Apple Watch May Detect A-fib in the General Population, but at What Cost?
Future studies will need to look at the impact on hard endpoints, clinical integration, as well as cost-effectiveness, experts say.
NEW ORLEANS, LA—A smartwatch-based app can detect A-fib among a population of all-comers, potentially escalating the diagnosis of the arrhythmia, according to a large study with a novel virtual design. The research has implications not only for the future of wearable technology in medicine, but also for physicians confronted by a “flood” of anxious patients, experts say.
“These findings may inform further clinical evaluation after notification when taking into consideration things like the history, physical examination, and other clinical observations,” said Marco Perez, MD (Stanford Medicine, Palo Alto, CA), who co-presented the results of the Apple Heart Study today in a late breaking session at the American College of Cardiology (ACC) 2019 Scientific Session. “We need rigorous investigation of this technology and its potential use in a clinical setting. We believe that the Apple Heart Study provides a solid foundation upon which further research in digital health can be conducted.”
Perez along with Mintu Turakhia, MD (Stanford Health), and colleagues enrolled 419,297 people with an Apple Watch and iPhone (Apple) to be monitored by an irregular pulse notification algorithm—of note, this is not the ECG feature that comes with the watch—between November 2017 and February 2019.
In all, 2,161 participants (0.52%) received a notification of an irregular rhythm, which was triggered if the sensor detected five out of six tachograms of an irregular pulse within 48 hours. Of these, 658 were sent a US Food and Drug Administration-approved ECG patch after a telehealth consultation with a study physician, and 450 returned it for analysis. A-fib was detected in 34% of the ECG patches, primarily in participants who were older, white, and male.
For patients wearing both an ECG patch and their watch simultaneously, the positive predictive values for the tachogram and the notification, respectively, were 0.71 (97.5% CI 0.69-0.74) and 0.84 (97.5% CI 0.76-0.92). Approximately 20% of these patients were determined to have A-fib 100% of the time, and 89% had A-fib durations of at least 1 hour.
In a survey of the patients who received a notification at 90 days, 57% said they had contacted a provider outside the scope of the study and 28% were started on a new medication. In the overall cohort, 1,038 adverse events were recorded, but most were not serious and only 16 were deemed to be app-related. Anxiety, not surprisingly, was a key adverse event documented in the study.
At this stage, Perez stressed that more information is still needed, especially regarding cost-effectiveness, before he would recommend use of this technology to everyone. “This study was really designed to help clinicians know what to do when somebody shows up with a notification,” he said. “Knowing that in the general population the rates of notification are very low, knowing that if somebody had a notification they probably have A-fib, and knowing that if they receive the [ECG] patch, they would identify A-fib 34% of the time, . . . then the clinician can make a judgement call.”
Big Data Understood
In a panel discussion following the presentation, Renato Lopes, MD, PhD (Duke Clinical Research Institute, Durham, NC), called the Apple Heart Study a “really impressive effort that I think helps us start to understand the challenges of really managing big data and more importantly transforming data that everybody has into meaningful information.”
Lopes added that the 34% rate of A-fib confirmed with the ECG patch could be deemed high, but this is likely “because you are getting A-fibs that would not be [detected] otherwise.”
And in fact, 15% of the patients who received a notification and were subsequently surveyed as part of the study follow-up acknowledged that they were aware that they had atrial fibrillation prior to entering the study.
The overall goal of the trial was to determine the performance of the irregular pulse algorithm, Turakhia stressed, but since paroxysmal A-fib “overwhelmingly” represented the majority of cases and one in five patients had permanent A-fib, “it's possible to make an immediate assessment with a single ECG.”
Like Perez, he stressed that this new information should be integrated into current clinical practice, not a substitute. “We know that the natural history of A-fib begins with really infrequent episodes, so where we are now is looking at how we incorporate this into our decision-making with that in mind. You have to anchor this fundamentally on the basics, the history, the physical, the context, possibly the context around the time of the notification—what were they doing?—then I think you have to look at that and look at their risk factors to determine how much additional testing would be appropriate.”
They plan to explore the new diagnoses, confirmed on ECG, that were made for these patients in the upcoming “end of study” data, Turakhia said.
To TCTMD, Rajat Deo, MD (The Hospital Of The University Of Pennsylvania, Philadelphia), an electrophysiologist who was not involved in the study, said Apple Heart is “a very strong first step. I’m very encouraged to see these findings.” The study has unearthed the potential of this technology for assessing the cardiac rhythms of the general population, he said. “Importantly, it also offers the opportunity to evaluate other kinds of cardiac rhythms. Right now, we're just talking about atrial fibrillation, but there clearly are other concerning heart rhythms that affect our patients. As a practicing clinician, I have now several cases in my own clinic where patients with A-fib have not felt well, and they've come in with tracings that suggests a non-A-fib type rhythm, another type of rhythm that requires clinical attention.”
That said, Deo would like to see the data from this study with more granularity. “I’d like to get a deeper dive into the sensitivity and specificity of the findings,” he said. “I’d like to understand in the 29% where there wasn’t that positive irregular wave form, what was going on there. Why did that not correlate?”
Understanding these nuances are important, he said, because most physicians now understand that they “should not be relying on this solely from a clinical decision-making standpoint.” But they also shouldn’t ignore this technology, especially if patients bring up ask about them, Deo added. “We should say that if you have a positive finding, let’s take a look at it and let's assess further.”
A-fib and Age
Panelist Jeanne Poole, MD (University of Washington, Seattle), noted that the issue of whether to anticoagulate asymptomatic A-fib patients is still up in the air. She also cautioned that this trial “has the possibility to lead a lot of patients potentially to being treated unnecessarily or prematurely or flooding doctors’ offices and cardiologists’ offices with a lot of young people, since the majority of the people who enrolled in this trial were young people.”
Highlighting that the portion of those aged 22-39 years receiving notifications was “very, very small” at 0.16%, Perez acknowledged that “stroke is important and we all think about it and worry about it. But I think what’s also important is that there are other things that A-fib is associated with [like] cardiomyopathy and heart failure. So finding A-fib, for example, in the young population might be important because if they really do have A-fib, maybe they don't need anticoagulation, but maybe there’s something else that's going on.
“I think that taking all of this into context, knowing that in the older population if you identify somebody with the risk factors and who now has A-fib, there may be a benefit here,” he continued. “But I agree with you, I think there’s a lot more research that we need to know in order to decide whether or not this population being detected with new technologies benefits from things like anticoagulation.”
Trial Design Implications
Also chiming in, panelist Patrick O’Gara, MD (Brigham and Women’s Hospital, Boston, MA), questioned whether the Apple Heart Study was “a proof-of-concept study or is this an incremental information to what we know about direct-to-consumer wearable devices?” He said he was left wondering: “How precise is precise? When you’re designing a trial where one of your goals is to see if you can scale things up to these massive numbers of patients, . . . how do you weigh the tension between pragmatism, scalability, and precision? It seems to me that there is a great enough deal of hype around this kind of technology that I wish for you to kind of put it into perspective as to how mature it is and where we should go with this inquiry regarding nearly a half a million patients.”
Turakhia responded that without a reference point for a study like this, they set out “to study this in a way we that we [thought] would afford significant generalizability to the total population or people who would be using these devices. We used that and back-calculated all of the engagement rates and participation rates to really determine that. So there are many lessons here about pragmatic operations, and in some ways this illustrates and compares and contrasts the benefits and the challenges with a traditional brick and mortar design versus a pragmatic study.”
At the same time, the researchers were able to “evaluate this technology really in the everyday scope of these participants lives and made it easy for them to engage and participate and access study visits,” he continued. “We found that the process was safe and we found it scalable. As we continue to learn, we think this will add a lot. One thing that excites us greatly is now we have footprint for how to think of evaluation of technology such as this and how to do that in a very pragmatic or scalable way.”
The Apple Heart Study was “just a little bit more than just a pilot,” Perez added. “It's giving us information now that physicians can actually use and take into account. That being said, I think this is really just the beginning. We were able to make some measurements here that are going to guide the future research designs that we need to answer more questions.”
Co-moderator Valentin Fuster, MD, PhD (Mount Sinai School of Medicine, New York, NY), said ultimately the study was one of feasibility. “In dealing with health, I am concerned about adherence here,” he said. “The adherence I predict will not be high, it will be low because we know with many other things in life that we ask people to adhere to, it really takes a short time and then it falls. I wouldn’t be very optimistic that this will change what A-fib is all about, which is a significant problem.”
Cost-effectiveness is also a huge issue, Fuster said. “Two-thirds of the world are middle- or low-income, and it’s hard for me to begin to predict that those technologies are going to be available and will be cost-effective in different parts of the world. I don’t want to be negative at all. I just want to be cautious that technology is advancing very significantly and this study you presented today is feasibility study and can help many people perhaps at high risk.”
For a general cardiologist today interpreting these results, O’Gara said jokingly that “most of us will have to change from flip phones. And then after that we have to call really a smart person like Mintu to try to help to put this into perspective.”
Ultimately, though, the issue of the “signal-to-noise ratio” is important, he noted. “We’ve heard a lot about quality of life and burnout and things like that, and certainly Dr. Poole addressed this as well about the noise that will come in when there's an irregularity detected on the watch that could range anything from [ventricular premature beats] to A-fib, but [the presenters] have been very circumspect in not overinterpreting where you are with this technology and where you plan to go. I understand very directly from you that these are limited observations, there are some issues, the generalizability factor is something that you brought up. I think you said for people who are wearing these watches and that is obviously a very select population. So kudos to you. It is limited, but this is a good step forward.”
C. Michael Gibson, MD (Beth Israel Deaconess Medical Center, Boston, MA), who will be leading the upcoming HEARTLINE trial to further evaluate the impact of the Apple Watch on all-cause death, all-cause stroke, and MI in patients 65 years and older, told TCTMD that this technology “is a consumer product at present with no medical indication. The next step is to move beyond that and see if there are any kind of true healthcare advantages.”
Turakhia M, Perez M. Results of a large-scale, app-based study to identify atrial fibrillation using a smartwatch: the Apple Heart Study. Presented at: ACC 2019. March 16, 2019. New Orleans, LA.
- The study was sponsored by Apple.
- Deo reports no relevant conflicts of interest.
- Gibson reports receiving research grant support and consulting monies in the past from Johnson and Johnson/Janssen.