accuracy, standards and INR meter design

[pellicle]

Hi

a user here frequently suggests that meters (such as the Coagucheck XS) are designed to be within 30% tolerance. I have always found that an interesting claim and have looked for supporting evidence for that since I first read it.

I have not found anything about the design of meters, but I have found some interesting things about the reference standards which I thought you (the reader) also may find interesting.

The idea of such "accuracy" seems to come from an ISO standard, the Coagucheck XS acceptance test documents say the following:

Additional evaluation criteria were set for information only according to the
draft ISO standard 17593 [1], which requires a
- mean bias of ≤0.3 INR (in the INR range 2.0 – 4.5) and
- a total of >90% of all data bias within +/- 0.5 INR (for INRs <2) or
within +/- 30% (for INRs >2)

I feel that what this suggests is that without other guidelines to refer to, Roche has chosen to adhere to the ISO17593 standard which it seems is not so tight. It has been suggested in the literature that this ISO standard is in need of review these days. A 30% range for INR greater than 2 would be between 1.4 and 2.6 With such tolerances it is no wonder that many clinics erred on the higher side (to prevent thrombosis at the risk of bleeds)

It has been said before that many things influence the numbers you get from a lab, everything from how the blood is handled to the 'reagents' used. This article supports that the 'reagents' used in the lab are important variables

In conclusion, our data show some important differences are present between INRs determined using Recombiplastin, Innovin and two conventional thromboplastins

I found slide 6 (page 6) of this presentation quite informative here, and it puts into historical context the advances of anticoagulation therapy. Page 10 is also interesting in the presentation of the data suggesting that Labs give higher (increasing with INR) reading than machines. Page 11 suggests that at INR lower than 3 the accruacy is good.

None the less studies comparing these Point Of Care meters has suggested very good results, much more accurate than 30% in the vast majority of cases.

As I posted in this thread the accuracy of home INR machines is significantly better than ±30%

For INR = 3 a 30% variance would be 2.1 to 3.9 yet the above chart shows nothing like that, with only the outliers even approaching that.

My view is that the testing of these meters has shown that they can be relied upon by many, but that there are outliers. So always get reference checks when you start home testing, and perhaps given the variablity of Lab results if you are concerned then get cross referenced results.

If anyone can find a reference to suggest that the current generation of coagulation meters were designed with 30% tolerance in mind I'd appreciate seeing that, however I feel that they are actually designed with much higher accuracy in mind and the compliance to standards is the issue - NB that the standards are too broad.

 

[marc_kowal]

I have a PDF that Alere sent me where it references the 30% variance that the W.H.O. allows. I can't post it here, but if someone would like to post it for me, shoot me a p.m. with your e-mail and I can forward it to you.

 

[Protimenow]

Yes, 30% does seem VERY high. Self-testers in particular should be very wary when the 30% tolerance is applied -- in theory, a meter that says 2.0 (whch my clinic now claims is the bottom of the desired range) could actually mean that an INR is 1.4 - well within the area for stroke risk. I can't imagine that the meter manufacturers are comfortable with even that loose a margin -- they would, of course, prefer to be within .1 or .2 of lab results. (Plus, for people who rely on a meter's accuracy, if a self-tester trusts that his or her meter, reporting 2.6 was correct, and actually had a TIA, like I did, with the hospital lab saying 1.7, these meter manufacturers certainly don't want that kind of legal liability).

30% is probably fairly easy to meet (as opposed, say, to 10%). It's probably a lot safer for the meter makers to DELIVER 30% while actually striving for 10% or so.

FWIW -- the InRatio and CoaguChek XS use indirect methods for calculating INR (as opposed to just looking for a clot to form). The meter that I have the most trust in is the Coag-Sense, which actually detects when a clot forms. The Coag-Sense seems to consistently score slightly lower than the hospital lab's blood draws and the CoaguChek XS and InRatio.

Also, as Pellicle noted, and as I've also said, results for the same blood draw may vary from lab to lab. Without a true level of consistency, we can probably just decide which method - which lab or which meter or both - offers the most accuracy.

A final thing -- and I've said it before --- trusting a meter in a doctor's office is no real guarantee that the DOCTOR's results are any more accurate than one using a similar meter at home. Hemochron MAY be more accurate than the other meters -- it was designed for use in operating rooms or medical treatment rooms and should provide pretty accurate Prothrombin times and INRs.

 

[lance]

Comparison testing

Comparison testing

Interesting link.
My experience with XS leads me to believe the referenced chart is accurate. Last week's lab comparison test: Lab was 3.1 and XS 2.9 results in keeping with comparison test results since I've been using XS. Same for spouse too. One comparison test showed an .08 difference--once in all these years. Lab test results took 4 days so once again XS was accurate.

 

[Jamey T]

My last lab testing from a week and a half ago, the lab blood draw was 2.8, my Inratio 2 said 3.2, and the labs new Inratio2 said 3.6. They have to do a bunch of testing comparing their Inratio 2 to blood draws, so I let them stick me. They were thankful, and I got more data.

 

[mecretired]

I have had a Coagucheck XS for about 3 yrs. I purchased it myself on eBay and have been very happy with it. I test weekly and go to the lab every 6-8 wks or so to verify home readings. Yesterday Coaguchek XS was 2.9. Lab was 2.9. The two have always been within .2 of each other. I occasionally wonder about the lab as the local hospital does the vein draw and sends it to a Quest lab about 8-9 hrs away. So it's very good to see my 2 readings are the same. &#55357;&#56836;

 

[tom in MO]

Nobody designs an analytical device to exactly meet a specification, to much risk of failure. The ISO standard has to be met, but the company's design team has specifications that are tighter than ISO so they can assure success at the ISO specifications. The internal design specifications are probably proprietary.

When I got my meter, the trainer said she had to test it's accuracy by comparison to another meter. They had to agree within 0.2 or 0.3 INR for my new meter to be acceptable. This is not a set accuracy based upon the reading. For example, a new meter can read 2.2 compared to the reference meter at 2.5 the difference is 0.3/2.2 = 14%.

 

[Protimenow]

Of course, this assumes that the reference meter reports values that are very close to the lab's (and assumes that the lab's results are consistent with those of other labs that test the same standardized samples). A result from a meter that is 14% lower than another meter that is similarly 14% below a lab that is also within a loose level of tolerance could result in a difference that is more than 30%.

It's too bad that the testing of INR or more art than science.

 

[tom in MO]

It is science just not simple. Most people are used to blood results that are a measured physical property like concentration. However, your INR is not a measured physical property like weight, density or concentration. It is a reaction rate. But then the reaction rate is normalized for the purposes of comparison. There is no reference standard for INR comparison (e.g. a blood sample prepared to a known INR level), the only comparison is to another method of measurment (i.e. the laboratory procedure.) I find it a hard concept to wrap my head around.

 

[pellicle]

Of course, this assumes that the reference meter reports values that are very close to the lab's

from http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1860570/
Accuracy and clinical utility of the CoaguChek XS portable international normalised ratio monitor in a pilot study of warfarin home‐monitoring

seem close enough to me ... although the study was only for 17 patients ...

Also, for what its worth the Coaguchek XS uses a different method of determining the INR, it uses an amperometric system, similar to that used in quite high end commercial lab machines. Older generation machines use other methods. One can only assume this is to make things more accurate (if not cheaper??)

 

[Protimenow]

In theory, any method that can provide for an accurate measurement of clotting is probably good. The original methods - which actually detected the formation of a clot, may have been the most accurate, but took too long to be of much use.

The World Health Organization (WHO) apparently has standardized on what's referred to as 'Tilt Table.' I am not exactly sure what this method involves, but I think that it uses a reagent of some type to activate the thrombin in the blood sample, and the blood/reagent mixture are put onto a glass slide (or something) and the thing is tilted like a teeter-totter (I think) until a clot forms and the blood stops moving on the 'tilt table.' This is a physical method for detecting clot formation. The Coag-Sense meter uses a spinning wheel and optical sensor -- when the wheel stops spinning, it's because a clot formed. The meter calculates time, and, based on time and reagent value, the INR is determined.

CoaguChek's amperometric method, which detects changes in amperages of the blood/reagent mixture, may be fine -- it seems to work well for many users. In my experience, with my own blood, the results were often considerably higher than the lab (or the Coag-Sense). However, this method has been tested hundreds, if not many thousands, of times, and is accurate enough to be used for INR measurement.

As I said in the beginning of this reply, any method that can reliably and accurately detect prothrombin time should be an acceptable basis for measurement. (Hey - if a color change worked, I'd be behind that, too).

 

[pellicle]

Hi
(just for discussion

In theory, any method that can provide for an accurate measurement of clotting is probably good. The original methods - which actually detected the formation of a clot, may have been the most accurate, but took too long to be of much use.

well perhaps, perhaps they were also irregular and unreliable? I don't know. Perhaps its just cheaper this way and equally reliable??

I'm for lower costs!

According to i-Stat ...

In a prothrombin time test, coagulation is initiated by mixing the sample with tissue thromboplastin. In traditional prothrombin time tests, complete activation is indicated when activated thrombin converts fibrinogen to fibrin and extensive or localized clots are detected mechanically or optically. The i-STAT PT/INR test is similar except that the endpoint is indicated by the conversion of a thrombin substrate other than fibrinogen. An electrochemical sensor is used to detect this conversion.

if anyone was interested ... moving on

CoaguChek's amperometric method, which detects changes in amperages of the blood/reagent mixture, may be fine -- it seems to work well for many users. In my experience, with my own blood, the results were often considerably higher than the lab

which is interesting because many studies seem to suggest that the Coaguchek reports an INR lower than a lab. However looking at the data above there are also data points that are higher than the labs.
Interestingly the variance gets lower as your INR gets lower ... which is good.

I believe from discussions here that there should be a way to 'test drive' meters to see if there is difference between each meters chemistry assumptions and one's own blood chemistry. Then a meter could be 'tuned' to the individual.

As everyone seems to react differently it could just be that meter A suits person X more than meter B.

Most people seem to be like iPhone / Andorid users and just love their team (without objective reasons).