Editor’s note: In part one of this two-part series, we explored the history of direct-to-consumer ECG technology and its utility. Here we present a product comparison and research validating direct-to-consumer ECGs.
A Comparison of the Devices
In the spring of 2019, AliveCor released the first six-lead ECG to the direct-to-consumer ECG market. The KardiaMobile 6L displays leads I, II, III, aVL, aVR and aVF via utilization of a rear electrode on the left leg. Previous Kardia products only used two front electrode finger pads. Alivecor was the first major company to begin selling pocket-sized ECGs, the KardiaMobile and KardiaBand, to the public. The “Mobile” is smaller than a credit card, has two finger pads and can clip onto a phone or slide into a wallet. The “Band” was a wristband accessory for the Apple Watch that was discontinued in August 2019. These products were originally designed and approved to help detect A-fib, an arrhythmia affecting 2.3 to 5 million Americans that increases the risk of stroke five-fold.
Alivecor also offers services to physicians. Doctors can subscribe to Kardia Pro for a live cloud dashboard of their patients’ Alivecor data. In fact, Kardia Pro is registered as a billable service by Medicare and Medicaid. Without the cloud service, doctors must manually receive individual readings sent from their patients.
Apple released two ECG features for the Apple Watch in December 2018, likely leading to the discontinuation of AliveCor’s KardiaBand. One is a new finger sensor built into the crown of the Series 4 (and Series 5, debuted September 2019). The second feature is a new algorithm — an app — to recognize patterns from the watch’s existing pulse reader on the wrist. This feature provides the capability to monitor heart rhythm in addition to heart rate, similar to the way a physician assesses a radial pulse on physical exam. The pulse reader, an optical sensor on the wristband rather than an electrode, has been a staple of the Apple Watch since the Series 1 in 2016. Apple currently sells the Series 5 and 3; users of earlier generations can use the pulse feature with a software upgrade.
|KardiaMobile||Two-sensor finger pad
Two-sensor finger pad with third rear sensor for leg
|One at a time: I, II, and “Anterior Precordial”
Six leads simultaneously: I, II, III, aVL, aVR, and aVF
|Apple Watch Series 5 and 4||Crown sensor and wristband optical sensor
Wristband optical sensor
None. Pulse pattern recognition
|WIWE||Two-sensor finger pad||I||~$340|
“Our patients are exposed to a growing number of healthcare devices that are available for purchase without a prescription,” Joseph Bumgarner, MD, told the American College of Cardiology (ACC) in a press release. He is affiliated with UNC Rex Hospital and completed his fellowship in electrophysiology at Cleveland Clinic. “What we don’t know is if these devices provide meaningful clinical information that patients and their physicians can interpret and use effectively,” he says.
The Apple Heart study, an ongoing project between Stanford University and Apple, is one study building on that.
The Early Research to Validate Current Uses
Before studies can begin to compare devices, they need to evaluate individual devices. The Apple Watch is FDA “cleared,” which accepts a baseline of safety and effectiveness compared to similar products. But clearance isn’t as rigorous as FDA “approval.” Both cleared and approved devices undergo continued evaluation analogous to post-approval trials of common prescription medications.
Ongoing Research on the Apple Watch: The Apple Heart Study
The Apple Heart Study evaluates how well the Apple Watch’s optical sensor and app can pick up patterns in heart rate such as the irregularly irregular rhythm of A-fib. Peter Kowey, MD, described the purpose of the study in an interview with the author: “How good is the optical sensor at recognizing a normal increase in heart rate versus abnormal? [This study is] really the validation of [identifying] arrhythmias using the Watch,” he says. The study does not assess the one-lead ECG on the crown, however. Kowey is a member of the scientific steering committee of the study and is the chairman emeritus of cardiology at Lankenau Heart Institute.
“The advantage of the app that uses the optical sensor is that it can check for an irregular pulse multiple times throughout the day in the background without needing the user to actively engage the application,” says cardiologist Marco Perez, MD, in a Stanford press release. Perez is an assistant professor of cardiovascular medicine at Stanford and is a principal investigator on the Apple Heart Study.
In November 2019, the study shared its official peer-reviewed publication in the New England Journal of Medicine and in a Stanford press release (Large-Scale Assessment of a Smartwatch to Identify Atrial Fibrillation). The sensor sent irregular rhythm notifications to only 0.5% of all participants.The study also found that the watch’s pulse detection had a positive predictive value (PPV) of 71-84% compared to simultaneous tachograms and ECG patches.
However, only 35% of users who had warnings actually later received confirmed diagnoses A-fib by one-week ECG patch later; this at least partially because A-fib is intermittent, so people may have merely lacked A-fib events during the confirmation follow-up. Overall, the 71-85% PPV means the watch is likely accurate, but the 35% later confirmation rate means it has not yet solved the grey area of consistently catching and diagnosing A-fib.
Thirty-five percent of the over 400,000 participants does translate to approximately 1,400 people with a new confirmed diagnosis of A-fib. It is unknown how many people would otherwise receive an A-fib diagnosis in their lifetime without the aid of the Apple Watch. Also, a PPV of 71-85% still means there were some false positives — around 200-400 people without A-fib received an abnormal rhythm alert and were unnecessarily sent to a doctor for follow-up.
Completed Research on the KardiBand
In 2018, Dr. Bumgarner published a completed study on the KardiaBand algorithm in the Journal of the ACC. Subjects received a KardiaBand ECG which had been immediately preceded by a 12-lead ECG for comparison. The study found that the KardiaBand had 93% sensitivity (SN) and 84% specificity (SP) for the rhythms it could classify on its own, but one-third of the time the device couldn’t decide between A-fib versus sinus rhythm (“unclassified”). Physicians could interpret most of these without the need of a 12-lead ECG. However, most unclassified Kardiaband rhythms (34 of 57, 60%) were sinus rhythm, meaning a patient would have had an unnecessary extra doctor interpretation, visit or ECG.
General users are left to decide whether further follow-up is necessary. Overall, 10% of 169 total recordings were not interpretable by the algorithm or a physician. Currently, the FDA clearance is for users with no prior history of A-fib or other arrhythmias, reflecting an informational rather than monitoring or diagnostic utility. Bumgarner’s study tried to expand on that, using subjects with a history of A-fib who were scheduled for cardioversion.
In March of 2019, a JAMA Clinical Update illuminated some unknown limitations of consumer handheld ECGs and suggested ways that physicians should follow-up on users’ findings. Electrical, sound, motion and inadequate skin contact can interfere with readings. Additionally, current devices are only designed to interpret A-fib for heart rates inside the 50-100/min range. This limitation could result in failure to detect A-fib with rapid ventricular response (RVR).
Further, A-fib alerts could be false-positives that are actually other irregular rhythms like premature beats or AV block. Dr. James E. Ip, MD, recommended in the article that, “if an [A-fib] alert is detected, confirmation with direct ECG measurement is essential.” Ip is a practicing electrophysiology specialist with Cornell University Medical Center.
A New World for Gathering Research Data
Personal Devices Transform Research Recruitment
Personal devices offer highways to big data not only for patients and doctors, but also for researchers. Technology like smartphones and smartwatches offer researchers a simple way to contact the public at a large scale for recruitment. The Apple Heart Study announced in early 2019 that it recruited over 400,000 people. Stanford published the study design and called it “the largest screening study on atrial fibrillation ever done.” All Apple Watch users over the age of 21 received an email offering them to participate. They were notified when the data collection concluded.
“This is a landmark study because of its size and because it’s basically a virtual study,” Kowey says.
The next products hope to broaden diagnostic capabilities beyond A-fib. Alivecor is focusing on detection of prolonged QT intervals, rhythms associated with hyperkalemia and electrocardiographic signs of heart attacks. Long QT intervals and hyperkalemia predispose patients to fatal heart rhythms.
Alivecor has partnered with a team from the Mayo Clinic to see if their devices could pick up STEMI heart attacks. In November 2018 at an American Heart Association conference, they presented an abstract with statistics for their early work. The abstract might suggest the devices could help rule out some heart attacks, but that’s not as helpful as accurately finding them. After a discussion with a biostatistician, I feel that the results may be rushed; the tables are difficult to trace to their original data and may contain miscalculations.
A cardiologist, Dr. Daniel Frisch, MD, also weighed in, saying, “my best interpretation of the abstract is that it is ‘hypothesis generating.’” Frisch is a clinical cardiac electrophysiologist at Jefferson University Hospital who uses Alivecor products with patients for clinical purposes and for research. “I am all for community outreach, but 12 lead ECGs are still the standard,” he says.
Through research, tech companies and physicians may find the proper utility for these direct-to-consumer devices, much like the way the utility of Viagra (sildenafil) to treat erectile dysfunction was discovered while studying its effect on hypertension and angina. Seamless accessibility and the promise of billable cloud-based monitoring demonstrate the endless potential of these devices. However, these devices currently wade in the dissatisfying cliché of “more research needs to be done.”