Vasdekis, Vassilis
Athens University of Economics and Business, Department of Statistics
Evidence synthesis methodologies become essential as more and more analyses are available for a specific research question. This dissertation has been focused on the evidence synthesis methods in healthcare, using randomized control trials (RCT) as a source of evidence. The first method described is the meta-analysis, an overall analysis to pool the treatment effect of two specific treatments being compared directly. The meta-analysis technique has two effect models, the fixed and the random effects, which their differentiation relies on a fundamental assumption over the uncertainty sources; the latter assumes between-study variance in additional to the within-study variance, which is the only source of variability in the fixed effect model. Furthermore, the indirect treatment comparisons (ITC) overcomes the limitation of the meta-analysis, making feasible the comparison of treatments without the requirement of them to be directly compared in an RCT. The ITC uses a common comparator, a treatment which has been compared with the other two treatments of interest, if both indirect and direct evidence are available a pooled estimation can be performed. The ITC and pooled effect methodologies can be considered as mixed treatment comparisons (MTC), however, since they are based on trivial mathematical equations they cannot exploit the geometry of the network made by the treatments connected. The last and most important evidence synthesis tool that has been presented is the network meta-analysis, the extension of meta-analysis. A network of multiple treatments, connected directly or indirectly by multiple studies is analyzed simultaneously by fixed or random effects. The dissertation is organized in two parts; the theory of these methods, accompanied with examples in the Bayesian and frequentist prospective for continuous outcomes, and an extensive application in network meta-analysis in patients with COPD, using a publication performed by Mapi [1]. The main scope of this thesis has been to present both in theory and application all the main steps of evidence synthesis and compare the estimations among different approaches and models. As a conclusion, the Bayesian and frequentist approaches deemed to result in approximately same estimations, with the random effects estimations in both cases providing more uncertainty around them.
Chronic Obstructive Pulmonary Disease (COPD)
Σύνθεση δεδομένων
Evidence synthesis
Meta-analysis
Indirect treatment comparisons (ITC)
Network meta-analysis
Έμμεση σύγκριση θεραπειών
Μετα-ανάλυση δικτύων
Χρόνια πνευμονική ανεπάρκεια