• Please Remember: Members are only permitted to share their own experiences. Members are not qualified to give medical advice. Additionally, everyone manages their health differently. Please be respectful of other people's opinions about their own diabetes management.

Libre HbA1c v Actual HbA1c

mashedupmatt

Forum Host
Relationship to Diabetes
Type 1
Interested to know how accurate everyone's Libre guesstimates are against actual blood results - mine was 5 mmol/mol higher on blood test to what Libre was saying - admittedly I don't pay too much attention to what it calculates but would be interested to know others experiences...
 
Last edited:
Libre always used to estimate much lower for me too.

Dexcom don’t even call it HbA1c any more, they call it a GMI (glucose management indicator)

And Medtronic just use Time in Range.

There was a compare and contrast of different formulae somewhere (with @Docb ’s experiments) I’ll see if I can find it

Edit: Here it is! The formulae variants are a few posts further on. Not sure what Abbott use though

 
Last edited:
Do you feed the libre software with actual HbA1c results? If not, I'm guessing that the Libre software uses some generic correlation, not taking into account the variability between individuals.

My predictions are based on correlations of my HbA1c results with my finger pricking. Its why I describe my predictions about my Hba1c results as being individual to me.
 
Do you feed the libre software with actual HbA1c results? If not, I'm guessing that the Libre software uses some generic correlation, not taking into account the variability between individuals.

My predictions are based on correlations of my HbA1c results with my finger pricking. Its why I describe my predictions about my Hba1c results as being individual to me.

I’m pretty sure Libre and Dexcom just use your 24hr average sensor glucose and apply an empirically derived (but not personal) algorithm.

Back in the day I used to use

HbA1c (%) = (AvgBG+2.52)/1.583

Though when using an earlier Contour meter it worked better for me if I added .5 or 1 to my AvgBG figure.
 
Any more want to post actual HbA1c result together with the Libre prediction? If we get a few more we should be able to see what the overall spread of the difference looks like.

to date we have

Libre Actual
45 47
51 47
41 39
35 45
 
If comparing/plotting the variations I think it is important to consider how much Libre data is being used to estimate HBA1c. It will provide an estimate based on 2 weeks of data whereas the blood test provides an indication of BG over the last 3 months.
 
From what I remember my last prediction was 48 came back at 46. Is this based of the 90 days data or the report say two weeks in libre view?
That is an interesting point and may explain some of the variability in prediction differences if each prediction is only based on the 2 weeks of data for each AGP period.
 
Any more want to post actual HbA1c result together with the Libre prediction? If we get a few more we should be able to see what the overall spread of the difference looks like.

to date we have

Libre Actual
45 47
51 47
41 39
35 45
Libre 39 - actual 44
 
Just had a play in LibreView and even if you select 90 day reports, when you go into AGP reports which gives you your predicted HbA1c it goes back to 2 week periods, so you would need to take the average of the last 6x 2 week predictions to get a better approximate. So when you look at your predicted HbA1c it is only the last 2 weeks. If you had a really good 2 weeks it is likely to under predict and if you had a poor 2 weeks perhaps over Christmas or when you had an illness, the prediction will likely be higher than actual result.
 
I’ve only had a Hba1c done once since I started using Libra just over 6 months ago
but the result was 48 and if I recall correctly the Libra predicted it as 46 (currently Libra prediction is 45) so will see what my result is in April when I will be getting my next Hba1c done
But for me the Libra has been a big help
 
This is brilliant. @helli - I've no idea how the libre does its estimating but if you don't enter any real results for it to fine tune predictions for any individual I reckon they must use a standard algorithm for everybody.

Libre Actual Difference
45 47 -2
51 47 4
41 39 2
35 45 -10
38 47 -9
48 46 2
39 44 -5
48 56 -8
46 58 -12

This is beginning to look interesting. Seems to be a bias towards underestimating, sometimes by quite a lot. Could do with more data. Any offers? Historic stuff would be OK.
 
@Docb But the Libre prediction is only based on the last 2 weeks worth of Libre data (ie the average glucose over that 2 week period) by the look of it, so not the full picture by any means.
 
Back
Top