A Practical Guide to Evidence Synthesis: Meta-Analysis, Network Meta-Analysis, Meta-Regressions and Beyond

Meta-analysis has become an indispensable tool in medical and dental research, providing an estimate of treatment efficacy or harm by combining data from all relevant studies identified by a systematic review. Meta-analysis produces evidence at the highest level of the study design hierarchy. Meta-analysis is a broad term for a set of statistical tools used to analyse data across a range of research questions. Different research questions have different criteria for selecting relevant studies and extracting and organising data in different forms and require different statistical approaches to synthesising the data. This article focuses on meta-analysis for comparing interventions. We use plain but precise language to explain the statistical concepts and assumptions underlying these methods. This article is divided into three parts. We begin with statistical models for and key assumptions of standard pairwise meta-analysis, comparing two treatment groups. We then extend our discussion to network meta-analysis, comparing multiple treatment groups. Finally, we discuss a few issues that arise in selecting studies and preparing data for a meta-analysis. Throughout, we use examples from periodontal regeneration; all analyses were conducted using the free statistical software R. We strongly encourage close collaboration between periodontal researchers and experienced statisticians when planning and conducting meta-analyses.