Does Libre 2 plus "correct" itself?

[Rosie9876]

For no discernable reason, my CGM quickly started climbing from about 6.0 -> To 9.4 with arrow pointing up, indicating higher to come. It then stayed there for a minute before descending bit by bit, now down to 6.1 -> , with most of the higher readings disappeared. I've seen corrections from low glucose errors before, but never this. Explanation?

Adding update: it's now deleted the lower graph and is at 7.1 from a circa 7.5 high. Maybe this is normal but I've never seen it before, but it seems weird - more sophisticated than I'd assumed.

 

[Jaylee]

Hi,

I’ve only just started using the L2+. (Little over 24 hours.)
Seems within 0.3Mmol of my meter. So far, pretty pleased..
I have had brief wild swings in the past with the libre 2 & even the the first incarnation, when used with a Bluetooth bridge making the L1 a CGM. Before settling back down?

Low glucose errors normally bounce back after the sensor has been compressed by lying on it..
(In my experience.) There is more of a lag with a real hypo.

 

[sgm14]

The software calculates your average blood sugar in 15 minute blocks and the graph appears to be drawn based on those averages and then adds a line to your current reading. So if your BS jumped to 9.4 but fell back quickly, your average for that period could be lower and probably closer to 6 and the peak would disappear from the graph once 15 minutes have past.

If you had of added a note at the time, this would show up on the LogBook as a coloured dot at the correct level but you would also see the graph drawn in black which in your case would be lower down and the dot would stand out on its own.

This is an example showing a bit of my LogBook which illustrates the issue. It was the middle of the night and I think the 'lows' were compression lows and my levels jumped back up very quickly after I stopped lying on the thing, so the averages did not really drop much below 6 despite the two red dots. (The fact that they are coloured red means they were below 3.9)



This can occur at any time your BS changes quickly, so is more noticeable at the two extremes.

 

[Westley]

I believe what you're seeing is the result of LibreLink's extrapolation algorithm,
and it is one of my biggest irritations with the software.

The sensor reads the glucose levels from the interstitial fluid under the surface of your skin, and changes in this lag behind the actual glucose value in your blood by around 15 minutes. So the value it senses is roughly what your blood glucose was 15 minutes ago. I expect the actual lag time varies quite a bit depending on hydration, movement, temperature etc.

However, this value is not what the reader shows you. Instead it attempts to tell you what your blood glucose is now by just making it up. To do this it uses the current and previous values to calculate the trend, and extrapolates that forward for the part it does not yet have real data for.
There are many different mathematical techniques for extrapolation, the simplest being linear (so if the current known value is 5.0 and it's changing by +1 every 5 minutes, it guesses that in 15 minutes it will be at 8.0). It could also be using something higher order, where it fits a curve to previous values instead of just a line, they don't share the algorithms they use.

What I hate about this is that it doesn't give you any way of seeing which part of the value is really from the sensor, and how much it has adjusted by extrapolating. When it's changing rapidly and you've just taken action to correct it, the guessed value can be quite wrong, because the software doesn't know that e.g. you've just eaten some glucose to correct a dropping low, so it keeps showing it as dropping for another 15 minutes, and if you took that at face value you'd overcorrect.
What would be more useful (and indeed what some of the other non-official apps do) would be if it showed the graph with the sensed values, and the extrapolated part as a dotted or differently coloured curve. Then you could make up your own mind, taking into account the information your sensor doesn't know about.

The whole way the software is written seems to treat the user as if they're a bit stupid and couldn't handle more information - See also the way it just stops giving readings at all when it is uncertain because they're changing rapidly, instead of giving the reading along with a warning.

 

[Rosie9876]

I believe what you're seeing is the result of LibreLink's extrapolation algorithm,
and it is one of my biggest irritations with the software.

The sensor reads the glucose levels from the interstitial fluid under the surface of your skin, and changes in this lag behind the actual glucose value in your blood by around 15 minutes. So the value it senses is roughly what your blood glucose was 15 minutes ago. I expect the actual lag time varies quite a bit depending on hydration, movement, temperature etc.

However, this value is not what the reader shows you. Instead it attempts to tell you what your blood glucose is now by just making it up. To do this it uses the current and previous values to calculate the trend, and extrapolates that forward for the part it does not yet have real data for.
There are many different mathematical techniques for extrapolation, the simplest being linear (so if the current known value is 5.0 and it's changing by +1 every 5 minutes, it guesses that in 15 minutes it will be at 8.0). It could also be using something higher order, where it fits a curve to previous values instead of just a line, they don't share the algorithms they use.

What I hate about this is that it doesn't give you any way of seeing which part of the value is really from the sensor, and how much it has adjusted by extrapolating. When it's changing rapidly and you've just taken action to correct it, the guessed value can be quite wrong, because the software doesn't know that e.g. you've just eaten some glucose to correct a dropping low, so it keeps showing it as dropping for another 15 minutes, and if you took that at face value you'd overcorrect.
What would be more useful (and indeed what some of the other non-official apps do) would be if it showed the graph with the sensed values, and the extrapolated part as a dotted or differently coloured curve. Then you could make up your own mind, taking into account the information your sensor doesn't know about.

The whole way the software is written seems to treat the user as if they're a bit stupid and couldn't handle more information - See also the way it just stops giving readings at all when it is uncertain because they're changing rapidly, instead of giving the reading along with a warning.

Thanks. This sounds plausible, and I agree with you, that more information would be useful.

 

[Jaylee]

From what I’ve gleaned over the years.
Sensors represent via the interstitial fluid under the skin what being supplied by the blood stream to it.
I’ve also read somewhere the body’s hydration (or dehydration.) can (or could.) affect these sensor readings??

Even finger pricking can throw up anomalies with the figures from samples taken from different parts of the body.
I’ve seen a lot of topics in the past on separate finger prick site differences taken in mere seconds..

My personal criteria for these sensors is, is it accurate over all to inform me of my day to day management?
& can it reliably give me advanced warning of an impending low..?