Calibration, plugins, problems Libre 3/3+/2+

[SeIex]

Hi,

not fluent writing English sorry upfront...

Since there is right now (#4098 #4093 #4092 and also older ones) a lot of calibration discussion, I thought to open this.

For me personally a calibration feature is essential to use a CGM, as I learned in the past years with Libre 3 (just starting with second Libre 3+) and a single G7 sensor.

In my opinion the "feature" "no calibration needed" (available as option by the manufacturer or not) is simple a marketing strategy to get insurance to prefer paying a CGM, as it is cheaper than paying a lot of expensive test strips.
In my case it was really hard to get insurance to pay for the G7, to which I change in a few weeks, instead of Libre, surely because of a price difference in their contracts.
I'm sure there are many cases where people are stuck with what isn't working for them, caused in financial reasons.

About accuracy:
A really good glucometer should mostly be, without user error, in a margin of 5-10 mg/dl or 5-10 % below/above 100 mg/dl and even mediocre ones should mostly stay inside a 15 units/percent.
I personally do often 2 blood measurements and if they are not really close, do a third or fourth one and calibrate the average, sometimes exclude an obvious false one.
Also I use two glucometers, check them against each other sometimes and also in the lab at my doc.
I tried a few glucometers and can tell that there are differences in accuracy AND repeatability for me.

Getting to CGM and accuracy, they are more like "best I can do is 30", surely not every sensor on every person is at any time this bad, but it is what studies and manufacturers tell - up to 30% is OKish and no error.

The important point is, that a CGM that measures interstitial fluid, is just guessing the really relevant value of blood glucose - it's almost like measuring the inside temperature of a fridge on its outside, could work with many parameters taken into account, but will not be precise for each and every fridge model.

And that is the point - everyone’s body acts a little different and those differences can't be covered by manufacturers calibration.

In my case with Libre 3 there is a margin of up to sure +/- 30%/units in one sensor over 14 days, usually starting too high and ending too low, overall there are better and worse ones.
A Sensor reporting 100 mg/dl at really 70 mg/dl blood glucose is maybe still better than nothing, but also not really useful and safe at all. I have had Libre sensors measuring 60 mg/dl (or even more?) too high or low, still working(!) and reacting on glucose changes, but delivering way too high/low values.
My G7 test sensor started also too high and was, after some calibrations, pretty accurate until day 8 or 9 and needed small calibrations again. I hope this will be the case for future sensors also, but reading about current quality issues, I'm hoping the best.
I'm pretty sure, with the upcoming G7 15 days version, this will be more critical also. I read of a 26% fail rate, not working 15 days, the study tells.

I also read of a study concluding, the Libre 3 MARD is even better than that of G7. Holy ...

In my case, calibration in xDrip is really important, maybe not for everyone, but at least not a forgettable minority at all. This is just not of interest for manufacturers financial success overall.
Speaking of minority - xDrip users are for sure a minority, but one that might be more interested and got more knowledge than average.
Without calibration, I just can't use the Libre 3, this will probably also be the case for the G7.
Also I'm pretty sure there is a (maybe not even small) number of cases where the CGM users just don't care about or feel/check wrongness of readings.
In my case, this would often have resulted in my BG being way too low for the first week and way too high for the last few days of a sensor.

Sadly right now there is no easy way to calibrate a Libre 3/3+/2+ (maybe other sensors too?) with xDrip, only tinkering with calibration plugins and/or overrides is possible.

As of calibration itself - I've had sensors that I think would have even needed a more point calibration - visible in glucose lines that change steepness on a specific level. But that would for sure get pretty complex.

It would be nice if I could stimulate a discussion with more people that maybe got similar experiences and really need calibration.

I fully understand the cautiousness on that topic for safety reasons. As I think nobody really knows exactly what Abbott is doing. And to ensure a low warning being triggered.
My humble observation is that the minute readings look pretty raw (calibrated by the manufacturer) on the live display of the Libre APP and also in the LibreView.csv/LibreLinkUp follower data, but are smoothed out in the stored 5 minute readings and I think also smoothed out about every 15 minutes, if there is a large deviation, complicated thinking, it looks like maybe it could be a corrected prediction. Or just correction of obvious incorrect raw data...
What I'm really unsure about, is what happens if the sensor pauses with error for a few minutes after unplausible rapid changes in readings. I personally can sometimes produce that by massaging the sensor. Sometimes, way too low readings stabilise again after that happened, I really can't say if this effect is sensor logic or body-tissue related.
Speaking of low alarm trigger - for me it sometimes looks like, although a sensor is reacting on BG changes at higher BG levels, that are way too low displayed, it can then at low values come to an almost flat line at about 50 mg/dl. I can't say what happens when really going hypo and don't want to test this intentionally. I think this is the security problem others see as well.
I personally have set an alarm high enough to not be in range of that risk within boundaries of maximum calibration values. But yes, this has to be considered and it maybe needs a better solution, how ever this could look like.

Would be nice to hear other user experiences or ideas on that topic.

Changing to G7 in a few 3+ sensors, as said above, I hope this will work better for me. (calibrating it with xDrip works fine, as tested allready, just some confusing log messages)

Greetings and thanks to all contributors!

[PhilTU]

A Libre 2/2+ sensor connected direct to xDrip+ can be calibrated based on either raw data or "glucose" values from the cut-down algorithm in OOP2. This does at least provide a y=mx+c approach with constraints on c for hypo safety. Plug-ins aren't required for this ?

Libre 3 is more complex as xDrip+ does not connect to the sensor, so there is no internal timing data and the preceding app or service may already have manipulated the data or delayed it. I note that juggluco does not provide calibration and the opinion of its author/developer is that in his experience it was not possible to improve on the outcome by calibrating in xDrip results sent from juggluco. If xDrip isn't connected direct to the sensor I think there's an argument that says it should just faithfully repeat what it is given.

BG meters usually test out with a MARD around 5% rather than CGMs at double that, obviously this is the mean and conceals a wider range of variability. I do understand why this greater accuracy, the timeliness and appropriate fluid sample use and the lived experience of the user and medical profession means that greater emphasis is put on the BG reading being "correct" although in time we may learn to use CGM data better and achieve better outcomes based on ISF measurement. Calibration will be even more difficult then as ISF sampling and testing isn't available to the user.

Timing remains a problem in attempting calibration, unless BG and ISF are at a steady state there's some need to "match" the BG value time to a corresponding CGM sample time, xDrip+ has something built in because the BG value for calibration seems to be entered sometime into the future - 10 minutes I think.

[SeIex]

Hi,

writing Libre 2+ in title was probably my fault, I remembered #3260 #2878 wrong, so this is depending on the region I think. I started CGM with a Libre 3 when it was pretty new, G7 was not available yet. So I never had a Libre 2.

My case and story is not ordinary in any kind, to much to write, pump currently not possible.
I personally need an alarm at 100 mg/dl at least, as I am allready to low at 80.
And this is not happening with a Libre without calibration for me, also I would have to ignore higher readings, because they are often way to high displayed the first 7+ days of a sensor.

After starting to calibrate daily, I was able to lower my A1C in 3 months by 0.5 and still avoid lows. Just because I could rely on readings. Calibrating once or twice and then going with it isn't working.

I know the opinion of the Juggluco developer on calibration, maybe his body just fits a Libre better than mine.

I compared roughly 700+ finger pricks of just a few months with Libre minute and 5 minute readings in an automated spreadsheet to prove my conclusion.

The point of timing and what is delivered looks, like I wrote, speaking of the minute values, pretty raw and jumpy at times. These also look pretty identical on Libre APP live readings and also Libreview export/LibrelinkUp follower. I don't think Juggluco is changing the data.
But who knows...

xDrip+ just displays the calibration 10 minutes later in the graph for the user to check the outcome (dot on line) I think. Used for calibration is the minute it is entered with the reading of that minute.
So if readings are jumpy, a calibration is pretty useless.
At this point it might be even less bad if a little smoothed 5 minute reading would be used.
Calibration at quick rising/falling readings is useless no matter how.

I tested some time and use minute readings in xDrip. I'm not sure if I've tried which reading is used for calibration when the minute readings aren't active.

[SeIex]

@PhilTU

Just saw that j-kaltes actually added a calibration to Juggluco.
At a first glance, there seems to be no calibration graph. That would make it very difficult to quickly identify plausibility, at least for me.

Reading again his old opinion:

For some time I have sent the sensor glucose values from Juggluco to xDrip and have entered all my finger pricks as calibration data in xDrip. As a results the readings of xDrip were much worse than the not manually calibrated values of Juggluco.

He won't have used the fixed slope Patched Libre source without a plugin, right?!

So maybe I'll have a look at it, not sure if I will change much for the last few Libre sensors, still hoping G7 will work for me and I don't have to change back again.
And Juggluco's forced landscape UI just isn't for me.

[PhilTU]

https://www.juggluco.nl/Jugglucohelp/calibrationhelp.html

I don't know what he did, perhaps the fixed offset slope=1 version. I see he has that as an option "change slope" above.

Juggluco is retiring the Libre 2 patched input to xDrip+

[SeIex]

Read that, so no more minute readings and even less calibration options in xDrip I guess.
I think he doesn't like the forced smoothing/delay in xDrip on that source, where he is right, that should really be switchable to under 10 minutes or off.
And also the not working backfill probably.

It could be possible to still get data out of Juggluco and into Patched Libre source with GlucoDataHandler, didn't try it yet.

And as of the 10 minute delayed calibration point in xDrip, tried that - if the finger prick is added as calibration 10+ minutes later, a later minute reading is used. So it depends on when it is set for calibration. Adding it much later isn't working.
And as I wrote, maybe the use of minute readings has an influence.