Assessing the diagnostic accuracy of wearable ECG monitors in detecting ischaemia
B Al-khayatt, A Sinha, M Li Kam Wa, H Morgan, M Ryan, H Rahman, D PereraAbstract
Background
Wearable ambulatory electrocardiography (aECG) monitors are commonly used to detect arrhythmia, with an array of devices marketed directly to the consumer. Their widespread availability and ease of use make them an attractive option for detecting transient ischemia but their accuracy has not been systematically compared thus far.
Purpose
To compare the diagnostic accuracy of three aECG devices in detecting myocardial ischaemia: the Apple Watch, Kardia ‘6-lead’ (6L) and Cardiostat patch. These devices represent a spectrum of aECG technologies, ranging from user-triggered to continuous monitors, as well as single to multi-lead devices.
Methods
Patients presenting with confirmed ST elevation myocardial infarction on 12-lead ECG and scheduled for primary PCI were eligible, if they were older than 18 years and provided verbal assent followed by formal post-hoc consent. Recruitment is ongoing with a target of 50 participants. Those in cardiogenic shock or presenting with cardiac arrest were excluded. De-identified ECGs were recorded on each device at two time points: at presentation and after revascularisation, defined either as the point when the ST segment became isoelectric or after 72 hours of convalescence, whichever occurred first. The degree of ST elevation visible was adjudicated by researchers blinded to clinical and 12-lead ECG data. Accuracy was judged against time-matched 12 lead ECGs.
Results
46 patients were recruited from a single tertiary cardiac centre, between December 2023 and January 2026. Paired (12-lead and aECG) data were available for 32 participants using the Apple Watch, 32 using the Kardia 6L and 20 using the Cardiostat patch. Not all participants completed recordings with all devices due to logistical constraints. The Kardia 6L demonstrated moderate diagnostic performance (AUC = 0.81) with 81% sensitivity and 54% specificity at detecting STEMI at a threshold of 1mm of ST elevation. Discriminatory performance improved further within the inferior STEMI subgroup (AUC = 0.90). Conversely, both the Apple Watch and Cardiostat performed poorly (AUC = 0.64 and 0.57 respectively). All three devices systematically underestimated STE compared to the 12-lead ECG; mean differences on Bland-Altman analysis were -0.55mm (95% CI -0.82 to -0.29) for the Kardia 6L, -1.52mm (95% CI -1.96 to -1.08) for the Apple watch and -1.36 mm (95% CI -1.99 to -0.73) for the Cardiostat. The Kardia demonstrated significantly less underestimation than both the Apple Watch (p < 0.01) and Cardiostat (p < 0.05).
Conclusion
aECG devices do not have comparable accuracy at detecting ischemia. The Kardia 6L performed best, possibly related to its additional limb leads. The clinical utility of multi-lead aECG devices needs to be assessed in prospective studies of patients prone to transient ischemia.AECG Figure