The explosion in smart devices — phones, watches, fitness gadgets and the like — has unleashed a wave of apps designed to manage chronic illnesses, detect behavioral diseases and manage pain. Most recently, Apple announced that apps due later this year will allow its Series 4 watches to perform electrocardiogram readings, or ECGs, and notify users of irregular heart rhythms.
The problem for consumers is knowing which apps — if any — actually work.
The Food and Drug Administration cleared the ECG app and irregular-rhythm notification feature on the Apple watch, but noted that the apps aren’t intended to replace traditional diagnosis methods. The agency said the ECG data displayed on the Apple watch is for informational purposes only and isn’t intended to be interpreted by the user without consulting a health-care professional.
Most apps on the market lack approval from the FDA, which hasn’t been able to keep up with the health apps being released, raising concerns that some apps could expose consumers to harm.
The FDA last September started working with Apple and eight other tech and medical-device companies — including Fitbit, Samsung and Verily Life Sciences, a subsidiary of Google parent Alphabet — to streamline approval of mobile medical apps. In the meantime, here is a status check on some of the areas where health-monitoring tools might make the most difference.
Measuring Heart Health
Smartphones and watches can collect data on the heart continuously, which promises to improve detection and treatment of heart disease.
For instance, devices that track heartbeat data can help doctors identify atrial fibrillation, a leading cause of heart failure and strokes. With a-fib, the upper two chambers of the heart beat erratically and at dangerously high speeds. Since symptoms come and go, it can be hard to detect, but watches worn for long periods have a chance of spotting it outside of a doctor’s office.
An ECG, the standard method of detecting a-fib, requires placing 12 electrodes on a patient’s body. The Apple Watch Series 4 will have electrodes built into the watch’s digital crown, which users touch for 30 seconds after opening the app to get an ECG reading, Apple says.
Another device, the KardiaBand watchband, also offers ECG capabilities and was cleared by the FDA in November. It was most effective in detecting a-fib when physicians looked at the results, rather than relying solely on the watch’s algorithms, according to a study by the Cleveland Clinic. The device’s instructions tell wearers to place their thumb over a spot on the watchband embedded with an electrocardiogram sensor, which records their heart rhythm.
But in the Cleveland Clinic study, about 35% of the recordings couldn’t be read by the watch’s algorithms, possibly because people didn’t press their thumbs down for the required 30 seconds. Electrophysiologists, however, looking at the same data, were able to accurately identify people with a-fib 100% of the time and people without a-fib 80% of the time. The electrophysiologists also beat the algorithm’s performance on recordings that it could read, correctly identifying people with a-fib 99% of the time, compared with the algorithm’s 93%.
“It’s a reminder for all of us in dealing with digital health that the patient is an active component of the equation and is part of the end results on how good these recordings are,” says Khaldoun Tarakji, an electrophysiologist who led the study. Dr. Tarakji says that with smartphone-based electrocardiogram monitors, the clinic can access recordings of patients’ heart rhythms no matter where they are. Doctors can also use the monitors to diagnose patients with intermittent episodes of a-fib, which are hard to catch, and follow up on patients who have had ablations, a procedure that removes diseased tissue from the heart to try to stop a-fib symptoms.
Another new tool on the market is machine-learning software called DeepHeart, which takes heart-rate, step-count and other data from an Apple Watch or similar device and calculates the risk that the wearer has one of several ailments, including a-fib, sleep apnea, hypertension or diabetes, according to Brandon Ballinger, co-founder of Cardiogram Inc., the company that built the software. DeepHeart’s diagnostic tests have FDA clearance, a status the agency assigns to tools that it determines are “substantially equivalent to another legally marketed device.”
For Apple Watch wearers, DeepHeart was 97% accurate in predicting a-fib for patients who had already been diagnosed (using an electrocardiogram, the standard for diagnosis) and were hospitalized for treatment, according to a study published earlier this year by Mr. Ballinger and researchers at the University of California, San Francisco. But it was only 72% accurate for patients who thought they had a-fib but hadn’t been diagnosed and hospitalized, the study said.
Outside the hospital, in what Cardiogram calls “the real world,” detecting a-fib is much harder, the company says. Motion, sweat and sunscreen can affect how successfully an Apple Watch, for instance, reads heartbeats. (Apple Watches have optical sensors which shine light into your wrist and measure how much light is absorbed. Between heartbeats, less blood flows, so less light is absorbed). Alcohol consumption and exercise also affect heart rate and can mask or mimic a-fib. Tests were conducted on earlier versions of the Apple Watch, through Series 2, but Cardiogram is now also compatible with Garmin and Android devices, Mr. Ballinger says.
Glucose Monitor
For more than 50 years, researchers have looked for ways to monitor glucose that don’t require people to prick their fingers and draw blood. Bodily fluids including urine, sweat, saliva, ocular fluids like tears, and interstitial fluids, which bathe cells, also contain glucose and are easier to get to than blood. But they can be challenging to work with.
Several companies are investigating minimally invasive or noninvasive glucose monitors, and a few have been approved by the FDA, but developing systems that don’t penetrate the skin has been challenging.
A glucose monitor from Dexcom Inc. that the FDA authorized for marketing in March uses a sensor about the width of a human hair that sits just under the skin and detects glucose in the interstitial fluid. It generates an electrochemical signal that’s read by a processor and converted into data that’s transmitted to a Dexcom receiver or a smartphone or watch. In 2016, a previous generation of monitors was recalled by the FDA because the receiver’s alarm didn’t sound when the glucose reading was high or low. But the company has since developed new technology, says CEO Kevin Sayer.
The new monitor uses some finger pricking, says Mr. Sayer, because “we’ve not seen anything sitting outside the body that delivers the accuracy that patients require.”
In a paper published earlier this year, Sunghoon Jang, chair of the department of computer engineering tech at NY City College of Technology at CUNY, surveyed a dozen emerging optical or electrochemical technologies that target both blood and other bodily fluids, including a contact lens that changes color depending on the level of glucose present in tears. But none of these techniques are commercially available, Dr. Jang says, showing how complicated it can be to develop reliable and affordable alternatives to more invasive ways of measuring glucose, which also continue to improve.
Tracking Blood Pressure
Researchers have long tried to improve on the traditional arm cuff to measure high blood pressure, which causes heart attacks and strokes.
By 2020, three billion people will have smartphones, “and a lot of people in this world have high blood pressure and don’t know it,” says Ramakrishna Mukkamala, a professor of electrical and computer engineering at Michigan State University.
Recently, a group he led created a way to take blood pressure with a phone, using the same principle as the blood-pressure cuff, which varies pressure on the arm. In a study published in March, the group used a modified smartphone case with two sensors, one that measured blood volume and one that measured applied pressure. Users steadily pressed their fingertips against the case to get a reading. The data was transmitted via Bluetooth to an app, which calculated blood pressure and displayed it.
In September, however, a proof-of-concept study done by Dr. Mukkamala and another set of authors showed that the same finger-pressing method can be applied to optical and force sensors that are already built into some phones — one sensor for taking selfies and one for displaying a 3-D touch feature.
The group has developed an iPhone app that guides fingertip placement and calculates blood pressure. Comparing the results against a traditional blood-pressure cuff, the app was less accurate than the arm cuff. But Dr. Mukkamala says it was comparable to a finger cuff, a device that’s been cleared by the FDA for measuring arm blood pressure but used primarily so far in research. Dr. Mukkamala hopes to market the phone technology, though he says it needs more work before it can be approved.
Another group is working on ways to test smartwatches that monitor blood pressure. At the National Institute of Standards and Technology, the Physical Measurement Laboratory is working with Tufts University’s School of Medicine to build a fake arm that reproduces the mechanical properties of blood pulsing through an artery and surrounded by human tissue. The arm could then be used to test new blood-pressure monitors that would be worn like a watch. NIST and Tufts expect to test optical sensors on the fake arm that block specific frequencies or colors of light. As the pressure changes, the color of light that is blocked also changes.
