Objective: To determine the relationship between serial international normalized ratios (INRs) in patients who have been undergoing long-term anticoagulation and the onset of warfarin-associated bleeding complications.
Methods: The study cohort consisted of 2391 patients treated in the Anticoagulation Service at Brigham and Women's Hospital, Boston, Mass, from April 1999 through July 2003. For each patient with a bleeding event, we selected 2 controls who were matched for age, sex, indication for warfarin therapy, and duration of enrollment in our Anticoagulation Service.
Results: Warfarin-related hemorrhage occurred in 32 patients (1.3%). The mean +/- SD INRs at the time of the bleeding event or matched patient's event date (5.9 +/- 5.9 vs 2.3 +/- 0.7; P<.001) and the mean+/-SD last INRs before the bleeding event or matched patient's event date (3.0 +/- 1.2 vs 2.1 +/- 0.8; P<.001) were higher in the patients than in the controls. The last INRs before the bleeding event were obtained an average of 11.6 +/- 17.8 (mean +/- SD) days before the event in the patients and 18.3 +/- 28.0 (mean +/- SD) days before the matched date in the controls (P = .22). The mean second-to-last INRs were similar in both groups (2.8 +/- 2.1 vs 2.3 +/- 0.8; P = .11). When the INRs were plotted in relation to the time before the onset of bleeding, a marked increase in the patients' INRs was observed shortly before the bleeding began.
Conclusions: Serial INRs are poor predictors of hemorrhagic events. There appears to be only a brief warning period during which a slightly elevated INR predicts an imminent bleeding event.