Head-to-head comparison of single-lead and six-lead electrocardiography of a wrist-worn device for atrial fibrillation detection

Abstract

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – EU funding. Main funding source(s): Funded by the European Regional Development Fund under agreement with the Research Council of Lithuania. Background The increasing numbers of available mHealth tools for electrocardiography (ECG)-based atrial fibrillation (AF) detection promote long-term screening. A common feature of smartwatches is lead-I-like ECG. However, only limited data exists directly comparing the performance of single-lead and six-lead wearable-recorded ECGs. Purpose To compare the accuracy of single-lead and six-lead ECGs of the same wrist-worn device for AF detection. Methods We included patients with AF which represent the main group for testing the diagnostic ability of wearable. In addition, authors selected control groups of stable sinus rhythm (SR) and SR with frequent premature contractions. Cardiac rhythm was monitored using an investigational wrist-worn device which provides six-lead ECG, similar to standard limb leads. To display a single-lead wearable-recorded ECGs, the same six-lead ECG tracings were trimmed to a width of lead-I-like ECGs. A validated Holter ECG device constituted a gold standard test for rhythm verification. Two independent diagnosis-blinded cardiologists evaluated reference, six-lead and single-lead ECGs as "AF", "SR", or "Inconclusive". A third cardiologist evaluated ECGs only in cases of physician disagreement. Results A total of 344 adult patients were enrolled in this study including AF group (121 patients) and control group of SR with or without frequent premature contractions (223 patients). Patients with missing (11/420; 2.62%) or insufficient quality (43/420; 10.24%) of wearable-recorded ECGs were excluded. AF detection based on single-lead and six-lead ECGs yielded sensitivity of 95.73% (95% CI 90.31–98.6%) and 99.16% (95% CI 95.41–99.98%), respectively. Specificity was 100% (95% CI 96.19–100%) for both single-lead and six-lead ECGs when differentiating between AF and stable SR. If patients with frequent premature beats were included in the control group, the specificity of single-lead and six-lead ECGs dropped to 95.81% (95% CI 92.31–98.07%) and 99.1% (95% CI 96.78–99.89%), respectively. False positive cases were more common for single-lead ECG (9/332) compared to six-lead ECG (2/341) (P=0.02). There was a strong association between reference ECGs and wearable-recorded ECGs (P<0.001): Cramer‘s V 0.91, (95% CI 0.82–1.0) for single-lead ECG and 0.98 (95% CI 0.89–1.0) for six-lead ECG. After including a control group of frequent premature contractions, single-lead ECG (12/344) was more frequently labelled "Inconclusive" than six-lead ECG (3/344) (P=0.01). Inter-rater agreement, measured as Cohen’s kappa, indicated great concordance in both methods but was higher for six-lead ECG (0.945, P<0.001) than single-lead ECG (0.887, P<0.001). Conclusions Six-lead ECG of a wearable device demonstrated higher diagnostic accuracy of AF detection than single-lead ECG when controlled by patients with frequent premature contractions. The performance of both methods was equivalent when controlled by patients with stable SR.