Averaging Glucose Meter Readings

[AdamJames]

For some purposes, when I want an accurate idea of blood glucose levels, I'll take 3 readings at a time in an attempt to overcome the inaccuracy of my meter (it claims to be accurate to +/- 20%).

Sometimes all three readings are the same or 0.1 mmol/l out, which is nice.

Sometimes however they are quite a broad spread, say for example 5.1, 5.5, 6.1. In this case taking an average seems like a very good idea.

However sometimes it seems more like there is a 'rogue result'. The most recent 3 readings I got were 9.4, 8.1, 8.1.

I've logged down the average as usual, but I strongly suspect that the two 8.1s mean that the 9.4 is likely to be a rogue result. Also in this case 8.1 is more the result I was expecting.

Does anyone have a best-approach to this kind of thing, either based on pure maths or similar experience with their own meter?

 

[Bluetit1802]

I do similar to yourself except I only take duplicate readings if I suspect the first is a rogue reading. If the first test is within the range I would expect for that time then I accept it.

Most meters now are accurate to within +/- 15%. The 20% range was discarded in 2016 . http://www.diabetes.co.uk/blood-glucose-meters/blood-glucose-meter-accuracy.html

 

[slip]

Does 9.4 or 8.1 make that much difference to you?

 

[Gabrielle_Tai]

Sometimes i wonder why do people even want to waste strips to get an average when meter was design as a rough outline of your current sugar level. I feel that the 15% different is acceptable even those on insulin, remember even lab result have accurate to within +/- 3% only.

 

[AdamJames]

Does 9.4 or 8.1 make that much difference to you?

For this type of test, yes. It's a 2-hour postprandial and could be very informative about whether the rise is within recommended guidelines or not (obviously how important the accuracy of the guidelines are is also open to question!).

It's also part of a test I do regularly and I want to spot patterns - a difference of getting on for 1.5 could easily obscure any patterns.

 

[AdamJames]

Sometimes i wonder why do people even want to waste strips to get an average when meter was design as a rough outline of your current sugar level. I feel that the 15% different is acceptable even those on insulin, remember even lab result have accurate to within +/- 3% only.

For many purposes I only take one reading.

For some purposes, one is useless. For example, getting to know if a new type of food is "safe" for you.

It's very easy for my meter to read, say a 6 before eating, then a 7, 2 hours later. But "truth" values could be say 4.8 and 8.4. Truth vs the readings could make all the difference between me believing that a certain meal is safe, and not.

I get around this by either taking 3 readings before and after food, or by re-testing the same food more than once.

My latest 3 readings are for another purpose altogether, where I want to spot trends in changes to glucose tolerance depending on what I've been eating / doing the day before.

 

[slip]

For this type of test, yes. It's a 2-hour postprandial and could be very informative about whether the rise is within recommended guidelines or not

So you'd do a triple test pre-prandial as well? I know some would say this is a bit excessive (as well as expensive! lol)

....and you've just answered....!

 

[AdamJames]

So you'd do a triple test pre-prandial as well? I know some would say this is a bit excessive (as well as expensive! lol)

....and you've just answered....!

I only do this triple-before-and-after at most once a week these days, sometimes only once every two weeks. It can cost as little as £1 depending on how cheaply I've managed to bag the strips for!

 

[LittleGreyCat]

For some purposes, when I want an accurate idea of blood glucose levels, I'll take 3 readings at a time in an attempt to overcome the inaccuracy of my meter (it claims to be accurate to +/- 20%).

Sometimes all three readings are the same or 0.1 mmol/l out, which is nice.

Sometimes however they are quite a broad spread, say for example 5.1, 5.5, 6.1. In this case taking an average seems like a very good idea.

However sometimes it seems more like there is a 'rogue result'. The most recent 3 readings I got were 9.4, 8.1, 8.1.

I've logged down the average as usual, but I strongly suspect that the two 8.1s mean that the 9.4 is likely to be a rogue result. Also in this case 8.1 is more the result I was expecting.

Does anyone have a best-approach to this kind of thing, either based on pure maths or similar experience with their own meter?

The reading of 9.1 seems reasonable considering the accuracy of the meter.

The examples you give are 3 the same, 3 all different, and 2 the same and 1 different. In that context 2 the same and 1 different fits in with the general pattern and I would average them as usual. It is reasonable to wonder if you got 1 accurate (or high) reading and 2 low ones.

Trying to measure post prandial rises is a lottery anyway. Your examples show a spread of 1.0 and 1.3 just from the same set of readings so trying to show that the 2 hour rise is at or less than 2 units can be very imprecise. I agree that taking 3 and averaging is a good approach, but I think that you should stick to that for consistency even where you get two results you like and one you don't.

 

[AdamJames]

I do similar to yourself except I only take duplicate readings if I suspect the first is a rogue reading. If the first test is within the range I would expect for that time then I accept it.

Most meters now are accurate to within +/- 15%. The 20% range was discarded in 2016 . http://www.diabetes.co.uk/blood-glucose-meters/blood-glucose-meter-accuracy.html

Yes I'm very tempted, if I get the first two readings to be the same, to leave it at that. I carry on anyway just so the process I use to get the results is always the same; it makes me feel that I'm comparing the same thing from one month to the next but I'm not so sure.

I'm also in two minds about whether to upgrade my old meter. On the one hand I can get cheap(ish) strips for it, on the other hand it's less accurate. I think I've got trust issues anyway, and even at +/- 15% I'd be wanting to take 3 readings.

My understanding of maths stops and "mean, median and mode". I think intuitively most people would see 8.1, 8.1 and 9.4 and think that something is awry with the 9.4, but I don't know if there's a best-practice for dealing with that? I might go back over all numbers for this type of test and see if discarding apparent-rogues results in something interesting showing up in trends.

 

[AdamJames]

The reading of 9.1 seems reasonable considering the accuracy of the meter.

The examples you give are 3 the same, 3 all different, and 2 the same and 1 different. In that context 2 the same and 1 different fits in with the general pattern and I would average them as usual. It is reasonable to wonder if you got 1 accurate (or high) reading and 2 low ones.

Trying to measure post prandial rises is a lottery anyway. Your examples show a spread of 1.0 and 1.3 just from the same set of readings so trying to show that the 2 hour rise is at or less than 2 units can be very imprecise. I agree that taking 3 and averaging is a good approach, but I think that you should stick to that for consistency even where you get two results you like and one you don't.

I'd go along with all of that. Just wondering if there was special maths for dealing with it, I guess there isn't.

I've also discovered that the postprandial rise is an enigma for other reasons - it can vary greatly depending on what I've been doing the day before (I think). Whatever the reason, all I'm really sure of is that even if I measure it accurately for exactly the same precisely-measured food, in the mornings, after the same amount of sleep, and with the same starting bg, sometimes the result I get suggests it's a miracle-food for safety, and other times it's a dangerous one.

But rather than this last observation making me give up with testing meals in frustration, I think it might help me work out what I can do to improve glucose tolerance. So far the combination of exercise, lowish calories and low fat the day before seem to reliably get the better results. There could be so many factors however that it's a minefield. One thing is for sure though, if there are factors that can influence my glucose tolerance, I'm going to need to keep an accurate log if I'm ever going to get to the bottom of it.

 

[Rachox]

Statistics?! You’ve got me!
I only do a repeat test if I get an unexpected result, either too high or too low.
I guess to remove doubt about testing a certain meal, you need to do it several times to get a good idea of your readings?

 

[bangkokdiabetic]

I only take 1 reading at a time I do have an older meter that still works which I use if I want to check accuracy this meter is much larger and since in This climate I rarely wear a jacket Nano one easier to carry on my belt together with pricker and tub of strips
As for averages I do not worry over the odd high I pay more attn. to my 7 day average my meter shows 7 14 30 90 day average after your hospital blood test is a 2-3months average I do not ignore the high but adjust lantus or diet as required

 

[slip]

There are all manner of things that affect your BG levels (not just the obvious) - start by looking at the alignment of the planets, wind direction and next weeks lotto numbers

 

[AdamJames]

Statistics?! You’ve got me!

So my cunning plan worked! I only started this thread in the hope that I could get you and Bluetit1802 talking about numbers again

Yes I think repeat testing is the best approach to checking for meal safety - more useful possibly than only testing the meal once and taking multiple readings either side, largely because it seems my body can process the same meal very differently under different circumstances. It's nice to have a best-and-worst case rough idea for any meal I think.

 

[AdamJames]

There are all manner of things that affect your BG levels (not just the obvious) - start by looking at the alignment of the planets, wind direction and next weeks lotto numbers

Haha! I need to add some more columns on my spreadsheet!

 

[ringi]

If you take 3 readings, just record the middle reading, unless one reading is very different to the other two, then discard the outliner. Best to take the 3 reading on different fingers, with at least one on the other hand.

However, I would rather test the same meal 3 or more times.

 

[AdamJames]

my meter shows 7 14 30 90

I was about to suggest you call an ambulance, then I finished reading the whole sentence!

 

[AdamJames]

If you take 3 readings, just record the middle reading, unless one reading is very different to the other two, then discard the outliner. Best to take the 3 reading on different fingers, with at least one on the other hand.

However, I would rather test the same meal 3 or more times.

Ha! I have actually done the make-sure-at-least-one-reading-is-on-the-other-hand thing a few times. I don't always however.

So you're suggesting using the "median" value, rather than the "mean" or "mode"? That does seem to get close to what one might intuitively think may be the "truth" for all cases:

6.1, 6.1, 6.1 (median = 6.1)

and

8.1, 8.1, 9.4 (median = 8.1)

and

5.1, 5.5, 6.1 (median = 5.5)

 

[slip]

It would be interesting to see if your median is more often than not the first test you do, you may find that statistically you don't need to do 3 tests..........

But I would have though the mean would be more 'accurate' as you are using 3 numbers which could be up to 15-20% off the actual result so averaging by mean should reduce the discrepancy?