Application of Inverse Probability Weighting to Clinical Trials of Treatments for Opioid Use Disorder

Public Deposited
Analytics

Downloadable Content

Download PDF
Request Version for Screen Reader
Creator
  • Ajazi, Elizabeth
    • Affiliation: Gillings School of Global Public Health, Department of Biostatistics
Abstract
  • Opioid use disorder (OUD) is a chronic condition that affects three million people in the United States and 16 million people worldwide. Randomized controlled trials (RCTs) are the gold standard for OUD, but there are multiple methodological complexities present in the analysis of randomized controlled OUD trials. This dissertation applies novel and advanced biostatistical methods to the analysis of OUD RCTs, with the ultimate goal of improving future research in substance abuse RCTs. Randomization is a key component of an RCT but high attrition and/or an inability to establish treatment therapy can pose a threat to its validity and is particularly a problem in OUD RCTs. An intention-to-treat (ITT) analytic approach preserves the randomized sample size since all patients are analyzed, while a per-protocol (PP) approach excludes patients who did not have sufficient exposure to the assigned study treatment. Regardless of an ITT or PP approach, patient failure to initiate treatment or non-adherence can compromise validity of results. Inverse probability weighting (IPW) is a method that can be used when non-adherence or missing data is prevalent in order to reduce bias in estimating the treatment effect. In OUD RCTs, IPW is a novel method currently not widely used. This dissertation introduces the application of a new IPW approach, inverse probability of induction weighting (IPIW), to Cox regression. Specifically, these methods will be applied to publicly available data from X:BOT, a RCT that was conducted to compare buprenorphine-naloxone (BUP-NX) versus injectable extended-release naltrexone (XR-NTX) in 570 patients.In Chapter 2, both ITT and PP approaches will be used through time-dependent non-proportional hazards models separately for outcomes of time to relapse, time to first positive opioid urine test, and time to overdose. Chapter 3 will focus on IPW methods in modeling time to relapse. IPIW will be introduced to account for differential induction status between treatment groups. Chapter 4 will outline a tutorial of how to compute IPIW that can be applied in many different research areas. The goals of this dissertation are to introduce, apply, and instruct how to create IPIW.
Date of publication
Keyword
DOI
Resource type
Rights statement
  • In Copyright - Educational Use Permitted
Advisor
  • Schwartz, Todd A
  • Marshall, Stephen W
  • Preisser, John S
  • Dasgupta, Nabarun
  • Monaco, Jane
  • Green Howard, Annie
Degree
  • Doctor of Public Health
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2021
Language

Relations

Parents:

This work has no parents.

Items

Thumbnail Title Date Uploaded Visibility Actions
file details Ajazi_unc_0153D_19929.pdf 2021-06-21 Public Download