With a multitude of sensors, wearables and mobile devices becoming available for use in clinical trials, sponsors should conduct systematic comparisons before designing protocols, according to experts at the annual SCOPE conference, who presented the work they’ve done to demonstrate the value of using digital monitoring in their studies, as well as the obstacles they encountered.
Choosing the right device from the outset, in a patient-centric manner, can directly affect adherence and the eventual quality and consistency of the data — as well as on the backend, where certain product choices can cause technical headaches.
Many devices, available at different price points, can return raw data in various formats, with some possibly inappropriate for use in clinical research, said Bhaskar Dutta, principal scientist of AstraZeneca’s Advanced Analytics Center.
Some can log missing data as a zero, potentially mirroring an actual measurement — and others don’t track heart rates over 100 beats per minute, Dutta said. In addition, certain devices may only report data on a daily basis, or even report some data types every minute and others every four, causing issues in sampling and final analyses.
AstraZeneca ran a human factors study of healthy volunteers evaluating the usability of six different body sensors and wearables over the course of a month, with some being worn for eight to 10 hours a day.
“How do you select the right device? That was a mind-boggling question for us… You don’t want junk in and junk out” during the course of your clinical study, Dutta said. And when asked if expensive medical devices delivered better quality data compared to consumer-grade activity trackers, Dutta said the short answer was no.
Considering the device’s size, weight, water resistance, durability and comfort are all essential — because every time a patient takes it off, there is a chance they do not put it back on, said Amir Lahav, from Pfizer’s rare disease research unit.
In addition, the processes for validating digital biomarkers and endpoints derived from remote sensors and wearables in clinical studies should be no less rigorous than any other biologically based measurement, said several experts at the Orlando, Fla. conference.
As the industry begins to pursue the benefits of digital health monitoring in clinical research — such as greater accuracy and richer datasets at a lower cost, as well as the ability to better focus on patient needs, reduce burdens and increase adherence — digital biomarkers and endpoints should not be treated as a company’s side project, or even as a technology in its infant stages. “We have to move beyond this concept of exploratory, exploratory, exploratory.
We have to get more robust about it,” said Rob DiCicco, VP of clinical innovation and digital platforms at GlaxoSmithKline. “We would not introduce a new tissue biomarker without a lot of methodology work up front.”
Mobile devices are being widely used in clinical research, but their utility in assessing benefits in interventional trials continues to be limited, DiCicco said. A lot of work remains to be done in standardizing data capture and use, and eventually having those methods pass regulatory muster.
Sponsors should approach the novel endpoint development process practically, develop technical standards and engage regulators throughout the process. But to move the field forward, industry collaborations and multi-stakeholder approaches will be needed to consolidate the evidence supporting specific, clinical outcomes, he said.
To define meaningful endpoints, sponsors should take a device-agnostic approach and validate them in a simulated living environment, said Lahav. Patients should also be asked to complete tasks relevant to daily life, not those that measure their ability to bring a finger to their nose, as they would in a clinic, for example.
“They need to be able to type and swipe on a mobile device,” Lahav said. “The activities should fit the era we live in.”
Sponsors need to work together to develop standard algorithms and data collection procedures in order to be validated by the FDA, Lahav said, lest the industry overwhelm the agency with too many different proprietary methods. “We can still compete on the drug part,” he said.
Christian Gossens, global head of early development workflows in Roche’s Pharma Research and Early Development Informatics group, demonstrated how gathering consistent, real-time sensor data on multiple sclerosis patients in a study became much more valuable than a traditional site visit.
Roche’s FLOODLIGHT study evaluated 60 MS patients, scheduling them for three site visits over 24 weeks, while also having them complete daily tests on their smartphones. Patients evaluated their mood, symptoms and disease impact, and tested their cognition, balance, and their ability to walk, pinch and draw shapes, mirroring other clinically validated endpoints. Passive information was gathered by the devices’ gyroscopes and accelerometers as well.
A normal clinical study schedule only offers so many windows on a patient’s disease progression, while daily monitoring offers a more granular picture, Gossens said. In addition, many people cannot remember more than the previous week with any amount of day-to-day detail.
For example, during a site visit one patient did not report a potential onset of a relapse from a few weeks prior — however, they did enter it into their daily smartphone diary, and the following data showed a worsening of certain symptoms.
A second case study in Parkinson’s disease patients identified a protocol breach, when a patient’s general practitioner prescribed them a new drug, resulting in a dramatic change seen in the daily data, he said.
“What we use this for at the moment is internal decisionmaking,” Gossens said, adding that the industry still needs to work out how to answer the regulatory questions regarding clinical validation. “It will be an interesting journey… We have strong confidence that the data here are real.”
Ieuan Clay, Novartis’ group leader for digital endpoints, described how digital sensor data can be used to guide clinical study design. While a typical trial evaluating limping and healing following knee surgery may run six months to a year, any differences in gait between the two legs begin to converge in about one month.
“That helped us design trials going forward, because we knew we had to gather more data points early on,” Clay said.
In addition, companies are looking at ways to recapitulate clinical measurements using mobile devices, and exploring the potential of siteless trials. For example, balance information can be derived from passive sensing of a patient sitting and standing over the course of their daily life, you don’t need to ask them to come in to an office to do a balance test, he said.
Gathering objective data from a natural living environment, such as while eating, walking, dressing or using the bathroom, can be much more useful, he said. In addition, it allows sponsors to gauge adherence in real time, and intervene before having to deal with data loss.
Lahav suggested sponsors re-evaluate spending billions on clinical trials that rely on mere snapshots of disease progression. “The clinic is not necessarily the best place to check your health status and quality of life,” he said.