Risk of Complicated Upper Gastrointestinal Bleeding With Direct Oral Anticoagulants: An Asian Population-Based Analysis.

BACKGROUND & AIMS: Asian data on complicated upper gastrointestinal bleeding (UGIB) risks among direct oral anticoagulants (DOACs) are limited.

METHODS

An industry-independent, population-based, retrospective cohort study was performed in Hong Kong during 2011 to 2020. Subjects with new exposure to apixaban, dabigatran, edoxaban, or rivaroxaban were included. Baseline demographics, comorbidities, drugs, and laboratory results were balanced between DOACs by inverse probability of treatment weighting. Subjects were followed for 2 years with non-UGIB deaths considered as competing risks. Subdistribution hazard ratios (SHRs) were generated by Fine-Gray model. False discovery rate (FDR) was controlled by the Benjamini-Hochberg procedure for multiplicity. The primary outcome was complicated UGIB, defined as any UGIB-related death, or UGIB with a hemoglobin drop >2 g/dL; blood transfusion; emergent endoscopic procedure; radiological/surgical intervention; or rebleeding within 30 days.

RESULTS

A total of 58,229 subjects with new DOACs exposure were included (22,359 apixaban, 20,639 dabigatran, 2068 edoxaban, and 13,163 rivaroxaban). Compared with apixaban, edoxaban (SHR, 1.891; 95% confidence interval [CI], 1.091-3.275; P = .025; FDR-adjusted P = .039) and rivaroxaban (SHR, 1.441; 95% CI, 1.047-1.983; P = .026; FDR-adjusted P = .039), but not dabigatran, were associated with significantly higher risks of complicated UGIB, with a number-needed-to-harm of 244 for edoxaban and 427 for rivaroxaban, respectively. In subgroup analysis, apixaban was associated with a lower bleeding risk in subjects with age ≥80 years or estimated glomerular filtrate rate <60 mL/min/1.73 m2. Low-dose apixaban was associated with a lower bleeding risk among all low-dose DOACs.

CONCLUSIONS

This Chinese population-based observational study demonstrated a lower 2-year risk of complicated UGIB in apixaban users compared with edoxaban and rivaroxaban, but not dabigatran, users among subjects without prior anticoagulant exposure.