Statistical methodologies for network meta-analysis with applications in veterinary medicine

No Thumbnail Available
Date
2023-12
Authors
Morris, Paul S
Major Professor
Advisor
Wang, Chong
O'Connor, Annette M
Ommen, Danica
Zimmerman, Jeffrey
Berg, Emily
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Research Projects
Organizational Units
Journal Issue
Is Version Of
Versions
Series
Department
Statistics
Abstract
Network meta-analysis (NMA) has become a popular method with which to conduct research syntheses pertaining to multiple interventions, as it allows for the comparison of interventions that have not been directly compared and typically improves precision of the comparative effect estimates relative to those produced via the individual trials. While the literature on NMA is well-developed, various problems have arisen in the process of conducting our own NMAs in the field of veterinary medicine that have not yet been addressed. Our research has focused on developing methods to fill these gaps. In Chapter 2, we propose an NMA model for an ordinal outcome that allows for outcome categorizations to vary across trials. It is common for the reporting of ordinal outcomes to differ from trial to trial, and the proposed model allows for data from all available trials to contribute to estimation of the comparative effects regardless of how the outcome was reported. In Chapter 3, we detail a procedure that can be used to design a connecting trial between two disconnected sub-networks such that the expected ranking of the best intervention is optimized. In Chapter 4, we propose using posterior model probabilities to select between component network meta-analysis models, a method that had not yet been implemented in the NMA literature. Our work therefore contributes to multiple facets of the NMA literature, including model formulation, utilization of the existing network to design a new trial, and model selection.
Comments
Description
Keywords
Citation
DOI
Source
Subject Categories
Copyright