GMI vs eHbA1c

[KC49]

Happy new year to everyone!

I hope everyone is having a great news years, I’m just having an ill timed effort to understand the new GMI version of the estimated HbA1c on my dexcom. I’ve had one proper blood test done since it changed over, and it was pretty accurate.

BUT. When I looked up the different conversions, from average glucose to GMI, at the lower end of the scale, it was telling me that an average glucose of 1.5mmol would produce the lowest end of a healthy HbA1c (of 4.0%, or 20 on the newer international standard) and I feel pretty sure that an average glucose that low is definitely not healthy.

so does anybody know at what point the new GMI stops being accurate?

 

[EllieM]

so does anybody know at what point the new GMI stops being accurate?

Interesting, I noticed that what I thought was hba1c on my clarity app had changed recently (I'm sure an average bg of 7.8mmol/L used to correspond to 6.5% and now 7.7 corresponds to 6.6, and now I realise that it's referring to GMI and not hba1c.

So it's now giving me a slightly higher figure than its old hba1c estimate, which was definitely less than the lab results.

Have you got a link to the conversion table?

When I google I get
Is the Glucose Management Indicator a Useful Tool to Measure HbA1c? - Endocrinology Advisor
GMI (%) is calculated using the formula GMI = 3.31 + 0.02392 × mean glucose in mg/dL, which is derived by regressing contemporaneously measured HbA1c values against mean sensor glucose levels. The study investigators sought to compare laboratory HbA1c results estimated from continuous glucose monitoring data with HbA1c calculated using GMI. GMI was assessed using data from 3 12-week randomized controlled trials of the glucose sensors Navigator 2 (n = 114) and Guardian 3 (n = 85), as well as data from Dexcom sensors. Trial participants were children and adults with type 1 diabetes and HbA1c levels between 7.5% and 10%.

The above link suggests that the data used to derive the formula was got from people with hba1cs between 7.5 and 10, so I wouldn't expect it to be accurate under 7.5%

However that article is from 2019 so a more recent one might have better data?

OK, a more recent article here, with a wider range of hba1c data, though it's saying that the result is pretty inaccurate in my opinion (within 0.5% for half the study? That's +/- 5.5mmol/mol)
Using the GMI to Estimate Your A1C: How Accurate Is It? | diaTribe

Edited to add, and a Happy New Year to you too.

 

[KC49]

Interesting, I noticed that what I thought was hba1c on my clarity app had changed recently (I'm sure an average bg of 7.8mmol/L used to correspond to 6.5% and now 7.7 corresponds to 6.6, and now I realise that it's referring to GMI and not hba1c.

So it's now giving me a slightly higher figure than its old hba1c estimate, which was definitely less than the lab results.

Have you got a link to the conversion table?

When I google I get
Is the Glucose Management Indicator a Useful Tool to Measure HbA1c? - Endocrinology Advisor
GMI (%) is calculated using the formula GMI = 3.31 + 0.02392 × mean glucose in mg/dL, which is derived by regressing contemporaneously measured HbA1c values against mean sensor glucose levels. The study investigators sought to compare laboratory HbA1c results estimated from continuous glucose monitoring data with HbA1c calculated using GMI. GMI was assessed using data from 3 12-week randomized controlled trials of the glucose sensors Navigator 2 (n = 114) and Guardian 3 (n = 85), as well as data from Dexcom sensors. Trial participants were children and adults with type 1 diabetes and HbA1c levels between 7.5% and 10%.

The above link suggests that the data used to derive the formula was got from people with hba1cs between 7.5 and 10, so I wouldn't expect it to be accurate under 7.5%

However that article is from 2019 so a more recent one might have better data?

OK, a more recent article here, with a wider range of hba1c data, though it's saying that the result is pretty inaccurate in my opinion (within 0.5% for half the study? That's +/- 5.5mmol/mol)
Using the GMI to Estimate Your A1C: How Accurate Is It? | diaTribe

Edited to add, and a Happy New Year to you too.

Here’s the webpage I was using to calculate different average glucose levels to GMI.

https://www.jaeb.org/gmi/

Whatever formula it uses must have a lower limit of accuracy- because it starts giving you a GMI higher than the average glucose.

For example: (copied and pasted)

Enter mean glucose and select the units:

Enter Mean Glucose:
Select Units: mg/dL mmol/L
The GMI for a mean glucose of 1.5 mmol/L is 4% ( 19.8 mmol/mol )
Formula to calculate GMI:
GMI(%) = 3.31 + 0.02392 x [mean glucose in mg/dL]
GMI(mmol/mol) = 12.71 + 4.70587 x [mean glucose in mmol/L]


I don’t have, have never had, an average glucose anything like that low, I was just curious what the lower limit of a’normal’ hba1c would be, and had to keep putting in lower and lower numbers until it came up with this result.

And no one is walking around living a happy life with an average glucose of 1.5mmol haha.

Occasionally, if I’m having a good day or two and have eaten decent food etc, I get a 2 day average of 5.2-5.5mmol and 95% time in range, but I don’t think there’s a chance in hell of me keeping that up for 90 days, to be able to compare it to an actual HbA1c. But I’d like to know what it would be, if living a perfect life were possible!

 

[EllieM]

OK, I've had a read of the study they used to calculate that formula.

The above graph is quite telling, It's not like there is a scientific principal saying that there is a linear relationship between average glucose and a1c, they've just drawn a straight line and calculated it's gradient from the data they have. And the data is non existent below an a1c of 5.5% and a mean glucose of about 100mg/dl (5.5mmol/L) and the straight line isn't looking that good an approximation below the low 6s.

So I would argue that this is a convenient construct that has no meaning once you get much below a bg of about 5mmol/L and an a1c of the low 6s.

 

[KC49]

OK, I've had a read of the study they used to calculate that formula.

The above graph is quite telling, It's not like there is a scientific principal saying that there is a linear relationship between average glucose and a1c, they've just drawn a straight line and calculated it's gradient from the data they have. And the data is non existent below an a1c of 5.5% and a mean glucose of about 100mg/dl (5.5mmol/L) and the straight line isn't looking that good an approximation below the low 6s.

So I would argue that this is a convenient construct that has no meaning once you get much below a bg of about 5mmol/L and an a1c of the low 6s.

only just saw this reply- thanks for the info!

interesting to see, and it makes more sense now. Hardly any data from lower average glucose results, and as you say, they don’t fit with the line they’ve decided on so much.

I guess I’ll have to wait and see if I can keep my average glucose around 6 mmol for long enough to compare it to a Hba1c from a blood test. I think I stand a better chance of it since I got an omnipod, now that my night time sugars don’t raise my averages. Time will tell!