How accurate are INR measurements?

This is a serious matter, and one which should be borne in mind when you look at your results .

It would appear that the "reliability" of INR results is about +/- 0.2 units, so that an INR of 2.0 could in fact be between 1.8 and 2.2, or perhaps even outside this range.

 This has major implications for changes in dosage based on INR results.

A really good article on this aspect is available at ClotCare online resource. Much of the information on that page is reproduced below, togther with some of my comments in blue:

At ClotCare, we often get questions from people who have obtained INRs from 2 sources in a short time period and want to know which source to trust. For example, maybe the lab reports an INR of 3.5 when a point of care (fingerstick or POC - Point Of Care) device reports an INR of 2.5. What do you do in this situation when one source indicates a high INR and one reports an INR in range?

In general, if performed correctly, the INR is usually accurate and reproducible, but there are a number of ways in which an incorrect INR can be obtained. If the INR result does not "fit" with previous values and/or what is expected, it is always reasonable to obtain another blood sample and repeat the test.

Lab reliability may differ from one lab to another: One lab (lab A) that I work with is very reliable and reproducible. For example, a few years back we had the lab perform duplicate INRs on 2 blood samples collected at the same time from a number of our warfarin-treated patients. The INR results on simultaneously collected blood samples did not differ by more than 0.1 INR units. Further, over the past 10 years we occasionally would re-test a patient with a POC device if we suspected the lab INR to be incorrect. We practically never see a substantial difference between the lab and the POC in those situations. Lastly, I have seen a number of our patients with INRs from lab A whose INR did not vary by more than 0.2 INR units over a period of several months. Therefore, I am convinced that INR results from lab A are very reliable and reproducible. On the other hand, a study that we performed in another setting using a different lab (lab B) found that approximately 10% to 15% of the INR results were wrong. 

The BOLD sentence above has been emphasised by me. It is a worry that some labs are much better than others in their reliability

What about lab vs. POC results? In general, some studies have considered INR results from 2 methods to be in reasonable agreement if the two results are within 0.5 INR units of each other (ex. 2.0 and 2.5 would be in reasonable agreement). 

We conducted 2 studies years ago to compare lab B INR with POC INR results.¹ In both studies, we did simultaneous POC and lab INRs - lab B was actually either of two labs (one at University Hospital and the other at the VA Hospital - both of which are affiliated with the UT Health Science Center in San Antonio, TX). If either method (POC or lab) told us that we needed to change the dose of warfarin in these very stable patients, we had the patients come back in 24 to 48 hours and we repeated both the POC and the lab INR; in no instance did the repeat INR results indicate that a dosage change was needed. 

Again this is not good - significantly different results and outcomes - The comment that treasonable agreement if results are within 0.5 INR units of eah other ( for example between between 2.0 and 2.5) reinforces the comment above that INR is really only reliable to about +/- 0.2 units! 

After the study was completed, we evaluated "erroneous" INR pairs (simultaneously obtained POC and Lab INR results). We considered paired INR results to be "erroneous" if the POC and lab INR differed by more than 0.5 INR units and if one method or the other would have led us to make a dosage change. (Paired INRs that differed by more than 0.5 INR units were not considered "erroneous" if neither value would have led to a dosage change - an example might be paired INR results of 2.2 vs 2.8). We identified 16 erroneous pairs of INRs in study #1 and 37 erroneous pairs of INRs in study #2. Among those erroneous pairs of samples, we found that 15 of 16 in study 1 and 26 of 37 in study 2 occurred with the lab INR method. In other words, the POC results were more reproducible and less likely to result in an incorrect dosage change.

My underline emphasis above. This is a quite remarkable result, and would indicate that the lab INR methods need significant attention

My conclusion from the above studies and experience with lab A is that the POC INR results were more reliable and reproducible than those obtained from lab B; but results from lab A are just as reliable as those from the POC device.

What can cause INR results to be incorrect?

There are a number of factors that can result in a given lab INR being inaccurate. A few examples are listed below:

• Under-filling the tube when blood is collected can lead to a higher INR result
• Using a blood collection tube that contains 3.8% citrate rather than 3.2% citrate can falsely increase the INR
• Performing an INR on a blood sample of a very anaemic patient 
• Entering the wrong ISI (International Sensitivity Index for the thromboplastin reagent that is used in the test) into the machine
• Having the laboratory device set at an incorrect temperature
• The patient taking other medications that may alter the test
• Having an interfering substance (such as certain antibodies) in the blood

With the POC INR, since it is performed on fresh whole blood, the problems listed above that relate to blood collection are not a problem; but the other potential sources of error may be. Further, there are other factors that may lead to an error in the POC INR result. For example:

• If a patient squeezes his/her finger too hard when obtaining a drop of blood for a POC test that may accelerate the blood clotting and give an INR that is low.
• If a patient takes too much time to apply the drop of blood, clotting may start before the blood is applied to the POC test strip and than may also give an INR that low.
• It also has been suggested that heparin or low molecular weight heparin may elevate the POC INR more than the lab INR because of differences in the method and the thromboplastins being used. The POC test is performed on whole blood while the lab INR is performed on plasma that has been separated from the blood cells.
• If the POC device is not on a flat and stable surface, that too may alter the results but the degree of alteration in the INR may depend on the underlying technology of the particular POC device. For example, one device used vibrating iron particle and recorded that the blood had clotted when the particles stopped vibrating. When the device was placed on a counter that was subject to mechanical vibration from other causes, we found that the INR results were often increased due to the additional "outside" vibration.
• Improper storage of the test strips also may lead to incorrect POC INR results.
• Some POC devices require that a computer chip be changed out with each new lot of test strips, failure to do so can yield incorrect INR results.

So, if both lab and POC may give bad results, which should I trust?

Unfortunately, the answer is not clear-cut. It is likely that pre-test issues (how the sample is collected and handled) play a substantial role in getting erroneous INR results.

With the POC INRs, you should use appropriate technique, store and care for the device and test strips properly, and be alert to changes in other medications that might alter the INR result. With lab-based INRs, you have to rely on the technique and attention to detail of a particular laboratory. Therefore, my own view is that clinicians should try to identify a laboratory that is reliable and use that lab as much as possible. Regardless of the method, if the INR result does not "fit" with what is expected, it is always reasonable to question the accuracy of the test and repeat the INR with a fresh blood sample.

And so, where do we go from here? It is not at all clear. INR tests are somewhat unreliable, and I guess all that we patients can do is bear this in mind, and if we have a test that appears to be unexpected, perhaps have a re-test.