I'm conducting a large systematic review where multiple arms are compared directly and indirectly. I want to integrate the direct and indirect evidence statistically in a network meta-analysis.
It also depends on the statistic approach you are going to use: frequentist or Bayesian. With Stata you could run a frequentist NMA with the appropriate commands, while for the Bayesian approach you would need R or any BuGS version.
The network suite in Stata written by Ian White is probably the "easiest" software package to use for conducting a network meta-analysis. It can produce a network map and most important can also evaluate the design-by-treatment inconsistency and side-splitting (i.e. the node-splitting) inconsistency. The only limitation is that it can only deal with continuous or binary data.You can check out the Stata journal for a recent article by White with regard to how to use network suite. If you wish to do a bayesian network meta-analysis, then you have to use BUGS as Ivan said.
If you would like to perform the analysis using the Bayesian approach, the most user friendly recently developed software is NetMetaXL. Its development has been supported by CADTH and in the following link you can find a publication/manual on how to perform the analysis. It is rather straightforward and easier to use/handle if you are not that familiar with R and BUGS.
https://www.cadth.ca/resources/hta-excel-tools
In addition, if you are familiar with R, there are 3 available packages to perform network meta-analysis: gemtc, pcnemeta and netmeta, Its of them has its advantages and disadvantages and depends again on the question whether you would like to use frequentist or Bayesian approach. Please have a look at: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4277278/ for a review comparing the 3 aforementioned packages.
I have found the GeMTC package for R the easiest to use (www.drugis.org/software/r-packages/gemtc)
There is a GUI that gets your data into the correct format for GeMTC. It is no longer in development, but is still available from www.drugis.org/software/addis1/gemtc. It allows you to enter data, run analyses and create graphics from a GUI so is great for getting started/understanding what is going on.
By far the easiest is MetaXL (www.epigear.com) Please note that this is unrelated to NetMetaXL noted in a previous post above. This software uses a GPM framework and thus avoids the various problems associated with assumptions under Bayesian and multivariate frequentist approaches.