More for the number nurds

Status
Not open for further replies.
Docb

Docb

Moderator
Relationship to Diabetes
Type 2
In my quest to help people get a perspective on test results and what you can get from them, I have done some more systematic tests.

They follow on from the tests I did some time ago when I jabbed all eight fingers and two thumbs in quick succession to see by how much they varied and reported on here. There I came to the conclusion that the number after the decimal point was a bit meaningless and that you could easily get a variation of +/- 1 around the average reading. I suggested that was as much to do with the variability in blood glucose around the body as "inaccuracy" in the meter.

This time I decided to see how many tests I could get from a single drop of blood. Might go some way to differentiating between sample variability and reader variability. So, bodged the finger to get a decent drop of blood, got a reading, replaced the test strip and tried again. On two occasions I got two readings and on the third occasion I got three. The three attempts were at different times - bedtime last night, one hour and a half after breakfast this morning and then again at 10:00 AM. I was aiming to get my blood glucose at different levels to see if that effected things.

The results were... Test 1: 6.2, 6.8
Test 2: 9.7, 10.1
Test 3: 7.6, 7.3, 7.2.

A quick eyeballing of the numbers suggests you cannot expect a reproducibility much better than half a unit and it is more sensible to assume a reproducibility of 1 unit.

Thought some of the number nurds might be interested. For everybody else, be thankful for the fact that there is kit around which can get some sort of instantaneous reading of blood glucose but don't get too wound up by changes of a unit or two. You can get that by repeating the test on the same drop of blood. The kit is very good, but not that good.
 
People's expectation that a instrument costing a few pounds will be 100% accurate or precise is misplaced, even instruments costing thousands will only have a certain degree of accuracy and the result will be an average of repeats from the same sample. Many tests would be done in triplicate and always include a negative and positive control sample.
Your experiment highlights the wisdom of not getting hung up on small differences.

We used to get very frustrated that an instrument which cost £8000 which measured the concentration of DNA extracted from either a blood or saliva sample was so variable and rarely gave the same result from 2 consecutive samples.
 
Funnily enough I was just thinking about this last night! I couldn’t believe that BG was at the same concentration throughout the body and was wondering about testing blood samples from different areas to prove this - another experiment for you @Docb ?

You are right, we shouldn’t get too hung up and worry about the difference between say 5.3 and 5.8. I suppose that is why TIR is looking like a more useful guide to overall benefits as that would translate into a lower HbA1c and reduce the risk of complications.
 
Been wondering about that myself @Eternal422. For example, what's the difference between fingers and toes.

TIR is great but only available to those with continuous glucose monitoring although that comes with extra complications. My thought is that if you have decent control, then when comparing finger bodged levels it is best to round them to the nearest whole numbers and ignore differences of 1 or 2, but maybe raise your eyebrows at 3 or more and seek an explanation.

If you look at my results above they translate to

Test 1: 6, 7
Test 2: 10, 10
Test 3: 8, 7, 7.

Tests 1 and 3 are effectively the same and test 2 different, but that's OK, my system was still dealing with breakfast when I did that test.
 
Looks like around +/- 5% to me.
 
There you go @Eternal422 ... little finger 7.3, little toe 7.2, so they are the same.

Not fond of percentages myself Eddy, especially if there is a possibility of a difference not being a function of the base value.
 
Docb said:
There you go @Eternal422 ... little finger 7.3, little toe 7.2, so they are the same.
Click to expand...
Interesting! So I presume blood circulates throughout the body quickly enough to spread the glucose evenly, which makes sense as the idea is to provide fuel for all cells.
 
I do love your experiments @Docb :D
 
Thank you @gll.

They are a bit of fun really, nowhere near rigorous enough to count as proper science, but if the ideas help others to understand what the numbers mean and prevent a bit of anxiety here and there, then they will have served some purpose.

Usual caveat required.... aimed at T2's, T1 is a whole different ball game.
 
Oddly, even though my BG results have been consistently very good, I do become anxious if any reading goes above 5.5. Not because it's 'bad' but because I am terrified that my pancreas may have decided to pack up and the fear of meds, injections, changing to type 1.

So far, so good though.

Interesting tests @Docb

I wonder what the value of calibration testing is too
 
Good point about calibration testing @Gwynn.

We get a lot of questions and comments on the forum about the "accuracy" of the various meters and systems. They are conversations I tend to keep out of because in order to determine accuracy you need to test against a common standard. Comparing the "accuracy" of libre vs finger prick on the basis of the results they report is technically unsound because rarely, if ever, are the tests done on a calibrated standard.
 
This is fun: https://www.bmj.com/content/368/bmj.m149

Your results may vary: the imprecision of medical measurements​


A little interactive tool for exploring the variability of some common lab tests.

Eg: Fasting BG. My last two reads have been 4.7 and 4.4 mmol/l. With the authors' default assumptions, these are effectively indistinguishable. Further, the latest 4.7 mmol/l read sits in the middle of a 95% CI spanning 4.0 mmol/l - 5.4 mmol/l.

So it's not just measurements with cheap personal devices that have pretty wide variability - real lab measures have this also.
 
Last edited:
Good find @Eddy Edson.

I would add somewhere my old bête noire of sampling error although it is encompassed somewhat in what they call biological error. Spent some time in and around QA and QC systems and methods where destructive tests would be done on a sample from a product line (you obviously cannot destructively test everything) and the results from those tests used to say something about the population from which those samples were taken.

Take the glucose blood test for example. First off you take a syringe full of blood and then take a drop of that syringe full and put it into some automatic analysis machine. In order to make some judgements about the health of the patient you need to have some idea of how good the measuring machine is, how representative of the syringe full of blood is the drop put in the machine and then how representative of the blood in the patients body is the syringe of blood.

Knowing little or nothing about the details of blood analysis, I would guess that the actual testing machine is very reproducible if tested against a standard. I would expect some variability between different samples taken from the syringe. That has nothing to do with the test machine but is all to do with inhomogeneity in the syringe full of blood. You then have variability in syringe fulls due to inhomogeneity in the blood of the victim.

One thing I am sure about is that the test procedures will be thought through and rigorous and protocols developed minimise these errors but at some point you have to say good enough because reducing the error requires more and testing which certainly increase cost but is mostly impractical.

The upshot is that unless you understand this sort of stuff, and most don't, it is easy to assume that test results are more precise than they are.
 
Eternal422 said:
Interesting! So I presume blood circulates throughout the body quickly enough to spread the glucose evenly, which makes sense as the idea is to provide fuel for all cells.
Click to expand...

I could change my BG results by warming my fingers. exercise, anything that changed my circulation or heart rate.
I would at any time, if you BG was changing, either from a meal, or from an insulin response, there could be a large lag from core blood, to any peripheral point, even cold feet, warm hands could in theory make a difference.
 
I found it interesting when they introduced the IFCC system of measuring and reporting HbA1c test results to the UK. At that time, the Wolfson Unit at B'ham University were made responsible to QA test all the labs participating, so the Blood Transfusion Donation service in Birmingham, specifically recruited diabetics to donate specifically for this purpose. To begin with this included Type 1s - so I volunteered - and was glad to do so cos I started doing that just after my 18th birthday then had to stop just after I was 22. So - they took the fabled whole armful (250ml is it?) each of which they divided up into 300 different samples - Wolfson kept 2, one of which they tested themselves and the other 298 they sent off to 298 different labs (some of which weren't in the UK) and though they didn't actually care what the result was per se, it had to agree with what their perfectly calibrated equipment said. Should there be any doubt, they still had the control sample in hand.

Anyway - after 2/3 years, they again stopped Type 1s taking part so since then, I have utterly lost track of who might be doing the QA at labs for any of their equipment - not a subject lost of us ever think about really - but as an ex Liability insurance geek, I just assume they do it on a strict and rigid schedule.

But - do they?
 
Status
Not open for further replies.
Back
Top