Diagnostic performance of single-lead electrocardiograms from a smartwatch and a smartring for cardiac arrhythmia detection

Background

Wearable devices are widely used for atrial fibrillation (AF) detection, yet most validation studies include only sinus rhythm or AF, likely overestimating diagnostic performance.

Objective

This multicenter study assessed the performance of automated AF detection and physician interpretation of single-lead electrocardiograms (SL-ECGs) from the Apple Watch and CART Ring.

Methodology

Participants underwent simultaneous 12-lead ECG and SL-ECGs from Apple Watch and CART Ring. Two cardiologists independently adjudicated all ECGs. Apple Watch and CART Ring classified recordings as “AF,” “Not AF,” or “Unclassified.” Diagnostic performance for automated AF detection was evaluated in “worst-case” (all SL-ECGs) and lenient (excluding unclassified SL-ECGs) scenarios. Physician interpretation of SL-ECGs was also compared to 12-lead ECG.

Results

Among 483 patients (median age, 66 years; 29% female), 196 (39%) had AF across 3 United Kingdom centers. A total of 2398 ECGs were analyzed. Interobserver variability was excellent (Cohen’s kappa: Apple Watch, 0.85; CART Ring, 0.84). In the “worst-case” analysis, CART Ring outperformed Apple Watch (sensitivity, 84.6% vs 69.1%; specificity, 89.9% vs 72.6%). Apple Watch had more unclassified SL-ECGs (20.1%) than CART Ring (1.9%). The lenient analysis showed an improvement in sensitivity (CART Ring, 84.8 %; Apple Watch, 86.4%) and specificity (CART Ring, 91.2%; Apple Watch, 91.7%). Physician interpretation improved diagnostic performance for AF and sinus rhythm but remained limited for other arrhythmias

Conclusion

Apple Watch missed approximately 1 in 3 episodes of AF and a high number of unclassified SL-ECG. CART Ring demonstrated superior performance. Physician interpretation significantly improved AF diagnosis but remained unreliable for other arrhythmias, emphasizing the need for cautious integration of wearable ECGs into clinical practice.