Interesting article in BMJ re libre accuracy

[Boo1979]

v small sample size (28) tho
2015 study published 2017/18
http://drc.bmj.com/content/bmjdrc/5/1/e000320.full.pdf

 

[Rokaab]

Yes interesting but as you mention a very small sample size, you can't really get any proper stats from a sample size that low

 

[Bluetit1802]

So basically, the Libre reads lower when glucose is in the lower range, and higher when glucose levels are in the higher range.
It also underestimates the effect of a meal - but how high was the glucose after the meal? What are the ranges? (sorry, I didn't read all the paper)

I have read on here that people have noticed it reads higher in the high range and lower in the low range. So that bit may well be correct.

 

[Rokaab]

I have read on here that people have noticed it reads higher in the high range and lower in the low range. So that bit may well be correct.

It does for me, if it shows below 5 I can't trust how accurate it is - when it shows 'Low' I can be anywhere between 2.4 and 5.6

 

[tim2000s]

Outside of roughly 4.5-10, Abbott recommends confirming readings with a blood test. That's because outside this range they're not very accurate. Like most CGMs as it happens.

The other key thing to look at is the Clarke Error Grid, which shows that use away from the arm (in this case in the abdomen) really doesn't look to be all that good an idea due to the dispersion of results.

 

[Boo1979]

The one nocturnal hypo its shown for me where Ive still been awake so able to do a comparison the libre has shown between 3.4 and 3.9: finger prick test has shown 4.3 - 5.6 - concerning but not a problem as I was able to do the comparison test. More worryingly tho, 2 different sensors have now recorded levels in the 2’s at times when I havnt been awake to do a comparison and on both occasions the next day has seen highs that dont match what Ive eaten / done

 

[bamba]

It may not be an issue with accuracy -
the libre is not measuring blood glucose but interstitial glucose as an indication of blood glucose.

 

[Deleted Account]

I also notice the article (with albeit a small sample) said that only 85.5% of the Libre readings were "within Clarke Error Grid". I assume this is a different method of accuracy calculation that the manufacturers use because the rep told me the Libre was inaccurate only 12% of the time.

 

[Bluetit1802]

I've seen some 2's with a majority of 3s when I've been asleep. After the initial shock I just learned to treat my asleep readings as inaccurately low. Possibly because of lying on the sensor, but probably not because shortly before I wake up (and this also includes waking during the night) I rise to the 5s and am still lying on it. I can't comment on the over 10 range as I've rarely seen those dizzy heights.

As for post meals, I have found the Libre and finger pricks give me a very similar rise from before to after. (even if the actual levels are a lot different)

 

[ringi]

It does for me, if it shows below 5 I can't trust how accurate it is - when it shows 'Low' I can be anywhere between 2.4 and 5.6

As soon as a cheap CGM (libre like pricing) sorts out the problem, and is 100% dependable at low ranges, the NHS will have no good reasons not to fund for CGM for everyone at risk of hypos, as the system would be approved by the DVLA, and hence save the cost of test strips.

I expect this is the key issue stopping the Libre being "standard issue" on the NHS as the cost is too high if people need to also use test strips to keep the DVLA happy.

 

[Boo1979]

I've seen some 2's with a majority of 3s when I've been asleep. After the initial shock I just learned to treat my asleep readings as inaccurately low. Possibly because of lying on the sensor, but probably not because shortly before I wake up (and this also includes waking during the night) I rise to the 5s and am still lying on it. I can't comment on the over 10 range as I've rarely seen those dizzy heights.

As for post meals, I have found the Libre and finger pricks give me a very similar rise from before to after. (even if the actual levels are a lot different)

For me it has to be a bit different as I am on Gluclizide so at significant risk of hypos from the meds - thinking “oh well its just the meter being wrong” isnt a very safe assumption to make. I would think thats even more the case for T1 & T2 folk on insulin. There needs to be increased accuracy to make the libre a good option.
If the libre was coming up with false lows all the time, then thats one thing but 3 nights out of 45 ( nearly7 weeks ) from 2 sensors out of 4 is something different which I cant safely ignore

 

[Boo1979]

I also notice the article (with albeit a small sample) said that only 85.5% of the Libre readings were "within Clarke Error Grid". I assume this is a different method of accuracy calculation that the manufacturers use because the rep told me the Libre was inaccurate only 12% of the time.

Customer services told me (when I queried a sensor that was showing some highs that were quite far off blood results) that a varience of 2-2.5mmol fits what their computer algorithms see as acceptable - that becomes a problem when we are talking hypo territory - they replaced the sensor because of the amount of varience it was showing

 

[tim2000s]

Customer services told me (when I queried a sensor that was showing some highs that were quite far off blood results) that a varience of 2-2.5mmol fits what their computer algorithms see as acceptable - that becomes a problem when we are talking hypo territory - they replaced the sensor because of the amount of varience it was showing

That really depends on the current glucose reading. They've always maintained that the error should be less than 15% of the current blood glucose reading.

 

[Boo1979]

The difference was between 8’s to 9 on blood reading and 11’s from the libre

 

[Mbaker]

I've posted several times about the inaccuracies I experienced with the Libre, maybe they could have a calibration protocol in their App, I was getting 2's.