Clinically significant differences in the International Normalized Ratio measured with reagents of different sensitivities

`The International Normalized Ratio (INR) system was introduced a decade ag o as a way of standardizing the results of prothrombin time testing for pat ients taking oral anticoagulants. A strong emphasis has been placed upon us ing thromboplastin reagents that are very sensitive to the effects of oral anticoagulants upon the prothrombin time [i.e. reagents with low Internatio nal Sensitivity Index (ISI)]. In order to assess how well the INR system fu nctions as currently used in clinical laboratories, we compared the INRs de termined using thromboplastins of differing ISIs in samples collected durin g a large clinical trial of oral anticoagulation for atrial fibrillation (S troke Prevention in Atrial Fibrillation III trial). Frozen plasma was subje cted to prothrombin time testing using thromboplastins with ISIs ranging fr om 0.97 to 2.49. INRs were calculated using machine-specific ISIs and Westg ard's rules were followed to maintain quality control. An unanticipated coa gulometer failure allowed a determination of the effect of machine recalibr ation upon the INR of control plasmas. The correlation between each pair of INRs obtained from 1181 plasmas was high (> 0.9), but the differences betw een reagents were statistically different from zero (P<0.001 for pairwise c omparisons). Plasmas had INRs within the therapeutic range (2.0-3.0) with o ne reagent but not with another in an average of 20% of instances. Among th e 20% discordant pairings, 43% (8.5% of the total tested) showed a differen ce in INR of more than 0.2 INR units above or below the target range. Low I SI thromboplastins did not perform better in this pairwise comparison than other reagents or the locally determined INR. Recalibration of a coagulomet er resulted in a significant change in the INRs obtained from control plasm as (P < 0.0001), which confirms and extends the observations of other autho rs concerning the sensitivity of the INR to coagulometer-related variables. There was a clinically significant difference in the INRs obtained with di fferent thromboplastins, and low ISI reagents did not perform better than o thers. Since the risk of thrombosis rises sharply below the lower limit of the currently recommended target ranges, consideration should be given to n arrowing the recommended range, or advising clinicians to aim for its mid-p oint. These findings illustrate the difficulties in imposing standardizatio n upon coagulation testing after a test is in widespread use. Blood Coag Fi brinol 10:215-227 (C) 1999 Lippincott Williams & Wilkins.