A watch to monitor atrial fibrillation?

A recent study analysed the performance of the 'Study Watch' device, that continuosly monitors atrial fibrillation and estimates its severity.

Study Watch: a wearable for atrial fibrillation management

Wearable devices (or wearables) will be making more valuable contributions in healthcare. In fact, wearable devices worn by patients will enable the real-time collection, analysis, and transmission of personal health data. There are numerous devices with health-related functions on the market today, but these are rarely integrated into clinical practice for the diagnosis and monitoring of diseases.

The 'Study Watch', on the other hand, already offers a practical solution for AF detection, assessment of disease severity, and on-demand recording of Lead-I ECGs in response to irregular rhythm notifications. This device could complement current disease management modalities by enabling the monitoring of AF over extended periods, beyond the possibilities offered by current non-invasive technologies.

In the clinical setting, the 'Study Watch' could be used for recording ECG events and continuous AF monitoring, facilitating screening, diagnosis, assessment of disease severity, optimisation of drug therapy and characterisation of episodes in AF patients.

Validation of a deep learning algorithm to enable AF detection with a wearable

Ambulatory ECG monitoring, a common method for monitoring the disease, may have limitations in accurately detecting and estimating the level of AF due to its intermittent nature. Implantable devices improve diagnosis, but are expensive and require invasive procedures.

Therefore, AF monitoring through wearable devices may be important to identify undiagnosed episodes of AF and support clinical management decisions, such as administration of anticoagulants, restoration of normal heart rhythm and control of ventricular rate.

In this context, the 'Study Watch' is an FDA-approved wrist-worn device that records single ECGs in response to events or symptoms and continuously monitors AF using a photoplethysmography-based algorithm. A study conducted by the Frankel Cardiovascular Center, University of Michigan Health, tried to validate the performance of this new algorithm against a reference ECG monitoring device in an outpatient setting over a 14-day period.

Monitoring studied for 14 days

The authors conducted a multicentre prospective study to evaluate the effectiveness of the 'Study Watch' and its photoplethysmography algorithm in detecting irregular heart rhythms suggestive of AF in subjects at risk of AF events in an outpatient setting. Patients wore the 'Study Watch' together with a continuous ECG monitoring device for 14 days.

The "Study Watch" is capable of acquiring a user-activated ECG and measuring physiological parameters, providing the ability to continuously monitor suspected AF events via the photoplethysmography sensor. Suspected AF events generate real-time notifications, encouraging the user to perform an ECG. Its operating model incorporates real-time prompts to record a single ECG when the algorithm suspects AF. Next, the ECG undergoes automatic classification, followed by a human review by certified telemedicine technicians. The complete report is then transmitted to the prescribing physician, thus reducing workload and false positives.

The study involved patients with a history of paroxysmal AF. The primary outcome was suspected episodes of AF in 15-minute intervals. Sensitivity and specificity of the study watch were the co-primary endpoints, assessed on analysable intervals. The analysis was conducted considering demographic subgroups and activity levels. The correlation between AF levels measured by the Study Watch and the reference device was assessed.

The study, conducted on 117 participants between September 2020 and May 2021, included 111 subjects in the per-protocol analysis. The majority wore ECG patches and the 'Study Watch' for 14 days, generating 91,857 evaluable intervals. Of the participants, 45% were women with an average age of 65±11 years. The average duration of use of the "Study Watch" was 18.3 hours per day.

Significant effectiveness in AF detection

The Study Watch photoplethysmography algorithm demonstrated a sensitivity of 96.1% and a specificity of 98.1% in the detection of AF. Performance was confirmed in demographic and physical activity subgroups. The algorithm identified subjects with ≥25% AF, demonstrating sensitivity of 90.6% and specificity of 60.8% at the subject level.

The estimation of AF severity based on photoplethysmography was highly correlated with the reference device, confirming the reliability of the Study Watch. The Study Watch ECG classification algorithm correctly identified all AF episodes, achieving a sensitivity of 100%, with a positive predictive value of 54.9% at the episode level.

The accurate estimation of AF severity offered by the device is an important feature. This finding is crucial for the optimal management of AF patients, as there is a correlation between the level of AF and symptoms, heart failure and stroke risk.

Subgroup analyses showed a lower sensitivity during vigorous physical activity. The presence of non-analysable intervals was limited (23.7% on average). False positives were mainly observed in subjects with low AF burden. In conclusion, the 'Study Watch' demonstrated an excellent performance in AF detection, confirming its usefulness in clinical practice, especially in high-risk patients requiring continuous ambulatory monitoring.

The authors concluded that the Study Watch emerged as a valuable tool for continuous AF monitoring in a home setting, not only for accurate detection but also for reliable estimation of AF severity. Its integrated approach, with the possibility to acquire ECGs in response to notifications of irregularities, shows promise in the advanced management of patients with a confirmed diagnosis of AF. Further prospective studies are needed to assess its impact on health and clinical outcomes.

  1. Poh MZ, Battisti AJ, Cheng LF, Lin J, Patwardhan A, Venkataraman GS, Athill CA, Patel NS, Patel CP, Machado CE, Ellis JT, Crosson LA, Tamura Y, Plowman RS, Turakhia MP, Ghanbari H. Validation of a Deep Learning Algorithm for Continuous, Real-Time Detection of Atrial Fibrillation Using a Wrist-Worn Device in an Ambulatory Environment. J Am Heart Assoc. 2023 Oct 3;12(19):e030543. doi: 10.1161/JAHA.123.030543. Epub 2023 Sep 26. PMID: 37750558.