Patients receiving warfarin and other anti-vitamin K agents are at risk of bleeding if the oral anticoagulant dose is too high, or thromboembolism if the oral anticoagulant dose is too low. Proper oral anticoagu-lant dosing requires an accurate prothrombin time (PT) measurement to generate an international normalized ratio (INR). Point-of-care (POC) instruments using fin-gerstick whole blood are generally convenient, easy to use, preferred by patients, and generate immediate results, allowing more rapid adjustments of anticoagu-lation therapy, compared with complex, central laboratory coagulation analyzers utilizing citrated plasma.1
Our objective was to investigate the accuracy of two widely utilized POC instruments by comparing the results generated from whole blood with results generated by a central laboratory instrument using citrated plasma.
Fifty-two consecutive patients receiving warfarin for long-term anticoagulation of atrial fibrillation, venous thromboembolism, or mechanical prosthetic valves as outpatients at a suburban satellite laboratory under-went venipuncture to obtain whole blood for POC instrument measurement of INR and to generate cit-rated plasma for central laboratory measurement of INR. Previously, we and others2 obtained similar
results from POC monitors employing both fingerstick and syringe-drawn venous whole blood. The study was conducted in accordance with institutional policies.
Immediately following venipuncture, whole blood was used to generate INR values using the CoaguChek S monitor, international sensitivity index (ISI)⫽2.0 (Roche Diagnostics, Indianapolis, IN, USA) and the Protime Microcoagulation System monitor, ISI⫽1.0 (International Technidyne Corporation [ITC], Edison, NJ, USA). Two devices from each manufacturer were employed simultaneously to generate INR results for each type of POC instrument. Concurrently, INR results were generated from citrated plasma with a central laboratory instrument, the STA-R®coagulation
analyzer (Diagnostica-Stago, Parsippany, NJ, USA), which utilized Neoplastine CI⫹PT reagent (ISI⫽1.29).
Point-of-care (POC) versus central laboratory instrumentation
for monitoring oral anticoagulation
David M Dorfmana, Ellen M Goonana, M Kay Boutiliera, Petr Jarolima, Milenko Tanasijevica
and Samuel Z Goldhaberb
Abstract: Point-of-care (POC) instruments employing fingerstick whole blood to
monitor patients treated with warfarin are a popular alternative to complex, central laboratory coagulation analyzers utilizing citrated plasma derived from venipuncture. We investigated the accuracy of two widely utilized POC instruments for oral anticoagulation monitoring compared with a central laboratory instrument. Instrument-to-instrument variation differed for the two POC instruments, which correlated with the central laboratory instrument, but exhibited positive bias of 0.24–0.35 INR units. Positive bias increased as the INR values increased. We conclude that clinicians should exercise caution when interpreting results generated by POC monitors, particularly at high INR values. A high POC measurement of INR does not necessarily warrant decreasing the warfarin dose. Instead, a predefined cut-off range for high INR values generated by POC instruments should mandate confirmatory testing with central laboratory instrumentation.
Key words: international normalized ratio (INR); prothrombin time (PT); warfarin
aDepartment of Pathology and bCardiovascular Division, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA
Address for correspondence: David M Dorfman, Department of Pathology, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115, USA. Tel: ⫹1 617 732 7581; Fax: ⫹1 617 731 4872; E-mail: firstname.lastname@example.org
Data were analyzed utilizing MedCalc version 220.127.116.11 statistical software (www.medcalc.be) for correlation analysis, least squares linear regression analysis, and to generate Bland and Altman plots.3,4
First, we correlated the data generated from each pair of POC instruments (Figure 1A). INR results generated using the two Roche CoaguChek instru-ments exhibited better correlation (correlation coefficient⫽0.9817, p⬍0.0001) than the two ITC instruments (correlation coefficient⫽0.8801, p⬍0.0001; Table 1). We also compared the results from the two sets of POC instruments (correlation coefficient⫽0.8968, p⬍0.0001, Table 1).
Next, we compared the INR values generated by the two sets of POC instruments with values obtained from the central laboratory coagulation analyzer
(Figure 1B). The mean INR measured with the STA-R central laboratory coagulation analyzer was 2.1, with a standard deviation (SD) of 0.6 and a range from 1.0 to 3.6. The POC instruments had higher mean INR values, 2.4 for the CoaguChek instruments and 2.5 for the ITC instruments, both with a higher SD (of 0.9) than the central laboratory coagulation analyzer (Table 1). The range of INR values was greater for the POC instruments: from 0.9 to 5.5 for the CoaguChek instruments and from 0.8 to 5.4 for the ITC instru-ments (Table 1). The correlation coefficients for the POC instruments compared with the central labora-tory STA-R instrument were comparable for the two manufacturers’ instruments: 0.9052, p⬍0.0001 for the CoaguChek instruments and 0.9067, p⬍0.0001 for the ITC instruments (Table 1).
Figure 2 depicts the differences in INR measure-ment between the POC instrumeasure-ments and the central laboratory instrument, using the method of Bland and Altman. The mean difference between the INR results generated with the CoaguChek instruments and the
Figure 1 (A) Comparison of CoaguChek S POC instrument INR values (left) and ITC POC instrument INR values (right) for two different devices. The correlation coefficient for the CoaguChek S comparison is 0.9817, p⬍0.0001, and for the ITC comparison it is 0.8801, p⬍0.0001 (Table 1). (B) Correlation of CoaguChek S POC instrument (left) and ITC POC instrument (right) INR values with those generated by the STA central laboratory instrument. The correlation coeffi-cients are 0.9052, p⬍0.0001, and 0.9067, p⬍0.0001, respectively (Table 1).
Table 1 Comparison of INR data from POC and central laboratory instrumentation. Correlation coefficient
versus STA-R Mean difference versus
Instrument Samples Range Mean⫾SD (central lab) STA-R (central lab)
CoaguChek 1 52 0.9–5.5 – 0.9817a, p⬍0.0001, 0.11a, 2 SD⫽ ⫺0.26 –⫹0.47 versus 95% CI⫽0.9682–0.9895 CoaguChek 2 ITC 1 versus 49 0.8–5.4 – 0.8801a, p⬍0.0001, 0.14a, 2 SD⫽ ⫺0.98 –⫹0.70 ITC 2 95% CI⫽0.7959–0.9309 CoaguChek 49 0.8–5.5 – 0.8968b, p⬍0.0001, versus ITC 95% CI⫽0.8233–0.9408 CoaguChek 104 0.9–5.5 2.4⫾0.9 0.9052, p⬍0.0001, ⫹0.24, 2 SD⫽ ⫺0.63 –⫹1.12 1⫹2 95% CI⫽0.8631–0.9349 ITC 1⫹2 98 0.8–5.4 2.5⫾0.9 0.9067, p⬍0.0001, ⫹0.35, 2 SD⫽ ⫺0.53 –⫹1.23 95% CI⫽0.8639–0.9365 aComparison between POC instruments from same manufacturer.
bComparison between POC instruments from different manufacturers.
Figure 2 Bland and Altman plots of INR measurements generated by the CoaguChek POC instruments (upper left) and ITC POC instruments (lower left) compared with the STA central laboratory coagulation analyzer. Both POC instruments exhibit a positive measurement bias compared to the central laboratory instrument. Also shown is the regression analysis for the same data (upper and lower right), showing a positive slope for the least squares linear regression analysis for both POC instruments, indicating that there is increasing positive bias with higher INR values (see text).
central laboratory STA-R instrument was⫹0.24 INR units, while the ITC instruments had a mean differ-ence of⫹0.35 INR units compared with the central laboratory instrument (Table 1). For most specimens, the CoaguChek INR results differed from the central laboratory INR results by⫺0.3 to⫹0.6 INR units, whereas the ITC INR results differed from the central laboratory INR results by a wider margin, ranging from⫺0.3 to⫹0.8 INR units. Both types of POC instruments exhibited increasing positive bias com-pared with the central laboratory instrument at higher INR values. The slope of the least squares linear regression analysis for the CoaguChek instruments compared with the central laboratory instrument was⫹0.42, p⬍0.0001, and the slope for the ITC instruments compared with the central laboratory instrument was⫹0.46, p⬍0.0001.
Our principal finding is that POC measurements of INR exhibit positive bias as INR values increase, despite overall good correlation with central labora-tory instrumentation for the entire range of INR meas-urements. This observation has potentially profound clinical implications. Without awareness of positive bias, clinicians might decrease the dose of oral antico-agulation based upon POC measurements yielding high INRs. This could lead to an increase in throm-boembolic events in cases when the POC INR meas-urement was spuriously high compared with the central laboratory. Although there is overall positive bias of INR values generated with POC versus central laboratory instrumentation, in some instances negative bias of INR values was also observed. Additional studies of a larger patient sample may be helpful to confirm these observations.
Our results are consistent with those of previous investigators who observed positive bias of INR val-ues in studies comparing POC versus central labora-tory instrumentation.5–7 Nevertheless, few clinicians
are aware of the implications and ongoing problem of obtaining high INRs on POC instruments compared with central laboratory instrumentation. Anticoagulation services rely upon accurate determinations of the INR in order to provide appropriate adjustment of warfarin dosing.8 In a consortium of Italian anticoagulation
clinics, with 2011 patient-years of treatment, 70 thrombotic events (3.5 per 100 patient years) were recorded in 67 patients. The frequency of thrombotic events increased as the INR decreased.9
The European Concerted Action on Anticoagulation (ECAA) developed approaches for the calibration of POC INR instruments utilizing citrated fresh plasma or lyophilized plasma; however, this approach does not work for all POC instruments.10,11 Ideally,
laboratories that employ POC instruments to monitor
anticoagulation would be able to calibrate these instru-ments as well as central laboratory instruinstru-ments employed for anticoagulation testing. However, even with calibration, there is increasing deviation of POC and central laboratory INR values with a positive bias for POC instruments, particularly when the INR exceeds 3.0.10,11
Based upon our findings, we have instituted a pro-tocol in which we do not report POC INR values that equal or exceed 4.0. In such cases, we require venipuncture to obtain citrated plasma for INR deter-mination by our central laboratory. Since instituting this program, we have performed 315 POC INR meas-urements, with 23 results (7.3%) that were ⱖ4.0 (mean⫾SD⫽5.2⫾1.3, range⫽4.0–8.0, with corresponding central laboratory mean⫾SD ⫽ 3.6⫾0.7, range 2.3–5.2), which exhibited a mean bias of⫹1.7 INR units compared with central labora-tory findings (2 SD⫽ ⫺0.9 –⫹4.3).
INR values generated by POC monitors exhibit posi-tive bias for INR values at the high end of the antico-agulation range. Clinicians should exercise caution when interpreting high INR results generated by POC monitors due to possible important discrepancies with plasma-based methods. Insisting upon a predeter-mined INR cut-off value for mandatory venipuncture and central laboratory INR determination may potentially decrease the frequency of avoidable thromboembolic events and improve patient safety. If there are clinical practices in which POC testing is used exclusively to monitor anticoagulation therapy, it would be useful to see whether this results in higher rates of thromboembolic complications, as we would predict.
1 Zimmerman CR. The role of point-of-care anticoagulation monitoring in arterial and venous thromboembolic disorders.
J Thromb Thrombolysis 2000; 9: 187–98.
2 Van den Besselaar AMHP, Meeuwisse-Braun J, Schaefer-van Mansfeld H et al. A comparison between capillary and venous blood international normalized ratio determinations in a portable prothrombin time device. Blood Coagul Fibrinolysis 2000; 11: 559–62.
3 Bland JM, Altman DG. Statistical method for assessing agree-ment between two methods of clinical measureagree-ment. Lancet 1986; i: 307–10.
4 Bland JM, Altman DG. Measuring agreement in method com-parison studies. Stat Methods Med Res 1999; 8: 135–60. 5 Gosselin R, Owings JT, White RH et al. A comparison of
point-of-care instruments designed for monitoring oral antico-agulation with standard laboratory methods. Thromb Haemost 2000; 83: 698–703.
6 Van den Besselar AMHP. A comparison of INRs determined with a whole blood prothrombin time device and two interna-tional reference preparations for thromboplastin. Thromb
Haemost 2000; 84: 410–12.
7 Kemme MJB, Faaij RA, Schoemaker RC et al. Disagreement between bedside and laboratory activated partial thromboplas-tin time and international normalized ratio for various novel anticoagulants. Blood Coagul Fibrinolysis 2001; 12: 583–91. 8 Grasso-Correnti N, Goldszer RC, Goldhaber SZ. The critical
pathways of an anticoagulation service. Critical Pathways in
Cardiology 2003; 2: 41–45.
9 Palareti G, Manotti C, D’Angelo A et al. Thrombotic events dur-ing oral anticoagulant treatment: results of the inception-cohort,
prospective, collaborative ISCOAT Study. ISCOAT study group (Italian Study on Complications of Oral Anticoagulant Therapy). Thromb Haemost 1997; 78: 1438–43.
10 Poller L, Keown M, Chauhan N et al. European concerted action on anticoagulation (ECAA): an assessment of a method for ISI calibration of two whole blood point-of-care PT moni-tor systems based on lyophilized plasmas using whole blood equivalent PT. J Thromb Haemost 2002; 1: 766–72.
11 Poller L, Keown M, Chauhan N et al. European concerted action on anticoagulation: comparison of fresh plasma and whole blood multicentre ISI calibrations of CoaguChek Mini and TAS PT-NC whole blood prothrombin time point-of-care monitors. Thromb Haemost 2002; 87: 859–66.