Apple Heart Study forms ‘solid foundation’ for AF detection via smartwatch
NEW ORLEANS — Anticipated results of the Apple Heart Study, which evaluated how well a smartwatch can identify and prompt subsequent clinical evaluation, were presented before a packed room at the American College of Cardiology Scientific Session.
The large-scale, app-based study — led by researchers at Stanford University School of Medicine — enrolled nearly 420,000 participants over 8 months.
“In the Apple Heart Study, we found a low proportion of notifications [of AF] across a diverse population,” Marco Perez, MD, associate professor of medicine (cardiovascular medicine) at Stanford University Medical Center, said during the late-breaking clinical trial presentation.
“These findings may inform further clinical evaluation after notification [of AF] when taking into consideration things like the history, physical examinations and other clinical observations. We need rigorous investigation of this technology and its potential use in the clinical setting. We believe that the Apple Heart Study provides a solid foundation upon which further research in digital health can be conducted.”
Participants in this study had to have an iPhone 5S or higher and an Apple Watch Series 1, 2 or 3. Of note, the study did not look at the Apple Watch Series 4, which features an electrical heart rate sensor that can take an ECG using a dedicated ECG app.
Virtual study design
The Apple Watch Study had a unique design.
The prospective, single-arm, open-label, nonsignificant risk study included 419,297 self-enrolled participants who were free from AF and atrial flutter at baseline. The participants’ mean age was 41 years and 42% were women. Enrollment was completed from Nov. 29, 2017, to July 31, 2018, with the last data collection on Feb. 25.
“Due to the wide consumer base of these devices, this provided us an opportunity to have a pragmatic design and to think of ways to have scalable study procedures such as virtual telehealth study visits,” Mintu P. Turakhia, MD, MAS, associate professor and executive director of the Center for Digital Health at Stanford University School of Medicine and a Cardiology Today Next Gen Innovator, said during the presentation.
The researchers evaluated a mobile app that uses the smartwatch’s existing light sensor — or photoplethysmography — to intermittently measure blood flow activity and detect subtle changes that might indicate an irregular contraction or heartbeat. The Apple Watch generates a tachogram, which Turakhia described as a plot of time between heartbeats. If an irregular tachogram is present, then more frequent tachogram collection occurs. The tachograms are then analyzed via algorithm to determine if an irregular rhythm is present and, if so, a notification is sent to the Apple Watch. The participant is notified if there are five of six confirmed incidences suggestive of AF.
“This tachogram serves as the building block of the entire five out of six algorithm” Turakhia said.
Upon notification of potential AF, the participant is prompted to contact a telehealth physician via the app, available 24/7 in the United States. Based on a video consultation on the participant’s device, he or she would be directed to seek urgent or emergency care and wear an ECG patch (BioTelemetry), sent by mail, for up to 7 days. After 7 days, participant and telehealth physician discuss the results.
In total, 658 participants were sent an ECG patch in the mail after notification of potential AF. Of those, 450 patches were returned and included in the analyses presented here.
The main study goals were to determine: the proportion of participants with an irregular pulse smartwatch notification who have AF on a subsequent ECG patch worn at home; how well the algorithm on the smartwatch matched the ECG findings; and the percentage of participants who received a notification who went on to seek medical help through the app.
The co-primary endpoints were AF longer than 30 seconds on an ECG patch among participants aged older than 65 years and simultaneous AF on both the ECG patch and the individual tachogram. Secondary endpoints included simultaneous AF on the ECG patch with notification on the Apple Watch and self-reported contact with a health care provider.
“This was a virtual study design, which allowed us to do this large study very quickly and very dramatically,” Perez said.
Low proportion of notifications
Of the 419,297 participants who self-enrolled in the Apple Heart Study, 0.52% —2,161 — received an irregular pulse notification. The number of initial irregular pulse notifications “ranged from 3.2% in the older population over 65 [years] to 0.16% in the younger population [younger] than 40,” Perez said. Additionally, men tended to have more notifications than women. Perez said 89% of participants who had any AF on ECG patch monitoring had at least an hour-long episode of AF.
Forty-four percent — 945 participants — who received a notification went on to attend the first telehealth study visit. The ECG patch was mailed to 70% of participants (n = 658) and 68% returned their patch for analysis. Mean time to hookup was 13 days and mean wear time was 6.3 days.
Of 450 participants who received an ECG patch to wear at home, 34.8% — 153 participants — had observed AF (97.5% CI, 27-43). Observed AF was “comparable across age groups, except in the young population, where only 18% had AF on patch monitoring,” Perez said. Again, AF on patch monitoring was higher among men.
When comparing pulse detections on the smartwatch with simultaneous ECG patch recordings among 450 participants who wore both at the same time, the positive predictive value for the tachogram was 71% and the positive predictive value for the notification was 84%. Thus, if a notification occurred, 84% of the time the ambulatory ECG monitor also showed AF, according to a press release.
“The notification’s positive predictive value of 0.84 supports the ability to correctly identify atrial fibrillation among those notified,” Perez said here.
All participants who received a notification of irregular heartbeat were asked to respond to a survey at 90 days to collect insights into the care pathway used in the Apple Heart Study; 64% responded. Of those, 15% admitted that they had AF before they self-enrolled in the Apple Heart study. There were 57% of patients who reported contacting a non-study provider. Of those, 28% started a new medication, 33% were referred to a specialist and 36% underwent additional testing.
The researchers noted several limitations of the study, including participants’ self- enrollment of inclusion criteria and self-reported outcomes, as well as higher-than-anticipated drop-offs after initial notifications.
“More than just a pilot study”
A panel discussion followed presentation of the Apple Heart Study, during which panelists considered generalizability of these results and the technology.
There is a “great deal of hype around this technology,” Patrick T. O’Gara, MD, from the division of cardiovascular medicine at Brigham and Women’s Hospital, the Watkins Family Distinguished Chair in Cardiology Professor at Harvard Medical School and former ACC president, said.
While questions of generalizability came up, Perez noted that “the study was really designed to help clinicians know what to do when somebody shows up with a notification. Even with progress made through this study, there are some concerns regarding unnecessary treatment with this technology.”
“The trial process provides a model for trials in the future, using patient-facing apps which may enhance enrollment,” Jeanne E. Poole, MD, physician at the cardiology clinic, director of cardiac electrophysiology and professor of cardiology and medicine at University of Washington in Seattle, said during the panel discussion. “I caution that listeners should not mistake that this trial has proven that we know how to treat subclinical asymptomatic AF. We have ongoing randomized trials that are evaluating this question. The ‘potential flood’ of patients contacting their physicians could lead to unnecessary testing and even inappropriate therapy.”
Detection of AF in young patients remains important even with the potential for unnecessary treatment, especially in the prevention of stroke and other events.
Considering these factors — and others — the Apple Heart Study, as a whole, provided more information for next steps and future directions with these technologies.
“This is a really impressive effort that helps us start to understand the challenges of managing big data and, more importantly, transforming data that everybody has into meaningful information,” Renato D. Lopes, MD, PhD, professor of medicine at Duke University School of Medicine, said. “[This] study adds several pieces to this complex puzzle."
Beyond the data, the Apple Heart Study design opens doors to further virtual studies.
“This was a large experiment in so many ways,” Turakhia said. “If you think of it, we really have no reference. What is the reference to do a virtual end-to-end study where your enrollment goals are ambitious for a consumer-facing technology [and] where you want to evaluate its performance but of course safety? ... This was an opportunity to study this in a way that we think would afford significant generalizability to the total population for people who would be using these devices.” – by Darlene Dobkowski, with additional reporting by Katie Kalvaitis
Turakhia MP. Joint American College of Cardiology and Journal of American College of Cardiology Late-Breaking Clinical Trials. Presented at: American College of Cardiology Scientific Session; March 16-18, 2019; New Orleans.
Disclosures: The study was sponsored by Apple. Perez reports he receives consultant fees/honoraria and research support from Apple. Poole reports she receives consultant fees/honoraria from Boston Scientific and Medtronic and research support from AtriCure, Biotronik and Kestra. Lopes reports he receives consultant fees/honoraria from Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi Sankyo, GlaxoSmithKline, Medtronic, Merck, Pfizer and Portola and research support from Bristol-Myers Squibb, GlaxoSmithKline, Medtronic and Pfizer.