Details of included studies|
numbers of included studies for each direct comparison; numbers of subjects for each treatment
study bias assessment
the quality assessment of included studies
Details relating to the network|
Network diagram including the number of trials included in each link
Sources of heterogeneity must be assessed and the impact of heterogeneity must be analysed.
Evaluation of the “confidence” in the network (amount of evidence, homogeneity, consistency)
How good a fit the chosen model is to the data set.
Details relating to effect estimates|
point estimates and confidence/credible intervals
95 % credible intervals/probability intervals must be included when reporting the effect estimate
absolute effect of each intervention [when reporting input parameters of economic modelling]
reporting of estimates and variances of the direct comparisons that form the indirect comparison
Comparison of results from direct evidence with results from NMA
Sensitivity analyses if necessary, and an explanation of the differences compared with standard MA
The key information needed is to provide separately the direct and indirect estimates, and try to provide the quality of evidence supporting each rankings [perhaps rankograms as well] and probabilities of each intervention being best.
For Bayesian mixed treatment comparisons: Probability Rankograms and Surface Under Cumulative Ranking Curve
Well reasoned sensitivity analysis, including/excluding different data sources.