Graphs to enhance understanding and improve interpretability of the evidence from network meta-analysis

Article type
Year
Authors
Chaimani A1, Mavridis D2, Salanti G3, Mavridis D4, Higgins J5, White I6, Spyridonos P7
1Paris Descartes University
2University of Ioannina
3Institute of Social and Preventive Medicine, University of Bern, Switzerland
4Department of Hygiene and Epidemiology, University of Ioannina School of Medicine, Greece
5School of Social and Community Medicine, University of Bristol, UK
6MRC Biostatistics Unit, Cambridge, UK
7Department of Medical Physics, University of Ioannina School of Medicine , Greece
Abstract
Objectives: To present and explain a series of graphical and numerical tools that can be used in network meta-analysis (NMA) in order to 1) present the evidence base, 2) evaluate the assumptions, and 3) present the results. We offer a Stata package with nine commands that can be used to produce the suggested graphs.

Description: This hands-on workshop begins with a brief presentation of the NMA model, its main assumptions and a demonstration of how network meta-analysis can be performed in Stata. We introduce network plots that offer an overview of the evidence base and its characteristics. We present graphs that can be used to evaluate important assumptions of network meta-analysis, specifically, the presence of important inconsistency (disagreement between direct and indirect evidence), small-study effects (by extending the standard funnel plot) and heterogeneity (prediction plots). Finally, we discuss several graphical options for the presentation of NMA findings and for treatment ranking (such as plots of the cumulative ranking curves). Stata commands are provided for all graphical tools. In order to benefit, participants will need to bring their laptops with Stata 12,13 or 14 installed.