Leveraging AI-ECG Technology for Early Notification and Tracking of AF Development (LATENT): A Randomized Controlled Trial

Abstract

BACKGROUND: Electrocardiographic analysis aided by artificial intelligence (AI) may identify atrial electrical vulnerability during sinus rhythm. The objective of this study was to determine whether AI–enabled ECG (AI-ECG) is effective for detecting atrial fibrillation (AF), atrial flutter (AFl), or prespecified pre-AF atrial findings in a randomized trial. METHODS: In this single-center, open-label trial, 14 665 adults 40 to 85 years of age without previous AF/AFl were randomized to an AI-ECG–guided strategy or usual care. Participants randomized to AI-ECG who were then identified by AI as having a high-risk sinus-rhythm ECG in the preceding year were offered Holter monitoring for ≤7 days. Each monitored high-risk intervention participant was paired with an age- and sex-matched control selected before control ECGs underwent AI analysis. Matched controls received identical monitoring to reduce ascertainment bias. Reporting cardiologists were blinded to AI results and randomized assignment. The primary outcome was a composite of newly detected AF, AFl, or pre-AF atrial findings within 90 days. Pre-AF atrial findings comprised frequent premature atrial complexes, short runs of premature atrial complexes, or nonsustained AF/AFl. RESULTS: Among 7337 intervention and 7328 control participants, 84 (1.1%) and 78 (1.1%), respectively, were classified as high risk. Of the 84 high-risk intervention participants, 48 (57.1%) underwent monitoring, as did 48 matched controls. The primary outcome occurred in 54 intervention participants (0.74%) versus 35 control participants (0.48%; odds ratio, 1.54 [95% CI, 1.01–2.37]; P =0.046). Hypothetical reclassification of 1 intervention event to a nonevent changed the primary analysis P value to 0.057. The intervention effect on detection of the primary composite outcome differed by baseline AI-ECG risk ( P interaction <0.001): 32.1% versus 2.6% among high-risk participants and 0.4% versus 0.5% among low-risk participants. AF/AFl detection (0.3% versus 0.2%) and antiarrhythmic therapy initiation at 180 days (2.3% versus 2.4%) did not differ significantly between the intervention and control groups. Participant-reported monitoring-related adverse events, including skin reactions and discomfort, occurred in 22.9% of monitored intervention participants versus 18.8% of matched controls. CONCLUSIONS: AI-ECG–guided monitoring resulted in a modest increase in the detection of new AF, AFl, or pre-AF atrial findings. However, these findings were fragile statistically and sensitive to reclassification of 1 event. LATENT (Leveraging AI-ECG Technology for Early Notification and Tracking of AF Development) demonstrates the clinical application of AI-ECG risk stratification to guide targeted rhythm surveillance, providing a foundation for larger trials evaluating its impact on patient outcomes. REGISTRATION: URL: https://www.clinicaltrials.gov ; Unique identifier: NCT06847932.