Back to search

University of Ioannina

Synthesizing randomized and non-randomized studies of interventions in meta-analysis models

Abstract

dc:description

Randomized controlled trials (RCTs) are considered the gold standard for evaluating the effectiveness of interventions, because randomization reduces confounding and ensures strong internal validity. However, in many areas the available RCT evidence is limited because trials often include selected patients, short follow-up, small sample sizes and few direct treatment comparisons. This is especially important in settings where randomized evidence is sparse or difficult to generate, such as rare diseases, oncology and precision medicine, and in decision-making contexts such as Health Technology Assessment (ΗΤΑ). Real-world evidence (RWE) can add useful information, but it also raises methodological challenges such as bias, confounding, heterogeneity and limited generalizability. This thesis studies statistical methods for including RWE in meta-analysis (MA) and network meta-analysis (NMA). Its goal is to provide a clear framework for when and how non-randomized studies of interventions can be used in pairwise meta-analysis and, especially, in NMA. The thesis focuses on key assumptions for valid analysis and the effect of remaining bias and confounding. We start with an introduction to evidence synthesis methods such as pairwise MA and NMA. Then we present a methodological scoping review of statistical methods for incorporating non-randomized evidence into evidence synthesis. The identified approaches include naïve pooling, design-adjusted models, meta-analytic predictive (MAP) priors, robust MAP and hierarchical cross-design models. We present the methods’ assumptions and compare them in terms of bias control, precision and usefulness in sparse data settings. The thesis also presents bridgemeta, an R package implementing selected Bayesian pairwise meta-analysis models for combining randomized and non-randomized studies of interventions. Overall, the thesis evaluates these methods both in theory and practice, with focus on the balance between gaining more information and increasing bias. Its final aim is to provide practical guidance for transparent and reliable evidence synthesis when RCT evidence alone is not enough.

Degree

thesis:*
Grantor dc:publisher
University of Ioannina
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Koutsiouroumpa, Ourania
  • Κουτσιουρούμπα, Ουρανία

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/62492

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Koutsiouroumpa, Ourania; Κουτσιουρούμπα, Ουρανία. Synthesizing randomized and non-randomized studies of interventions in meta-analysis models. University of Ioannina, 2026. http://hdl.handle.net/10442/hedi/62492