Methods for including information from multi-arm trials in pairwise meta-analysis.