Libre accuracy and predictive ability

[SimonP78]

Apologies for slightly confusing matters by saying I don't use libre, it's the official app I don't use.

Sorry I've no idea what you mean by time offset and working in the past?
Importantly, no CGM will attempt to tell you what your BG is as it has no access to blood. Medical device regulators would not allow a device that produces actionable information that does not reflect the source of the data. All CGMs report only on ISF.

I didn't explain myself very well, the value of interest to most users is actual blood glucose rather than ISF glucose level, so when one comes from doing finger prick testing and knowing what the value is now (well sort of as I guess there's also a slight delay in peripheral vs central glucose level responses), the display is telling you your "blood glucose" is such and such, but that's actually what it was 5 min (or however long) ago. I realise it's not actually blood glucose being measured, but the algorithms are giving a value which isn't direct ISF glucose concentration but rather an estimate of what actual blood glucose would be to generate a given ISF glucose concentration. Or is there no concentration difference (aside from the time-dependence) so there is no ISF to blood step required?

The confusion I think, is that some people think the libre extrapolates forward to give an estimate of current blood glucose (derived from current ISF, which lags actual blood glucose.) This is what I was asking about, but I agree that the regulators probably wouldn't be too happy with a device that is providing an extrapolated estimate rather than something based on the current ISF value. Nevertheless such an extrapolation method if it could be made to work would be an interesting thing.

That's a good question - how long the lag? I've seen 2-45 mins depending on where you look and it can depend on where the CGM sensor is placed but mostly on the individual. The lag is due to movement of glucose from blood to ISF so if you have circulatory or cardio-vascular issues, ion imbalances, an infection or other illnesse. the lag may change.

That sounds like you think it's relatively constant for a given person-sensor position combination - I was curious as to whether it might change with temperature, hydration and heart rate (so basically things that change peripheral blood flow and blood turn-over in general.)

What is always the case is that lag time is minimal when glucose levels are relatively consistent. It's when they change over a short period of time that there is a noticable difference between BG and ISF levels.

Is it that the lag is minimal or the effect of any lag is lower with a smaller delta? I'm guessing it's the latter unless there's a concentration gradient effect at play?

 

[Jasmin2000]

Apologies for slightly confusing matters by saying I don't use libre, it's the official app I don't use.

I didn't explain myself very well, the value of interest to most users is actual blood glucose rather than ISF glucose level, so when one comes from doing finger prick testing and knowing what the value is now (well sort of as I guess there's also a slight delay in peripheral vs central glucose level responses), the display is telling you your "blood glucose" is such and such, but that's actually what it was 5 min (or however long) ago. I realise it's not actually blood glucose being measured, but the algorithms are giving a value which isn't direct ISF glucose concentration but rather an estimate of what actual blood glucose would be to generate a given ISF glucose concentration. Or is there no concentration difference (aside from the time-dependence) so there is no ISF to blood step required?

No, the algorithm does not give an estimate of what BG would be at all - it gives a reading of what ISG actually is. The device is not able to estimate BG as it has no access to the blood and no basis for any estimation.

Apart from the lag, there is a minimal difference between the IFG and BG in a stable state, but when rapid changes are occurring the difference between ISG and BG becomes apparent.

The confusion I think, is that some people think the libre extrapolates forward to give an estimate of current blood glucose (derived from current ISF, which lags actual blood glucose.) This is what I was asking about, but I agree that the regulators probably wouldn't be too happy with a device that is providing an extrapolated estimate rather than something based on the current ISF value. Nevertheless such an extrapolation method if it could be made to work would be an interesting thing.

CGMs do not extrapolate to estimate BG, they have no access to BG and importantly, they measure ISG glucose levels following BG levels. To give a forward estimate they would need to predict the future!

BG and ISF measurements are both real time so if I get a two sequential BG readings that are going down rapidly, I will get some carbs. But the IFG is registering this drop 5-20 mins later and I lose that reaction time. BG tells you the car is crashing into a wall and you can slow down, but ISF tells you that the car has already hit the wall. This is why I use ISF to review the week but BG to get actionable information.

That sounds like you think it's relatively constant for a given person-sensor position combination - I was curious as to whether it might change with temperature, hydration and heart rate (so basically things that change peripheral blood flow and blood turn-over in general.)
Is it that the lag is minimal or the effect of any lag is lower with a smaller delta? I'm guessing it's the latter unless there's a concentration gradient effect at play?

Not sure if the lag is affeceted by technical issues with the sensor - probably if it means the glucose reaction on the sensor is slow.

Is the lag affected by the delta? Interesting question and it depends on the speed of glucose diffusion from blood plasma to the ISF, as well as how quickly glucose is taken up by tissue cells. We need a bunch of people to take different doses of pure glucose and see how long it takes to peak in the blood and on the sensor - any volunteer?

 

[SimonP78]

No, the algorithm does not give an estimate of what BG would be at all - it gives a reading of what ISG actually is. The device is not able to estimate BG as it has no access to the blood and no basis for any estimation.

CGMs do not extrapolate to estimate BG, they have no access to BG and importantly, they measure ISG glucose levels following BG levels. To give a forward estimate they would need to predict the future!

I'm sorry to labour the point, but I don't really see why not - if the concentrations reach equilibrium between the blood and ISF, then estimating a past BG from current ISG is surely possible (i.e. looking backwards to work out what BG would have resulted in the current ISG)?

Assuming that is possible, the issues that remain are around how to predict the future from the point of view of ISG (which may be as simple and inaccurate as extrapolating forward to remove the lag, or something more complex like taking into account carbs and insulin on board and their effects); plus what the prediction window needs to be (i.e. what is the lag between BG and ISG and therefore how far forward does the extrapolation need to go)?

I'm not saying it's easy nor that it will necessarily work every time (e.g. it certainly rained more today than expected, but overall it wasn't too far off), but I don't see any reason why this is not possible, but perhaps I've missed something?

BG and ISF measurements are both real time so if I get a two sequential BG readings that are going down rapidly, I will get some carbs. But the IFG is registering this drop 5-20 mins later and I lose that reaction time. BG tells you the car is crashing into a wall and you can slow down, but ISF tells you that the car has already hit the wall. This is why I use ISF to review the week but BG to get actionable information.

I suppose this depends on rate of change of BG (or ISG) and typical lag time (perhaps mine is small, which is why I find it works reasonably well) - I find using the value and trends from the CGM works unless I've got lots of insulin on board or am exercising.

Re lag time variability, I assume there must be some research somewhere (quite possibly linked from the DynamicISF page mentioned above by @CheeseSeaker ), I shall go and Google!

 

[Jasmin2000]

I'm sorry to labour the point, but I don't really see why not - if the concentrations reach equilibrium between the blood and ISF, then estimating a past BG from current ISG is surely possible (i.e. looking backwards to work out what BG would have resulted in the current ISG)?

You're saying that If BG and ISG are in equilibrium then ISG can effectively be called BG as if it were a Glucometer?
So when I look at my Libre I can say I have a BG of 97 although I'm really looking at the ISG which came from the blood 5-20 minutes ago. That only works if I have a BG of 97 forever because any BG change will no longer be in equilibrium and will only be shown on LIbre after 5-20 mins.

There are also other reasons this doesn't work,
-- sensors / software are approved by the MHRA only to measure ISG and cannot report BG. Abbott's instructions call the output 'senor glucose' and describe how it measures ISG with good reason - it's the law.
-- if a sensor tried to become a predictor of BG, this would assume that BG can be reproducibly determined from a subsequent ISG with high accuracy - and you only need to do exercise to see that BG and ISG diverge quite quickly (discussed above) and irreproducibly because we are all different.
-- any device that claims to predict results would need to undergo strenuous clinical trials to determine accuracy and reproducibility and for a sensor it isn't worth the extra millions and the risk of life altering errors e.g. hypos, increases.

Assuming that is possible, the issues that remain are around how to predict the future from the point of view of ISG (which may be as simple and inaccurate as extrapolating forward to remove the lag, or something more complex like taking into account carbs and insulin on board and their effects); plus what the prediction window needs to be (i.e. what is the lag between BG and ISG and therefore how far forward does the extrapolation need to go)?
I'm not saying it's easy nor that it will necessarily work every time (e.g. it certainly rained more today than expected, but overall it wasn't too far off), but I don't see any reason why this is not possible, but perhaps I've missed something?

Answered above.

I suppose this depends on rate of change of BG (or ISG) and typical lag time (perhaps mine is small, which is why I find it works reasonably well) - I find using the value and trends from the CGM works unless I've got lots of insulin on board or am exercising.

Exactly - and this is where the CGM fails for me so I use BG becasue it gives me info 10-20 mins earlier.

Re lag time variability, I assume there must be some research somewhere (quite possibly linked from the DynamicISF page mentioned above by @CheeseSeaker ), I shall go and Google!

There are several factors involved which may be on different timescales. According to this excellent article, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2903977/ the lag during rapid changes of blood glucose is likely due to the magnitude of concentration differences in various tissues at a time of rapid change, so we know that rapid BG changes do affect the lag.

 

[EllieM]

Just to say that the moderator team have moved the posts preceding this one from the thread
https://www.diabetes.co.uk/forum/threads/has-anyone-quit-using-freestyle-libre.202578 so that posters can have a detailed discussion of libre accuracy without derailing that thread.