There is no doubt the age of wearable monitors is upon us. Sales of physical activity monitors have increased dramatically from 15 million in 2014 to 65 million in 2017, according to Statista.com. A 2017 study by Juniper Research estimates that the use of fitness trackers will double by 2021. Another 2017 study, this one by Researchscape, found that 51 percent of Americans reported using fitness trackers at least once a day. It is no surprise then that clinical researchers are eager to take advantage of this potential treasure trove of data. “The use of wearable devices in clinical trials is turning the corner from an exciting, but untested, idea to an established tool in researchers’ kits,” said Adam Pellegrini, general manager and senior vice president at Fitbit Health Solutions. Pellegrini cites the inclusion of Fitbit in more than 675 peer-reviewed publications. The popularity of Fitbit products and other wearables likely assures participant compliance.
The challenges for researchers looking to harness this data, however, are manifold: managing the large sets of data generated by wearables, keeping pace with wearable device innovations, validating the accuracy of devices used in studies (especially clinical trials aimed at FDA-approval), and devising ways to translate clinical research into easily-digested results physicians can use to improve outcomes. Researchers working with digital biomarker data every day say the challenges pale in comparison to their potential power to transform everything from better chronic disease management to the improved probability of successfully shepherding a drug candidate through clinical trials.
Fitbits for All of Us
Before researchers can fully realize the power of these data, they must tackle the most basic one—discovering which biomarkers generated by wearables are the most informative. In January, researchers with the National Institutes of Health’s (NIH) All of Us trial announced the Fitbit Bring-Your-Own-Device Project. The goal of the trial is to realize the potential of personalized medicine by finding correlates between genetics and a variety of health measures. So far, 104,000 participants—of an eventual 1 million—have completed the All of Us initial protocol, which includes providing consent to access their electronic health records, completing surveys, and providing physical measurements and biological samples. Researchers will also collect Fitbit data from participants who already own the devices and, in the future, provide the devices to some of those who do not. “Our purpose is to find new digital biomarkers,” said Chris Lunt, chief technology officer for All of Us. “We are living in a day where the affordability and the acuity of these devices are really good and we can take advantage of an existing commercial product that allows us to get information on people’s activity level.” Right now, he said, researchers are exploring data quality and, in the future, they will address how to best share the data with researchers. “There is a lot we are trying to learn.”
According to Fitbit’s Pellegrini, the obvious benefits of using wearables in clinical research are the reduced costs and the improved participant engagement. But, he added, they also have the potential to make the science better by eliminating bias. “Health data from wearable devices can also help researchers overcome the challenges of subjective bias when participants log their own data.” Wearables can also generate novel endpoints. “Until the advent of wearable devices, many types of health data, such as the quantity and quality of sleep, were not accessible to study participants without visiting a sleep lab.” Use of the devices also has the potential to expand access to hard-to-reach populations. “Wearables can help to facilitate virtual trials… so people can participate from the comfort of their own home.”
Wearables can also improve the information collected by allowing researchers to ask more granular questions. “We can now collect far more data points that contribute to one variable,” said Scott Collier, Ph.D., a professor of cardiovascular exercise science at Appalachian State University. Collier and his colleagues have studied the impact of consumer physical activity monitors on human physiology research. He spends much of his time comparing the accuracy of the latest wearable technology to gold-standard laboratory devices. When it comes to heart rate monitoring, for example, Collier said scientists can now look at that measure in a person’s own environment. “A lot of different things contribute to heart rate that we have never paid attention to before.” These might include time of day, exercise levels, and situational stress. Combining heart rate with other biomarkers, such as activity tracking data and pulse wave velocity—an indication of the stiffness of arteries that is measured by the latest home body weight scales such as PhysioWave Pro—will allow for more precision care. “We can manage chronic conditions better because we can account for life stressors.”
Biomarkers of Mental Health
At HealthRhythms, a company with a software platform that allows physicians and clinical researchers to look at various biomarkers relevant to mental health, CEO Tanzeem Choudhury, Ph.D. and her team are trying to merge the insights that smartphone technology can provide with those of clinicians and therapists. Choudhury is also an associate professor of information science at Cornell University where she runs the People-Aware Computing Lab. “This is a domain where most diagnosis is done either through observation or patient-self-reporting,” she said.
Choudhury’s goal is to make diagnosis more objective and quantifiable. “That’s where biomarkers come in.” HealthRhythms uses smartphone sensor data to automatically measure patterns of behavior. In 2017, the company received a $2.1 million grant from the NIH to continue investigating smartphone-based behavioral health interventions. Currently, the company’s applications focus on measuring biomarkers of depression, including how a person uses their phone, their number of steps, location (at home, work, or out in the community), and sleep time. These data help generate scores of sociability, mobility, and activity.
The company is currently performing a clinical trial with researchers at the University of Utah to look at interventions based on these data, which can inform researchers if people are becoming more isolated, if their energy levels are decreasing, and if they are sleeping poorly—all indications of worsening depression. “We can actually track change in mental health,” Choudhury noted, adding that the challenge she and other clinical researchers face is how to turn vast amounts of continuously collected data into packaged results physicians can—and will—use the data. The question, she said, is how can these data be delivered in the most clinically efficient way so that doctors can use it to improve the efficiency of their practice. “Most physicians don’t want to see data unless something is wrong,” Choudhury said. “The delivery has to be targeted.” Collier even suggested that digital biomarkers may require new training for medical professionals. “This is more data than we have ever had,” he said.
Counting Steps in Patients
Digital biomarkers also promise to change in-patient care. An ankle-worn step tracker by modus health, called the StepWatch, received FDA clearance more than 15 years ago. “Our goal is to become part of routine clinical care worldwide,” said CEO Teri Rosenbaum-Chou, Ph.D. More wearable device makers are likely to pursue FDA clearance in the future, opening up new research opportunities. While the FDA determined the StepWatch is safe for use, and some devices, depending on their classification, can be used in the context of clinical decision-making, they are not, however, used to diagnose or treat disease. In 2018, the AppleWatch with EKG and abnormal heart rhythm indicator received FDA clearance, followed in 2019 by the Study Watch with ECG by Verily—a subsidiary of Alphabet, Google’s parent company—that is available only to researchers.
StepWatch’s most recent version is Bluetooth-enabled, which makes it more attractive for research studies. In fact, the company customized a mobile app to use with its StepWatch4 for use in a small clinical trial looking at intravenous infusion of a drug to treat children with Duchenne muscular dystrophy. A 2017 study led by researchers at the University of California, Los Angeles showed the StepWatch accurately measured “community ambulation” in this population in which strides per day and frequency of strides correlated with condition status. Rosenbaum-Chou said the StepWatch uses highly accurate algorithms for quantifying walking. “Accuracy and reliability are very important to digital biomarkers because filtering out the noise is required to capture clinically relevant change.”
The StepWatch was developed using patients who had difficulty walking, patients in which other consumer-available activity trackers often did not record steps taken. These monitors have the potential to improve the efficiency with which nurses and other caregivers in hospitals triage patients after surgery. Both in hospitals and after discharge, patients recover from surgery more quickly if they are able to walk around. “The more patients walk before they leave the hospital, the lower the possibility of readmission,” said Rosenbaum-Chou.
Saving Drug Discovery Millions
Most people would agree that getting more value for healthcare dollars in the U.S. includes reducing the cost of bringing individual prescription drugs to market, which was estimated at $2.7 billion in 2017 by the Tufts Center for the Study of Drug Development. Pharmaceutical companies are keen to reduce the money spent on drugs that don’t gain FDA approval. Chris Benko, CEO of Koneksa Health, spoke at a Keystone Symposium on Digital Health in January 2019 in which he made the case that digital biomarkers are the key to making drug trials more efficient, weeding out costly underperformers more quickly, and potentially saving the industry millions—or even billions—of dollars.
Koneksa’s platform allows its client researchers to remotely capture real-world data—including physical activity, blood pressure, ECG, spirometry, and other digital measures—from patients in real time from remote biometric trackers, mobile devices, and health apps. Koneksa also works to develop devices and validate the measures to be used in a study, delivering to clients a device kit for each study participant, then collecting and analyzing the data. This individualized approach is necessary, Benko said, because all digital biomarkers are not created equal. The intended use of the data defines the rigor of biomarker validation needed—the commercial fitness tracker data intended for personal use are different than the data intended for a drug approval, which needs to meet the FDA requirements for human experimentation.
“In almost every case, we invest in studies that collect the human data to characterize a particular biomarker and determine if such a biomarker is fit-for-purpose,” Benko noted. The fit-for-purpose principle means that biomarker development and validation experiments are tailored to answer specific questions asked by drug development studies. The study participants take experimental medicines while digital biomarker data is being collected to see if, for example, there is a response to treatment or there is a safety signal. Later, the data needs to be presented to the regulators. Eventually, Benko predicted, traditional assessments presented to the FDA will be replaced by digital ones because they are collected in a real-life setting and more accurately reflect how patients feel 24/7.
The importance of validation depends on how data will be used, said André Henriksen, a research fellow in medical informatics at UiT The Arctic University of Norway, Tromsø. In some research, the wearables are not replacing research-grade instruments, but being used as additional sources of data. “With the built-in sensors, we can track activity, heart rate, sleep, location, and more, and future devices are bound to include even more sensors with added capabilities,” Henriksen said in an e-mail. The pace of technology is one thing researchers like Henriksen must confront. Although several validation studies on consumer-based fitness trackers are published each year, very often these studies do not include recent models. But, he said, “without these objectively conducted validation studies, we cannot truly know how accurate the devices are.” For researchers like Henriksen, the possible utility of digital biomarkers outweighs the challenges. “There is a large potential here, and I think we will see a huge increase in wearable usage in research going forward.”