An R package for performing network meta-analysis and network meta-regression with aggregate data, individual patient data, or mixtures of both.
Network meta-analysis (NMA) combines (aggregate) data from multiple studies on multiple treatments in order to produce consistent estimates of relative treatment effects between each pair of treatments in the network TSD2multinma.
Network meta-regression (NMR) extends NMA to include covariates, allowing adjustment for differences in effect-modifying variables between studies TSD3multinma. NMR is typically performed using aggregate data (AgD), which lacks power and is prone to ecological bias. NMR with individual patient data (IPD) is the gold standard, if data are available.
Multilevel network meta-regression (ML-NMR) allows IPD and AgD to be incorporated together in a network meta-regression methods_paper,Phillippo_thesismultinma. As in IPD NMR, an individual-level regression model is defined. AgD studies are then fitted by integrating the individual-level model over the respective covariate distributions. This correctly links the two levels of the model (instead of "plugging in" mean covariate values), avoiding aggregation bias. Population-adjusted treatment effects TSD18multinma can be produced for any study population in the network, or for an external target population.
Models are estimated in a Bayesian framework using Stan Carpenter2017multinma. Quasi-Monte Carlo numerical integration based on Sobol' sequences is used for the integration in ML-NMR models, with a Gaussian copula to account for correlations between covariates methods_paper,Phillippo_thesismultinma.