Predicting Bleeding Risk in Patients With Atrial Fibrillation Who Are Taking Direct Oral Anticoagulants

Is there an accurate tool to assess bleeding risk in patients with atrial fibrillation (AF) when starting a direct oral anticoagulant (DOAC) to minimize the risk of stroke?

Family physicians must consider the balance between thromboembolic disease prevention and major bleeding risks when starting anticoagulation in patients with AF.¹ Although key guidelines recommend the CHA₂DS₂-VASc (congestive heart failure, hypertension, age 75 years or older [doubled], diabetes, stroke/transient ischemic attack/thromboembolism [doubled], vascular disease, age 65 to 74 years, sex category [female]) score to stratify stroke risk, there is less consensus on a risk score for bleeding.^1,2 The HAS-BLED (hypertension, abnormal renal and liver function, stroke, bleeding risk, labile international normalized ratio, elderly [older than 65 years], drug and alcohol use) score is used most often, but it was created for calculating the risk of bleeding events while using warfarin.³ With the use of DOACs increasing in the management of AF, a bleeding risk calculator tailored for patients taking DOACs would be beneficial in guiding patients and clinicians through the decision to start anticoagulation.

The ABH (age, history of bleeding, and nonbleeding-related hospitalization) score was created to better predict the risk of major bleeding (ie, bleeding that is fatal, symptomatic, causes a significant decrease in hemoglobin level, or requires a transfusion) in patients taking DOACs compared with non-DOAC–specific bleeding risk scores, such as HAS-BLED, ATRIA, HEMORR₂HAGES, ORBIT-AF, and CHA₂DS₂-VASc.^4,5 A 2023 retrospective observational study in one Michigan health care system compared the ABH risk score with the non-DOAC–specific clinical prediction rules. The measure of accuracy was the area under the curve (AUC; range 0–1); a value of 0.5 means that a model predicts no better than chance, whereas numbers closer to 1 indicate better overall accuracy. The study found that two of the non-DOAC–specific clinical prediction rules, the ORBIT-AF (AUC 0.62) and ATRIA (AUC 0.61), best predicted bleeding risk over 1,100 days. The ABH and the non-DOAC–specific rules did not reliably predict bleeding at 2,100 days (AUC < 0.6).⁵

In 2024, a French multicenter prospective cohort study compared 13 bleeding risk scores in 994 patients with AF taking DOACs. The authors concluded that all of the scores showed poor to moderate accuracy for predicting major bleeding (AUC < 0.6).⁶

These trials highlight the lack of accurate bleeding prediction tools for patients with AF taking DOACs. Given this gap, a group of researchers developed the DOAC score to more accurately predict major bleeding risk for patients with AF taking DOACs (Table 1 and Table 2).³ The score was derived using data from the RE-LY trial, a multicenter clinical trial that compared stroke, systemic embolism, and bleeding risks for patients with AF taking dabigatran (Pradaxa) and warfarin, and the GARFIELD-AF registry, which included patients taking edoxaban (Savaysa), apixaban (Eliquis), and rivaroxaban (Xarelto). The scores were validated in external cohorts using the COMBINE-AF and RAMQ trials. The DOAC score assigns points for patient age; underweight status; creatinine clearance; stroke, transient ischemic attack, and systemic embolism history; diabetes; hypertension; antiplatelet use; nonsteroidal anti-inflammatory drug use; bleeding history; and liver disease. It had an AUC of 0.73 in RE-LY, 0.71 in GARFIELD-AF, 0.67 in COMBINE-AF, and 0.65 in RAMQ and outperformed HAS-BLED in each cohort.³ The study team did not compare the DOAC score to other clinical prediction rules. Of note, the DOAC score is only intended for use in patients 65 years and older.