AIMS--To examine the magnitude of thromboplastin and coagulometer interactions on the precision of International Normalised Ratio (INR) values when the manufacturers' recommended instrument specific International Sensitivity Index (ISI) values are adopted for the INR calculation. METHODS--The variability of INR values obtained from four automated phototopical coagulometers frequently used in North American laboratories was studied. When used with five commercial thromboplastins of moderate to high sensitivity (ISI values 0.92-1.97), 20 prothrombin time results were generated for each of a population of 98 patients on established warfarin treatment. RESULTS--The mean INR values of the patients ranged from 2.05 to 2.81, depending on which reagent/coagulometer combination was used. Within patient variation increased as the median INR value increased. The mean coefficient of variation of within patient INR values was 10%; the mean coefficient of variation of the prothrombin time results in seconds and prothrombin time ratio were 21 and 18%, respectively. CONCLUSIONS--There was considerable bias in the estimated ISI values of the thromboplastins compared with the manufacturers' instrument specific ISI value. Despite this apparent imperfection, our study clearly showed that the INR is preferable to other prothrombin time reporting formats for assessing the degree of anticoagulation for patients on warfarin treatment.