Comparison of Risk-Scoring Models to Predict Gastrointestinal Bleeding in Patients With Direct Oral Anticoagulants

Background and aim: The risk of gastrointestinal bleeding (GIB) remains a concern with the use of direct oral anticoagulants (DOAC). We evaluated the efficacy of four risk-scoring models (HAS-BLED, ATRIA, VTE-BLEED, and ORBIT) in predicting GIB according to the concomitant use of antiplatelet therapy in DOAC users.

Methods: Patients prescribed DOAC between December 2014 and October 2020 were enrolled in two university-affiliated hospitals. The performance of the four models was compared based on the concomitant use of antiplatelet therapy. The primary outcomes were likelihood ratios and the area under the receiver operating characteristic (AUROC) curve to predict GIB.

Results: A total of 4494 patients were included in the study. The AUROC values for the entire cohort were 0.643 (95% CI: 0.601-0.686) for HAS-BLED, 0.693 (95% CI: 0.649-0.737) for ATRIA, 0.708 (95% CI: 0.665-0.750) for VTE-BLEED, and 0.709 (95% CI: 0.667-0.751) for ORBIT. The AUROC for all scoring models increased in patients without antiplatelet therapy compared to the entire cohort and patients with antiplatelet therapy. The specificity and diagnostic accuracy for all scoring models increased in patients without antiplatelet therapy compared to patients with antiplatelet.

Conclusions: Our results confirmed that current risk-scoring models for predicting GIB perform better in patients without antiplatelet therapy than in those on concomitant antiplatelet therapy. This suggests that future risk prediction models should consider the concomitant use of antiplatelet therapy for diagnostic accuracy.