A novel software algorithm compatible with a wide range of smartwatches and wearable fitness trackers successfully detected atrial fibrillation (AF) in users of such devices with a positive predictive value of 98% in the Fitbit Heart Study.
"This is a higher positive predictive value than seen in previous studies of other algorithms used in a similar way, and suggests that such an approach could be used for the large-scale identification of undiagnosed AF," said study author Steven Lubitz, MD, MPH, associate professor of medicine at Massachusetts General Hospital, Boston.
"We found that individuals with irregular heart rhythm detected on this algorithm tend to be at elevated risk of having AF on subsequent ECG patch monitoring and have a considerable AF burden that warrants clinical assessment," Lubitz told theheart.org | Medscape Cardiology.
"I would say that our results show that if someone wearing one of these devices has a notification of an irregular heart rhythm, then they do have a high likelihood of having AF. If this occurs, they should contact their doctor and request an ECG workup," Lubitz said.
"At a population level, these algorithms now give us an opportunity to detect undiagnosed AF in a large number of individuals who use these devices. They can then seek medical care to reduce the downstream morbidity from arrhythmias," he added.
Results from the Fitbit Heart Study were presented at the American Heart Association (AHA) Scientific Sessions on November 14.
The trial is similar to the Apple Heart Study, which was reported in 2019 and showed similar results.
Lubitz explained that undiagnosed atrial fibrillation may cause morbidity that could be prevented with early detection. Smartwatches and fitness trackers are common, and many have optical sensors to measure heart rate. Software algorithms that passively analyze pulse data can infer the presence of AF, but correct classification is critical to minimize false-positive notifications and downstream adverse events, he said.
Lubitz and colleagues have developed a novel software algorithm with frequent overlapping pulse tachygram sampling. The algorithm is owned by Fitbit, and the company is seeking U.S. Food and Drug Administration approval for its use. Other similar algorithms are already available — such as the one used in the Apple Heart Study, which is now being used quite extensively.
The current study tested the positive predictive value of the Fitbit algorithm with a range of wearable devices for detecting undiagnosed AF.
Participants for the study were recruited and enrolled electronically and remotely via in app notifications and via the Internet.
Individuals participating in the trial who received an irregular heart rhythm detection were notified by an in-app notification and email and invited to schedule a visit with a telehealth provider via their Fitbit app on their smart phones. They were sent an ECG patch monitor, which was self-applied and worn for 1 week and then returned via mail. They were then invited to schedule a second telehealth visit to discuss the result of the patch monitor.
All participants who had an irregular heart rhythm detected were invited to complete a 90-day postnotification survey. And all participants in the study were invited to complete a survey at the end of the study.
Patients were included if they were age 22 years or older, were a U.S. resident, had a compatible Fitbit device (fitness tracker or smartwatch), and a smartphone with an installed Fitbit app. Excluded patients had prior self-reported AF or atrial flutter, were receiving therapy with oral anticoagulants, or had a pacemaker or defibrillator.
Lubitz reported that the algorithm continuously samples the pulse data in 5-minute blocks that overlap each other by 50%. "If 11 out of 11 consecutive 5-minute blocks are irregular, then that signals an irregular heart rhythm detection. By design, that means that at a minimum that the algorithm requires at least 30 minutes of irregular rhythm to detect AF," he noted.
The algorithm operates only if the participant is inactive, and that is judged by accelerometers on the device. The algorithm resets with a normal 5-minute tachygram.
The study enrolled 455,669 participants. Of these, 4728 had an irregular heart rhythm detection notification, and 1671 completed the first telehealth visit, with 1409 eletrocardiography (ECG) patches shipped. Of these, 1162 were returned and 1057 were included in the ECG monitor analysis, which showed AF in 340 individuals.
The end-of-study survey was completed by 24,532 participants; 1504 completed the 90-day survey after having had an irregular rhythm detected, including 225 with irregular rhythm detected during ECG endpoint analysis.
Irregular Rhythm Detected in 1%
Results showed that about 1% of participants overall had an irregular pulse detection. That fraction was greater among men (2.1%) than women (0.6%) and in those age 65 years or older (3.6%) compared with those under 65 years (0.7%).
Among individuals who had an irregular heartbeat detection and subsequently wore an ECG patch monitor, 32% had AF confirmed on the monitor.
"That's a marked enrichment compared to other trials of screening for AF, which use ECG patch monitors without any type of pulse irregularity pre-screening," Lubitz stated. "Previous studies without pre-screening have shown rates of less than 5%."
The primary endpoint was the positive predictive value of the irregular heart rhythm detection algorithm for concurrent AF on the ECG patch monitor. The positive predictive value was 98% and was stable across sex and age.
"This means that when the algorithm fired while an individual was wearing an ECG patch at the same time, it correctly detected AF 98% of the time," Lubitz explained. "This is the highest predictive value reported to date in studies with a similar design," he added.
"The figure of 32% comes from the percentage of people who had an irregular rhythm detected on the device and AF was subsequently detected on an ECG patch. The irregular rhythm could have occurred several days before the patch was worn. This lower figure is not surprising as AF may not always be present. It can come and go, and this type of AF is more difficult to pick up," Lubitz noted.
The patients who had AF on the patch monitor also had a relatively high burden of AF (7%), Lubitz reported. "That means the patient is in AF for 7% of the time." In contrast in previous studies in which ECG patch monitors have been used without this type of pre-screening, the burden is often about 1% or less, "so we are definitely picking up a high-risk population," he said.
"The median duration of the longest episode during the ECG patch monitoring was 7 hours, showing a substantial burden of AF," he commented.
Lubitz explained that if AF is confirmed, then the patient would undergo a standard assessment for their risk for stroke, which will lead to a decision on whether to prescribe an anticoagulant.
"We know that there is a relation between the amount of time a patient is in AF and the risk of stroke, but what is not well understood is where the threshold lies. But what we can say from this study is that the AF episodes that were detected are not trivial," he stated.
Asked if the technology is applicable to groups most at risk for AF (as these sorts of wearable devices are used mainly by the young), Lubitz replied that many older people also wear these devices — 13% of the participants in the current study were older than 65 years of age.
Comparison With Apple Study
Designated discussant at the late-breaking science session, Mintu Turakhia, MD, director of the Stanford Center for Digital Health in California, noted that the Fitbit Heart Study was very similar to the previously conducted Apple Heart Study, but the Fitbit app/algorithm can be used with both Apple and Android devices.
He pointed out that the two studies enrolled a similar number of participants, but the Fitbit study included more women (71%) than the Apple study (42%).
Although the Apple study showed a lower percentage of notifications (0.5% of the total population enrolled compared with 1% in the Fitbit study), the percentages of those older than 65 years in whom an irregular heartbeat was detected was similar in both studies (about 3%).
The positive predictive value for simultaneous AF was higher in the Fitbit study than the Apple study (98% vs 84%), but the AF yield on subsequent ECG patch testing was similar in the two studies.
"So, both of these algorithms are identifying individuals with AF early in the course on the disease," Turakhia said.
But he cautioned that there was a problem of engagement, with only 35% to 44% of those receiving a notification of an irregular heart rhythm going forward to request an ECG patch in the studies. He also noted that the Apple study did not lead to an increase in outpatient visits for AF. "This tells us that more work needs to be done here," Turakhia concluded.
Will It Change Outcomes?
Discussing the study at an AHA news briefing, Sana M. Al-Khatib, MD, professor of medicine at Duke University Medical Center, Durham, North Carolina, said, "We know that if we screen patients, we will detect AF. We also know that patients with AF have worse outcomes. But what we don't know is whether treating these patients would change those outcomes. This study does not tell us if detecting AF in this way leads to any changes in management of these patients or improvements in outcomes."
She added: "In this study, they really lost a lot of participants throughout the study. While a lot of people had an irregular heartbeat detected, not many ended up having an analyzable ECG on patch monitoring, so it is important to keep that in mind when interpreting the results."
However, Al-Khatib concluded, "I do think that this study shows that something like a Fitbit can be used to look for irregular heartbeat."
Noting that the algorithm can be used only when people are inactive, she questioned whether it would pick up sympathetically driven AF, which is mostly triggered with activity.
Lubitz responded that most of the available algorithms require that participants be inactive at the time their pulse is sampled. "That is not unique to this particular algorithm. It is more a limitation of the current technology. But one could imagine scenarios or algorithms in the future that can interpret data during periods of activity."
Moderator of the news briefing, Elaine Hylek, MD, professor of medicine at Boston University School of Medicine in Massachusetts, said doctors were trying to figure out how to use the information from these studies.
"What population should we be offering this type of screening to? Is it the older patient with diabetes, heart failure and hypertension — the group that is going to be at highest risk?" she asked.
She also pointed out that it wasn't certain what a short run of AF meant in terms of the clinical relevance. "Paroxysmal AF is still, I would say, an unstudied area. We can't necessarily extrapolate that to the persistent and permanent AF group, which is the population that has given rise to our stroke risk models in large part."
But Hylek added: "As a consumer myself, I would be thinking that if I see some AF on a wearable device, I am going to ask myself whether I should be exercising more and losing weight or decreasing my salt as blood pressure is such a huge trigger for AF. So, I think that is quite exciting."
Lubitz explained that the study was aiming to show whether the algorithm would be able to accurately detect AF among existing users of wearable devices. "And we think the result of that is yes."
But he added that he could also imagine a scenario where physicians are prescribing wearable technology or payors are supporting the reimbursement of wearable technology to detect AF. "But those questions have yet to be answered," he said.
The Fitbit Heart Study was funded by Fitbit. Lubitz reports ho noraria from Bristol-Myers Squibb/Pfizer and Blackstone Life Sciences and research grants from Bristol-Myers Squibb/Pfizer, Bayer AG, Boehringer Ingelheim, IBM, and Fitbit.
American Heart Association (AHA) Scientific Sessions 2021. Presented November 14, 2021. LBS4. Abstract
Lead Image: Dreamstime
Medscape Medical News © 2021 WebMD, LLC
Send comments and news tips to firstname.lastname@example.org.
Cite this: Fitbit Algorithm Successfully Detects AF - Medscape - Nov 14, 2021.