Validation of an algorithm for atrial fibrillation detection with an analog smartwatch: prospective interventional clinical study (Preprint)

Atrial Fibrillation (AF) affects about 4% of the World's population and is one of the major causes of stroke, heart failure, sudden death, and cardiovascular morbidity. It can be difficult to diagnose when asymptomatic or in the paroxysmal stage, and its natural history is not well understood. New wearables and connected devices offer an opportunity to improve on this situation.To validate an algorithm for the automatic detection of AF from a single-lead electrocardiogram (ECG) taken with a smartwatch.Eligible patients were recruited from 4 sites in Paris, France. Twelve-lead reference ECGs and single-lead ECG were captured simultaneously. The ECGs were reviewed by independent, blinded board-certified cardiologists. The sensitivity and specificity of the algorithm to detect AF and normal sinus rhythm (NSR) were calculated. The quality of single-lead ECGs (visibility and polarity of waves, interval durations, heart rate) was assessed by comparison to the gold standard.Two hundred and sixty two patients were included in the final analysis: 100 AF, 113 NSR, 45 Other arrhythmia, 4 presented unreadable ECGs. Mean age was of 74.3 years ± 12.3 in the AF group versus 61.8 years old ± 14.3 and 66.9 years old ± 15.2 in the NSR and other arrhythmia groups respectively. 6.9% (18/262) were classified as "Noise" by the algorithm. Excluding "Other" arrhythmia and "Noise", the sensitivity to detect AF was of SeAF/NSR = 0.963 (0.894), and specificity of SpAF/NSR = 1.000 (0.967). Visibility and polarity accuracies (1-lead ECG vs 12-lead ECG) were respectively: P-waves: 96.9%/100%, QRS-complexes: 99.2%/98.8%, and T-waves: 91.2%/99.5% . P-wave visibility accuracy was of 99% (99/100) in AF patients and of 95.7% (155/162) when excluding AF patients. The absolute values of the mean difference in PR duration and QRS width were below 3ms, and more than 97% of these differences were below 40ms. The mean difference between the HR calculated by the algorithm and the device 1-lead ECG read by cardiologists was 0.55 bpm.Withings algorithm demonstrated great diagnostic performance for AF detection. The smartwatch single-lead ECGs also demonstrated good quality for physician use in daily routine care.ClinicalTrials.gov registration number: NCT04351386.