AI heart-tracing trial of 14,665 adults finds slightly more atrial fibrillation caught, but result is fragile

Only 84 of 7,337 intervention participants (1.1%) were flagged as high risk by the AI, and 48 of them (57.1%) underwent Holter monitoring of up to 7 days. Each was paired with an age- and sex-matched control who received identical monitoring, which was meant to reduce ascertainment bias.

The authors describe the increase in detection as modest and statistically fragile: the odds ratio was 1.54 (95% CI, 1.01-2.37; P=0.046), and reclassifying one intervention event as a nonevent moved the P value to 0.057. The trial measured detection, not patient outcomes, and the authors call for larger trials.

The LATENT trial tested whether AI analysis of ordinary sinus-rhythm ECGs could identify people with atrial electrical vulnerability who would benefit from targeted rhythm monitoring. Adults without previous AF or flutter were randomized to an AI-ECG-guided strategy or usual care in a single-center, open-label design.

Intervention participants whose ECG from the preceding year was judged high risk by the AI were offered Holter monitoring for up to 7 days. Each monitored participant was paired with an age- and sex-matched control, chosen before control ECGs underwent AI analysis, who received identical monitoring.

The composite outcome covered newly detected AF, flutter or pre-AF findings such as frequent premature atrial complexes, short runs of them, or nonsustained AF or flutter. The intervention effect differed by baseline AI-ECG risk.

The authors conclude the result shows a clinical application of AI-ECG risk stratification for targeted surveillance and offers a foundation for larger trials of patient outcomes. The trial is registered as NCT06847932.