The International Normalized Ratio (INR) for Monitoring Warfarin Therapy: Reliability and Relation to Other Monitoring Methods

Methods

One hundred plasma samples were collected from 79 patients between 8 April and 20 May 1992 at seven weekly sessions of the outpatient anticoagulant clinic of the University of California, San Diego, Medical Center. All samples except one were obtained from patients who had taken warfarin for longer than 14 days. All patients gave informed consent. After the Coumatrak prothrombin time was measured, 4.5 mL of venous blood was collected in a siliconized BD Vacutainer tube (Becton-Dickinson; Rutherford, New Jersey) containing 0.5 mL of 3.2% sodium citrate and sent to the medical center's Special Coagulation Laboratory. All subsequent handling and testing were done by the same experienced medical technologist who followed the routine techniques used in this laboratory. Platelet-poor plasma was prepared within 2 hours of the time of sampling by centrifugation at 3000 rpm for 15 minutes at room temperature. In the early afternoon, when all samples for a given day had been processed, prothrombin times were done on each sample with six thromboplastin reagents. The plasma samples were then stored at −70°C for subsequent measurement of the P&P percent activity, specific prothrombin (factor II) activity, and native prothrombin antigen.

This protocol was adopted after a preliminary experiment was done in which blood samples from five patients were divided into two aliquots and allowed to stand for either 2 hours or 5 hours before separating the plasma and measuring the prothrombin time. The prothrombin time values, as measured with three different thromboplastins, did not differ significantly for the paired samples (data not shown). In addition, virtually identical P&P test results were obtained on fresh plasma samples from five patients and on these samples after they had been stored frozen for 1 week.

Control Pooled Plasma

The normal plasma used to obtain the prothrombin time ratio for plasma prothrombin times was the routine control pooled plasma of the Special Coagulation Laboratory. This plasma is prepared by pooling plasma obtained by the technique described above from blood collected in 1/10 volume of a buffered citrate anticoagulant (0.06 mol/L sodium citrate plus 0.04 mol/L citric acid) from between 15 and 20 healthy donors. It is stored in small aliquots at −70°C. The prothrombin time and activated partial thromboplastin time of the plasma from each donor had to be within 2 standard deviations (SD) of the mean of the individual plasma samples from the donors that were included in the laboratory's previous batch of control pooled plasma.

Coagulation Assays

Capillary Blood Prothrombin Time (Coumatrak)

This test was done by two experienced practitioners on finger-stick blood. The thromboplastin used in the cartridge for the Coumatrak instrument had an ISI of 2.04. The value for a normal prothrombin time was set by the instrument at 12 s, and the prothrombin time ratio and INR were provided by the instrument as a direct readout.

Prothrombin Time Assays

The prothrombin times were done on an Electra 800 (Medical Laboratory Automation; Pleasantville, New York). The six thromboplastins, made available by sales representatives of the manufacturers, were Thromboplastin C (Dade C; ISI, 2.88), Thromboplastin C-plus (Dade C-plus; ISI, 2.03), and Thromboplastin IS (Dade IS; ISI, 1.27) from Dade division, Baxter Scientific Products, Miami, Florida; Simplastin Excel (Excel; ISI, 1.99 or 2.04 for two lots used) and Simplastin Excel-S (Excel-S; ISI, 1.34) from Organon Teknica, Durham, North Carolina; and Ortho Brain Thromboplastin (Ortho; ISI, 2.30) from Ortho, Raritan, New Jersey.

Prothrombin time ratios for assays done with each thromboplastin were calculated using a normal value obtained with each thromboplastin for each run with the control pooled plasma. The mean clotting times (± SD, n = 7) of the control pooled plasma with the different thromboplastins were as follows: Dade C, 11.8 ± 0.2 s; Dade C-plus, 12.2 ± 0.2 s; Dade IS, 13.7 ± 0.2 s; Excel, 11.7 ± 0.1 s; Excel-S, 13.4 ± 0.2 s; and Ortho, 12.7 ± 0.3 s. The INR was calculated as INR = PTR^ISI with the manufacturer's value for the photo-optical ISI as given in the package insert.

Prothrombin-Proconvertin Test

This test was done according to the manufacturer's instructions with Simplastin A (Organon Teknika), which is the source not only of the thromboplastin (ISI, 1.70) and CaCl₂ but of an optimal concentration of factor V and fibrinogen. The test plasma was diluted in saline 1:10, which sensitized the assay to small changes in the prothrombin, factor VII, and factor X activity of the sample. Clotting times were determined with an ST-4 semi-automated coagulometer (Diagnostica Stago; Parsippany, New Jersey). A log-log plot of clotting times against dilutions of control pooled plasma in saline between 1:10 and 1:80 yielded linear standard reference curves. The clotting time of the 1:10 dilution of the control pooled plasma was assigned a value of 100% P&P activity, and clotting times of test samples were converted into percent of the control pooled plasma activity.

Specific Prothrombin Assay (Factor II)

Prothrombin activity was assayed by a one-stage assay in which a mixture of 100 µL of a prothrombin-depleted human serum/barium-adsorbed bovine plasma reagent [8] and 100 µL of a 1:10 to 1:40 dilution of test plasma were clotted by the addition of 200 µL of a thromboplastin C reagent containing CaCl₂. Clotting times were converted to percent normal plasma prothrombin activity from a log-log standard curve prepared with dilutions of control pooled plasma.

Native Prothrombin Antigen Assay

Plasma native prothrombin antigen concentration was measured with native prothrombin antigen enzyme immunoassay kits provided by Organon Teknika. Color was measured at 450 nm with a Thermomax enzyme-linked immunosorbent assay reader (Molecular Devices; Menlo Park, California).

Results

Relations between the Prothrombin Time Ratio and International Normalized Ratio Values Obtained with Six Thromboplastins and with the Coumatrak Monitor

The thromboplastins used in this study had ISI values ranging from 1.27 to 2.88 (see Methods). Dade Thromboplastin IS (Dade IS; ISI, 1.27) was the most sensitive of the thromboplastins and was therefore selected as the standard against which we compared data obtained with the other thromboplastins.

In an initial analysis, the prothrombin time values obtained with the Coumatrak capillary monitor and with Dade IS were plotted both as prothrombin time ratios Figure 1, A) and as INR values (Figure 1, B). Because of the different ISI values for the Coumatrak monitor and Dade IS, the slope of the linear regression line for the data plotted as prothrombin time ratios, 0.36 (r = 0.87, n = 100), clearly differed from 1. The slope of the regression line of the data plotted as INRs, 0.84 (r = 0.90, n = 100), was closer to 1, which is the theoretically expected slope when plotting INR values against each other.

The linear regression lines obtained by plotting the prothrombin time ratios obtained from all of the other thromboplastins and from the Coumatrak monitor as a function of the prothrombin time ratios obtained with Dade IS are shown in Figure 2 A. The slopes of the linear regression lines varied between 0.38 to 0.88. The individual prothrombin time ratios obtained for a given sample varied with the sensitivity of the thromboplastin. Thus, a sample that gave a prothrombin time ratio with Dade C (ISI, 2.88) of 1.5 gave a prothrombin time ratio with Dade IS (ISI, 1.27) of 2.3.

The prothrombin time ratios that yielded the regression lines of Figure 2A were then converted into INRs and plotted against the INRs obtained with Dade IS Figure 2B. The slopes of the resultant linear regression lines varied between 0.96 to 1.97, with a correlation coefficient range of 0.96 to 0.98. Thus, converting prothrombin time ratios to INRs—by use of the ISI given in the thromboplastin package inserts and clotting times obtained on the same sample with the same photo-optical instrument—failed to yield an expression of prothrombin time test results independent of the thromboplastin used.

An INR range between 2.0 and 3.5 has been recommended as the therapeutic range for patients receiving warfarin in the United States [9], whereas a higher range with upper limits between 4.5 and 4.8 are used in some European countries [10, 11]. Table 1 shows the relation of selected INR values obtained with Dade IS, with the other thromboplastins, and with the Coumatrak monitor as calculated from the linear regression lines of Figure 2B. At a Dade IS INR of 2.0, the INRs for the plasma prothrombin times obtained with the other thromboplastins were all also essentially 2.0, whereas the INR obtained with the Coumatrak monitor was 2.3. At the upper limit of recommended therapeutic ranges, the Coumatrak monitor INRs corresponded closely to the INRs obtained from plasma prothrombin times done with Dade IS. In contrast, the INRs obtained with three thromboplastins with high ISI values substantially exceeded recommended upper limits as defined with Dade IS thromboplastin.

Relation between Prothrombin-Proconvertin Percent Activity and International Normalized Ratio Values

For the P&P test, a therapeutic range of 10% to 30% activity was originally recommended [12], and an optimal level of 15% was later recommended for patients with arterial diseases [13]. In our initial analysis, the reciprocal of the INR values obtained with Dade IS were plotted as a function of percent P&P activity Figure 3A. A linear relation between 1/INR and percent P&P activity did not hold for P&P values of less than 10%, reflecting the fact that clotting times in the P&P test are converted to percent activity from a log-log plot. The linear regression lines obtained from similar plots of the reciprocal of the INR values obtained with all thromboplastins (excluding Coumatrak) are plotted against percent P&P activity in Figure 2B. Plotting the INR as a reciprocal markedly diminished the ability to detect the variation in slope of the linear regression lines of the INRs so readily recognized in Figure 2B. This change stems at least partly from the poorer correlation coefficients obtained for the regression lines of Figure 3B of between 0.84 to 0.94 (n = 90) than for the correlation coefficients of the linear regression lines of Figure 2B. The relation between percent P&P activity and selected mean INR values, as calculated from the regression lines of Figure 3B, is given in Table 2. A P&P activity range of 30% to 13% corresponded to an INR range of 2.0 to 3.0.

Relations among Values for International Normalized Ratios, Native Prothrombin Antigen, and Specific Prothrombin Activity

Because determinations of residual specific prothrombin activity [14] and native prothrombin antigen [7, 15, 16] have been proposed as being better techniques for monitoring oral anticoagulant therapy than the prothrombin time [14, 15], we also examined the relation between these measurements and INR values. We confirmed an earlier report of a good correlation between residual plasma native prothrombin antigen levels as measured by enzyme immunoassay and residual specific prothrombin activity as measured by a one-stage coagulation method [17] (r = 0.92, n = 89; data not shown).

The reciprocal of Dade IS INRs is plotted against plasma native prothrombin antigen levels in Figure 4A and against plasma prothrombin activity in Figure 5A. The linear regression lines from similar plots with each thromboplastin (excluding Coumatrak) are plotted in Figure 4B and Figure 5B. The range of correlation coefficients for these regression lines were as follows: from 0.66 to 0.83 for native prothrombin antigen and from 0.76 to 0.88 for specific prothrombin activity. Table 2 summarizes, as calculated from Figure 4B and Figure 5B, the relation between selected mean INR values and residual native prothrombin measured as antigen and activity. A mean INR range of 2.0 to 3.0 corresponded to between about 40% to 20% residual native plasma prothrombin.

Discussion

Although reporting prothrombin times as an INR was proposed as a way to ensure reliable oral anticoagulant therapy [1], INR values obtained under different laboratory conditions may not be equivalent. Variation has been attributed primarily to the use of different automated coagulation analyzers [18-22]. In our study, in which the same automated instrument (Electra 800) was used, converting to INRs the prothrombin time ratios obtained on the same plasma samples with six thromboplastins of varied sensitivity (ISI values, 1.27 to 2.88) did not reliably yield equivalent values (Figure 2B.

The variation in INR values for the thromboplastins with lower sensitivity (and therefore the higher ISI values) at least partly reflects the exponential function of the ISI in calculating an INR, which magnifies errors derived from any source as the ISI value increases. In contrast, two high-sensitivity thromboplastins, with ISI values approaching 1.0, gave virtually identical INR values (Figure 2B and Table 1).

It was somewhat reassuring that the plasma prothrombin time INRs obtained with all thromboplastins were acceptable in the lower portion of a recently recommended therapeutic INR range of 2.0 to 3.5 [9]. Unfortunately, the same was not true for values in the higher portion of the therapeutic range. At an INR of 3.5 obtained with the thromboplastin with the highest sensitivity (ISI, 1.27), three low-sensitivity thromboplastins yielded INRs exceeding this upper limit of the therapeutic range (Table 1). At an INR of 4.5 obtained with the highest-sensitivity thromboplastin, which is an INR within some accepted European therapeutic ranges for arterial thromboembolic disease [10, 11], the same three low-sensitivity thromboplastins yielded INRs that would definitely have led clinicians to reduce the dosage of warfarin prescribed (Table 1).

Thus, our data provide additional support for the recommendation [23] that all laboratories use thromboplastins with high sensitivity (low ISI values) for doing prothrombin times. Until this recommendation is widely adopted, we believe that clinicians monitoring therapy with plasma prothrombin times should, whenever possible, refer patients to a laboratory using a low ISI thromboplastin. This suggestion is particularly important for those patients in whom antithrombotic protection requires keeping the INR at the upper limit of a therapeutic range.

Capillary blood prothrombin time monitors are being increasingly used in outpatient anticoagulation clinics, whereas plasma prothrombin times are virtually always used to monitor therapy in hospitalized patients or patients being followed by private physicians. In our study, the INRs obtained on capillary blood correlated well with the INRs obtained from plasma prothrombin times done with high-sensitivity thromboplastins except for a positive bias at the lower limit of the therapeutic range (Figure 2B and Table 1), which others have also noted [24, 25]. As long as the clinician is aware of this positive bias, serious problems should not arise from reporting results as INR values within an institution using the Coumatrak capillary prothrombin time monitor for outpatients and a plasma prothrombin time with a high sensitivity thromboplastin for inpatients.

The P&P test was introduced in Scandinavia in the 1950s [6] to monitor oral anticoagulant therapy with a recommended therapeutic range of 10% to 30% [12]. The P&P test may be valuable in monitoring anticoagulant therapy in patients with a strong lupus anticoagulant prolonging a patient's baseline prothrombin time. Because the INR is determined from a prothrombin time ratio obtained by dividing the patient's prothrombin time on warfarin by a normal reference prothrombin time, not by the patient's baseline prothrombin time, the patient's INR will reflect a combined effect of warfarin and the underlying lupus anticoagulant on the patient's prothrombin time. Thus, the use of the usual INR therapeutic range could result in inadequate anticoagulation.

The percent P&P activity is a better measure of the effect of warfarin on the vitamin K-dependent factors affecting the prothrombin time in a patient with a strong lupus anticoagulant, because the 1/10 dilution of the test sample in the P&P test eliminates the inhibitory effect of a lupus anticoagulant upon the clotting time. As done in our study with the commercial reagent available in the United States in 1992, an 11% to 30% P&P range corresponded to mean INR values of 2.0 to 3.5 (Table 2). Thus, relating a P&P test result in a patient with lupus anticoagulant to the patient's corresponding INR value should allow the clinician to select a target INR for subsequent monitoring.

Hemorrhage and thrombosis in patients receiving an oral anticoagulant have been reported to be more closely related to specific plasma prothrombin activity [14] and to residual native prothrombin antigen [15] than to the prothrombin time test result. As first described by others [16, 17], native prothrombin antigen levels and specific plasma prothrombin activity were closely correlated. In our study, an INR range of 2.0 to 3.5 corresponded to a reduced normal plasma prothrombin content of about 40% to 15% of our normal reference pool when measured both as native prothrombin antigen and as specific residual prothrombin activity (Table 2).

Recent in vivo [8] and earlier in vitro [26] experiments provide convincing evidence that the antithrombotic effect of warfarin is strongly dependent on a reduction of plasma prothrombin activity. Therefore, we interpret our data to mean that, under most circumstances, reducing plasma prothrombin activity to 20% to 40% of normal is both required and sufficient to support the antithrombotic efficacy of warfarin. This information should be of value to a physician who may wish to confirm, by a specific prothrombin activity assay, the adequacy of oral anticoagulant therapy in particular situations as, for example, in deciding when after starting warfarin it is safe to withdraw heparin in a patient with extensive venous thrombotic disease.

Our data add to the evidence that the clinician may not yet assume that the use of the INR normalizes prothrombin time results obtained with different thromboplastins. A high-sensitivity thromboplastin with a low ISI value will give a more reliable INR, and, whenever possible, clinicians monitoring patients with plasma prothrombin times should refer patients to laboratories using such a thromboplastin.