Petty, George W.

Background We sought to determine the occurrence of major bleeding events in a population-based cohort of stroke patients with atrial fibrillation in a population-based cohort who were risk-stratified at baseline, one and five years, following initiation of warfarin treatment. Methods The resources of the Rochester Epidemiology Project Medical Linkage System were used to identify acute ischemic stroke patients with atrial fibrillation undergoing warfarin treatment for secondary stroke prevention from 1980 to 1995. These patients were risk-stratified according to the HAS-BLED and HEMORRHAGES scores prior to warfarin initiation following stroke onset; these scores were reassessed one and five years later. Major bleeding was considered in patients with fatal or clinically overt bleeding associated with either transfusion of >2 units of blood or >2 g/dl decrease in hemoglobin. Results One hundred patients (mean age, 79.3 years; 68% women) were studied. Median duration of warfarin treatment was 2.4 years (interquartile range, 0.4-6.6) and duration from stroke onset to death was 3.7 years (1.1-8.8). Major bleeding events occurred in 41 patients at a median duration of 19 months (5-58) following warfarin initiation. Sixteen events occurred within a year of treatment, 25 events from years 1-5, and 14 events in patients treated for >5 years. Patients with baseline HAS-BLED scores ≥2 had higher life-time risks of major bleeding events compared with those ≤1 (95% vs 5%, p=0.002), while those with HEMORRHAGES score ≥2 had higher life-time risks of major bleeding events compared with those ≤1 (90% vs 10%, p=0.030). Periodic reassessments of these risks were associated with a higher detection of major bleeding events. Patients with an increase in the HAS-BLED scores had a higher subsequent risk of major bleeding events compared with those with no change (year one, 64% vs 34%, p=0.043; year five, 75% vs 29%, p=0.018). Similarly, patients with an increase in the HEMORRHAGES scores had a higher risks of major bleeding events compared to those with no change (year one, 57% vs 33%, p=0.063; year five, 71% vs 10%, p<0.001). Conclusions Periodic assessments of bleeding risks using prognostic scales are associated with a higher detection of major bleeding events. These scales should be periodically reassessed in the years following warfarin initiation to determine if the benefit of ongoing use outweighs the bleeding risk.