Glucose Management Indicator (GMI) on Libre reports - conflicting info?

[Vectian]

When you look at your CGM data on LibreView, it gives a Glucose Management Indicator (GMI) figure which it says is "Approximate A1C level based on average CGM glucose level". It also gives a Average Glucose level. When I put the average glucose number into an average glucose to A1C converter (e.g. https://www.diabetes.co.uk/hba1c-to-blood-sugar-level-converter.html ), the resulting A1C is much lower than the GMI figure on Libre. For example, average glucose of 6.2 gives and A1C of 42, but using the converters gives and A1C of 36.7, a pretty big difference.

Which is correct? According to the GMI, to reach an A1C of 38 ("normal" non diabetic level that I was a couple of years ago) would need an average glucose of 5.4, which seems pretty much impossible given that everyone rises well above that every time they eat.

 

[John Gray]

Which is correct?
Probably neither! The figures are based on interstitial fluid, not on blood glucose. They may, coincidentally, be similar.

My Libre 2 "Estimated A1c" shows 49 mmol/mol, my latest blood tests results are 44. That's within about 10%.

 

[rebrascora]

The predicted HbA1c on Libre is usually a few mmols lower than my actual HbA1c and I think most people find that although @John Gray above seems to be the opposite of that. The feature is just there to give you a guide.
Contrary to your comment, there really isn't actually much difference between 37 (rounding 36.7 up) and 42 and even the HbA1c blood test will have an error factor, which I would imagine would be about +/- 1-2mmols with that alone and that is a laboratory test.
Libre is not accurate enough to give you better data. Also the conversion table you are using to translate average BG into HbA1c is just a guide and not totally reliable/accurate either. Just too many variables which affect the conversion from one body to another.

 

[PerSpinasAdAstra]

The Libre 2 gives me two different estimated figures - the A1c displayed in the app based on the past 90 days worth of gathered data, and the GMI figure in LibreView where you can specify the date range. These two numbers don't match for me. Apparently they use different formulas, and no formula used to estimate HbA1c is perfect. A few of the formulas used are listed in this thread:

Diabetes apps

Hi I use MySugr app for recording my bs readings however only thing I'm not happy with is that unless you post so many entries in a 7 day period you don't get an estimated HbA1c. This to me is a bit unfair. Surely if I'm recording values every day even if it's only once a day that I test I...

forum.diabetes.org.uk

My Libre GMI at the beginning of March was 39, so not impossible. Since losing all the weight (and due to the meds I'm on) my fasting BG is generally below 5 (The Libre shows it hitting 4 sometimes while I'm asleep) and presumably the hours of fasting overnight balance out the hours of higher BG after eating to lower the overall average.

 

[littlevoice359]

It is quite possible the Libre GMI number uses the same methodology as Dexcom Clarity. If so, this link to a post I made last year may be helpful.

https://forum.diabetes.org.uk/board...ue-shown-on-the-dexcom-clarity-report.106799/

 

[Vectian]

The predicted HbA1c on Libre is usually a few mmols lower than my actual HbA1c and I think most people find that although @John Gray above seems to be the opposite of that. The feature is just there to give you a guide.
Contrary to your comment, there really isn't actually much difference between 37 (rounding 36.7 up) and 42 and even the HbA1c blood test will have an error factor, which I would imagine would be about +/- 1-2mmols with that alone and that is a laboratory test.
Libre is not accurate enough to give you better data. Also the conversion table you are using to translate average BG into HbA1c is just a guide and not totally reliable/accurate either. Just too many variables which affect the conversion from one body to another.

42 is the upper limit and almost into prediabetes, whereas 37 is comfortably in non-diabetic range, about 0.8 lower on the average. I read that every amount the A1C can be lowered gives significant benefits in overall health, even into the higher end of "normal" range.

 

[Vectian]

The Libre 2 gives me two different estimated figures - the A1c displayed in the app based on the past 90 days worth of gathered data, and the GMI figure in LibreView where you can specify the date range. These two numbers don't match for me. Apparently they use different formulas, and no formula used to estimate HbA1c is perfect. A few of the formulas used are listed in this thread:

Diabetes apps

Hi I use MySugr app for recording my bs readings however only thing I'm not happy with is that unless you post so many entries in a 7 day period you don't get an estimated HbA1c. This to me is a bit unfair. Surely if I'm recording values every day even if it's only once a day that I test I...

forum.diabetes.org.uk

My Libre GMI at the beginning of March was 39, so not impossible. Since losing all the weight (and due to the meds I'm on) my fasting BG is generally below 5 (The Libre shows it hitting 4 sometimes while I'm asleep) and presumably the hours of fasting overnight balance out the hours of higher BG after eating to lower the overall average.

Is that metformin?

 

[PerSpinasAdAstra]

Is that metformin?

I'm currently on 2000mg Metformin per day and 10mg Dapagliflozin per day. I'm unsure of their exact impact on my fasting BG. In a person with impaired fasting BG Metformin will definitely push it down as it inhibits glucose output from the liver, but as Metformin is not known for causing hypoglycaemia it clearly can't push it down below a certain level. Dapagliflozin also is not known to cause hypoglycaemia.

I'm extremely curious to find out how much effect the weight loss alone has had on my fasting levels, and am considering some experiments in future to find out, but right now I'm dutifully taking the meds to see how low my next HbA1c will be, due in June.

 

[helli]

What do you mean by “correct”?
They are predictions and estimates and as such they are both correct.
They are not telling you exactly what your HbA1C would be if you had a test today because that is done through a blood test which measures something else to Libre’s tests.

I consider the value from Libre or from other CGMs or anything else that use an average BG as another indicator of how well I am managing my diabetes. Useful to track but not the same as a HbA1C.
For me, the Libre estimate is more than 20% lower than a blood test HbA1C.

 

[Vectian]

I'm currently on 2000mg Metformin per day and 10mg Dapagliflozin per day. I'm unsure of their exact impact on my fasting BG. In a person with impaired fasting BG Metformin will definitely push it down as it inhibits glucose output from the liver, but as Metformin is not known for causing hypoglycaemia it clearly can't push it down below a certain level. Dapagliflozin also is not known to cause hypoglycaemia.

I'm extremely curious to find out how much effect the weight loss alone has had on my fasting levels, and am considering some experiments in future to find out, but right now I'm dutifully taking the meds to see how low my next HbA1c will be, due in June.

Well I had A1C of 97 6 and a half weeks ago, was put on insulin as they didn't know if it is T1 or T2 and the test take 6 weeks (!) still not in. I took it for a month but it really messed me up and I was confident that I am T2, so came off insulin 10 days ago and much better BS since. With restrictive diet, intermittent fasting and exercise I am now usually about 5 overnight, and rarely about 8 at any time, so weekly average of 6.2 with no meds at all. So may be something similar if you stop. I am wondering whether it's worth going on a low dose metformin to bring it down further. Do you get any side effects from it?

 

[Vectian]

What do you mean by “correct”?
They are predictions and estimates and as such they are both correct.
They are not telling you exactly what your HbA1C would be if you had a test today because that is done through a blood test which measures something else to Libre’s tests.

I consider the value from Libre or from other CGMs or anything else that use an average BG as another indicator of how well I am managing my diabetes. Useful to track but not the same as a HbA1C.
For me, the Libre estimate is more than 20% lower than a blood test HbA1C.

Correct as in most accurate, as there is quite a discrepancy. If the estimated A1C on Libre is lower than the actual one, then you would need an average BG of 5 or so to get in the non-diabetic range, that can't be right when the lowest you can safely go is 4.

 

[rebrascora]

If the estimated A1C on Libre is lower than the actual one, then you would need an average BG of 5 or so to get in the non-diabetic range, that can't be right when the lowest you can safely go is 4.

That is not strictly true. Non-diabetic people drop below 4 occasionally particularly in the depths of sleep and 3.5 is technically the hypo level I believe, but 4 is used for those of us on insulin to help give us some wriggle room and preserve our hypo awareness. You could spend several hours in the high 3s and 4s overnight and get an average of 5, if your mealtime spikes were not too high and short lived.

 

[helli]

Correct as in most accurate, as there is quite a discrepancy.

They are estimates using different algorithms and we are all different.
One maybe close to the blood test HbA1C for one person and the other may be closer for another person.

And they are both based on readings which must comply to a standard to be within 15% of the true BG 95% of the time. So you are building estimates upon estimates to get a value which, itself is not guaranteed to 100% accurate.

I learnt many years ago that managing diabetes is not precision engineering - our bodies do not work that way and we don’t know everything about how they work.
As and engineer myself, this was a hard lesson to learn.

The only way to find out if you are in remission (assuming you have type 2 diabetes), is to take the blood test.
You can use the estimates and predictions to track your journey but not provide accuracy or even “closest to accuracy”.

 

[everydayupsanddowns]

I’ve always found those average glucose to estimated HbA1c formulae slightly undercook things for me.

I’ve also had occasions where a low-ish average fingerstick result over 30-60 days or so wasn’t matched by a lower HbA1c.

I find them vaguely helpful these GMI indicators, but I’d never rely on one, and to be honest these days I find Time in Range gives me a much more nuanced assessment of how things have been going in recent weeks / months 🙂

 

[PerSpinasAdAstra]

I am wondering whether it's worth going on a low dose metformin to bring it down further. Do you get any side effects from it?

I had some nausea and 'digestive discomfort' for a few weeks starting Metformin but nothing since. I also had a metallic taste in my mouth for the first couple of days. No side effects at all that I can perceive from the Dapagliflozin, at least for me. Metformin might possibly have an impact on the body's adaptations in response to exercise though, so for this reason, for the next few years, I'd rather be off it or at least reduce it, if possible - Link and Link

I've followed the Roy Taylor 'remission' game plan to get fat out of my liver and pancreas through weight loss, seen great results in lowering BG levels (with a load of other positive changes) and now I'm very keen to prevent the fatty liver problem from recurring by all means possible. As low skeletal muscle mass is a risk factor for Non Alcoholic Fatty Liver Disease (NAFLD), and as I'm currently capable of doing a mighty total of 3 push-ups, I'm going to try to build whatever little muscle might be possible for me. It appears that Metformin might be getting in the way of that. If not for that effect I think I'd be happy enough to stay on it - no side effects for me personally and it lowers insulin resistance, which is likely helping to get my BG down faster after eating, even if I might possibly no longer need it at my current dose to keep my fasting BG levels down.

After my next HbA1c and if it's low enough I'm thinking of maybe asking the doctor to reduce my Metformin dose and getting it as separate tablets to the Dapagliflozin (I currently take Xigduo, which is a combination tablet). I could then use some Libre 2s over a period of a couple of months and experiment with the Metformin dose to see how low I can go without raising my BG too much. After a few years of lifting weights and trying to get much fitter I'd then reconsider the Metformin dose again.

 

[Vectian]

I had some nausea and 'digestive discomfort' for a few weeks starting Metformin but nothing since. I also had a metallic taste in my mouth for the first couple of days. No side effects at all that I can perceive from the Dapagliflozin, at least for me. Metformin might possibly have an impact on the body's adaptations in response to exercise though, so for this reason, for the next few years, I'd rather be off it or at least reduce it, if possible - Link and Link

I've followed the Roy Taylor 'remission' game plan to get fat out of my liver and pancreas through weight loss, seen great results in lowering BG levels (with a load of other positive changes) and now I'm very keen to prevent the fatty liver problem from recurring by all means possible. As low skeletal muscle mass is a risk factor for Non Alcoholic Fatty Liver Disease (NAFLD), and as I'm currently capable of doing a mighty total of 3 push-ups, I'm going to try to build whatever little muscle might be possible for me. It appears that Metformin might be getting in the way of that. If not for that effect I think I'd be happy enough to stay on it - no side effects for me personally and it lowers insulin resistance, which is likely helping to get my BG down faster after eating, even if I might possibly no longer need it at my current dose to keep my fasting BG levels down.

After my next HbA1c and if it's low enough I'm thinking of maybe asking the doctor to reduce my Metformin dose and getting it as separate tablets to the Dapagliflozin (I currently take Xigduo, which is a combination tablet). I could then use some Libre 2s over a period of a couple of months and experiment with the Metformin dose to see how low I can go without raising my BG too much. After a few years of lifting weights and trying to get much fitter I'd then reconsider the Metformin dose again.

I wonder if some of the effects of metformin would linger if you tapered off and stopped taking it. Like if it has made your body more insulin sensitive and reduced liver glucose production then your body having healed to an extent would carry on like that (providing that you didn't mess it all up by eating the wrong things obviously). I have found that intermittent fasting and even 10mins of brisk walk 30mins after each meal really helps to reduce insulin resistance, which was getting quite bad when I was injecting insulin (which I didn't need).

 

[Lucyr]

The formula Libre uses is published in the user manual, it’s an estimate

 

[Lucyr]

Also, are you entering the details correctly in the diabetes.co.uk website? I.e entering the 90 day average bg reading from libre and comparing against the libre estimate? They give identical estimated a1c for me.

 

[Lucyr]

My guess is that you’re misinterpreting the a1c % here in the first screenshot as an average bg. It’s not it’s an a1c estimate.

The average bg that you use to estimate a1c is the number at the bottom of the second screenshot, ensure it’s set to 90 days.

The third screenshot proves that the website you links gives the same estimated a1c

 

[Vectian]

My guess is that you’re misinterpreting the a1c % here in the first screenshot as an average bg. It’s not it’s an a1c estimate.

The average bg that you use to estimate a1c is the number at the bottom of the second screenshot, ensure it’s set to 90 days.

The third screenshot proves that the website you links gives the same estimated a1c

No, below that there is a converter from average glucose to A1C, other converters on other sites give me the same figures:

Then compared with the estimated A1C in LIbreView (based on the GMI) it's quite a bit higher, gives 42 A1C there but 36.7 above.