So we've introduced the concept of a quick point of care test for diabetes, using a glucometer to diagnose diabetes in a dog, and that's quite a significant thing to do because you are committing. This dog and this owner to a lifetime of insulin injections. So I think it's worthwhile just taking a moment to just stop and the second part of the talk, talking about point of care there, specifically glucometers in diabetes and how we should use them.
Point of care tests have existed for, for, for a long time in diabetes, and I'm going to start with a, a, a little introduction to the general concept of point of care tests, . And then we'll talk about how we assess these points of care tests and some of the types available to us. So the first thing to say is the most important point of care test is a good history and a good clinical examination.
I hope that goes without saying, in, in this context, but I think it's important that we, we remember that everything that we assess in terms of blood work and so forth needs to be reflected in the history and the clinical examination. They need to tie up, and if they don't, we need to find out why they're not. Point of care tests are faster.
They're often cheaper, and they use, because you do them immediately, the sample quality tends not to be so much of an issue. On the other hand, These point of care meters have to be maintained. And if they're not maintained, their reliability can become an issue.
In some instances, we have to spend a lot of money if on the, on the point of care meter, and particularly if you're using one point of care meter for individual tests, then you can end up with lots of these point of care meters. And finally, because you're not involving the main laboratory, the interpretation of the results is all down to you. So there are benefits and disadvantages of point of care tests that we need to think about when we go forward and start using them.
Of course there are other factors in the choice of laboratory tests, the, the animals clinical er situation, the medical situation. The practice policy, what the client wants to pay for, what, how quickly the client wants results, how quickly, we need the results, and of course this sample size from the animals. So if we have the, the, the, the, the, the mother of this litter, then we can take a sample on.
If we want to take a a glucose from the puppies, then probably a point of care meter starts to become, much more useful because they've got very much lower sample volumes. Point of care meters are a growing market. There are new technologies coming on all the time.
The quality is improving in the most case, costs are falling, and there's definitely consumer demands for ever faster results, and to be honest, part of those consumers are also vets wanting ever faster results. But when we take on a point of care test. We should assess how good it is.
And, and there's a number of issues that we need to think about either for assessing it ourselves or for reading and checking that the assessment has been done properly. With point of care tests, the veterinary surgeon is responsible for the quality of the laboratory results. So any machine that you use needs to be cleaned, calibrated, and this should be recorded as a matter of practice policy.
And you should also make sure that any machine that you use has had some validation work done. There is an assumption, I think, in the veterinary profession. That if someone is selling a machine for the veterinary use, that it has been validated.
Unfortunately, that is not always true. And it's important that when we look at these, machines, we ask questions about their assessment and their validity. There are guidelines produced by the American College of Veterinary Clinical Pathology looking at quality assurance in these machines, and, if you, if you fancy a read of that, the reference is there.
It's quite a long detailed document. It's written for clinical pathologists, it's probably far more than you would ever need to do in practice, but it's useful to know that this work has been done. So how good are the glucometers because they are a cornerstone of what we're doing in, in veterinary work.
It's an important question this because these glucometers really do affect decisions. I, you will change your insulin dose, you will, maybe change the food, you might change the, type of insulin, there, there's, there's, there's a lot resting on these figures. and here are two glucometers, both.
Of which have had the same blood sample applied. One says 4.7, 1 says 5.8, which one is right?
Or are they both right? Does it matter? It's important, therefore, that we have some sort of in-house check, as well as checking the validation, because it's important that we understand our own machines.
Some of these machines are poorly correlated with laboratory reference methods. That's not to say it's necessarily bad or wrong, but they will give you different results. Whereas some are very good, and are very well tolerated with reference, laboratory results, and then they're more interchangeable with those reference results.
So, when we get a new device into, into the hospital, we always check. We don't necessarily do every check under the sun that we need to, we could do, but you should always check new devices for yourself to give yourself the confidence and to understand what where the differences are. So when we got our pet tracker advice, we were a device, we, we then compared it to our existing Alpha Track 2 and Alpha Track 3, and with our laboratory method.
And in doing that, therefore, We're going to have to know about what are we actually looking for in these figures. And that means we have to know some of the terms that we, we use in assessing this data and a bit of limited bit of statistics in looking at these values that we get. So when you assess a test, you're going to compare it, first of all to the gold standard, the reference measure.
But then you can also compare to itself to see how precise it is, how repeatable it is, and whether the results are linear. Now, most companies will have done the, the, these, these two things to a greater or lesser extent. And you also, as a, as a practice, need to assess the usability of the machine, the robustness of the machine as well, the other, what might be called non-statistical, aspects of using these machines.
And you need to understand the terms that are given to you when someone tries to, to sell you these tests. So in, in comparing to a reference method, the truth, we talk about, the precision, the repeatability being the variability of the measured thing. And then comparing that to the truth, that is the accuracy and the bias.
And that, and that, these are terms that reflect compared to each other and compared to the the, the, the, the truth as it were. You also need to look at how sensitive the assay is to pick up very low levels, and the terms that will often be used is the limit of the blank or the limit of detection. And these are terms that refer to how low this device is reliably able to tell the difference between no.
In this case, glucose, i.e. A blank glucose, versus, the, the smallest amount of analyte that is, reasonably detectable.
The limit of quantification is actually the ability to tell apart two values that are close together that, to see how close you can get them, and still tell them apart reliably, and that's the limit of quantification, . And for the most part, the limit of quantification and the limit of detection, represent the lower part of the curve when you measure the measured value against the true value. And then there's this linear interval, which is the line over which when you as you increase the amount of the the glucose, that you get a linear response in the measured value.
Clearly, if you take a pot of syrup with lots of glucose syrup in it, and you try and measure that on a glucometer, it won't. Give you an accurate result. The value is too high.
Similarly, if you get a very, very low concentrations, you won't get accuracy down there. And what we're aiming for is, is that the linear phase, and the linear phase for most glucometers, the pet track included, will be from about 2 to about 30 something like that. Then what we're interested in doing is looking at the correlation between the actual result and the truth, and, and.
That is something that we can do in different populations. So we may take the, the normal non-diabetic population, we may take, a diabetic population at the point of diagnosis, and you may take a group of dogs that have low glucose, and, look at these, values independently, or, or, or altogether. So, in, in our little study that that that that we did, we looked at normal glycaemic dogs, and we started by comparing 2 m that should be roughly the same, the alpha track 2 and the alpha track 3.
And what you can see here is they're not completely the same. They're very close. The difference is trivial, but they're not exactly the same.
And I put that up simply so that you have an idea about what we would consider pretty close correlation. The R squared there of 0.8 shows that it's not perfect correlation.
What happens when we ran the pet tracker against the Alpha track 2 and the Alpha Track 3, you can see that the alpha track 2, the correlation was quite a little bit less good, whereas with the Alpha track 3, the pet tracker and the alpha track 3 were really still very closely correlated. That does, of course, assume that the alpha track is correct and so actually, the better way about looking at this is to do what's called a bland Altman plot. Now that's looking at the difference between two machines that may or may not be the gold standard.
And therefore it doesn't assume that the alpha track is a gold standard, and it could be either. So we try, we try ban Alman plot and what you're looking at there is the average of the two machines on one axis, with the difference between the two machines on the other axis. And when you look at that on this, then you can see that there are differences between the values from these two machines.
90% of those differences lie between 0.5. Millimoles per liter and minus 0.5 millimoles per liter.
So that we can say that on average, that most values that you get from the pet tracker will be between plus and minus 0.5 of the alpha track and vice versa, which means that that that that the clinical decision making may be unlikely to be affected at this, at this level. But note also, there are occasionally values that may be as much as 1.4 millimoles difference.
And this is still satisfactory, but it's important to be aware that if you compare two devices, you may get that sort of order of magnitude, even in normal glycaemic dogs. We can look at other aspects, more asset assessments can be done. You can look at regression analysis, and we'll come, come on to that.
We can look at how often are the results classified differently, and that's a cap and that can be shown on what's called a Parkes consensus error grid. And that gives you an idea about how badly they are misclassified. So just to, to, illustrate that point, the pet tracker has been, used as a, in a clinical evaluation, a much more detailed clinical evaluation than we've done, from Cornell University, looking at the, .
The, regression between the standard reference method and the pet tracker, and you can see here in the, the orange and the blue are cats and dogs, and you can see here that the, the, the closeness of these results is really very, very good indeed. Because these, these are, these are idealized situations and. These may not always lead to to the same results if you do it in your practice with the time difference between sampling, for example, that may be important, but nevertheless, at its best, this device does almost as well as the the reference method.
And that's followed through when you look at what this consensus error grid, which is where things in zone A are are results that are different. But it wouldn't matter. BCDE courage becomes more and more different and more and more clinical effect.
So it really doesn't matter at the at the at the at the low end if if an animal is is recorded, sorry, at the high end if if an animal is recorded as 25 on one machine and 30 on the other, it's not going to alter your assessment of, of the, values, but, it could matter an awful lot if you thought that a dog had a blood glucose of 12 when it actually had a blood glucose of 6. So this consensus error grid is a tool that people use to assess machines and using this, the Cornell Group were able to demonstrate that nearly all bar one, I think value was in Zone A, i.e.
It didn't matter that the change, that the difference was, was, was there, it wouldn't have altered. Your clinical impression, your clinical decision making. And those lines are drawn by experts who who've done a lot of this sort of diabetic work and so forth.
of course, you know, those lines are arbitrary, to an extent and therefore, therefore, some people, might, might still make different decisions. But it's not just the the, the differences between machines we have to think about. We can also make these machines perform badly if we use them in the wrong way or with the wrong sort of animals.
And when it comes to glucometers. The big, big variable is the amount of red blood cells, which can have a significant impact on the measurement. So it's important that you assess the animal for anemia.
And, and, and it may be nothing more than lifting up the gums and lifting up the lip and looking at the gums and, and, and checking that the dog or cat is not anemic. But that, that, that is an important thing to do. The second source of variation is contamination with tissue fluid.
When you do an ear prick and you then squeeze, if you squeeze too hard, then you will not get a capillary blood sample, but a capillary blood sample with tissue fluid included. And that may have a different glucose value to the blood, and that may create a false glucose value. So it's really important when you do these er ear pricks or or prick the side of the pad that you do so carefully, that you make sure you squeeze gently, and if you don't get blood, don't carry on squeezing, but rather find another site and repeat the process.
There is known to be some differences between venous and capillary blood in terms of glucose. Some research says it's minor, some research says it's a bit more significant. Then there have been some studies looking at the site of capillary blood samples.
That's tended to show that there's no difference between the sites that we, when we do these blood glucosis, but I think it's important to be aware that consistency. Better where you can. For me, I tend to use the ear, I tend to warm the ear very, very well with my with my hands, and, and then I, I, I, I do, I do a prick towards, towards the, the center of the ear.
