Anticoagulant-associated bleeding: real-world data-driven pharmacovigilance and insights for bleeding management.

BACKGROUND

Anticoagulant drugs are essential for managing thrombotic diseases but lead to increasing bleeding risk. The choice of suitable anticoagulant drugs requires an assessment of the patient's thrombotic risk factors and underlying ischemic pathology.

METHODS

This study constructs a closed-loop framework integrating qualitative literature review and quantitative data analysis to evaluate bleeding risks in anticoagulant therapy. The qualitative review covers molecular mechanisms, risk factors, and nursing practices related to anticoagulant drugs. Quantitatively, we analyzed the bleeding events from the FDA Adverse Event (AE) Reporting System (FAERS) database (2004-2024), using disproportionality analysis with Reporting Odds Ratio, Proportional Reporting Ratio, Bayesian Confidence Propagation Neural Network, and Empirical Bayes Geometric Mean to identify significant bleeding signals. For each anticoagulant drug, the 10 strongest signals were extracted and anatomically profiled to construct a drug-specific haemorrhage atlas. Logistic regression analyses were applied to investigate the effects of age, sex, reporter, and comorbidities on the occurrence of anticoagulant drugs-associated bleeding. To address the critical issue of bleeding risk management in the clinic, we further discussed the multidisciplinary nursing across the anticoagulation therapy cycle.

RESULTS

The study revealed significant variations in bleeding risks among different anticoagulants. Anticoagulant-associated bleeding comprised 110,701 reports from FAERS database. Elderly patients (≥ 65 years old) supplied 63.2% of cases; 32% of events emerged within 30 days of therapy initiation. Hospitalisation was documented in 46.8% and death in 20.7%. Spinal subdural haemorrhage yielded the highest signal (ROR 144.79, 95% CI 28.09-746.33), followed by adrenal, mesenteric, and intracranial haemorrhages among all anticoagulant drugs. Gastric/lower gastrointestinal bleeding appeared in the top-10 list of every oral anticoagulant, while retroperitoneal bleeding was a shared high-signal events among heparins and parenteral direct thrombin inhibitors. Betrixaban exhibited extreme RORs for traumatic (ROR 18 059.67, 95%Cl: 8829.21-36 940.05) and intracranial (ROR 643.97, 95%Cl: 249.82-1660) bleeding. Each drug also exhibited distinct bleeding characteristics and differences in organ system distribution. Multivariable logistic regression across four hierarchical models identified agents with distinct, model-independent bleeding risk profiles relative to warfarin.

CONCLUSIONS

This large-scale pharmacovigilance study provides the most comprehensive, publicly available atlas of anticoagulant-related bleeding signals. The pronounced inter-drug heterogeneity and persistent long-term risk mandate individualised choice of agent and prolonged monitoring to ultimately improve the safety-efficacy balance of antithrombotic therapy.