EHA2026: Risk stratification and anticoagulation data

By Laura Boyd

Key insights from the 31st Annual Meeting of the European Hematology Association (EHA2026 Congress) highlighted advances in thrombotic and bleeding risk stratification, including validation of thrombotic risk models, machine learning–based bleeding prediction tools, and comparative anticoagulation outcomes in hematologic populations.

Giulia Bassi (Fondazione IRCCS San Gerardo dei Tintori, Monza, Italy) presented results from the international PREDIC-TO study evaluating venous thromboembolism (VTE) risk models in diffuse large B-cell lymphoma (DLBCL). Among 1,504 patients, VTE occurred in 14.6%, with a median time to event of 72 days, indicating early clustering of thrombotic events after diagnosis.

VTE was associated with inferior overall survival (OS; 38.5 vs 172.3 months). In comparative analyses, Model IX demonstrated the highest discriminatory performance for both VTE (c-index, 0.63) and OS (c-index, 0.67), outperforming Khorana, ThroLy, and R-IPI scores.

Additional analyses identified factors including deep vein compression, BMI >30, and hemoglobin level as stable predictors of VTE risk, suggesting the presence of residual multidimensional risk not fully captured by existing models. Bassi concluded that these findings support Model IX as the “most informative risk stratification tool in this setting.”

Tamar Tadmor (Bnai Zion Medical Center, Haifa, Israel) shared findings from a machine learning–based bleeding risk model (ADAPT-BLEED) in anticoagulated patients. Among 153,540 patients, 2.61% experienced major bleeding within 12 months. The XGBoost model achieved an area under the curve of 0.732, significantly outperforming established scores including HAS-BLED (0.638), ORBIT (0.633), ATRIA (0.653), and the DOAC score (0.656).

The derived point-based ADAPT-BLEED score retained 93.1% of the full model’s discriminatory performance and demonstrated a 9.7-fold gradient in bleeding risk between lowest- and highest-risk groups. A digital application enables rapid bedside risk estimation to support clinical decision-making, with Tadmor concluding that “this approach provides an actionable and scalable tool for personalized bleeding risk assessment.”

Cher Ying Foo (Rochester General Hospital, USA) presented real-world data comparing direct oral anticoagulants (DOACs) with warfarin in patients with atrial fibrillation and hematologic malignancies. Among 52,373 patients, warfarin was associated with a higher risk of the primary composite outcome of ischemic stroke, intracranial hemorrhage, or arterial embolism compared with DOACs (4.7% vs 2.0%; HR, 2.08; P<0.001).

Rates of ischemic stroke (2.4% vs 0.7%; HR, 3.13), arterial embolism (0.6% vs 0.2%; HR, 2.84), and intracranial hemorrhage (2.0% vs 1.3%; HR, 1.41) were all higher with warfarin, while major gastrointestinal bleeding rates were similar between groups. These findings were consistent across key subgroups, including older patients and those with comorbidities. From this study with the largest cohort to date, the authors noted that these data “support DOACs as the preferred anticoagulation strategy in this high-risk population.”

Collectively, these studies highlight ongoing efforts to refine risk stratification and guide anticoagulation strategies in hematologic populations to support individualized clinical decision-making.

Developed by EPG Health for Medthority, independently of any sponsor.