Aristotle University Of Thessaloniki (AUTH)
Development and comparison of statistical methodology in network meta-analysis of diagnostic test accuracy studies applied to cardiovascular diseases
Abstract
dc:descriptionNowadays, meta-analysis (MA) has evolved into a popular, influential research design with a substantial impact on the formulation of medical practice and health policy.Many systematic reviews (SRs) and MAs of diagnostic test accuracy studies (DTA) are published every year providing direct comparisons between the different tests of interest. Network meta-analysis (NMA) expands the scope of a conventional pairwise MA by analyzing simultaneously both direct and indirect comparisons across trial as well as indirect comparisons. The problem lies in the fact that the NMA methods for comparing interventions cannot be applied to DTA studies due to the double outcome of these studies (sensitivity & specificity), the different design of the primary studies and the application of the diagnostic tests on the same participants which violates the principle of independence. In this doctoral thesis, an overview of systematic reviews (OoSRs) was first conducted in DTA studies related to cardiovascular diseases (CVDs) for the period 2010-2020. The purpose of this study was to describe them and assess the quality of their abstracts according to PRISMA-DTA for abstracts reporting guidelines. Specifically, 72 SRs were found, half of which were published in general medical journals and the other half in special journals. In addition, most of the reviews were in the field of acute myocardial infarction and coronary artery disease with imaging tests being the most frequently encountered diagnostic tests. Moreover, most journals were found to adopt the PRISMA statement and have structured abstracts, while overall completeness reporting was significantly higher in abstracts of studies published in general journals compared to specialized journals. Furthermore, this thesis presents a scoping review of methodological articles that have been published and deal with the models and methods that have been developed for DTA-NMA by applying and comparing them using data of 32 application papers on different medical fields. In detail, 11 models based on Bayesian statistics and two hierarchical meta-regression approaches were found which do not take into account the correlation between tests applied to the same participants. Comparison of all methods showed that sensitivity estimates are comparable between meta-regression approaches and DTA-NMA models while specificity estimates are higher in meta-regression. In addition, in this PhD thesis, an OoSRs was performed on studies related to the diagnosis of pulmonary embolism using imaging tests in order to apply and compare as many of the above DTA-NMA methods as were feasible according to the nature of the data. Finally, 13 SRs and 33 DTA studies were found which included four imaging tests. The results of the NMA showed that different methods can potentially lead to different estimates of diagnostic accuracy. The choice of method for synthesizing DTA studies via NMA depends on the type of diagnostic tests (categorical or continuous), the design of the primary studies, and the researchers' familiarity with Bayesian statistics.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Pagkalidou, Eirini
- Παγκαλίδου, Ειρήνη
Subjects
dc:subject × 14- Μετα-ανάλυση δικτύου
- Έμμεσες συγκρίσεις
- Μελέτες διαγνωστικής ακρίβειας
- Καρδιαγγειακά νοσήματα
- Network meta-analysis
- Indirect comparisons
- Diagnostic test accuracy studies
- Cardiovascular diseases
- Άλλες Ιατρικές Επιστήμες
- Ιατρική και Επιστήμες Υγείας
- Ιατρική, άλλοι τομείς
- Other Medical Sciences
- Medical and Health Sciences
- Medicine, miscellaneous
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/53551