A New Risk Scheme to Predict Warfarin-Associated Hemorrhage: The ATRIA (Anticoagulation and Risk Factors in Atrial Fibrillation) Study

Study Questions:

Can we predict hemorrhage risk associated with warfarin use?


The authors studied data from the ATRIA (Anticoagulation and Risk Factors in Atrial Fibrillation) study, which followed 13,559 adults having nonvalvular atrial fibrillation who were enrolled in Kaiser Permanente of Northern California health care system. This included all subjects with International Classification of Diseases-Ninth Revision, Clinical Modification (ICD-9) diagnosis codes for atrial fibrillation between July 1, 1996 and December 31, 1997, with follow-up through September 30, 2003. The authors obtained data from clinical databases and identified hemorrhage events validated using a review of medical records. They then developed a hemorrhage risk stratification score using a Cox regression model with candidate variables selected using bootstrapping approaches. The final model was internally validated with split-sample testing, and compared to six previously published hemorrhage risk schemes.


The authors of this study followed 9,186 patients with atrial fibrillation, contributing 32,888 person-years of follow-up (median warfarin duration 3.5 years [interquartile range, 1.2-6.0 years]). During follow-up, they observed 461 first major hemorrhage events (1.4% annually). The final model included five independent variables, weighted by regression coefficients: anemia (3 points), severe renal disease–defined as glomerular filtration rate <30 ml/min or on dialysis (3 points), age ≥75 years (2 points), prior bleeding (1 point), and hypertension (1 point). Cumulative point score was associated with major hemorrhage rates ranging from 0.4% (at 0 points) to 17.3% (at 10 points). The c-index for the continuous risk score was 0.74 and it was 0.69 across three categories of risk (low risk 0-3 points; intermediate risk 4 points; high risk 5-10 points). This was higher than the other previously published risk schemes. There was a net reclassification improvement with this newly derived risk score compared with all six previously published schemes (ranging from 27-56%).


The authors concluded that this simple five variable risk score is effective in quantifying the risk of warfarin-associated hemorrhage in this large community-based cohort of adult patients with atrial fibrillation.


As larger, more complex, and more detailed clinical databases become available, as well as improved statistical methods, better and more precise epidemiologic studies become possible. The authors of the current study have contributed significantly to the important clinical dilemma of assessing the potential harm of a medical therapy that has a quantifiable proven clinical benefit. Often, the clinical decision regarding the appropriateness of instituting warfarin therapy is made on the basis of an estimation of the potential clinical benefit. In the case of atrial fibrillation, this means an estimation of the patient’s risk of stroke without anticoagulation. Warfarin therapy is deemed appropriate when the risk of stroke without anticoagulation exceeds the risk of bleeding with anticoagulation. While increasingly precise tools have been developed and widely disseminated for estimating the risk of stroke without warfarin therapy, much less attention has been paid to the other side of the equation—estimating the risk of bleeding with warfarin therapy. The current study has developed, validated, and compared with prior schemes, a clinically useful and relatively easy to calculate tool for estimating the individualized risk of bleeding with warfarin therapy in adults with atrial fibrillation. This should greatly enhance clinical decision-making surrounding the use of warfarin therapy for stroke prophylaxis in atrial fibrillation.

Clinical Topics: Anticoagulation Management, Prevention, Hypertension

Keywords: Renal Dialysis, Follow-Up Studies, California, Warfarin, Risk Factors, Glomerular Filtration Rate, Hypertension, Hemorrhage

< Back to Listings