multinma: a comprehensive R package for Bayesian network meta-analysis with aggregate data, individual patient data, or a mixture of both

The multinma R package provides a comprehensive and user-friendly suite of functions for performing network meta-analysis (https://dmphillippo.github.io/multinma/), with models estimated in a Bayesian framework using Stan. The package includes functions that streamline the setup of networks, model specification and fitting, and facilitate model comparison and diagnostics. Fixed and random effects models, and node-splitting and unrelated mean effects inconsistency models are all supported. Any user-specified regression model for covariates can be provided using standard R formulas. After models have been fitted, functions are available for producing posterior summaries of relative effects, rankings, and absolute predictions, and creating flexible graphical outputs using ggplot and ggdist. A wide range of arm- and contrast-based data types and likelihoods are supported, including parametric and flexible semi-parametric models for time-to-event data that were added in a recent update. Models can be fitted with aggregate data (AgD) and/or individual patient data (IPD), all within a common user-friendly syntax. When a mixture of IPD and AgD studies are available these are synthesised coherently and without aggregation bias using multilevel network meta-regression.
This session will provide an overview of the multinma package and its functionality, using real worked examples. We begin by demonstrating an AgD network meta-analysis with a network of treatments for plaque psoriasis and a binary outcome. With the same example, we then show how to incorporate IPD available from a subset of studies in a multilevel network meta-regression, adjusting for effect-modifying covariates to produce population-adjusted estimates in a target population of interest for decision-making. Lastly, we demonstrate the analysis of time-to-event data with progression free survival on a network of treatments for newly diagnosed multiple myeloma. We conclude with a discussion of future planned updates.