Glycemic markers
A1C, estimated average glucose and GMI — three numbers that describe glucose exposure in three different ways, each with its own equation, its own units, and its own published error. Conversion in both directions, in mg/dL and mmol/L.
Three numbers, three different measurements
A1C is a laboratory measurement of the proportion of your haemoglobin that has become glycated. Because haemoglobin sits inside red blood cells that circulate for roughly 120 days, it is a physical record of glucose exposure over the previous two to three months, weighted towards the most recent weeks.
eAG is not a measurement at all. It is that A1C run through a regression - estimated average glucose in mg/dL equals 28.7 times A1C minus 46.7 - fitted by Nathan and colleagues in 2008 across 507 people in the ADAG study. It answers the question "what average glucose did people with this A1C tend to have".
GMI runs the other way, from continuous-monitor data. Bergenstal and colleagues published it in 2018: 3.31 plus 0.02392 times CGM mean glucose in mg/dL. They coined the term specifically to stop that figure being called an estimated A1C, because in the same person the two commonly differ. Three names, three equations, three things - and the commonest error in this area is treating any two of them as interchangeable.
The units, which are the other half of the confusion
Glucose is reported in mg/dL in the United States and India, and in mmol/L in the UK, Australia and Canada. The molar factor for glucose is 18.0182, so the two are one division apart, and every tool here shows both.
A1C itself has two scales. The NGSP percentage most readers know, and IFCC units in millimoles per mole, which is what a UK, Australian or most European laboratory prints. The NGSP/IFCC master equation converts between them: IFCC equals the NGSP percentage minus 2.15, times 10.929. So 6.5% and 48 mmol/mol are the same result, and someone comparing reports across countries is usually looking at one measurement in two costumes.
One subtlety this site is open about: the ADAG paper published a separate regression for the mmol/L column. This site derives mmol/L by converting its own mg/dL figure so the two columns agree with each other exactly, which leaves it under 0.1 mmol/L from the path the ADA's printed table follows. The methodology page shows both rather than pretending the choice does not exist.
The tolerance, and what it means for small changes
NGSP certification allows a certified A1C method to differ from the reference method by roughly half a percentage point. Against the ADAG slope of 28.7 mg/dL per point, that is about 14 mg/dL of estimated average glucose - larger than most of the differences people scrutinise between consecutive results.
Two things follow directly. A change of 0.2 or 0.3 points between tests sits inside measurement tolerance, and the honest reading is that nothing has clearly changed. And a value a tenth either side of a boundary is exactly where one number tells you least, which is why the ADA asks for a confirmatory test there rather than treating a boundary crossing as settled.
This is also why the arithmetic here always shows its working. Every conversion on these pages is one published equation applied once, with the source and the date it was last checked printed next to the result.
What this cluster does not publish
There is no "how to lower your A1C" content anywhere on this site, despite the search volume. That intent is treatment advice, and what moves a glycemic marker, by how much, and whether it should be moved at all are clinical questions that depend on a diagnosis, other results, medications and hypoglycaemia risk - none of which a calculator can see.
There is no at-home test-kit recommendation and no affiliate relationship with any laboratory, supplement or telehealth service; a test in this repository fails the build if one is ever added. And there is no page that tells you what your number means about you. The ADA defines ranges; this site names which range a figure falls in, in the ADA's own wording, and stops.
What is left after those exclusions is arithmetic, provenance and the size of the error - which is, on a subject this consequential, the part a website can actually be trusted with.
The per-value pages, and why the family stops where it does
Alongside the three calculators, this cluster publishes a page for each individual A1C value people actually search for. Each one carries the conversion, the range it falls in, the neighbouring values for contrast, and hand-written context on the figures that get looked up most.
The family is deliberately smaller than it could be. It was originally specced at tenth-of-a-point steps from 4.5 to 9.0, which would have been 56 pages; measured search demand per value exists only between roughly 5.2 and 6.5, so the family is now tenths across 5.0 to 7.0 plus half-points to 14.0. Every remaining page answers a question somebody asks. Curating a family down is cheaper than defending pages that exist because a loop ran.
A machine check enforces the other half of that discipline. Sibling pages are compared by Jaccard similarity over five-word shingles, and the build fails above a threshold calibrated against the measured distribution rather than guessed at. If two pages in the family read as near-duplicates, the fix is more distinct writing or fewer pages - never a looser threshold.
In-depth guides
Calculators for this topic
An A1C calculator converts a hemoglobin A1C percentage into an estimated average glucose, in both mg/dL and mmol/L, using the ADA equation eAG = 28.7 x A1C - 46.7 published by Nathan in 2008. It converts in either direction, shows IFCC units, and states which range the ADA defines your figure as falling in.
This converts an average blood glucose reading into the A1C percentage it corresponds to, by inverting the ADA equation from the ADAG study. Enter a meter or CGM average in mg/dL or mmol/L and it returns the equivalent A1C in both NGSP percent and IFCC units, with the ADA range named.
GMI estimates what an A1C would look like given your CGM mean glucose, using the equation GMI = 3.31 + 0.02392 x mean glucose from Bergenstal 2018. Enter your mean glucose and, optionally, your laboratory A1C to see the gap between them, which the paper treats as informative rather than an error.
Common questions
Get a quarterly nudge to re-check your numbers
Four emails a year: a reminder to re-take your numbers, and what changed in the guidance behind these tools. No spam, unsubscribe any time.
Double opt-in: you will get one confirmation email and nothing else until you click it.