Development of a biometrics-based platform to predict relapse

About This Grant

Strides have been made in addressing the causes of opiate/opioid use but we remain in crisis mode as the rate of fentanyl- and prescription analgesic-related overdose (OD) deaths remains high. The alarming number of OD deaths despite many efforts to curb use renders this a serious public health problem. The concern is that while many treatment-seeking individuals can successfully become abstinent, relapse remains the number one challenge. Individuals are at greater risk for an overdose when they relapse and so a major gap in the field is that we are unable to predict when a slip or relapse is more likely to occur (in real time) to save lives. The past few years have witnessed an explosion in the popularity of wearable biosensors that inform individuals of many physiologic and behavioral metrics with the goal of improving overall health, sleep and fitness. Consumer Sleep Trackers are the most common and use different approaches that have promise in health and fitness applications. Despite recent advances in other areas of medicine and even in some Substance Use Disorders, there is a paucity of clinically available digital biomarker-based tools designed to target individuals who use opiates and/or opioids. Many tools are in the research and development pipeline and have potential but while wearable biosensors have been used to document opioid-related health issues (e.g., acute administration, withdrawal, and toxic overdose), none to date have been directed at predicting relapse to opiate/opioid use. The present R21 Development Project will use Machine Learning (ML) approaches to combine data from two different experiments and two different populations to develop algorithms that will integrate the unique physiological, behavioral and subjective mood state that exist when individuals begin to crave opiate/opioids, which often heralds the onset of relapse. This project will be the first to establish the foundation for this approach as it is impossible to conduct clinically relevant outpatient trials that call for relapse. GPS location services will play a critical role in developing the model, but that will need to be tested in a subsequent project. Instead, we will conduct one study in a natural setting “studio” apartment environment, which is where many individuals use opiates and opioids. Aim #1 will focus on using commercial biosensors to quantify the physiological, behavioral and subjective "fingerprint" profile of opiate/opioid cue-related craving in individuals who are diagnosed with Opiate Use Disorder. Aim #2 will be conducted in healthy individuals who will be given test doses of a variety of psychoactive drugs to provide additional data for the ML model by adjusting for possible confounds due to polydrug use even when individuals have abstained from opiate/opioid use. By combining the data from both aims into ML models, we will create a database of the key features of relapse to opiate/opioid use and use this to develop rules that will detect the early signs of relapse. Should the data prove to be robust, the next steps will be to deploy the devices and use similar analytical methods with individuals on an outpatient basis.