Evidence synthesis methods for continuous outcomes

Abstract

Evidence synthesis methods have become an indispensable tool for clinicians, policy makers and researchers across various fields in this era of evidence-based research. The standard pairwise meta-analysis enables the synthesis of quantitative findings from multiple studies to provide an overall estimate of the studied topic and guide decision-making further, while network meta-analysis enables additional comparisons of (medical) conditions that have not been directly compared in a head-to-head setting. This thesis considers both statistical techniques focusing on continuous outcomes and provides novel solutions to the issues of a) unavailability of individual participant outcome data by generating and analysing pseudo observations and b) missing outcome participant data by proposing a Bayesian one-stage approach employing the pattern mixture model. For the former, a user friendly web-based tool is created to encourage technical and non-technical researchers to take up the most appropriate method for meta-analysing continuous outcomes measured in a pretest-posttest design, without requiring any programming skills from the user.