Data documenting the performance of the PT/INR line correction method for reconciling INR discrepancies between central laboratory coagulation analyzers using different thromboplastins during the evaluation of a portable Coagulometer

The data presented here was produced as part of an evaluation of the performance of the CoaguChek XS point-of-care coagulation analyzer, which is discussed in the research article “POCT PT INR – Is it adequate for Patient Care? A Comparison of the Roche Coaguchek XS vs. Stago Star vs. Siemens BCS in Patients Routinely Seen in an Anticoagulation Clinic” (Baker et al., in press) [1]. An effort to reconcile discrepancies in the patient INR result distributions from different central lab instruments (Stago Star and Siemens BCS) with the PT/INR line method is described (Poller et al., 2010, 2011; Ibrahim et al., 2011) [2], [3], [4]. While regression analysis of the ECAA Poller calibrant data provided a linear PT/INR line for all methods, Pearson's chi-squared and one-way repeated measures ANOVA analyses showed that central lab INR measurements continued to exhibit measurement site dependence after the PT/INR line correction was applied. According to paired t-test analysis, only the human thromboplastin dependent methods (CoaguChek XS and Siemens BCS both before and after PT/INR line correction) showed statistically significant agreement (p-value >0.05).

Described in the methods section Data source location UTMB (Galveston, TX), ARUP (Salt Lake City, UT), CPL (Austin, TX) Data accessibility Raw sample data in Table 1 and ECAA standards data Ref. [1] Value of the data The data indicate that although agreement between central lab methods was improved by the PT/ INR line correction, the INR results continued to exhibit statistically significant site dependence.
The raw data here may provide insight pertinent to INR standardization efforts. Although the PT/INR line correction method is designed to minimize INR variation due to differences in instrumentation, thromboplastin, and local ISI correction strategy, the data indicate a need for method innovation. Additionally, a method [5] which allows for standardization of INR results obtained from disparate sample types (e.g. whole blood and plasma) would be beneficial.

Data
Fig . 1 shows the European Concerted Action on Anticoagulation (ECAA) Poller calibrant-generated PT/INR lines for the Stago Star Evolution instruments at the University of Texas Medical Branch (UTMB, Fig. 1A), the Associated Regional and University Pathologists, Inc. (ARUP, Fig. 1B), and the Siemens BCS XP at Clinical Pathology Laboratories, Inc. (CPL, Fig. 1C). The calibrant data provide a linear fit to the model for all methods, as evidenced by their coefficients of determination (R 2 4 0.998). The generation of the calibrations plots and PT/INR line method is described briefly in the methods section and detailed more fully by Poller et al. [2][3][4].
The PT/INR line fit was employed to correct central lab INR results for 100 warfarin therapy and 20 control samples examined as part of a CoaguChek XS point-of-care device study, as detailed in Table 1 [1]. Fig. 2A shows the frequency distribution of the uncorrected INR values for the central lab and CoaguChek XS coagulometers. Visual inspection of the histograms in Fig. 2A    (the POCT device is only approved for analysis of whole blood samples thus could not be calibrated using the plasma-based ECAA Poller standards [5]). Visual inspection of Fig. 2B reveals no obvious improvement in the net agreement between methods, however, Pearson's chi-squared testing of the frequency distribution data shown in Table 2    showed that although the PT/INR line correction method failed to provide inter-method agreement at the p ¼0.05 threshold, improved agreement was observed.
One-way repeated measures ANOVA was performed to investigate correspondence between methods both before and after PT/INR line correction. Because the raw and PT/INR line corrected data failed the Mauchly's Test for Sphericity (p ¼6.17e-59 before correction and p ¼1.30e-37 after correction), a Greenhouse-Geisser correction was applied to all ANOVA models. The p-values for one-way repeated measures ANOVA of the raw and corrected INR data for all methods were 3.52e-24 and 8.91e-14, respectively, indicating improved but statistically significant inter-method variation. In order to identify the source of the variation, paired t-test were performed both before and after PT/INR line correction, as shown in Table 3. While the p-values show that inter-method agreement improved for all methods following PT/INR correction, only the human thromboplastin based methods (BCS and uncorrected CoaguChek XS) showed agreement exceeding the p ¼0.05 threshold.

Experimental design, materials and methods
The study protocol was approved for verbal informed assent by the Institutional Review Board of the University of Texas Medical Branch (UTMB) and Roche Diagnostics USA (Indianapolis, IN). Warfarin therapy patients were recruited from March to May 2015 during routine visits to monitor their PT/INR at UTMB's anti-coagulation clinic and all POCT measurements were performed on the  The 20 normal (non-warfarin) samples and 100 warfarin treatment samples were included in this investigation. All warfarin therapy samples were obtained from patients who have been on stabilized warfarin therapy for at least one month to allow both short half-life (Factor VII) and long half-life vitamin K-dependent clotting factors to attain therapeutic equilibria. Patients included in this study are expected, based on their medical history, to be negative for anti-phospholipid antibodies, have hematocrit values in the 25-55% range, and have taken no additional anticlotting medication such as aspirin.

PT INR measurement
PT INR measurements were performed using the POCT CoaguChek XS system (UTMB anticoagulation clinic), core laboratory Star Evolution (at UTMB and ARUP), and BCS XP (at CPL). INR measurements were obtained for all 120 samples, using all methods except for one BCS sample. The CoaguChek XS system detects coagulation in a drop of whole blood while the central lab methods employ citrated plasma (as described above). The CoaguChek XS and the BCS XP use a human recombinant thromboplastin while the Star Evolution uses rabbit brain thromboplastin (STA Neoplastin). A single CoaguChek XS device and only one lot of test strips were used for this study.

Calibrant standards and controls
Poller calibrants (European Concerted Action on Anticoagulation or ECAA PT/INR plasma set, batch 1591-6) were purchased from Hart Biologicals Ltd. (Hartlepool, England). The frozen standards and controls were stored and prepared according to vendor specifications prior to analysis in duplicate on the three central laboratory analyzers.

PT/INR line correction
INR data from the Stago and BCS XP were corrected using the regression equation generated from ln(PT observed) vs ln(INR certified) calibration plots obtained from central lab analysis of the ECAA Poller calibrant [2][3][4]. The certified Poller INR values were obtained from the manufacturer and the values for the specified thromboplastins were employed (See Table 2A Table 1) were generated using the rearranged equation x ¼(−b/m) þ(1/m)y by inputting y ¼ln(sample PT) and raising x ¼ln(INR corrected) to the natural exponent (e ln(INR corrected) ¼INR corrected).

Statistical Analysis
Pearson's chi-square testing, paired t-tests, and one-way repeated measures ANOVA were performed using R 3.4.