OBJECTIVE To determine the relative contribution of changes in the plasma warfarin level to variation in the serial steady-state prothrombin times. METHODS This was a prospective observational cohort study performed at two outpatient anticoagulation clinics. Serial prothrombin times and paired plasma total warfarin levels were determined in a convenience sample of otherwise healthy patients who required long-term oral anticoagulation therapy with warfarin. RESULTS Serial measurements were obtained from 129 patients, 60 of whom provided three or more serial samples. Analysis of covariance showed a highly significant (p = 0.0001) relationship between the anticoagulant effect and the logarithm of the warfarin concentration (R2 = 0.75), with 15.3% of the total variance attributable to the effect of warfarin and 31.1% attributable to individual variation in sensitivity to warfarin. In an analysis of the subjects who had three or more serial measurements, the mean weighted correlation coefficient for the relationship between the logarithm of the warfarin concentration and the anticoagulant response varied widely, from strongly negative to strongly positive, and as the range of observed prothrombin times increased, stronger positive correlation was observed. CONCLUSIONS In this cohort, the plasma warfarin level was a strong predictor of observed changes in serial prothrombin time measurements. However, the correlation between clotting times and warfarin levels varied widely among subjects, particularly when the range of observed prothrombin times was moderate. This suggests that in these subjects, other factors, such as measurement error or pharmacodynamic changes, played a major role.