The United States has experienced a four-fold increase in medical opioid use over the last two decades. Medical opioid abuse and addiction have increased and afflict 5% and 2% of US adults respectively; heroin addiction afflicts 0.5% of US adults. In 2016, >39,000 US individuals died of a medical opioid, illicit opiate or synthetic opioid overdose, an increase of 18% from 2015. While opioid addiction treatment admissions and treatment capacity have increased many-fold over the last two decades, patient needs greatly exceed capacity. Medication assisted treatment (MAT) approaches for opioid use disorder include three approved pharmacotherapies combined with psychosocial and supportive therapies. Active investigations are evaluating established and novel therapies for opioid detoxification prior to MAT. MATs have established but limited efficacy in treating opioid use disorders compared to placebo pharmacotherapy. Patients treated with MATs have mortality rates during and after treatment that exceed population mortality rates. Improving efficacy through personalized treatment is essential to optimize MAT resources and reduce mortality. “Biosignatures of opioid addiction treatment success” offers a data science solution that aligns with NIDA priorities and Institute of Medicine (IOM) guidance to expand use of MAT, and with HHS priorities to support cutting edge research on pain and addiction. Our overall aim is to decode the predictors of MAT success and enable health care providers to improve treatment efficacy using accurate forecasting. In Aim 1, we will create a platform for applying learning algorithms to existing and future opioid addiction treatment data. We will organize and merge clinical, treatment, and outcome variables from clinical trials of opioid addiction treatment publicly available through the NIDA Data Share resource. We will import these data into a database optimized for high dimensional data analysis. Workflows will be developed to apply multiple Bayesian learning algorithms to the data. In Aim 2, using the biosignature learning platform, we will identify sets of variables that together predict opioid addiction treatment success. We will apply this platform to each trial independently, and then to multiple trials in an integrative analysis. The learned biosignatures will be ranked by how well they predict treatment success. The best models will be incorporated into a proof-of-concept calculator. The calculator will provide treatment success scores based on the characteristics of new patients. We will present the prototype to multiple stakeholders for assessment. At the end of Phase I, we will have created a biosignature learning platform and a proof-of-concept opioid addiction treatment success calculator. We plan to fully develop these components with additional datasets and variables in Phase II. Our commercial goal is to develop licensable and easily deployable algorithms for healthcare networks treating opioid addiction. The algorithm will help providers understand profiles of patients likely to be successful with MAT and to personalize treatment strategies to maximize abstinence.
Public Health Relevance Statement: In this project, we are using algorithms to learn the predictors of opioid addiction treatment success from publicly available clinical trial data. Healthcare providers can use these models to improve their opioid addiction treatment programs.
Project Terms: Abstinence; addiction; Admission activity; Adult; Aftercare; Algorithmic Software; Algorithms; Attitude; base; biosignature; Buprenorphine; Characteristics; clinical translation; Clinical Treatment; Clinical Trials; Clonidine; Data; Data Analyses; Data Science; Data Set; Databases; Development; Disease; Dose; experience; Fentanyl; Formulation; Foundations; Future; Goals; Health Personnel; Healthcare; Heroin; Heroin Dependence; heroin use; high dimensionality; illicit opioid; improved; Individual; individualized medicine; Institute of Medicine (U.S.); Investigation; Learning; Letters; Licensing; lofexidine; Machine Learning; Medical; medication compliance; medication-assisted treatment; Methadone; Modeling; mortality; Naltrexone; National Institute of Drug Abuse; novel therapeutics; Opiate Addiction; Opioid; opioid abuse; Opioid agonist; Opioid Antagonist; opioid mortality; opioid overdose; opioid use; opioid use disorder; opioid withdrawal; overdose risk; Pain; Patients; personalized medicine; Pharmacotherapy; Phase; Phenotype; Placebos; Population; Prevalence; prevent; profiles in patients; prototype; Provider; psychosocial; Published Comment; Relapse; Research; Resource Sharing; Resources; sharing data; success; Supportive care; Tramadol; Treatment Efficacy; Treatment outcome; treatment program; treatment strategy; United States; Veterans
