VITAL-AF (AI-ECG)
2026
Risk-Guided Atrial Fibrillation Screening with Artificial Intelligence-Enabled Electrocardiogram Models: a VITAL-AF Trial Analysis
In 16,937 primary care patients aged ≥65 years without known atrial fibrillation, single-lead ECG screening targeted by an AI-ECG risk model increased 2-year incident AF diagnosis compared with usual care in control practices.
Result
Primary endpoint: 2-year incident AF diagnosis rate (per 100 person-years), screening vs control across AF-risk deciles
2-year AF discrimination AUROC: CHARGE-AF 0.711, ECG-AI 0.784, CH-AI 0.788. In the top CH-AI decile, AF diagnosis 10.07 vs 7.76 per 100 person-years (difference 2.32, 95% CI 0.01–4.63; P<0.05), number needed to screen 43 per year.
- Subtopic
- Atrial fibrillation
- Design
- Secondary analysis of the VITAL-AF cluster RCT (16 practices), USA
- Relevance
- Supporting: context rather than must-know.
- Source
- Vedage NA, et al. J Am Coll Cardiol. 2026;87(14):1798-1813.doi 10.1016/j.jacc.2026.01.087
Discussion & critique
Reframes a screening trial with limited age-based yield: an AI-ECG risk model identified the decile in whom single-lead ECG screening actually raised AF diagnoses, giving a practical number needed to screen. A post-hoc analysis confined to the 16,937 of 30,630 participants with a prior ECG; the screening effect was seen only in the top decile, with a confidence interval that nearly touches the null, and the authors note the trade-off between screening efficiency and population coverage.