Objective: To define risk factors for complications that occur during warfarin therapy.
Design: Retrospective cohort study.
Setting: Five anticoagulation clinics.
Patients: Nine hundred twenty-eight consecutive patients receiving 1103 courses of warfarin.
Main outcome measures: Hemorrhagic and thromboembolic complications.
Results: In 1950 patient-years of follow-up, there were 1332 bleeding events (4 were classified as fatal, 31 as life-threatening, 226 as serious, and 1071 as minor). The cumulative incidence of fatal bleeding was 1% at 1 year and 2% at 3 years. The cumulative incidences of first episodes of life-threatening and serious bleeding at 1, 2, 4, and 8 years were 1%, 2%, 5%, and 9% and 12%, 20%, 28%, and 40%, respectively. Of 156 patients who had a serious or life-threatening hemorrhage, 32% suffered a recurrence, typically within 1 year. Independent predictors of a first episode of serious bleeding included a mean prothrombin time ratio (PTR) of 2.0 or more during the course of treatment (relative risk, 3.0; 95% CI, 1.9 to 4.7); recent initiation of warfarin therapy (relative risk during the first 3 months compared with the rest of the first year, the second year, and anytime thereafter, 1.9 [CI, 1.3 to 3.0], 3.0 [CI, 1.8 to 4.8], and 5.9 [CI, 3.8 to 9.3], respectively); variability of the PTR over time (relative risk for the highest compared with the lowest tertile, 1.6 [CI, 1.2 to 2.7]); and the presence of 3 or more comorbid conditions (RR, 1.4 [CI, 1.1 to 2.5]). Age, reason for anticoagulation, use of interfering drugs, and hypertension were not associated with risk for bleeding. The risk for a thromboembolic complication at a PTR of less than 1.3 was 3.6 (CI, 2.1 to 6.4) times higher than at a PTR of 1.3 to 1.5.
Conclusions: The incidence of warfarin-associated bleeding may be reduced by attending to modifiable risk factors (that is, highly variable PTRs and values greater than 2.0), frequent monitoring early in treatment, and careful patient selection. Older age, in and of itself, is not a risk factor.