Isenegger Corinne, Mannhart Diego, Weidlich Simon, Brügger Jonas, Serban Teodor, Jordan Fabian, Krisai Philipp, Knecht Sven, Schaerli Nicolas, Subin Behnam, Mosher Luke, du Fay de Lavallaz Jeanne, Schaer Beat, Mahfoud Felix, Kühne Michael, Sticherling Christian, Badertscher Patrick
Department of Cardiology, University Hospital Basel, Basel, Switzerland; Cardiovascular Research Institute Basel, University Hospital Basel, University of Basel, Basel, Switzerland.
Department of Cardiology, University Hospital Basel, Basel, Switzerland; Cardiovascular Research Institute Basel, University Hospital Basel, University of Basel, Basel, Switzerland.
JACC Clin Electrophysiol. 2025 May 29. doi: 10.1016/j.jacep.2025.04.027.
Multiple smart devices can record single-lead electrocardiograms (SL-ECGs) with automated rhythm classification. The impact of pre-existing baseline ECG anomalies on the accuracy of automated rhythm classification remains largely unknown.
This study sought to compare the presence of predefined ECG anomalies and their impact on rhythm classification ability of 5 commercially available FDA and CE-marked wearable smart-devices.
This prospective study included consecutive patients undergoing electrophysiological procedures at a tertiary referral center. Each participant obtained a 12-lead ECG followed by SL-ECGs with 5 different smart devices (AliveCor KardiaMobile, Apple Watch 6, Fitbit Sense, Samsung Galaxy Watch 3, and Withings ScanWatch). Two independent cardiologists performed manual rhythm classification and assessed the following ECG anomalies: ventricular pacing, conduction delay, low voltage, artifacts, and premature atrial or ventricular complexes.
A total of 256 participants were included (29% female, mean age 66 years) generating 1,280 recorded SL-ECGs. Of these, 242 SL-ECGs (19%) were classified as inconclusive by at least 1 smart device. The presence of any ECG anomaly was significantly higher in inconclusive vs conclusive SL-ECGs, with 74% vs 42%; P < 0.001. ORs with 95% CIs for inconclusive classification by ECG anomaly were ventricular pacing 6.35 [3.84-10.61], conduction delay 2.42 [1.82-3.22], low voltage 2.37 [1.75-3.21], minor artifact 1.72 [1.17-2.51], major artifact 10.62 [6.78-16.99], premature atrial complex 2.23 [1.29-3.74], and premature ventricular complex 1.94 [1.29-2.89]. Notable differences were found between the assessed smart devices.
Automated rhythm classification is highly susceptible to baseline ECG anomalies. This study provides insights into the most appropriate patient population for smart device-based arrhythmia monitoring and offers guidance for selecting the optimal smart device tailored to individual patient characteristics.
多种智能设备可记录单导联心电图（SL-ECG）并进行自动心律分类。既往基线心电图异常对自动心律分类准确性的影响在很大程度上仍不清楚。
本研究旨在比较预定义心电图异常的存在情况及其对5种获得美国食品药品监督管理局（FDA）和欧洲合格认证（CE）的商用可穿戴智能设备心律分类能力的影响。
这项前瞻性研究纳入了在一家三级转诊中心接受电生理检查的连续患者。每位参与者先进行一次12导联心电图检查，然后使用5种不同的智能设备（AliveCor KardiaMobile、苹果手表6、Fitbit Sense、三星Galaxy Watch 3和Withings ScanWatch）记录SL-ECG。两名独立的心脏病专家进行人工心律分类，并评估以下心电图异常：心室起搏、传导延迟、低电压、伪迹以及房性或室性早搏复合体。
共纳入256名参与者（29%为女性，平均年龄66岁），记录了1280份SL-ECG。其中，至少有1种智能设备将242份SL-ECG（19%）分类为不确定。不确定的SL-ECG中任何心电图异常的存在率显著高于确定的SL-ECG，分别为74%和42%；P<0.001。按心电图异常分类为不确定的95%置信区间的比值比为：心室起搏6.35[3.84-10.61]，传导延迟2.42[1.82-3.22]，低电压2.37[1.75-3.21]，轻微伪迹1.72[1.17-2.51]，严重伪迹10.62[6.78-16.99]，房性早搏复合体2.23[1.29-3.74]，室性早搏复合体1.94[1.29-2.89]。在评估的智能设备之间发现了显著差异。
自动心律分类对基线心电图异常高度敏感。本研究为基于智能设备的心律失常监测最合适的患者群体提供了见解，并为根据个体患者特征选择最佳智能设备提供了指导。
