Within-individual biological variation of INR (CV(B)) was assessed in 245 selected stable warfarin-treated patients monitored by three thrombosis centers. Selection criteria were: treatment period of six months or longer before the observation period; at least six consecutive INRs within the therapeutic range of 2.0 - 3.0; interval between consecutive INR measurements of two weeks or longer; no change in warfarin dose; no changes in the patient's circumstances which may influence the INR, such as intercurrent diseases, invasive procedures, starting or stopping drugs interacting with warfarin. The minimum, maximum and mean within-individual coefficient of variation CV(B) of the INR measurements in the 245 selected patients were 0.4%, 14.5%, and 9.0%, respectively Analytical performance goals for the INR measurement (imprecision) could be derived from the mean CV(B). For a therapeutic range of 2.0 - 3.0 with warfarin, the desirable and optimum imprecision of INR determination is