Development of a Novel Composite Stroke and Bleeding Risk Score in Patients With Atrial Fibrillation: The AMADEUS Study

Study Questions:

What are the independent predictors of the overall clinical outcome of patients with atrial fibrillation (AF), including both stroke/thromboembolism and/or major bleeding?


The investigators used data from the vitamin K antagonist (VKA) arm (n = 2,293; 65% men, mean age 70 ± 9) of the AMADEUS trial, which was a multicenter, randomized, open-label, noninferiority study that compared fixed-dose idraparinux with VKA in AF patients. They defined two composite endpoints: Endpoint 1 was the sum of stroke/thromboembolism or major bleeding. Endpoint 2 was defined as the sum of stroke, systemic or venous embolism, myocardial infarction, cardiovascular death, or major bleeding.


The independent predictors for composite endpoint 1 were age (p = 0.014), previous stroke/transient ischemic attack (p = 0.049), aspirin use (p = 0.002), and time in therapeutic range (p = 0.007). For composite endpoint 2, similar predictors were evident, plus left ventricular dysfunction (p = 0.011). Based on the regression models, two novel composite risk prediction scores were developed and externally validated in a “real-world” cohort of 441 anticoagulated outpatients with AF. Both composite scores 1 and 2 demonstrated numerically highest discriminatory performance (area under the curve [AUC], 0.728; 95% confidence interval [CI], 0.659-0.798 and AUC, 0.707; 95% CI, 0.655-0.758, for endpoints 1 and 2, respectively), and positive net reclassification when compared to currently used risk models (CHADS2, CHA2DS2VASc, HAS-BLED), but the differences were not statistically significant.


The authors concluded that they have developed and validated two novel composite scores for stroke/thromboembolism/bleeding, which offer good discriminatory and predictive performance.


The authors have developed and validated a novel composite risk prediction score for stroke/thromboembolism/bleeding that offers good discriminatory and predictive performance in the derivation and external validation cohorts. However, these composite risk scores did not perform better than the easier and more practical ‘traditional’ stroke and bleeding risk scores (CHADS2, CHA2DS2VASc, HAS-BLED), which allow personalized balancing of risks and clinical utility. Furthermore, the new composite scores require complex calculations and are less likely to be useful in everyday clinical practice, on ward rounds, or in busy outpatient clinics. It appears that the continued use of existing individual stroke and bleeding scores would allow more practical and personalized balancing of stroke/thromboembolism versus serious bleeding in a particular patient, rather than relying on a complex multivariate formula to derive a composite thromboembolism and bleeding risk score.

Keywords: Vitamin K, Thromboembolism, Risk, Myocardial Infarction, Stroke, Ischemic Attack, Transient, Ventricular Dysfunction, Hemorrhage

< Back to Listings