GMI versus estimated A1C on SugarMate -- How do I interpret this number?



Just noticed that GMI is listed on SugarMate rather than estimated A1C. How do I interpret this number? Is it any different than eA1C?

Why A1c can be different from eA1c and GMI

The GMI is supposed to be based on the average A1c most people would get with that average BG. Apparently the math was done using actual CGM data and A1cs of a bunch of people, so it is supposed to be more representative. If you click on the little “learn more about GMI” thing, the math for it is
GMI (%) = 3.31 + 0.02392 x [mean glucose in mg/dL]

I’ve seen eA1c from average BG math done using a formula like
A1c (%) = (46.7 + [mean glucose in mg/dL]) / 28.7

So yes, the math is a little different. Right now my Sugarmate is calculating my GMI to be 7% and my eA1c to be 6.8%

I’ve personally found that my lab A1c tends to be closer to the eA1c that Sugarmate generates than to the GMI, but I keep the GMI as the stat on my screen since it tends to run a little higher for me, and serves as encouragement to try harder lol


Strange. Liam’s A1C is reflecting higher than his GMI. 6.7% A1C, 6.8% GMI


That is pretty strange! Are they both set to the same time frame?
I did realize I previously had GMI set to 30 days, now when they’re both 90 days its A1c 6.8%, GMI 6.9%

I would expect to see differences like that at different avg BG based on the math, but given how close the %s are Liam and I can’t be too far off on BG averages :stuck_out_tongue:


So I did a quick graph of the average BG vs GMI and eA1c and right around where Liam and I sit is actually where eA1c and GMI cross each other! Pretty interesting to see it graphed out actually. I guess this does help me see the difference between eA1c and GMI more properly though


I am not sure I would consider that higher. While the A1c number is higher, I am assuming that those numbers are equal when the error is factored in.


Aha, that makes sense then. Yes, both mine were set to 90-days. Makes sense the percentages would be different given different time-frames for each.


No, the GMI was higher (what was strange was that @glitzabetes GMI was lower.) My thought was…given the same math, they should have followed the same results (the GMI being higher in both cases.) My A1C that was showing was 6.7, and my GMI showing was 6.8. But we figured out why. She had hers set to two different timeframes. Makes a lot more sense now.


My GMI was also higher actually, missed this part in translation. I think what threw me off was you saying

When really, looking at the numbers, we were both saying the GMI was higher than the A1c :stuck_out_tongue:


?? I’m talking about this statement.


GMI = 7%
A1c = 6.8%

7 > 6.8



:slaps forehead:

D’uh. Sleep deprivation?


And to add fuel to the fire, when they study error in the HbA1c test they get a measured error of 0.19 on any one measurement using the HPLC method, which is much better than the machine at the endo. So any measurement of A1c carries at least a variation of 0.19 and could be as big as 0.85 depending on the test type. Here is a short excerpt from the paper talking about what contributes to the error:

The reproducibility ( s x), or variation arising from different operators, instruments, and laboratories is then given by…and calculated as 0.19. Accordingly, among the five laboratories, 57% of total variance is between-laboratory, 40% is attributable to laboratory-sample interaction, and 3% is attributable to repeatability

So 6.8 +/- 0.19 = 6.7


Nice! Very representative results.


@TiaG, my understanding on the genesis of GMI is that users of CGMs were unhappy about the fact that their real A1c was different from the eA1c estimated from CGM results—hence the creation of the GMI. Now nobody complains that their GMI is different from the A1c they got at the endo’s.


so the GMI usually isn’t different, or it still is and just no one complains cause it doesn’t have the letters “A1C” in it :laughing:?


It still is different. It is almost exactly the same number as before, as per @glitzabetes graph. But it does not have the “A1c” label, so no one complains…


To me, this whole thing is “much ado about nothing”. GMI, A1C…whatever…they’re so close, I’m not going to lose any sleep worrying about which is “the most” accurate representation of the true number.


Between 6-8% the differences between eA1c and GMI are pretty negligible. Outside of that range they do start to differ but… at the end of the day it’s all just trying to predict what your lab A1c will be, and we know there are a variety of factors that can influence this apart from average BG, so it’s likely that neither of them will be correct lol


That’s absolutely correct. Here is the recent paper that elaborates on GMI, eA1c, and A1c.

Could not agree more. Reading through the paper, it’s very difficult not to be a bit cynical about the entire endeavor.