E-Thesis 41 views 16 downloads
Bayesian evidence synthesis methods for two diagnostic tests: application to Alzheimer’s disease dementia / Athena McBride
Swansea University Author: Athena McBride
-
PDF | E-Thesis – open access
Copyright: The Author, Athena McBride, 2025. Licensed under the terms of a Creative Commons Attribution-Only (CC-BY) license. Third party content is excluded for use under the license terms.
Download (7.17MB)
DOI (Published version): 10.23889/SUthesis.69487
Abstract
This thesis considers a range of methodological challenges related to the synthesis of comparative diagnostic accuracy studies that evaluate two tests in the same patients, and aims to address them through the development of novel meta-analysis methodol-ogy in a Bayesian framework. The novel methods...
| Published: |
Swansea, Wales, UK
2025
|
|---|---|
| Institution: | Swansea University |
| Degree level: | Doctoral |
| Degree name: | Ph.D |
| Supervisor: | Owen, Rhiannon K. ; Fry, Richard ; Bujkiewicz, Sylwia ; Quinn, Terence J. |
| URI: | https://cronfa.swan.ac.uk/Record/cronfa69487 |
| Abstract: |
This thesis considers a range of methodological challenges related to the synthesis of comparative diagnostic accuracy studies that evaluate two tests in the same patients, and aims to address them through the development of novel meta-analysis methodol-ogy in a Bayesian framework. The novel methods are applied to a real-world example in Alzheimer’s disease dementia, for which test comparisons are important in opti-mising diagnostic pathways and improving detection. Firstly, the thesis introduces complex dependence structures present in meta-analyses of comparative diagnostic accuracy studies; in particular, within-study associations arising between sensitivi-ties and specificities when patients undergo both tests of interest. This thesis assesses the impact of accounting for within-study dependencies on key test accuracy param-eters by fitting a meta-analysis model that treats the two tests as independent to simulated data in which the associations are known. Ignoring within-study depen-dencies is shown to lead to underestimation of joint sensitivity and specificity, which measure the agreement between tests and enable modelling of diagnostic test com-binations and pathways. This motivates the need for methodological development to jointly model the accuracy of two diagnostic tests and the associations between them. Novel Bayesian meta-analysis models for synthesising evidence on the ac-curacy of two diagnostic tests are developed, capturing within-study dependencies using bivariate copulas. Motivated by an example in Alzheimer’s disease dementia, the bivariate copula framework is shown to lead to improved model fit compared to the approach that does not account for within-study associations. The bivariate copula models are extended to incorporate individual participant data on combined test performance, capturing within-study dependencies through trivariate copulas. These methods can be used to inform optimal combinations of diagnostic tests for health care policy and decision-making. |
|---|---|
| Item Description: |
ORCiD identifier: https://orcid.org/0000-0003-1564-0740 |
| Keywords: |
diagnostic test accuracy, meta-analysis, test comparison, health technology assessment, copula, Bayesian analysis |
| College: |
Faculty of Medicine, Health and Life Sciences |
| Funders: |
Health Data Research UK (HDR UK), HDR UK Studentship N1WA1 |