Point-of-care testing in oral anticoagulation: what is the point?

Warfarin is commonly used for oral anticoagulant therapy (OAT) in patients predisposed to thromboembolism (TE) or being treated for thrombosis. The efficacy of warfarin therapy is followed by monitoring the prothrombim time (PT) and international normalized ration (INR), which were, until recently,

Approximately 1% of the population or two million people in the United States receive OAT. (1) As warfarin is the only oral anticoagulant currently available, patients who are at high risk for thromboembolism due to atrial fibrillation (AFIB), mechanical heart valve replacement, or congenital thrombophilia are necessarily treated with this drug. Warfarin is, unfortunately, a less-than-ideal drug due to its narrow therapeutic window, a wide variability of response from patient to patient, and its sensitivity to foods and drugs, which can markedly increase or decrease the level of anticoagulation per given dose. Hence, patients treated with warfarin always have been forced to comply not only with a difficult dosage regimen, but also to undergo testing of their prothrombin time to calculate the INR from weekly to once every couple of months to ensure that their regimen is adequate.

Such testing--until recently--has taken place in the setting of a primary care physician or specialist's office, or at an anticoagulation clinic. Consequently, the turnaround time for the results depended on whether the clinic was connected to a laboratory or the samples needed to be sent out to a reference site. Practically speaking, in many cases the patient would be at home once results were analyzed by his physician, and would have to be contacted about how to proceed with his dosing schedule. Due to the inherent risks of warfarin, namely bleeding and thromboembolism, and the cumbersome nature of drug monitoring, it has been found in the United States that only an estimated one-third of people with AFIB who are eligible for warfarin therapy actually receive it. (2) Among patients taking warfarin for any indication, the proportion of time they achieve INR within therapeutic range varies from 35% to 70%, depending on how closely and by whom they are followed. (2) These unacceptably variable levels of care demanded changes in how patients are monitored, leading to the development of POCT monitors for measuring the INR. Patient self-monitoring (PSM) using POCT instruments has recently been shown to enable mechanical valve recipients to maintain tighter control of their OAT, at the same time reducing the rate of complications associated with warfarin therapy. (3,4)

Indications for oral anticoagulant therapy

There are numerous indications for OAT. Perhaps the most prevalent of these is AFIB. A common disorder, the occurrence of AFIB increases with age, reaching figures as high as 10% among people 75 years and older. (5) Thromboembolism directly related to AFIB may account for as much as 50% of all cerebrovascular attacks (strokes), with a calculated risk of thromboembolism of 4.5% of patients per year. (6) In other words, of all patients with AFIB, approximately 5% will suffer a stroke each year. Preventing such morbidity and mortality through effective anticoagulation is clearly desirable.

Another group consists of patients who have had one or more heart valves replaced with a mechanical valve. Such devices require lifelong systemic anticoagulation due to a risk of thromboembolism of 8.6%. (6) There is also a smaller but growing population of patients who have already had one or more venous or arterial thromboembolic events and are placed on either short- or long-term anticoagulant therapy.

Monitoring warfarin therapy

Patients on warfarin require daily monitoring of their response to the drug at the start of treatment until the PT/INR stabilizes in the therapeutic range, and once every month or two thereafter in particularly compliant and stable patients. Warfarin inhibits the in vivo conversion of vitamin K to its active form, leading to the depletion of the vitamin K-dependent clotting and anticoagulant factors (II, VII, IX and X, Proteins C and S). (5) Warfarin has a very narrow therapeutic window, and there is no reliable way to dose it (i.e., with a set mg/kg dosing chart) given the notable variability from patient to patient. In addition, the activity of the drug within the body is susceptible to interaction with other drugs and to a number of foodstuffs, the most notable of which are "leafy green" vegetables. Broccoli and kale, for example, increase vitamin K in the blood, counteracting the effect of a given dose of warfarin. Due to its action on the vitamin K-dependent factors, warfarin most prominently affects the initiation (formerly called the extrinsic or tissue factor pathway) of the clotting cascade, which is assessed by the PT. Warfarin has almost no effect on the activated partial thromboplastin time (APTT), except in cases of severe overdosage. The PT assay is conducted on blood collected in 3.2% sodium citrate and centrifuged to achieve platelet-poor plasma. A standardized thromboplastin (TP) reagent is then added, and the time to the detection of a fibrin clot is measured. The results are reported in seconds, and normal ranges vary widely, depending on the reagent-instrument combination used. (5) Inherent differences in the sensitivity of TP reagents led in the 1980s to the development of the INR to standardize PT results of the growing number of patients on warfarin. The INR is the ratio between the sample PT over the mean normal PT (MNPT) raised to the power of a calculated "international sensitivity index (ISI)" of the TP reagent:

INR = (patient PT / MNPT)[.sup.ISI]

The INR normal range is 0.9-1.13, and increasing INR values correspond to increased anticoagulation. (7) Most people on warfarin are required to maintain INR values between 2-3, or 2.5-3.5 for OAT due to a mechanical valve.

The ISI of a particular system is determined by calibration of the TP in use with an international reference preparation (IRP) of TP from the World Health Organization (WHO). Results from the TP reagent to be used vs. those of the WHO reagent are plotted on a log (PT) vs. log (PT) plot to determine if there is an adequate correlation between the reference standard and the new reagent. The within-laboratory imprecision of the calibration should be no more that 3%. (7) Insensitive reagents have a higher ISI (2-3), meaning that they will not produce a prolonged PT until the level of clotting factors is markedly decreased. More sensitive reagents have lower ISIs (~1), and react to smaller changes in clotting factor levels. Ideally, each new TP reagent would be tested vs. an IRP measured by the WHO. In practice, however, most laboratories use a secondary IRP (which was itself measured against the WHO standard) to calibrate new TP reagents. Studies have shown that the total coefficient of variation of INR estimates measured with a secondary IRP ranges between 8.6% at an INR of 2 to as much as 11.8% at an INR of 4. (8) It should not be forgotten that the INR is a measurement specifically created for and, therefore, relevant only to those patients on warfarin.

Also, until recently, most patients on warfarin therapy were monitored either by their primary care physician or by a combination of nurses, clinical pharmacists, and physicians at anticoagulation clinics. Such clinics have become more prevalent as multiple studies showed poor OAT control among patients not treated at specialist facilities. As noted, patients followed by their "regular" physicians while on warfarin have been found to have within therapeutic range INR values only about one-third of the time. (2) Thus, most of the time they are at risk of complications from either bleeding or thromboembolism, rather than benefiting from the drug. In comparison, patients followed in anticoagulation clinics achieve therapeutic range INRs from 65% to 80% of the time on average. (2,9,10) Specialized care is also associated with decreased complications, which translates into fewer emergency room visits and hospitalizations, leading to an estimated savings per 100 patient years of more than $150,000. (11) Such results have been achieved due not only to the expertise of the staff at anticoagulation clinics, but also to centralization of testing. Patients could have samples drawn while waiting to see the physician or nurse; at the same time, their specimen was sent to an in-house/hospital laboratory and processed immediately. For the most part, physicians could receive results and make treatment decisions with the patient still in the clinic, or be able to contact the patient the next day to ensure tighter control. In fact, anticoagulation clinics have become one of the most renowned users of POCT for INR, allowing the patient to be tested in the office and have an INR result in less than five minutes.

Point-of-care testing monitors for INR

INR measured by POCTs is analogous to the now ubiquitous use of hand-held glucose monitors by diabetic patients. While a few INR systems exist, they all function on the same basic principles: collecting a drop of blood and applying it to a test strip penetrated with TP reagent, then monitoring the sample movement across the strip as the blood reacts with the reagent and clots. Some monitors follow clotting time via optical detection of capillary blood flow, while others detect peristaltic movement of the sample. A third method involves the detection of the oscillation of iron oxide particles, which mix with the blood sample on a test strip placed in the monitor over an electromagnet, creating a pulsating magnetic field. The movement of the iron oxide particles within this field is monitored by a photodetector until it stops once the sample clots. (12)

POCT monitors vary in their ability to accept different sample types. Some are restricted to capillary blood, whereas others accept venous or capillary blood and even citrated plasma. This becomes an important distinction, as monitors that accept citrated plasma can be calibrated vs. TP reagents used in a local reference laboratory, potentially allowing clinicians to calibrate or coordinate INR results of a patient's home monitor vs. those from a hospital or clinic laboratory. Testing of different monitors suggests that they are unaffected by variations in hematocrit ranging from 0 (pure plasma sample) up to 56%, as well as by the platelet count. (12) All monitors produce results within about three minutes from the time the sample is drawn, a distinct advantage over a turnaround time of hours or days, depending upon the remoteness of the patient and his physician from the laboratory. A computer-chip test card within the monitor analyzes the data and converts it to an INR. The monitors rely on test strips, and these so-called dry-reagent systems have clotting times that are longer than the conventional laboratory wet-chemistry techniques, due to differences in the kinetics of solid-solid vs. liquid-liquid interactions. Hence, the analyzer is required to "transform" the data to a coordinate system, which has been standardized during manufacture to a particular wet-chemistry reference method. Calibration factors to make this transformation are encoded magnetically on the computerized test cards. The cards also enable the monitor to compensate for lot-lot variation of the TP reagent and potentially allow them to be calibrated with a local laboratory's reference method, rather than to the manufacturer's standards. (12)

Performance of POCT monitors

Given that numerous POCT monitors are available, how is one to choose among them? There can be, unfortunately, rather large variations in the clinical performance of monitors that initially appear equally accurate and precise. There are only a few studies comparing all existing monitors; most only compare two or three machines and/or manufacturers. An exception is a paper by Gosselin, et al (13), which examines the accuracy and precision of nine different monitors from six manufacturers. This is a well-constructed study, as it subjects the monitors to testing with two TP reagents, one more and one less sensitive, and considers the accuracy and precision of each monitor's performance with each reagent. In addition, the performance of each monitor vs. reference laboratory values for samples is recorded using Bland-Altman or bias plots, as opposed to calculating correlation of the results via log-log plot regression analysis. Correlation is a measure of the similarity of relation between data points (i.e., that the data points follow a normal distribution or lie upon a particular slope) but does not capture the agreement between two methods because the normal distribution by the new method may lie many standard deviations away from that of the old method. When a new methodology is introduced, it is important to know how much it differs from the old, and whether that difference is pronounced enough as to be clinically significant. One can demonstrate the lack of agreement between methods by calculating bias, which is estimated by the mean difference of results from the new minus the old method and the standard deviation of the differences. A Bland-Altman or bias plot is created by plotting that difference +/- 2SD over the average of results measured by both the old and new methods across the range of data points. Such a plot shows how much and in what direction the results of the new method vary from those produced by the older standard. (14) Bias plots illustrated by Gosselin, et al, show that all of the POCT monitors tested produced accurate results in comparison to the reference laboratory INR.

Although there is a tendency for both testing methods to be more variable as the INR increases, the POCT monitors seem more prone to a positive bias at the high end of the INR scale. When the coefficient of variation (CV) of the results was compared for the two TP reagents (one with low and one with high ISI), however, differences between the monitors became apparent. Assuming a desired CV of [less than or equal to]3% for calibrating INR values, only 3/9 monitors reached this level of precision with the more sensitive TP reagent, while 7/9 were sufficiently precise using the less sensitive one. It may be recalled that a more sensitive TP reagent captures smaller changes in clotting factor levels, potentially allowing finer tuning of anticoagulation. While this study demonstrated that all monitors were similarly accurate when tested vs. the reference laboratory INR, an interesting study from Israel (15), hints that accuracy as measured between different methods does not necessarily equate with accuracy in clinical practice.

Approximately 80 patients on stable OAT at the Sheba Hospital anticoagulation clinic had their INR measured concomitantly on one Hemachron Jr. POCT monitor and the reference laboratory. Two weeks later, a similar group of 111 patients followed the same procedure, using a CoaguChek S monitor. Results from the POCT monitors were given to a physician (unaware of the reference laboratory result) who was instructed to follow an established dosing algorithm based solely on the results of the POCT monitors. Changes in dosing based on the POCT results would have led to unjustified dose increases in 22% of the patients tested with the Hemochron Jr, vs. 8% of the patients tested with the CoaguChek S instrument. The overall time within therapeutic range INR that was achieved with the Hemochron Jr. monitor was 62%, vs. 90% with the CoaguChek S. These data clearly demonstrated that it is imperative to verify that monitors are not only statistically reliable when compared with a reference method, but that they are also reliable on clinical grounds.

They also bring to the fore the point that simply comparing single INR values from POCT monitors vs. reference laboratory methods is an inadequate estimation of clinical performance of these monitors, since OAT is managed over a range of acceptable INR values. This evaluation of two POCT monitors used Anderson's criteria to test the agreement between POCT and laboratory INR measurements. According to the criteria, agreement is achieved if any one of these three clinically relevant conditions is met:

* both measurements are within the patient's therapeutic range;

* both are either above or both below that range; and

* that the measurements were within 0.4 units of each other. (16)

Another interesting study by Poller, et al (17), considered the variation inherent in POCT monitor manufacture. It compared 14 CoaguChek instruments to 14 Thrombolytic Assessment System (TAS) monitors. Although produced by different manufacturers, both measure INR via the reaction of a sample with iron oxide particles. The same operator in the same laboratory using identical duplicate blood samples drawn from two normal individuals and four patients on stable OAT tested the instruments from each manufacturer. Therefore, each of the 14 machines from each manufacturer tested 12 samples to determine the within and between monitor variability. While testing conditions were the same, the results were so markedly different that the authors chose to use pseudonyms "Brand A" and "Brand B" to report results. (17) They decided that since all POCT monitors have not undergone such testing it would be unfair to bias consumers for or against one particular manufacturer. Testing of four duplicated patient samples on "Brand B" showed significant variation in INR values: from 1.27 to 1.79, 2.16 to 3.50, 2.03 to 3.87, and 3.03 to 4.12. "Brand A" duplicate INR tests did not show major differences. Had a patient or anticoagulation clinic chosen to purchase a machine from "Brand B," they would be unable to determine whether the variation in results was due to true variation in drug levels, or whether it were merely due to performance variability. Unnecessary and potentially dangerous dose changes could be made due to such imprecision.

Ease of use

An additional factor which must be considered is how easy the monitors are to use and whether there is any variation in their operation in real life. A study conducted at four Italian medical centers involving 80 patients followed them using INR results obtained at home on their POCT monitor, and compared the results with those obtained using blood samples drawn and processed at one of four central laboratories. (18) The patients were screened to determine if they were both mentally astute and physically dexterous enough to be able to handle the process of self-testing. After an initial training period during which the patients were taught how to use the POCT monitor and troubleshoot any difficulties encountered, significant differences in the variation of INR measurements developed between the reference laboratory and the POCT monitor. Total agreement between measurements taken by the patients vs. INR values determined in the central laboratory (both values being either within, above, or below therapeutic range INR) varied between 50% at one center up to 89% at another. Of course, these results reflected the inherent difficulty one encounters in measuring INR by any method. Given, as we have seen, that the CV of an assay conducted under near-optimal conditions in a laboratory can be as high as ~12%, it is perhaps not surprising to see the low levels of agreement between samples measured by POCT monitors vs. the laboratory. An interesting correlate to this study would have been a comparison between results achieved after drawing identical blood samples and distributing them among the four laboratories to see if any greater agreement between INR results could have been achieved. While actual measurement of the INR may be complicated, the patients using the POCT monitors had no difficulty performing the test 80% of the time. Of the difficulties that were encountered, more than half were due to insufficient amount of blood obtained from the fingerstick, followed by difficulty in placing the drop of blood in the testing cartridge. (18)

POCT cost-effectiveness

Two studies by a group of cardiothoracic surgeons in Germany, which collectively followed over 3,000 patients with prosthetic heart valves, demonstrated the remarkable effects patient self-monitoring can have in OAT. The initial study, published in 2001 under the eponym ESCAT [I.sup.3], showed that patients who used POCT monitors remained within therapeutic range INR (2.5-4.5 in this patient population) 78.3% of the time vs. 60.5% for patients followed by their general practitioner. Patients were able to do this by monitoring their INR much more often: those using POCT monitors performed a total of 23,693 measurements vs. 4,599 measurements on an equivalent number of patients using conventional testing. In patients with artificial valves, the complications of oral anticoagulation most commonly occur during the first six months of therapy, and then decline to a low, steady state thereafter. Patient self-monitoring appeared to abrogate complications during this initial peak risk period. POCT also significantly reduced the number of serious complications from bleeding and thromboembolism, as patients using the monitors achieved tighter control of their OAT. The second portion of the study, ESCAT [II.sup.4] was undertaken to see if patients using POCTs could be safely maintained at a lower therapeutic range of INR (2.0-3.0 vs. 2.5-4.5). Self-monitoring enabled them to reach this lower target range, facilitating their achievement of the lowest rate of complications from bleeding or thromboembolism of any study group (0.87% per year vs. 2.9% per year for POCT users with INRs from 2.5-4.5 vs. 4.7% per year for patients followed by a general practitioner). (3,4) These are interesting studies, which provide the first evidence of a true benefit to patients who are able to use self-monitoring to manage OAT.

The population of patients who may benefit from POCT monitors is limited by a number of factors, some associated with the patient's physical and mental capacity, and others related to the inherent cost of such testing. Studies such as ESCAT I and II have provided sufficient evidence of the overall cost effectiveness of POCT monitors in measuring the INR, so as to lead Medicare to agree to fund the purchase of POCT monitors for patients who have a mechanical heart valve. As a recent policy statement by Medicare reads (19), this is the only population for whom evidence-based medicine has proven the benefit of this testing device.

POCT monitors are becoming a common method of managing patients taking warfarin. With their use already well established in anticoagulation clinics, they will likely become increasingly popular among private patients, as the indications for underwriting their use grow broader. Time and use will reveal which among the many POCT monitors available are truly the best. Certainly, we can expect to see overall improvements in the care of chronically anticoagulated patients, if problems with accuracy and precision are carefully monitored.

References

1. Tripodi A, Chantarangkul V, Mannucci P. Near-patient testing devices to monitor oral anticoagulant therapy. Br J Haematol 2001;113:847-852.

2. Samsa G, Matchar D, et al. Quality of anticoagulation management among patients with atrial fibrillation. Arch Intern Med 2000;160:967-973.

3. Kortke H, Korfer R. International normalized ratio self-management after mechanical heart valve replacement: is an early start advantageous? Ann Thorac Surg. 2001;72:44-48.

4. Koertke H, Minami K, et al. INR self-management permits lower anticoagulation levels after mechanical heart valve replacement. Circulation. 2003;108[supp II]:II-75-II-78.

5. Ahya S, Flood K, Paranjothi S, et al. The Washington Manual of Medical Therapeutics. 30th ed., Lippincott Williams & Wilkins; 2001.

6. Schulman S. Care of patients receiving long-term anticoagulant therapy. N Engl J Med. 2003;349:675-683.

7. Kitchen S, Preston E. Standardization of Prothrombin Time for laboratory control of anticoagulant therapy. Sem Thromb and Hemost. 1999;Vol 25; No 1.

8. van den Besselaar A. Accuracy, precision and quality control for point-of-care testing of oral anticoagulation. J Thromb and Thrombol. 2001;12(1):35-40.

9. Oral Anticoagulation Monitoring Study Group. Point-of-care Prothrombin Time measurement for professional and patient self-testing use. A multicenter clinical experience. Am J Clin Pathol. 2001;115:288-296.

10. Ansell J, Oertel L, Wittkowsky A. Point-of-care testing: patient self-testing and patient self-management. In: Managing Oral Anticoagulation Therapy, Clinical and Operational Guidelines, 2nd ed. Aspen Publishers; 1997:44:1-6.

11. Chiquette E, Amato M, Bussey H. Comparison of an anticoagulation clinic with usual medical care. Arch Intern Med. 1998;158:1641-1647.

12. Oberhardt B, Dermott S, et al. Dry reagent technology for rapid, convenient measurements of blood coagulation and fibrinolysis. Clin Chem. 1991;37:520-526.

13. Gosselin R, Owings J, White R, et al. A comparison of point-of-care instruments designed for monitoring oral anticoagulation with standard laboratory methods. Thromb Haemost 2000; 83:698-703.

14. Bland J, Altman D. Statistical methods for assessing agreement between two methods of clinical measurement, The Lancet. 1986;307-310.

15. Loebstein R, Kurnik D. Potential dosing errors using portable Prothrombin Time monitoring devices. Blood Coag Fibrinol. 2003;14:479-483.

16. Anderson D, Harrison L, Hirsch J. Evaluation of a portable Prothrombin Time monitor for home use by patients who require long-term oral anticoagulant therapy. Arch Intern Med. 1993;153:1441-1447.

17. Poller L, Keown N, Chauhan N, et al. European concerted action on anticoagulation (ECAA): International Normalized Ratio Variability of CoaguChek and TAS Point-of-Care Testing Whole Blood Prothrombin Time Monitors. Thromb Haemost. 2002;88:992-995.

18. Cosmi B, Gualtiero P, Moia M, et al. Accuracy of a portable Prothrombin Time monitor (Coagucheck) in patients on chronic oral anticoagulant therapy: a prospective multicenter study. Thromb Res. 2000;100:279-286.

19. Burken M, Whyte J. Home International normalized ratio monitoring: where evidence-based medicine is exemplified in the Medicare coverage process. J Thromb and Thrombol. 2002;13:5-7.

Brenda Katz, MD, and Marisa B. Marques, MD, are on staff at UAB Coagulation Service, Division of Laboratory Medicine, Department of Pathology, University of Alabama at Birmingham.

By Brenda Katz, MD, and Marisa B. Marques, MD