Discrepancies in international normalized ratio results between instruments: a model to split the variation into subcomponents.

BACKGROUND Observed differences between results obtained from comparison of instruments used to measure international normalized ratio (INR) have been higher than expected from the imprecision of the instruments. In this study the variation of these differences was divided into subcomponents, and each of the subcomponents was estimated. METHODS Blood samples were collected at 4 different patient visits from each of 36 outpatients who were receiving warfarin treatment and were included in the study. INR was determined on 1 laboratory instrument (STA Compact®) and 3 point-of-care instruments (Simple Simon®PT, CoaguChek®XS, and INRatio™). All 4 INR instruments were compared in pairs. Linear regression was used to correct for systematic deviations. The remaining variation of the differences was subdivided into between-subject, within-subject, and analytical variation in an ANOVA nested design. RESULTS The mean difference between instruments varied between 1.0% and 14.3%. Between-subject variation of the differences (expressed as CV) varied between 3.3% and 7.4%, whereas within-subject variation of the differences was approximately 5% for all 6 comparisons. The analytical imprecision of the differences varied between 3.8% and 8.6%. CONCLUSIONS The differences in INR between instruments were subdivided into calibration differences, between- and within-subject variation, and analytical imprecision. The magnitude of each subcomponent was estimated. Within results for individual patients the difference in INR between 2 instruments varied over time. The reasons for the between- and within-subject variations of the differences can probably be ascribed to different patient-specific effects in the patient plasma. To minimize this variation in a monitoring situation, each site and patient should use results from only 1 type of instrument.