An overview of network meta analysis in health care data: From manual methods to generative AI

Abstract

Network meta analysis(NMA) takes the concept of pair wise meta analysis further by incorporating both direct and indirect treatment effects. The basic requirement is that the included studies should have at least one treatment arm per study that is also present in at least one of the other included studies. This ensures that all treatment arms are looped within a network. In this paper we review the general concepts and assumptions of NMA. A generalized workflow and steps for conducting NMA is presented. We highlight two different frameworks that are widely used, namely the Bayesian and Frequentist based analysis. We briefly delve into the different software and techniques prevalent in this field. The subtle aspects of categorical, continuous and time-to-event endpoints are highlighted. We also discuss the challenges and areas of current research in this field. Finally, we discuss the advances in the application of Network meta analysis in healthcare.