By Deborah Borfitz

February 1, 2024 | A clinical-grade smartwatch does a good job of detecting possible atrial fibrillation (AFib), which could fill the need for a noninvasive option for monitoring patients for more than 30 days. Current available electrocardiography (ECG) technologies, such as Holter monitors and wearable event recorders, are both more time-limited and resource-intensive, according to cardiac electrophysiologist Hamid Ghanbari, M.D., assistant professor of internal medicine-cardiology at University of Michigan (U-M) Medical School. 

Ghanbari led a study, recently published in the Journal of the American Heart Association (JAHA, DOI: 10.1161/JAHA.123.030543), validating a deep learning algorithm in the Verily Study Watch for flagging heart irregularities. The prescription-only device was cleared by the U.S. Food and Drug Administration (FDA) in 2019 for continuous monitoring of AFib in healthcare facilities and recording single-channel ECGs that are event- or symptom-triggered.   

The Verily Study Watch, like its consumer-grade counterparts Apple Watch and Fitbit, detects subtle changes in heart rhythm via photoplethysmography (PPG), which involves sending light pulses through the skin into the blood vessels. It also has an ECG feature whereby users can opt to see if they have heart abnormalities via a pair of electrical contacts. 

Where the Verily Study Watch differs is its PPG-based algorithm, which succeeded in demonstrating it could correctly identify AFib in most participants with comparable levels of false negatives and positives as other PPG-based devices in clinical use. That gives it a major leg up over consumer-facing devices that have been generating a concerning number of false positives and needlessly sending many people to their doctor, says Ghanbari. 

Consumers often purchase smartwatches for cardiac monitoring purposes without giving much thought to how the readings might impact their care and the utilization of limited healthcare resources, he notes. Such patients frequently visit the University of Michigan Health Frankel Cardiovascular Center with what turns out to be a false-positive indication of AFib and it both increases their anxiety and overwhelms physicians with data they’re uncertain how to properly use. 

U-M has a digital health clinic that has started using wearable devices as part of the routine clinical management of patients with AFib, Ghanbari reports. Patients are given a device to use for the duration of their ECG monitoring period, with the incoming data interpreted by an expert clinician to decide if they have AFib.   

The devices on hand do not include the Verily Study Watch, he adds. Its developer, Alphabet’s Verily Life Sciences, is focused on conducting heart research in lieu of marketing the device to clinicians who are arguably the ultimate end users. 

Multistage System 

In the latest multicenter study, conducted between September 2020 and May 2021, 117 participants with intermittent episodes of AFib were monitored with both the prescription Study Watch and a reference device (Zio XT wearable heart monitor patch) for two weeks to evaluate the algorithm’s sensitivity and specificity for detection of AFib in 15‐minute intervals. Interval‐level sensitivity was reported to be 96.1% with a specificity of 98.1% for AFib detection. 

The smartwatch continuously scanned for the possibility of AFib and, when it was suspected, notified users of the irregular rhythm and prompted them to take a single-lead ECG using the Study Watch to confirm the ECG signal, explains Ghanbari. To get a baseline ECG on everyone, all study participants were asked to take an ECG daily. Incoming data from the ECG readouts were reviewed by a cardiographic technician before being reported to participants’ physician. 

The study provides a “framework” for how a medical-grade wearable device could be integrated into the diagnostic process for AFib and aid in clinical decision-making, he says. An experienced physician would ideally review the ECG data generated by the device, but cardiographic technicians are a good alternative since their job duties often include the monitoring and interpretation of readouts from standard ECG equipment.  

AFib, which is known to increase the risk of heart attack and stroke, is generally diagnosed when patients experience symptoms—e.g., palpitations, rapid heartbeat, lightheadedness, shortness of breath, and fatigue—and seek care, says Ghanbari. As part of the process, they are often given an in-office test using a 12-lead ECG machine that takes only 10 seconds to estimate heart rate using sticky electrode patches applied to the chest. 

But longer-term monitoring with a Holter monitor or event recorder is often required to make a diagnosis, he adds. If patients have fainting episodes or palpitations, and other tests haven’t provided any answers, doctors might also resort to an invasive procedure such as an implantable loop recorder that records the heartbeat continuously for up to three years. 

The consumer wrist-worn devices for AFib detection using PPG-based algorithms perform only periodic checks when the user is stationary and have been FDA-cleared only for prediagnostic uses and not for clinical decision‐making. They also have no system in place to deal with the false-positive results that are generated so that the devices might be incorporated into the care pathway of patients, says Ghanbari.

Going Digital 

While it is “always better” for individuals to be stationary during an ECG test, the Verily Study Watch performed quite well when participants were mildly or moderately active, reports Ghanbari. Performance was modestly affected with more vigorous activity (e.g, running) but did not affect the overall detection of AFib. 

PPG being a light-based technology, the algorithm had a slight decrease in performance among individuals with darker skin tones compared to people with very light skin tones, he says. But since clinically significant AFib is typically of long duration, the algorithm was ultimately able to detect AFib in all patients. “Even though it may miss some of those intervals [of AFib], that does not affect the time to diagnosis.”  

Conceptually, as depicted graphically in the JAHA paper, ECG and PPG data could be sent through a patient app to a cloud-based system where automated rhythm classification is reviewed by human experts. Validated findings would then be shared with prescribing physicians through reports. 

U-M’s digital health clinic is brand-agnostic, Ghanbari points out, and will serve as the intake point for patient data produced by multiple wearable health devices. Specialists there will be generating reports that get communicated back to patients and the physician involved in their care. 

Digital devices are “sure to be an integral part of how we will manage atrial fibrillation moving forward,” says Ghanbari. He has several clinical trials underway, and more planned, examining use of the Verily Study Watch in the clinical management of patients with a variety of cardiovascular conditions, including AFib as well as heart failure, Parkinson’s disease, and cancer.