PPG-based smartwatch algorithm detected AF with 99.2% accuracy and estimated AF burden with near-perfect correlation (r=0.999) to ECG patch in 728 AF patients.
Does a continuous PPG-based smartwatch algorithm accurately estimate atrial fibrillation burden compared to a single-lead ECG patch in patients with atrial fibrillation?
A continuous PPG-based smartwatch algorithm provides highly accurate and consistent atrial fibrillation burden monitoring compared to a reference single-lead ECG patch.
Abstract Background Wearable devices utilizing photoplethysmography(PPG)-based algorithm smartwatch can effectively monitor atrial fibrillation (AF). However, evidence of continuous PPG-based smartwatch for AF burden monitoring is rather limitied. Purpose This study aims to evaluate the accuracy and consistency of a continous PPG-based smartwatch algorithm for AF burden monitoring compared to a single-lead ECG patch in a large sample of Chinese patients with AF. Methods In an observational study, we enrolled AF patients at a single site centre from January to April 2024 to evaluate the performance of PPG and a PPG-based algorithm in estimating AF burden. Using an single-lead ECG patch as the reference device, we validated the algorithm’s ability to detect AF and estimate AF burden in 30-second intervals. Results The study included 728 patients with AF (67.3% male，median age of 62.0 years). The average monitoring duration was 20.4 ± 4.5 hours. After dividing the monitoring time into non-overlapping 30-second intervals, 1,440,826 paired ECG and PPG recordings were generated. The overall valid recording rates were 96.23% for ECG and 62.50% for PPG. At the patient level, PPG demonstrated an accuracy of 99.2%, sensitivity of 98.6%, and specificity of 99.5%. At the segment level, these metrics were 94.0%, 92.0%, and 96.4%, respectively. Across different time periods, PPG’s accuracy, sensitivity, and specificity all exceeded 98%. For patients with at least one AF episode, the AF burden estimated by wrist-based PPG (W-PPG) strongly correlated with that calculated by patch-based ECG (P-ECG) (r = 0.999). Bland-Altman analysis revealed high agreement between PPG and ECG in AF burden estimation, with a mean difference of -0.335% and 95% limits of agreement ranging from -5.333% to 4.662%. Consistent results were observed across subgroups, with most measurements (96.4%) falling within the 95% confidence interval, indicating excellent agreement. Conclusion Our findings demonstrate that smartwatches equipped with PPG-based algorithms exhibit high accuracy and stability in continuously monitoring AF burden compared to ECG patches, highlighting their potential for AF diagnosis and management.
Kong et al. (2025) studied this question. PPG-based smartwatch algorithm detected AF with 99.2% accuracy and estimated AF burden with near-perfect correlation (r=0.999) to ECG patch in 728 AF patients.