In North America, 81% of adults own a smart phone and 17% own a wearable technological device such as a smartwatch. These devices continuously and passively collect data from users including pulse rates, location, and three-dimensional motion. The application of the data for medical purposes is the fundamental principle of mobile health. With advances in technology, leading companies including Amazon, Uber, and Apple are now entering this burgeoning field.
In the Apple Heart Study, recently published in NEJM, investigators designed a pragmatic trial to test the ability of an Apple smartwatch application to identify some of the estimated 700,000 patients in the United States with undiagnosed atrial fibrillation. Notably, the study strengths were its practicality and size: almost 420,000 patients were recruited remotely with consent, data, and communication transferred through the downloaded application, making it a “siteless” study. The watch utilized optical sensors to intermittently generate 1-minute “tachograms,” that were subsequently analyzed as regular or irregular. If an arrhythmia was suspected, the participant was prompted to initiate a telemedicine visit for a more detailed evaluation.
The goal of the Apple study was large-scale passive screening for a common disease and the results are both remarkable and disappointing. The study size, siteless design, and ability to engage the patient in their own care are remarkable. But only 0.52% of participants received irregular pulse notifications, likely reflecting the 41-year-old mean age of study participants. Furthermore, only 20.8% of notified patients sought further testing, and atrial fibrillation was confirmed in approximately one third of those patients, thereby bringing into question the clinical value.
Although groundbreaking, the study raises new questions about technology in healthcare. How do health care professionals deal with such large volumes of data? How do clinicians interpret findings using non-validated algorithms? How will patient privacy be maintained?
The following NEJM Journal Watch summary explains the study and results in more detail:
The Era of App-Detected Atrial Fibrillation Is Coming, but Evaluations Are Ongoing
Harlan M. Krumholz, MD, SM reviewing Perez MV et al. N Engl J Med 2019 Nov 14 Campion EW and Jarcho JA. N Engl J Med 2019 Nov 14
The size and methods of this Apple Watch study are groundbreaking, but it raises more questions than it answers.
The Apple Watch has an optical sensor that can detect heart rates, thus introducing the possibility of detecting atrial fibrillation (AF). The industry-sponsored, prospective, open-label, siteless, pragmatic Apple Heart Study tested an algorithm to identify AF (NCT03335800). The 419,297 adult U.S. participants enrolled via an app, owned Apple Watches and iPhones, and had no prior AF.
During the study, 2161 people were notified of an irregular pulse, of whom 79% were excluded for various reasons, including 1216 who failed to attend a telemedicine visit. The researchers urgently contacted 20 people: 18 with AF and a rate >200 beats/minute, 1 with a pause >6 seconds, and 1 with nonsustained ventricular tachycardia >6 seconds.
For confirmation, electrocardiographic patches were mailed to 658 participants with nonurgent symptoms. Participants began wearing the patches about 13 days after the notification, for about 6 days. Of 450 people who returned the patches, AF was confirmed in 153 (34%); 20% had continuous AF. The yield was higher in older than younger people. Of 293,015 participants who never received a notification and who returned an end-of-study survey, 3070 reported new AF diagnoses.
Comment: With a new algorithm and a later version of the Apple Watch, which allows for the recording of a real-time, single-lead ECG, the ability to track AF might have improved. Regardless, this study is groundbreaking because of its massive size and siteless method. However, as an assessment of this approach, the study has many limitations. Large percentages of patients were lost to follow-up. Also, the confirmation method's sensitivity is unknown — many patients might have had intermittent AF. Overall, the algorithm caught some AF episodes and missed others. As clinicians, we should not rely on the Apple Watch, but neither can we ignore these results. We certainly need more study about how to optimize such tools and, especially, how to best respond to brief, intermittent, subclinical AF episodes.
Browse more From Pages to Practice »
Ahmad is a 2019-2020 editorial fellow at the New England Journal of Medicine. He is from Toronto, Canada where he is completing his training in pulmonary medicine at the University of Toronto.