Diagnostic accuracy comparable to that of cardiologists when signal is good
An automated algorithm for smartwatch devices accurately identified atrial fibrillation (Afib) optically, although adequate signal quality was often an issue, a prospective study found.
The photoplethysmographic (PPG) pulse wave analyzed by a novel automated algorithm on a commercial smartwatch and showed a sensitivity of 93.7% (95% CI 89.8-96.4), a specificity of 98.2% (95% CI 95.8-99.4), and an overall accuracy of 96.1% (95% CI 94.0-97.5) in identifying Afib compared with cardiologists' assessment of single-lead ECG from a Kardia Band on an Apple Watch.
The positive likelihood ratio was 52.1 (95% CI 23.5-160.7) and negative likelihood ratio 0.06 (95% CI 0.1-0.04), reported Marcus Dörr, of University Medicine Greifswald in Germany, and colleagues in JACC: Clinical Electrophysiology.
PPG is an optical approach to detect volumetric changes in blood in peripheral circulation. New smartwatches like the FDA-approved Apple Watchand AliveCor devices are able to detect arrhythmias when a user touches the built-in electrodes.
A passive, automated technology "could potentially raise rhythm screening to the next level as compared to previous approaches with smartphones by enabling convenient long-term screening or extended non-invasive rhythm analysis in individuals that are not suitable for other screening methods," the study authors wrote.
Early Afib detection and higher detection rates could lead to reduced stroke risk and initiation of the appropriate treatment. Improved and longer screening efforts have been known to lead to substantially higher detection rates, Dörr and colleagues noted.
Previous trials have found that using an external rhythm recording device led to more than a five-fold higher detection of Afib following a 30 day screening period for cryptogenic stroke patients. Invasive cardiac monitoring also led to high Afib detection rates. Although effective, these approaches are not always applicable for all Afib patients because of inconvenience and lack of reimbursement, among other reasons, the researchers wrote.
The researchers evaluated 508 patients (225 women and 237 with Afib) with a mean age of 76.4. They assessed and compared cardiologists' diagnosis using mobile internet-enabled electrocardiography (iECG) from a Kardia Band on an Apple Watch (using the method approved by the FDA for Afib screening) against that of automated diagnosis with an investigational algorithm on PPG data from the Samsung Gear Fit 2 smartwatch.
For comparison, the researchers also compared cardiologists' diagnosis to Kardia device's own automated iECG interpretation algorithm for identifying Afib. Sensitivity of the algorithm was 99.5% (95% CI 97.5-99.9), specificity was 97.4% (95% CI 94.7-98.9), positive predictive value was 97.6% (95% CI 95.0-99.0), negative predictive value was 99.2% (95% CI 97.3-99.9), and overall accuracy was 98.0% (95% CI 96.9-99.2).
Notably, 21.8% of the datasets were not suitable for PPG analysis (typically due to motion artifacts), and 15.1% were also uninterpretable by the automated iECG algorithm.
Looking ahead, further research is needed to determine "whether smartwatches may be a useful tool for convenient long-term AF [Afib] screening in patients with an increased pretest probability for AF must be evaluated in larger population-based samples of at-risk patients," the researchers wrote.
The U.S. Preventive Services Task Force does not back ECG screening for Afib in asymptomatic seniors or for cardiovascular disease in asymptomatic people, citing lack of evidence to support a clinical benefit and risks for low-risk people.
The trial was supported by the University Hospital Basel and Preventicus GmbH.
Dörr disclosed a relationship with Preventicus.
JACC: Clinical Electrophysiology
Source Reference: Dörr M, et al "(WATCH AF) trial smart(WATCH)es for detection of atrial fibrillation" JACC Clin Electrophysiol 2018; DOI: 10.1016/j.jacep.2018.10.006.
Read the original article on Medpage Today:Smartwatches Can Spot Afib With Automated, Optical Monitoring