Arrhythmias & electrophysiology
Bayesian Machine Learning Model Guiding Iterative, Personalized Anticoagulant Dosing Decision-Making: ENGAGE AF-TIMI 48 Trial Analysis
Authors' conclusion
Adele represents a state-of-the-art Bayesian framework, using clinical factors and PK data, to enable longitudinally adaptive, patient-centric, preference-weighted predictions.
Study context
- Population
- Edoxaban-treated patients with AF with pharmacokinetic data
- Participants
- 5,380 patients
- Comparison
- Adele vs standard Kaplan-Meier estimators
- Outcome
- Predictive accuracy (concordance index) vs Kaplan-Meier estimators
- Follow-up
- 3 years
Methodological detailssecondary analysis of the ENGAGE AF-TIMI 48 trial, Bayesian competing-risk, multistate hazard model, machine learning
- Secondary analysis of the ENGAGE AF-TIMI 48 trial
- Bayesian competing-risk, multistate hazard model
- Machine learning
Performance
As reported in the abstract
3-year prediction improvement (concordance index) vs Kaplan-Meier
CV death
+12.1%
Disability
+11.8%
Major gastrointestinal bleeding
+13.7%
Ischemic stroke
+3.3%
Further findings
Dynamic risk prediction
- Future event probabilities shifted following intracranial hemorrhage or ischemic stroke (IS)
Patient example A (80-yr female; 52 kg)
- Preference-weighted event rates lowest at 15 mg and 30 mg, irrespective of preference to avoid CV death, major bleeding or disability
Patient example B (72-yr male; 80 kg)
- Preference-weighted event rates lowest at 60 mg for avoiding death or IS/intracranial hemorrhage; lowest at 15 mg for disability avoidance
Original publication
Bayesian Machine Learning Model Guiding Iterative, Personalized Anticoagulant Dosing Decision-Making: ENGAGE AF-TIMI 48 Trial Analysis. JACC Adv. Published 9 December 2025. doi:10.
Summarised from the published abstract; every number is checked against it