An R package for performing network meta-analysis using INLA.
Guenhan, B.K., Friede, T., Held, L. (2018) A design-by-treatment interaction model for network meta-analysis and meta-regression with integrated nested Laplace approximations. Res Syn Meth. 2018;1-14. https://doi.org/10.1002/jrsm.1285
Sauter, R. and Held, L. (2015). Network meta-analysis with integrated nested Laplace approximations. Biometrical Journal 57 1038--1050.
Jackson, D., Barrett, J. K., Rice, S., White, I. R. and Higgins, J. P. (2014). A design-by-treatment interaction model for network meta-analysis with random inconsistency effects. Statistics in Medicine 33 3639--3654.
Rue, H., Martino, S. and Chopin, N. (2009). Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 71 319--392.
Dias, S., Welton, N. J., Sutton, A. J. and Ades, A. (2011). NICE DSU Technical Support Document 2: A Generalised Linear Modelling Framework for Pairwise and Network Meta-analysis of Randomised Controlled Trials. Last updated September 2016.
Dias, S., Sutton, A. J., Welton, N. J. and Ades, A. E. (2013). Evidence synthesis for Decision Making 3: Heterogeneity--Subgroups, Meta-Regression, Bias, and Bias-Adjustment. Medical Decision Making 33 618--640.
Network meta-analysis is a generalization of pairwise meta-analysis to analyze networks of trials comparing two or more treatments simultaneously (Dias et al, 2011). Bayesian hierarchical models are commonly used for network meta-analysis (Dias et al, 2011). The default choice for performing inference within such models are Markov Chain Monte Carlo (MCMC), for example using BUGS-variants programs such as JAGS. A deterministic approach to do fully Bayesian inference for latent Gaussian models (LGMs) are integrated nested Laplace approximations (INLA) (Rue et al, 2009) which is a fast and accurate alternative to MCMC. INLA methodology is implemented as an R package INLA (<www.r-inla.org>). Sauter and Held (2015) has shown that INLA can be used for fitting many NMA models including fixed effect and consistency models, node-splitting models.
This package extends the INLA implementation of Sauter and Held (2015) to Jackson model (Jackson et al, 2014) and network meta-regression and extracts the features needed for NMA models from INLA R package and presents in an intuitive way (Guenhan et al, in preparation). Currently, contrast-based network meta-analysis using trial-arm level data for datasets with binary, continuous, and survival outcomes are supported. Note that the installation of R package 'INLA' is compulsory for successful usage. The 'INLA' package can be obtained from <http://www.r-inla.org>. We recommend the testing version, which can be downloaded by running: source("http://www.math.ntnu.no/inla/givemeINLA-testing.R").
Type vignette("nmaINLA") to how to use this package.
The development version of nmaINLA is available on GitHub <https://github.com/gunhanb/nmaINLA>.