Research progress on bleeding risk assessment models in anticoagulant therapy

Area of Science:

• Cardiovascular Medicine
• Pharmacology
• Clinical Risk Stratification

Background:

• Established bleeding risk assessment models (RAMs) like HAS-BLED were designed for warfarin, with uncertain applicability to non-vitamin K antagonist oral anticoagulants (NOACs) and venous thromboembolism (VTE).
• Accurate bleeding risk stratification is crucial for optimizing anticoagulant therapy, balancing thromboembolic prevention with bleeding complications.

Purpose of the Study:

• To review recent advancements in bleeding risk stratification for atrial fibrillation (AF) and VTE patients treated with anticoagulants.
• To evaluate the performance of traditional and novel RAMs, including drug-specific adaptations and biomarker-driven approaches.

Main Methods:

• Systematic review of recent literature on bleeding risk assessment models in AF and VTE populations.
• Analysis of model performance metrics (e.g., AUC, C-statistic) and clinical applicability across different anticoagulant regimens.
• Examination of emerging biomarker-based tools and dynamic risk modifiers.

Main Results:

• Traditional RAMs show moderate predictive accuracy in NOAC users (AUC: 0.55-0.74), with some models like HEMORR₂HAGES performing better.
• Biomarker-integrated models (e.g., ABC, DOAC score) and context-adapted VTE models (e.g., IMPROVE, RIETE) demonstrate improved risk stratification.
• Heterogeneity in study designs and endpoint definitions limits current model generalizability, highlighting the need for refined approaches.

Conclusions:

• Current RAMs have variable performance, necessitating next-generation models integrating dynamic factors and biomarkers for precision anticoagulation.
• NOAC-specific tools are more suitable for AF patients, while context-adapted models are preferred for VTE populations.
• Future research must focus on real-world validation, machine learning, and standardized bleeding definitions to enhance anticoagulation safety and efficacy.