Validation of the DOAC Score among patients on vitamin K antagonists

Background The DOAC Score stratifies bleeding risk in patients with atrial fibrillation (AF) taking direct oral anticoagulants (DOACs). The discriminative ability of the DOAC Score among patients with AF taking vitamin K antagonists (VKAs) is unknown. Purpose This study evaluates the discrimination of the DOAC Score among patients with AF taking VKAs. Methods Two data sources were used for validation: COMBINE-AF and GARFIELD-AF. COMBINE-AF included patients with AF randomized to warfarin in the clinical trials of RE-LY, ARISTOTLE, ROCKET-AF, and ENGAGE AF-TIMI 48. GARFIELD-AF is a real-world registry, and patients who were prescribed with VKAs at enrollment were included. The DOAC Score consists of a 0 to 10 scoring system, was externally validated among patients with AF taking DOACs, and includes commonly obtained clinical variables (age, weight, estimated glomerular filtration rate, antiplatelet or nonsteroidal anti-inflammatory use, hypertension, diabetes, history of bleeding, liver disease, or stroke/transient ischemic attack/embolism). Patients were stratified into five clinical risk categories by the DOAC Score (very low [score 0-3], low [score 4-5], moderate [score 6-7], high [score 8-9], very high [score 10]), with one-year major bleeding rates determined for each risk group. Discrimination of the DOAC Score was estimated using Harrell’s C-statistics and compared to the HAS-BLED score with DeLong’s test. Results A total of 28,818 patients and 20,183 patients from COMBINE-AF and GARFIELD-AF were included and treated with VKAs. Major bleeding by one-year occurred in 994 (3.4%) patients in COMBINE-AF and 313 patients (1.6%) in GARFIELD-AF. Higher rates of major bleeding occurred in higher risk categories for patients in COMBINE-AF: very low (1.8 events per 100 person-years [events/100 p-y]), low (3.0 events/100 p-y), moderate (4.5 events/100 p-y), high (5.4 events/100 p-y), and very high (7.5 events/100 p-y) (Figure). A similar pattern was observed among patients in GARFIELD-AF: very low (0.8 events per 100 person-years events/100 p-y), low (1.5 events/100 p-y), moderate (2.2 events/100 p-y), high (3.2 events/100 p-y), and very high (7.6 events/100 p-y). Discrimination of the DOAC Score was moderate and higher than the HAS-BLED score in both COMBINE-AF (C-statistic: 0.62 vs 0.59, P<0.001) and GARFIELD-AF (C-statistic: 0.65 vs 0.62, P<0.001) (Table). Conclusions Among patients with AF taking VKAs, the DOAC Score was able to risk stratify patients for major bleeding risk, demonstrated moderate discrimination, and had improved discrimination compared to HAS-BLED in a large clinical trial cohort and a real-world registry. The DOAC Score could be considered an alternative to HAS-BLED for bleeding risk stratification.