Assessment of smartwatch-based electrocardiogram (ECG) abnormality detection among childhood cancer survivors.

Abstract

10069 Background: Childhood cancer survivors (CCS), exposed to cardiotoxic cancer therapies, are at lifelong risk for premature cardiovascular disease. However, adherence to guideline recommended screening is low later in life, long after cancer treatment. Novel wearable technologies offer a potential low-cost, easy screening method for arrythmias in CCS, including AF, bradycardia, and tachycardia. However, their utility in population level screening for CCS is unknown. Methods: We collected paired single lead smartwatch with FDA cleared ECG functionality and 12-lead 10 second ECGs from adult participants in the St. Jude Lifetime Cohort Study (SJLIFE) who were treated for their primary cancer 1962-2012 and survived ≥5 years. Participants completed an in-person comprehensive examination including a standard 10 second 12-lead ECG recording as well as single lead rhythms were collected using a smartwatch. The smartwatch provided automated annotations including sinus rhythm, AF, bradycardia (HR<60 bpm), tachycardia (HR>100 bpm), and inconclusive rhythm. We compared the smartwatch generated statements to reference diagnostic statements generated by GE MUSE system. Results: There were 598 same day ECG pairs in 580 participants (83% White, 14% Black, 50% male, and mean age(SD) 37(10) years). The mean(SD) times between smartwatch and reference ECG recordings were 32(50) minutes. The heart rate from reference ECG and smartwatch ECG had a Pearson Correlation of r=0.85 (p<0.001). The smartwatch ECGs presented statistically significantly (p<0.001) higher heart rates compared to reference ECGs with mean heart rate difference (95% confidence interval) of 1.2 (0.5-1.8) beats per minute. Standard 12-lead ECGs processed by GE MUSE annotations included 478 (80%) sinus rhythm and 120 (20%) as ‘no sinus rhythm’ with 1 (0.2%) AF, 57 (9.5%) sinus bradycardia, 18 (3.2%) sinus tachycardia and 28 (6.4%) other rhythms. The smartwatch assigned sinus rhythm to 590 (98.7%) of these ECGs. The detailed rhythm annotations between reference ECG and smartwatch ECGs are summarized in Table 1. The smartwatch detected only 3.5% (2 of 57) of reference bradycardia, none of the 18 reference tachycardia and 1 AF event. Overall, among 120 ECGs labelled as ‘no sinus rhythm’ by the reference device, only 3 (2.6%) of them were also labeled as ‘no sinus rhythm’ by the smartwatch. Conclusions: The smartwatch considered in this study produces heart rate that is not clinically different than heart rate calculated by a reference 12-lead ECG. However, the ECG abnormalities identified by reference ECGs were typically missed by the smartwatch. Comparison of GE MUSE and smartwatch ECG statements. Smartwatch Statements Sinus Rhythm Atrial Fibrillation Low Heart Rate High Heart Rate Inconclusive/Poor GE MUSE Statements Sinus Rhythm 478 0 3 0 2 483 Atrial Fibrillation 1 0 0 0 0 1 Bradycardia 55 0 2 0 0 57 Tachycardia 18 0 0 0 1 19 Other Rhythm 38 0 0 0 0 38 Total 590 0 6 0 3 598