• Guest - w'd love to know what you think about the forum! Take the 2025 Survey »

Reliability of glucose monitors

In the industry we calibrate instrumentation - all the time There is metric and imperial and atmospheres and inches of mercury.

All sorts of things end up in all sorts of numbers.

We used to joke about calibrating instruments in pumpkins per furlong/fortnight.

The numbers mean nothing until you set your zones,

So if one meter reads different to another just get used to how much you need for that meter to bring you back into your good control zone.

The lower the reading the more accurate the meter will probably be.

MY view remains most people would find colour coding easier to use than numbers.

For example It is the top of the blue zone so I will jab 30 units or Wow I am in the red but just in the red I will jab 40.

We do it all the time when we drive the car we look at the speedo and say **** I am speeding and liftum foot we dont calculate anything we see it and wing it. We have zones set up in our head it indicates numbers but we read zones.

We do it with everything even under the shower to get the water temp right. Numbers just confuse the issue.
 
In the shower and speedo situation we have continuous feedback (heat feeling, speedo dial) to allow us to return to the acceptable range. As T1s we have no such feedback so we must measure in order to gauge the magnitude of the correction required. No choice. This is a basic rule of being T1. It also applies to any control system with feedback intervals slower than the critical time to effect a response. Showers and speedos are not that type of control system.

And by the way most speedos are numeric. If yours is just colour coded I don't want to be on the same road as you! ;-)
 
You are just being difficult @Spiker you know what I mean.

You glance at a speedo and it registers in your mind as too high

You do not think mmm it says 73 and the SL is 60 and if I take 60 away from 73 that leaves the 13 so I am driving at SL +13 now how much fine is SL +13. We bracket everything we see..

I have a motorbike with an analogue speedo and a car with an analogue speedo and a car 12 months old and I have no idea what the speedo looks like I would have to go and have a look.

The shower is too hot or too cold and you tweak the taps ' til you are happy

You would have to be pedantic to use a thermometer and calibrated taps.

Many speedos have colour coded numbers my Iveco motorhome for one so you don't chug around in top gear when you should have changed down. Holden cars here would beep as you crossed from under a preset speed to over that preset. Bracket and control is all around us.

There is even a joke about an amp that was so loud the volume went from 0 to 11 and not as per normal 1 to 10. The are not calibrated in dB(A) decibels "A" weighted just in pumpkins and everyone is happy with their volume controls.

Pedants get carried away with the detail and get bogged down and miss the main game. So it is good to keep things simple so they have less to worry about.

 
Calibration of glucose meters are hit and miss.
From this website, it states.

ISO guidelines state that blood glucose meters should provide results which fall between the upper and lower error bounds, given in the table below, 95% of the time:

  • Within ± 0.83 mmol/L of laboratory results at concentrations of under 4.2 mmol/L
  • Within ± 20% of laboratory results at concentrations of 4.2 mmol/L or more
Which is quite a reasonable figure, considering you have the basic meter accuracy, which is going to be a percentage, plus an offset, (eg +/- 5%, +/- 0.1 mmol/L for example) then on top you have the variability of the test strips, which are manufactured to another tolerance, on top of the meter tolerance. So they could be +10% in one batch, and -10% in the next batch, with the meter accuracy on top of that.
(I have made up the figures here, apart from the ISO guidelines quoted - could be any combination)
It would be reasonable to assume the meter should be fairly consistent from one reading to another, but needs to be kept at room temperature, as it's going to filled full of cheap electronics.
Test strips of better made now too, (no need to code mine anymore, batch to batch variance must be negligible, but only within the accumulated tolerance)
But you could still get the swing from the meter+the test strips in one batch, to the meter-the test strips in the next.

One thing that springs to mind as well is blood temperature.
I know it's supposed to be constant, but not when it's on your fingertips. Mine can be freezing at times, and the blood must be cooled.
We should try a few as a group, cold finger, warm up our hands, try it hotter, and see if there is any change.
Should still be in range though.

edit - just checked the spec on my meter.

Nexus meter system Operating condition: 10°C to 40°C (50°F to 104°F), below 85% R.H. (no condensing). The Meter, Test strips and Control solution should all be at room temperature (20°C-25°C / 68°F-77°F) before testing.

Not a great range to test in..
 
Last edited by a moderator:
Spiker said:
This is known I think as "the engineer's problem". What do you do if your data is unreliable? The engineer's answer, the real world answer, is you use the data anyway, and you adapt through experience. In our case you learn your meter, learn its variability, its likely range, its uncertainty. The uncertainty and variability in meters is no greater than the uncertainty and variability in the injection process, portion sizes, nutrition analysis, day to day changes in physiology. If you are a scientist you are paralysed by all these compounding errors. If you are an engineer you rely on the fact that over the long term they all average out. We need to be engineers.

Even bad data is worse than no data.

Shout out to @Engineer88
Click to expand...

Love this so much @Spiker :D @hale710 What do you think?
 
First result in, as my fingers were a bit chilly, and it's a good time to test, as I've been sitting around after breakfast.

No frozen peas, so after 1 minute holding a bag of frozen fish, 6.3
Then 1 minute in water slightly too hot to be comfortable 7.7.

I did nothing in between, apart from turn on a tap, same meter, same batch of strips I've been using for a while, and normally would expect the 6's between the two numbers, at a temperature between the two.
And I did dry my hands.
 
I just test until I get the one I want (that was a joke)

We all know consecutive tests produce different results even without temp changes etc I asked for and was sent the test solution and mine worked fine on the test solution but when they found out how old my meter was they sent me a new one free.

The meters are rough enough, it is just the numbers don't make sense with all the intrinsic variations.
 
To use an example in regard to accuracy in BG meters , their impact on bolusing and why a colour coded system wouldn't work for. type 1 , I'm using Douglas's figures above:

Let's say it's mid morning at which my insulin to carb ratio is 1:3 g . I have 2.5 u insulin on board. I decide to snack with 25g carbohydrate . I anticipate a quiet morning of pootling around not being that active.

First reading 6.3 Calculated insulin for this, for me, would be 5.5 u.
Second reading of 7.7 , the calculations for insulin are 6.5u.
A difference of 1 unit of insulin.

Therefore I would choose 6 units . And know from experience that this would keep me in my desired range of 6 to 6.5 . This would be impossible to calculate with a colour coded system meter. . If I went for the higher amount of insulin, I would probably go too low. The lower amount of insulin would make my target BG higher than I would like. Taking into account that my basal has a slight peak around 5 hours after injecting, if I was to have this snack later rather than now, my calculations would be slightly different.

In theory :D . In actuality any number of variables could slew my eventual BG level, sudden bursts of unanticipated activity, sudden onset of illness, wind changes direction, angering the D gods ....:eek:

But I do need to start those calculations somewhere, and finding a middle ground between those two readings makes all the difference.


This BTW is a hypothetical situation, and yes, I do know about insulin stacking, and choose to go against the advice because I know my insulins, and how it works for me. I wouldn't advise anyone else to do as I do.

So this is just to illustrate how I would deal with inaccuracies in readings though it's rare I would take two or more consecutive readings, that would only happen if one was wildly unexpected , and I hope it shows how just going by colour would not be of any use at all to me.the inaccuracy is manageable, if you know and have a feel for your meter. With any change in meter I do parallel tests to compare and adjust.

It's entirely possible though that a colour coding system with the capability of target ranges being inputted for the individual work work for a Type 2 who was only concerned with remaining within that target range and didn't use insulin.

Signy
( very bored hanging around for a delivery, time of delivery given as being between 7am and 7pm :rolleyes::banghead: )
 
forge said:
I just test until I get the one I want (that was a joke)

We all know consecutive tests produce different results even without temp changes etc I asked for and was sent the test solution and mine worked fine on the test solution but when they found out how old my meter was they sent me a new one free.

The meters are rough enough, it is just the numbers don't make sense with all the intrinsic variations.
Click to expand...

Not normally that much difference
That is exactly an exact 20% on the mean of 7.
-10%, and +10% on two identical consecutive readings.

It's the same batch of strips, I would expect a low variation in a single batch.
To be honest, I would expect them to be running at least 6 sigma on the whole process.
My meter is usually repeatable.
I suspect it's a combination of the temperature, and the opening of the capillaries.
That also suggests the blood at the finger tips may be different to that at my core.
Re-inforced by my experiments with running up and down stairs, which again would affect my capillary blood flow.
 
douglas99 said:
To be honest, I would expect them to be running at least 6 sigma on the whole process.
Click to expand...
No way. This is not air traffic control or a hospital system.

There is no point making the meter that much more accurate than all the input variables that all have uncertainties on the same order of magnitude as the meter.
 
douglas99 said:
No frozen peas, so after 1 minute holding a bag of frozen fish, 6.3
Then 1 minute in water slightly too hot to be comfortable 7.7.
Click to expand...

If you don't do at least 3 tests for each scenario you are wasting your time because the normal test-to-test variation in the meter swamps anything else you are trying to find.
 
Heathenlass said:
But I do need to start those calculations somewhere, and finding a middle ground between those two readings makes all the difference.
Click to expand...
Exactly. This is the "engineer's problem", and ours too. We must do our best with the data available to us.
 
Spiker said:
If you don't do at least 3 tests for each scenario you are wasting your time because the normal test-to-test variation in the meter swamps anything else you are trying to find.
Click to expand...

one data point is not acceptable for any test. generally If you wanted to do something like this you should really be waering a CGM, be having a canula with blood removed for lab testing and be testing with a meter. So really, I suggest you dont bother with such silly 'home' tests because its telling you sweet FA.
 
Spiker said:
No way. This is not air traffic control or a hospital system.

There is no point making the meter that much more accurate than all the input variables that all have uncertainties on the same order of magnitude as the meter.
Click to expand...

a 6 sigma process means that most of the test strips will be in spec.
Don't forget, these things should be made in a sterile clean room, as they have direct contact with your blood.
There's no point in throwing away any of your production, when you're making billions of these things, and they're very simple things to be repeatable on.
 
A single point calibration check only tells you if the unit is good at that point. If the meter fails that it is definitely crook
 
forge said:
A single point calibration check only tells you if the unit is good at that point. If the meter fails that it is definitely crook
Click to expand...

That's academic really,I'll bet most people haven't even done a single point check.
 
Well there ya go. I did, and it cost me nothing optium supplied it by courier after a phone call request.by me. And as I said they even took my old meter back and gave me a new one despite the old one not failing the one point test.
 
I'm T2, on 3 metformin daily, have Contour AND Contour Next meters.
Testing same drop of blood at same time:
Meter readings the same only 2% of tests.
Meter readings differ by 1 to 2 mmol/L 90% of tests
Meter readings differ by 0.2 to 0.9 mmol/L 8% of tests
More worrying, the Contour meter ALWAYS gives the lower reading.
All fasting readings between 6 & 8.5 mmol/L
Any comments or observations?
 
Back
Top