Libre 2 Accuracy issues - help?

[SimonP]

Ignore this post, after re-reading the posts above it seems that since I first used the app they have started extrapolating to predict what the value will be now without the interstitial lag. I'd not realised as I use XDrip+.

The BG values reported by Libre are NOT 10 to 15 minutes behind.
The algorithm used to convert Interstitial Fluid readings to BG take this into consideration.

Are you sure about this? When I first got one I don't recall being told this, nor did I see it in the app at that point. But that was a long time ago! They may well have changed the algorithm/app since I used.

I'm interested because XDrip+ does assume there's a lag, but the lag is only 8min iirc, I'm curious as to whether the raw data coming from the Libre device is actual raw data (i.e. something proportional to the value just sensed) or manipulated to extrapolate the expected interstitial value at 10,15,20min (whichever it is) in the future. I assume the former, and that the app does the extrapolation (which makes me wonder how Juggluco produces matching values, so I've read, I'll have to have a look at the code.)

Does the official app move the line if it gets it wrong? [Having had another look at the responses above, I see it does indeed do this and redraw the line, interesting]

 

[Bubbleblower]

is the libre particularly sensitive to positioning on the back /side of arm?

Yes, it is very much so, I suspect especially if your BMI is low. I also suspect one reason the libre is so unreliable is because it often gets loose; if my sensor eventually falls off the glucose readings will be all over the place before that.

The BG values reported by Libre are NOT 10 to 15 minutes behind.
The algorithm used to convert Interstitial Fluid readings to BG take this into consideration.

And then the average lagtime is -only- 2 minutes and 24 seconds for adults.

I try and apply a sensor a minimum of 12 hours before its activated,

Same here, but that didn't seem to help much, so I tried letting it bed in for longer, which should help against the foreign object response. But if I have it bed in longer I still get this response when I finally activate it and I can't explain that.

 

[helli]

Are you sure about this? When I first got one I don't recall being told this, nor did I see it in the app at that point. But that was a long time ago! They may well have changed the algorithm/app since I used.

This was an update a few years ago.

 

[rebrascora]

I also suspect one reason the libre is so unreliable is because it often gets loose; if my sensor eventually falls off the glucose readings will be all over the place before that.

I started using an adjustable (plastic watch face style) arm band to provide additional support for my Libre and I have not had one come loose since..... but you are right that readings will be unpredictable once it starts to come loose. Now I just use the arm strap for the first few days to improve the Libre's adhesion during the early hours and days after application AND when I am in the shower, as it acts as a visual reminder of which arm my Libre in on so that I don't inadvertently try to scrub it off. 🙄 A hot shower and vigorous scrubbing with an exfoliating mitt are not good friends to Libre sensors and it doesn't seem to matterthat I get into the shower thinking "I must be careful of my Libre sensor and it is on this particular arm this fortnight.... I will forget.... just like I regularly try to scan the wrong arm for the first few days after a changeover. That arm strap really helps protect my sensor in the show just by being able to see it, so that i
t has become part of my routine.

Now Libre removal is always a painful limpet prising experience on day 15 or 16 after application.

 

[Satan’s little helper]

Yes, it is very much so, I suspect especially if your BMI is low. I also suspect one reason the libre is so unreliable is because it often gets loose; if my sensor eventually falls off the glucose readings will be all over the place before that.

You suspect? My “BMI”whatever that is.) is low. But I can still maintain a limpet stick with a Libre.
The OP hasn’t to date mentioned a lose Libre.

 

[Satan’s little helper]

I started using an adjustable (plastic watch face style) arm band to provide additional support for my Libre and I have not had one come loose since..... but you are right that readings will be unpredictable once it starts to come loose. Now I just use the arm strap for the first few days to improve the Libre's adhesion during the early hours and days after application AND when I am in the shower, as it acts as a visual reminder of which arm my Libre in on so that I don't inadvertently try to scrub it off. 🙄 A hot shower and vigorous scrubbing with an exfoliating mitt are not good friends to Libre sensors and it doesn't seem to matterthat I get into the shower thinking "I must be careful of my Libre sensor and it is on this particular arm this fortnight.... I will forget.... just like I regularly try to scan the wrong arm for the first few days after a changeover. That arm strap really helps protect my sensor in the show just by being able to see it, so that i
t has become part of my routine.

Now Libre removal is always a painful limpet prising experience on day 15 or 16 after application.

Yes, I don’t get any shower products near a Libre, the armpit is miles away.
I’m bemused by the ”dark arts” posted using one. These mothers can stick for me.

 

[Bubbleblower]

You suspect?

Here are some of the studies that suspicion is based on:

Assessment of factors that determine the mean absolute relative difference in flash glucose monitoring with reference to plasma glucose levels in Japanese subjects without diabetes

Accuracy and precision of flash glucose monitoring sensors inserted into the abdomen and upper thigh compared with the upper arm

A review of the foreign-body response to subcutaneously-implanted devices: the role of macrophages and cytokines in biofouling and fibrosis.

Accuracy of a CGM Sensor in Pediatric Subjects With Type 1 Diabetes. Comparison of Three Insertion Sites: Arm, Abdomen, and Gluteus

Plus my own experience FWIW.

 

[rianos]

Hi
I recently finished a free trial of the libre 2 sensor and was impressed - all went well. I've now been prescribed them on the NHS and here is where my problems begin

Sensor 1 I had to abort as was getting spuriously low readings - false hypo alarms at night. Abbot are replacing.

Sensor 2 I've had one since yesterday. This seems to be consistently high eg my accucheck meter is showing 6.1 and the sensor is 8.3; miles outside the expected /allowed error margin

For any longer term users... is the libre particularly sensitive to positioning on the back /side of arm? Am I just unlucky with these two sensors. Certainly it's dampened my initial excitement of having real time insights

I found I was having the same issues as you. My trial and first few sensors were pre app update when you still had to manually scan and I didn’t have any issues but since the update i’ve found the libre quite frequently is 2-2.5+ above my readings on a glucometre but that it also tells me i’m hypo when i’m nowhere near it. This has rendered it totally and utterly pointless for me as it seems like a random number generator. This was during periods of stable blood sugars etc.

I did a trial of dexcom G7 and found it MUCH more accurate, the graph was smooth without the constant spikes that libre 2 seemed to suggest i had. Also the dexcom works 30 minutes after applying and immediately it gave a reading exactly the same as my glucometre. The app/widget/ alarms are also infinitely better than the librelink so i’ve decided to self fund. It’s a real shame I have to pay £140 a month for something that actually works as I find the libre unfit for purpose ever since the update.

 

[WoogieBoogie]

I found I was having the same issues as you. My trial and first few sensors were pre app update when you still had to manually scan and I didn’t have any issues but since the update i’ve found the libre quite frequently is 2-2.5+ above my readings on a glucometre but that it also tells me i’m hypo when i’m nowhere near it. This has rendered it totally and utterly pointless for me as it seems like a random number generator. This was during periods of stable blood sugars etc.

I did a trial of dexcom G7 and found it MUCH more accurate, the graph was smooth without the constant spikes that libre 2 seemed to suggest i had. Also the dexcom works 30 minutes after applying and immediately it gave a reading exactly the same as my glucometre. The app/widget/ alarms are also infinitely better than the librelink so i’ve decided to self fund. It’s a real shame I have to pay £140 a month for something that actually works as I find the libre unfit for purpose ever since the update.

I'm glad not just me. I'll keep going for now but this morning it's still, as you put it well, a random number generator eg fasting finger prick 5.2, libre 8.2. Maybe coincidental that it's since cgm update but does seem suspicious

 

[Bubbleblower]

I apologize for saying a libre could get unreliable if it gets lose and will be more carefull what I say next time.

 

[rebrascora]

I apologize for saying a libre could get unreliable if it gets lose and will be more carefull what I say next time.

Not sure why you are apologizing. It is also my experience that Libre becomes unreliable when it starts to come loose and totally understandable because the filament will be disturbed.

 

[SimonP]

Just to add to the conversation about the sensors, some work, some don't work so well.

I use XDrip+, and do calibration finger pricks to ensure the calibration stays in line during the life of the sensor (it varies quite a lot from looking at the calibration graph), but it's now got to the point with my current sensor where it's unwilling to generate a sensible calibration as the sensor is under-reading so much (XDrip+ is correctly unwilling to generate a calibration line that at the off-scale low value from the sensor would place you above a hypo level, i.e. you'd never get alerts/see that you're hypo as it would flat line at something that looked fine)

This is caused by my current sensor reading much lower than my actual blood glucose, which is what I've found in the past (the reverse may also happen but I don't notice as it can be calibrated and I'm not so bothered about accuracy at off-scale high conditions! There is probably some interesting data diving to be done looking at how calibrations change, and whether sensor failure can be predicted....)

Anyway on the last two scans with the official Libre reader I'm getting the following:
Libre: 4.2; Finger prick: 8.8 (mmol/l)
Libre: 5.1; Finger prick: 8.2 (mmol/l)
<Edit> Just did a third one</edit>
Libre: 4.9; Finger prick: 8.7 (mmol/l)

So it's off by quite a large margin. I'm just waiting a little while to do a third blood test then will phone to request a replacement.

In my opinion there should be a requirement to do calibration finger prick tests and their inclusion in the official app (to make it easier to see the time offset and ensure that high rate of change regions are discounted), as otherwise you can blithely believe you're doing a good job, when in fact you're running high continuously.

The sensors do work for me as I use XDrip+ and can calibrate to make even faulty sensors last a bit longer, but otherwise I'd also be rather annoyed about performance.

 

[Val1133]

When I first started using the Libre 2, I was also finger pricking 4 times a day and found the difference confusing and annoying. I now only rely on the Libre 2 and love that it is now CGM. I use it as a rough guide to let me know what my bloods are doing generally. When they seem very high or very low, then I do a finger prick test just to check.

Just now my Libre 2 is reading 12.4 and rising slowly. It is just 1 hour after my porridge breakfast and I have been sitting at my desk, so I am happy with this. (before breakfast it was 8.1). My finger prick test is showing 10.2. I am also happy with that because it is showing a lower figure than the Libre 2. Because I am taking so long to type this, my CGM now shows 12.0 11.9 11.8 11.7 11.9 11.6 and the arrow is pointing right 🙂

I am much more relaxed with the CGM now and have accepted the differences/inaccuracies.

 

[SimonP]

As long as you do some finger prick tests so you're aware of what the difference is, then I think is approach is ok (though it would be better IMO if the gadget actually took these into account so you didn't need to do the mental arithmetic/get data reported to your medical team that is actually correct.)

 

[MrsDavey]

This is a really useful conversation! I am losing confidence in Libre. I have an iPhone and during the upgrade I lost the use of the sensor then the alerts and finally all of the data on the app.

Since then I have also had several defective devices despite being an experienced user of libre. I recently had a HbA1c test at my diabetes clinic and found it was much higher than the CGM estimate. I’m testing my glucose using the blood test strips and have found that there can be some quite big differences between the actual blood and CGM (and not for the reasons they suggest!).

This evening I put on a new libre and it was faulty. I report them each time but I feel like they must think I am trying it on because I am calling so often. Also it takes nearly an hour each time. I probably need to delete and reinstall the app to fix some of the issues however this will mean losing all the data again.

I feel like there has been a dip in quality recently and it’s affected my confidence in the devices. Would be really helpful to hear other people’s experiences.

#libre2 #appupgrade #libreerrormessages

 

[MrsDavey]

Just to add to the conversation about the sensors, some work, some don't work so well.

I use XDrip+, and do calibration finger pricks to ensure the calibration stays in line during the life of the sensor (it varies quite a lot from looking at the calibration graph), but it's now got to the point with my current sensor where it's unwilling to generate a sensible calibration as the sensor is under-reading so much (XDrip+ is correctly unwilling to generate a calibration line that at the off-scale low value from the sensor would place you above a hypo level, i.e. you'd never get alerts/see that you're hypo as it would flat line at something that looked fine)

This is caused by my current sensor reading much lower than my actual blood glucose, which is what I've found in the past (the reverse may also happen but I don't notice as it can be calibrated and I'm not so bothered about accuracy at off-scale high conditions! There is probably some interesting data diving to be done looking at how calibrations change, and whether sensor failure can be predicted....)

Anyway on the last two scans with the official Libre reader I'm getting the following:
Libre: 4.2; Finger prick: 8.8 (mmol/l)
Libre: 5.1; Finger prick: 8.2 (mmol/l)
<Edit> Just did a third one</edit>
Libre: 4.9; Finger prick: 8.7 (mmol/l)

So it's off by quite a large margin. I'm just waiting a little while to do a third blood test then will phone to request a replacement.

In my opinion there should be a requirement to do calibration finger prick tests and their inclusion in the official app (to make it easier to see the time offset and ensure that high rate of change regions are discounted), as otherwise you can blithely believe you're doing a good job, when in fact you're running high continuously.

The sensors do work for me as I use XDrip+ and can calibrate to make even faulty sensors last a bit longer, but otherwise I'd also be rather annoyed about performance.

Completely agree. My recent HbA1c was much higher than the estimate. It’s like trying to hit a moving target because I have been trying to avoid the lows.

 

[MrsDavey]

Are you aware that the "error tolerance" is only for "normal" levels and that CGMs have a tendency to be more inaccurate when high or low?
As I mentioned, these limitations are not mentioned by Abbott (or any other CGM manufacturer) but are repeated many times on the forum.
You may find some useful information if you use the Search function at the top of every page and search for "Libre". I am sure there are more limitations than those I have already mentioned here.

Completely agree, the irony is that the CGM is less accurate when levels are outside of normal range and varying which is when we rely on them the most ‍♀️

 

[MrsDavey]

Will be interesting to see whether Libre 3 sees any improvement although I assumed algos same? Anyone know when Libre 3 will supersede 2. I guess ultimately depends on cost for NHS /business decision by Abbot to eventually deprecate older one

I have been using Libre3 and actually found them less reliable than the Libre2 after it because a CGM. False low alerts, big differences between actual and CGM estimated blood sugar. Really disappointing.

 

[helli]

The estimate on Libre uses a different algorithm to calculate HbA1C.
In my experience, it is a useful additional metric to track but not one to compare with the real HbA1C.
For me, the Libre estimate is consistently 20% lower which is too far out to think of it as the same thing.
I pay more attention to TIR anyway regardless what my DSN tracks.

 

[void]

Here are some of the studies that suspicion is based on:

Assessment of factors that determine the mean absolute relative difference in flash glucose monitoring with reference to plasma glucose levels in Japanese subjects without diabetes

Accuracy and precision of flash glucose monitoring sensors inserted into the abdomen and upper thigh compared with the upper arm

A review of the foreign-body response to subcutaneously-implanted devices: the role of macrophages and cytokines in biofouling and fibrosis.

Accuracy of a CGM Sensor in Pediatric Subjects With Type 1 Diabetes. Comparison of Three Insertion Sites: Arm, Abdomen, and Gluteus

Plus my own experience FWIW.

Thanks for these links. I was particularly interested in this:

Accuracy and precision of flash glucose monitoring sensors inserted into the abdomen and upper thigh compared with the upper arm - PubMed

Nowadays, most Belgian patients with type 1 diabetes use flash glucose monitoring (FreeStyle Libre [FSL]; Abbott Diabetes Care, Alameda, California) to check their glucose values, but some patients find the sensor on the upper arm too visible. The aim of the present study was to compare the...

pubmed.ncbi.nlm.nih.gov