Giant Study Suggests Apple Watch Accurately Catches Atrial Fibrillation

This story has been updated.

Early results from a giant study of Apple Watch users show the wearable device appears to detect atrial fibrillation (AF) with a high degree of accuracy, and about half the users who got an AF alert said they contacted a doctor.

Overall, however, the number of users who got an AF alert was tiny—just 2161 or 0.5% out of more than 419,000 who enrolled in the Apple Heart Study, which kicked off the 68th Scientific Session of the American College of Cardiology Saturday in New Orleans, Louisiana. The study gave wearable technology equal billing at a meeting where drug studies typically dominate on the opening day.

The study also highlights the potential for virtual trials to quickly track and enroll patients—in just 8 months, researchers at Stanford University used telehealth to engage with more than 400,000 participants, who already owned their own Apple Watch (series 1-3) and compatible iPhone. Of note, the study did not evaluate the latest Apple Watch model, series 4, which has a built-in electrocardiogram (ECG) app.

Early detection of AF a common heart rhythm disorder, can potentially prevent blood clots, strokes, or other heart complications. Given the Apple Watch’s footprint—one estimate put lifetime sales at 46 million—the potential to find undiagnosed AF is significant. An estimated 6 million Americans have AF, but many do not know they have it. Apple funded the study.

The Apple study aimed to compare how well the Apple Watch’s algorithm detected AF compared with an ECG patch worn at home, as well as what share of patients who received an AF alert contacted a physician. More broadly, authors said the study aimed to tell clinicians what to do when patients show up in their offices after getting an AF alert from their Apple Watch.

“These parameters help us understand how we, as clinicians, should think about these notifications,” said Mintu Turakhia, MD, associate professor of cardiovascular medicine at Stanford School of Medicine and a principal investigator, in a statement.

The watch and its app use a light sensor technology to measure blood flow at discrete intervals and detect any changes, which could suggest an irregular heartbeat. It measures the interval between beats, called a tachogram. If 5 of 6 of these intervals over 48 hours falls outside the algorithm, the app triggers an alert to the user.

In this study, those who got an alert were asked to have a video consultation with a doctor, who decided whether the person should wear an ECG patch for 7 days. One-third of those who were sent the patch based on their AF reading on their watch were found to have it on the patch as well; the study authors said this is normal, as AF can “come and go.”

“It’s not surprising for it to go undetected in subsequent ECG patch monitoring,” Mintu Turakhia, MD, associate professor of cardiovascular medicine at Stanford School of Medicine and a principal investigator, said in a statement. “So, while only 34% of people who were still having [signs of] AFib on the ambulatory ECG, that doesn’t mean that 66% didn’t have afib. It just means that afib may not have been there at the time.”

During a press conference, Turakhia said the lack of false alarms during the study was significant, as this had been a concern going into the study. Researchers did report anxiety among 16 users as an adverse event.

What’s noteworthy is when users wore the patch and the watch at the same time, the positive predictive value for tachogram was 71%, and the predictive value of the notification was 84% (meaning if the watch notified the user of AF, so did the patch).

About half those receiving alerts got in touch with a study doctor, although some contacted their own doctor. In a follow-up survey, 57% said they got in touch with a clinician after getting a notification. Because the purpose of the study was to evaluate the safety of the Apple Watch algorithm in users who did not have known AF, the study excluded those already taking anticoagulants.

As large as the study was, researchers fell short of their goal of 500,000 enrollees, and slightly less than 25,000 were over 65 years of age, fewer than they had hoped.

The authors repeatedly acknowledged the limits of the study—it was not a randomized control trial (RCT), relied on self-reports, and participants enrolled themselves through Apple's website. Still, they say, the study forms a foundation for further work. During a discussion on the future of digital health after the main presentation, a Johnson & Johnson panelist announced plans for an upcoming RCT, which will also involve a novel pragmatic design based on claims data.

In an interview with The American Journal Managed Care®, co-author Marco Perez, MD, also a cardiac electrophysiologist at Stanford, acknowledged that the paradigm of the traditional RCT has kept some from understanding what the study was about and what the Apple Watch purports to do. It’s not a diagnostic device, Perez said, nor does it claim to be.

“This is not a medical device. This is not a continuous monitor,” said Turakhia during the press conference. The idea behind the study, he said, was to gauge whether the watch had the technical capacity to passively monitor users, and to assess safety and performance. While the early results are promising, he said, “We’re in the first half of the first inning.”

Cardiologists are keenly interested in wearables for their potential to shift patients from episodic to continuous care, with devices doing the monitoring between visits. No doctor can be with a patient 24/7, but devices can help prevent major medical events that cause significant harm and healthcare costs.

A panel discussion held immediately after the study presentation featured some skepticism about what the results meant. Turakhia was asked if this was a “proof of concept” study, and he said there was really no reference point for a virtual study like the one undertaken. One lesson learned: although the researchers tried to exclude people who already had AF, 15% of the participants who signed up lied and took part anyway. Another panelist asked what the types of events accounted for the remaining 16% of notifications—ongoing evaluations show these are other types of heart rhythms.

The later panel on the future of digital health—which featured a representative from Apple—was more enthusiastic about the study, particularly the virtual format. And while some noted the study was limited to those who could afford an iPhone and an Apple watch, others said new research models could take these digital tools to the most remote locations in the world. Bray Patrick-Lake, MFS, director of Stakeholder Engagement at the Duke Clinical Research Institute said she works with a practice group of young stroke patients and, “None of them would say they would rather not have known.”

Harlan Krumholz, MD, a cardiologist at the Yale School of Medicine, said during the discussion that people want digital tools that require no lessons or instruction manuals. At the press conference after the main presentation, cardiologist Matthew Martinez, of Allentown, Pennsylvania, said while the watch is not a diagnostic device, more of his patients are asking about it, especially if they face out-of-pocket costs for expensive mobile ECG devices. An older patient recently learned she had AF after getting an alert from an Apple Watch she received as a gift from her adult son, he said.

Martinez and Annabelle Volgman, MD, agreed the watch is not designed to replace traditional medical devices. But Volgman said given the lack of standards for widespread screening for AF, the Apple Watch could potentially detect a problem in someone who had no other known risk factors.

“We need data, and this is an incredible amount of data to look at,” she said.

Reference

Turakhia M, Perez M, Desai M, et al. Results of a large-scale, app-based study to identify atrial fibrillation using a smartwatch: the Apple Heart Study. Presented at the 68^th American College of Cardiology Scientific Session, New Orleans, Louisiana; March 16-18, 2019. Abstract 19-LB-20253.