Value of Sleep Metrics in Predicting Opioid-Use Disorder Treatment Outcomes: Leadership and Data Coordinating Center

Opioid-use disorder (OUD) is a major public health problem, affecting over 2.7 million people and resulting in over 90,000 deaths in 2020 in the U.S., as well as broadly impacting the mental and physical health of individuals suffering from an OUD and the communities and families of affected individuals. Treatment effectiveness for OUD depends on retention in care. Unfortunately, a systematic review reported that the median retention rate at 6 months for Medications for Opioid Use Disorder (MOUD) programs across 19 studies was only 58%. Sleep disturbances have been identified as predictors of treatment attrition and are related to OUD through bi-directional pathways involving pain, stress, and emotional dysregulation, and could serve as future intervention targets to improve OUD outcomes. However, there is a limited understanding of the predictive value of specific measures of sleep and circadian rhythm during early recovery, and almost no data on how sleep and circadian parameters interact with other risk factors for MOUD outcomes. We have assembled a team of sleep scientists, addiction medicine specialists, biostatisticians, and clinical trialists and will leverage the exceptional resources of Brigham and Women's Hospital's Program in Sleep Medicine Epidemiology, Sleep Reading Center, and Division of Biostatistics at Harvard Pilgrim Health Care Institute to lead the Leadership and Data Coordinating Center (LDCC) for this multi-site study. The LDCC will develop a Common Protocol for the collection of standardized data for predicting MOUD outcomes across four Research Centers and will lead, coordinate, and implement all aspects of this common protocol, providing comprehensive, responsive, and innovative data and project management. It will facilitate recruitment and data sharing and support the rigorous collection and analyses of comprehensive sleep measurements. In addition to risk factors suggested by existing prediction models and HEAL initiative common data elements, sleep will be assessed by EEG -- a biomarker of neurophysiology and psychiatric diseases -- and will be measured on two nights approximately one month apart from 400 patients enrolled in a MOUD program. Sleep macro- and micro-architecture will be derived using centralized sleep scoring and advanced EEG quantitative signal analysis. Sleep-disordered breathing, periodic limb movements, sleep-wake patterns and circadian risk factors will be measured by polysomnography, validated questionnaires and multiple day Fitbit trackers. The primary endpoint will be treatment retention at 6 months after study enrollment. Secondary outcomes will include time to treatment drop out, and illicit drug use and non-medical opioid use; opioid craving; withdrawal symptoms; alcohol use; pain intensity and interference; physical functioning; sleep disturbance and quality measured at 6- month. We will adopt an ensemble algorithm (the Super Learner) to develop the prediction model that finds the optimal combination of a collection of statistical and machine learning algorithms. This data-science based plan and our multi-disciplinary expertise will ensure that the goals of this study are met.