DIA 2019
A hidden gem: Emerald device wirelessly detects clinical trial patients using radio signals and AI
The Emerald device was developed by Massachusetts Institute of Technology (MIT) Professor Dina Katabi – who presented at the DIA Annual Meeting last month – and her students at MIT’s center for wireless networks and mobile computing. Based on their work, Katabi cofounded Emerald Innovations, through which the device is available to pharmaceutical companies for use in research.
Katabi described Emerald as a Wi-Fi like device “that sits in the background and analyzes the surrounding radio signals with advanced machine learning algorithms.” From these wireless signals, it can extract various physiological metrics related to gait, mobility, breathing, sleep stages, sleep apnea, and others, using machine learning algorithms – and it can do so through walls.
The device is termed an 'invisible' in that it uses wireless signals to monitor health metrics without requiring patients to wear any sensors or devices, which Katabi said can create compliance and adherence problems.
“This is particularly important since patients already have medication adherence problems. Asking patients to do extra work could potentially exacerbate such problems,” she told us. Conversely, invisibles allow patients to go about their normal routines without any added burden – providing a unique opportunity to support at-home clinical trials.
In the next couple years, Katabi said it is likely that the device will be integrated in more trials and research, potentially being used to manage chronic diseases, such as COPD and CHF, enabling early detection and tracking changes in disease conditions.
Novartis collaboration: Exploring the benefits and limitations
Over the past year, Katabi and her team at MIT collaborated with Novartis to explore the potential use of the technology in clinical trials. Individuals were studied for multiple days to compare data measured from Emerald with existing standards.
Jason Laramie, global head translational medicine data sciences at the Novartis Institutes for BioMedical Research said the basis for the collaboration was to evaluate new technologies that could generate meaningful data about clinical trial participants, without adding to patient burden.
The system was then deployed as part of a non-interventional trial with healthy volunteer subjects to see how the device would perform in a clinical trial setting. As Laramie explained, with the potential to passively measure various physiological parameters, the device could enable researchers to measure quality of life measurements in the comfort of the patient’s own home.
“This could allow us to tailor clinical trial designs that make participation in a trial easier,” Laramie told us. “The next steps, for us, is to explore possible matches for this technology and its benefits and limitations to our portfolio to see if there are places where it may make sense to apply.”
Speaking to the development process, Laramie said prototyping was done in a similar manner to how the company has completed technology assessments in the past: “We started by standing up the device and the infrastructure internally to understand more of the process and to work out any challenges,” he explained. “Internally, we do some data generation and algorithm development to ensure we can handle the data coming from this device and process it into meaningful measures for us.”
From wearables to invisibles: Advancing data collection in clinical trials
Among the benefits over existing methodologies are improved biomarker development, cost reduction, as well as the ability to detect behavioral variations and provide context for other measurements.
By monitoring patients in their homes, Emerald is able to provide more patient-centric endpoints and real-world evidence (RWE) representative of a patient’s actual quality of life, said Katabi. For example, diseases affecting mobility, such as Parkinson’s and Multiple Sclerosis, use a six-minute walking test in the clinic as an endpoint – which is not reflective of a patient’s daily life.
“In contrast, observing the actual walking, balance and activity levels of patients in the home, is much more representative of the disease impact on the patient, and is consequently of great interest to patients, payers, and the FDA.”
The use of Emerald also could help develop more sensitive biomarkers. “The inability to observe meaningful and fine-grained metrics has limited some therapeutic areas to using endpoints based on coarse and infrequent events such as mortality, hospitalizations, et cetera,” explained Katabi.
Mortality, for instance, is a common metric in heart diseases, though due to the rarity of the event, these trials take long and require large populations. Alternatively, Katabi said Emerald can observe “more fine-grained improvements” such as increased mobility and activity levels.
Additionally, due to its ability to provide continuous, in-home monitoring, the use of invisibles could reduce the length of trials and the number of patients required, ultimately reducing trial costs.
“Further, it reduces the need for repeated clinical site visits, which both lowers the direct costs to pharma companies from such site visits, and the indirect cost due to churn and recruitment difficulties imposed by the overhead of site visits on subjects,” added Katabi.
Another challenge to conducting clinical trials is the occurrence of “confounding factors” affecting outcomes: “For example, patients participating in a clinical trial might experience changes to their work and sleep schedule leading to significantly less sleep and excessive stress, and therefore more sensitivity to side effects and poorer drug response,” said Katabi.
These, in turn, have direct implications for efficacy and safety. Detecting such changes in behavior, Emerald enables clinical trials to use these variations to slice data, she explained, and garner insight into patient responsiveness.
Katabi also said the device provides context which is lacking from patient reported outcomes (PROs). “Emerald can provide vital context, such as the activity levels before the measurement is taken, thus improving the interpretability and usability of the data,” she explained. “Similarly, a patient might self-report increased fatigue on certain days, which when augmented with contextual information from Emerald about sleep and activity, could be explained by poor sleep or engaging in tiring activities.”
