Causal Inference and Principal Stratification: Competing Risks, Bounds, and Surrogates Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 22, 2019
Creator
  • Long, Dustin
    • Affiliation: Gillings School of Global Public Health, Department of Biostatistics
Abstract
  • Establishing statistical methods for quantifying the effects of interventions to prevent infectious diseases is the overall objective of this research. The principal stratification framework is frequently implemented to make causal comparisons where naive methods fail. For HIV vaccine trials, estimates of the causal effect of vaccine on viral load or post-infection survival is challenging using standard methods because all individuals do not become infected during the trial. In this scenario, the "principal" effect, which is the causal effect within a principal stratum, is the effect of vaccine on viral load for subjects who would be infected during the trial regardless of treatment assignment. Without strong assumptions, the principal effect is not identifiable and usually requires bounding, or sensitivity analysis, of the principal effect often resulting in bounds that are often large and uninformative. Methods for estimating, i.e., bounding, the principal effect of treatment on competing risks outcomes have not been developed. Furthermore, situations where bounds on the principal effect can be improved by using baseline covariates have not been investigated. The principal stratification framework can also be used to determine surrogates of vaccine protection, i.e., biomarkers measured during a trial that are correlated with the desired outcome (infection). Repeated low-dose challenge studies are often used to evaluate potential vaccines. While these studies more accurately mimic exposure, the assessment of the potential surrogates greatly depends on the study design. Evaluation and comparison of different study designs have not been performed. Therefore, we propose to 1) develop methods to analyze the principal effect of treatment on competing risks outcomes, 2) investigate the improvement of the bounds on the principal effect based on baseline covariates, and 3) evaluate designs of repeated low-dose challenge experiments to assess surrogates of vaccine protection.
Date of publication
Keyword
DOI
Resource type
Rights statement
  • In Copyright
Advisor
  • Hudgens, Michael
Degree
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill
Graduation year
  • 2012
Language
Publisher
Parents:

This work has no parents.

Items