different units. The percentage is the NGSP scale, used in the United States and India among others; the millimoles per mole figure is the IFCC scale, used in the United Kingdom, Australia and much of Europe.

The NGSP/IFCC master equation converts between them: IFCC mmol/mol equals the NGSP percentage minus 2.15, all multiplied by 10.929. So 5.7% is about 39 mmol/mol, 6.5% is about 48, and 7.0% is about 53. If you are comparing an old report with a new one from a different country, this is usually the whole of the confusion.

The translation into an average glucose

The ADAG study, published by Nathan and colleagues in Diabetes Care in 2008, compared laboratory A1C against continuous and repeated fingerstick glucose measurement in 507 people and fitted a line: estimated average glucose in mg/dL equals 28.7 times A1C, minus 46.7. That single equation generates every row of every A1C-to-glucose chart on this site.

Read the slope for what it says: each whole percentage point of A1C corresponds to about 28.7 mg/dL of average glucose, so each tenth of a point corresponds to roughly 2.9 mg/dL. The intercept of 46.7 is not a physiological quantity - an A1C of zero is not a real state - it is where the fitted line crosses the axis.

The word estimated is doing real work. In the ADAG data, people who shared an A1C had mean glucose values that differed meaningfully from one another. eAG describes what the average person with your A1C had, and you are not the average person with your A1C.

One detail this site is explicit about: the ADAG paper also published a separate regression for mmol/L. This site displays mmol/L by converting its own mg/dL figure, so that the two columns you see agree with each other exactly; the ADA's printed mmol column follows the paper's separate mmol regression instead. The two paths differ by well under 0.1 mmol/L, and the methodology page shows both rather than pretending there is only one.

GMI is a different number with a different equation

If your average glucose came from a continuous glucose monitor rather than a laboratory draw, the equation designed for it is not the ADAG one. Bergenstal and colleagues published the Glucose Management Indicator in 2018: GMI as a percentage equals 3.31 plus 0.02392 times mean glucose in mg/dL. A mean of 150 mg/dL gives a GMI of about 6.9 percent.

The paper introduced the term specifically to stop the CGM-derived figure being called an estimated A1C, because the two commonly differ in the same person, sometimes substantially. That difference is not an error in either number. Its size and direction is a real observation about you, and a more interesting thing to bring to an appointment than either figure alone.

Running the equation backwards, from a meter average

The same ADAG equation inverts: A1C equals your average glucose in mg/dL plus 46.7, all divided by 28.7. An average of 154 mg/dL gives 6.99, which is 7.0 percent at the precision anyone reports. If your readings are in mmol/L, multiply by 18.0182 first so the whole calculation stays on one equation instead of two.

The output is only ever as good as the average you feed it, and this is where home calculations usually go wrong. A handful of fasting readings is not an average of your day - it systematically samples your lowest values. Testing mostly after meals biases the other way. Neither is dishonest; both produce an average of the moments you chose to measure rather than of the day you actually had.

A continuous monitor worn for a full fortnight gives a genuinely representative mean, and for that data the better-suited equation is GMI rather than the inverted ADAG one. If all you have is a meter, the inverted equation is still worth running - just hold the answer loosely, and remember that the laboratory measurement is the one that counts when a number is going to inform a decision.

The ranges, and who they belong to

The ADA defines three ranges on the A1C scale: below 5.7%, the 5.7% to 6.4% band it calls prediabetes, and 6.5% and above, which it uses as one of its diagnostic criteria for diabetes. Those are definitions of numeric ranges. Where a result sits relative to them is information; a diagnosis is something a clinician makes, and the ADA normally asks for a repeat abnormal test, or one abnormal test alongside unambiguous symptoms, before one is made.

This site writes those ranges in that form deliberately - the range belongs to the body that defined it. You will not find a page here that tells you what your number means about you, because that reading depends on your history, your other results and your medications, none of which a calculator can see.

The tolerance the measurement itself carries

NGSP certification allows a certified A1C method to differ from the reference method by roughly half a percentage point. Set that beside the slope of the ADAG equation and something uncomfortable follows: half a point of assay tolerance corresponds to about 14 mg/dL of estimated average glucose, which is far larger than the differences people scrutinise between two results.

Two consequences are worth carrying. A change of 0.2 or 0.3 points between tests sits inside measurement tolerance and is not clearly a change at all. And a value sitting a tenth either side of a boundary - 5.6 against 5.8, or 6.4 against 6.5 - is precisely where a single number tells you least, which is why the ADA asks for confirmation there rather than less.

If you are comparing results, compare like with like: the same laboratory where possible, since analytical methods differ, and noting whether either result was a point-of-care fingerstick rather than a venous draw.

What one A1C cannot show you

It cannot show variability. A1C is an average, and averages hide shape: two people with identical A1C values can have very different daily glucose ranges, one steady and one swinging widely. Nothing in the number distinguishes them, which is part of why CGM metrics exist alongside it.

It can also be moved by things that are not glucose. Anything that alters red blood cell lifespan - anaemia, recent blood loss, transfusion, some haemoglobin variants, pregnancy, chronic kidney disease - shifts A1C independently of glycemia, in either direction. Where any of those apply, the number needs interpreting rather than reading, and that interpretation belongs to a clinician who knows the rest of the picture.

And it cannot show a trend. One value is one point. The most useful thing you can do with an A1C result is write it down with its date and the laboratory that produced it, so that the next one has something honest to be compared against.