Network metaanalysis

Many packages fit the second stage of a twostage IPD network metaanalysis, such as network in Stata, and netmeta in R. A nice overview of various packages is given here.

Bayesian onestage and twostage approaches to network metaanalysis are provided here, and could be adapted for IPD situations.
â€‹

For combining IPD and aggregate data, multilevel network metaregression (MLNMR) can be seen as a natural extension of IPD network metaregression for incorporating aggregate data.

In brief, the approach simultaneously fits a participantlevel model (to the IPD trials) and an aggregatelevel model (to the aggregate data trials). The approach recognises that aggregate data arise from averaging over a population of individuals (trial participants), and so the included aggregatelevel model arises from averaging (i.e. integrating) the included participantlevel model over the population in each aggregate data trial.

Notably, MLNMR reduces to a onestage network IPD metaregression when every trial has IPD, and reduces to a network metaanalysis of aggregate data when no adjustment is to be performed (i.e. the second stage of the twostage approach).

MLNMR models can be fitted using the R package multinma.