Last week, at the Connected Health Conference in Boston, researchers from Fitbit presented some data that didn’t get as much attention as they should have. Part of that was due, perhaps, to how the study was presented—not in a high-profile, main-stage session, but rather on a sheet of cardboard during the meeting’s “poster session.” That’s where you typically find hordes of post-doctoral researchers desperately searching for an audience for their work, as conference-goers mill around between panel discussions. Still, this is where you’ll often find some of the best stuff at medical meetings.
In this case, Fitbit, the San Francisco wearable pioneer, presented data on an algorithm it had developed to detect a certain dangerous heart arrhythmia, called atrial fibrillation, using a technology already built into its wristband trackers: photoplethysmography, or PPG.
Fitness trackers have long used PPG devices to monitor pulse rates. The tiny sensors, which consist of infrared light-emitting diodes (LEDs) coupled with a sensitive light detector, measure infinitesimal gradations in light in human tissue, due to changing blood volume in the microvasculature as blood circulates through the body—a process that follows in rhythm with the beating of the heart. (For those interested, there’s a great description of how the technology works here.) And while PPG itself is actually some 80 years old, Fitbit’s team has developed an algorithm that can, if its latest findings hold true, accurately detect persistent atrial fibrillation (AFib) in a person wearing one of its wristband trackers—and, importantly, not over-detect it (that is to say: not flag it in a normal beating heart).
The Fitbit team tested its algorithm on heart-tracking data from four cohorts: two groups of test subjects as they sat quietly in place (which included 16 patients with persistent AFib matched with 21 heart-healthy subjects), plus two groups of sleeping volunteers (the first with persistent AFib, as above, and the second with normal sinus rhythms). The researchers then sliced up all that heart-tracking data into overlapping one-minute and five-minute stretches in order to see how well they could peer into any given testing “window” with their algorithm and detect the arrhythmia.
The data presented last week, though not peer-reviewed, suggests they can do that remarkably well—with the algorithm, in the five-minute testing windows, detecting AFib with a 99.3% rate of sensitivity and a 0.8% rate of “false positives.” In other words, it nearly always found an arrhythmia when it was there, and almost never found it when it wasn’t.
What makes these findings so intriguing—and potentially important—is that atrial fibrillation is a major risk factor for stroke. In AFib—which affects as many as 3% of the population, with most cases occurring in those over 65—the electrophysiology of the heart is out of whack, and its two upper chambers (the atria) fibrillate, or quiver, instead of contracting fully. That sometimes allows small amounts of blood to pool in the chambers rather than being pushed through. And when blood remains in one place over time it tends to clot. Such clots, in turn, can break off and get stuck in the narrow blood vessels of the brain, causing a stroke. Those with AFib have a four- to five-times-higher risk of stroke than those with normal heart rhythms.
One key limitation of the recent Fitbit study, says Dr. Vincent Thijs, an expert on stroke at the University of Melbourne’s Florey Institute of Neuroscience and Mental Health, is that it demonstrates only that the tracker and algorithm can detect persistent or permanent atrial fibrillation, rather than paroxysmal AFib (or short, sudden spasms). The latter are often undiagnosed for years and are thought to lead to many strokes as well. But Thijs believes that wearable devices that can reliably detect paroxysmal AFib aren’t far away. “This space is moving quickly,” he says.
Dr. Venkatesh Raman, an interventional cardiologist at MedStar Georgetown University Hospital, agrees. Raman, who is co-Principal Investigator for the ongoing Fitbit-sponsored research on AFib (but who was not an author of the study presented at Connected Health), says he thinks this will one day change the practice of cardiology. “So it is kind of funny how we practice outpatient care,” he says. “I might see a patient once or twice a year—If they’ve been in the hospital, I might see them a few more times—but the vast majority of their life occurs outside our office setting. But if we have these devices, we have the ability, perhaps, to collect [the data that we need.]” And rather than use a “one-size-fits-all paradigm for who and what needs treatment,” he says, physicians can treat their patients based on real-world, 24/7 data—which means they can treat them on a smarter, more individualized basis.
Indeed, what makes AFib such a good test case for the digital health revolution is that, once caught, it can often be treated cheaply and effectively with oral anticoagulants. So discovering someone who has an undiagnosed arrhythmia isn’t a useless exercise: Rather, it might just save a life.
This essay appears in today’s edition of the Coins2Day Brainstorm Health Daily. Get it delivered straight to your inbox.
