Table 1 Glossary of terms

Pairwise comparisons: All treatment comparisons in the network irrespective of the availability of data. A network with T treatments has T(T-1)/2 pairwise comparisons. Depending on whether there are studies reporting the studied outcome, the pairwise comparisons can be distinguished into observed for this outcome, observed for other outcomes, and unobserved.

Direct evidence: The evidence available (statistical information derived from data) about a pairwise comparison that is available from direct, within study information about that comparison.

Indirect evidence: The evidence available (statistical information derived from data) about a pairwise comparison that is not available from within study information, i.e. is obtained indirectly via a common comparator or chain of comparisons.

‘Only direct’ estimate: Relative treatment effect estimated in an network meta analysis that is derived only from direct evidence.

‘Only indirect’ estimate: Relative treatment effect estimated in an network meta analysis that is derived only from indirect evidence.

Mixed estimate: Relative treatment effect estimated in an network meta analysis that is derived from both direct and indirect evidence.

Network meta-analysis estimates: Estimates of relative treatment effects derived from network meta analysis; these can be distinguished into ‘Only direct’, ‘Only indirect’ and Mixed estimates.

Within-study assessment of bias due to missing evidence: Bias arising from missing results due to selective outcome reporting i.e. results being reported, but not others, within studies published or otherwise known to exist.

Across-study assessment of bias due to missing evidence: Bias introduced from missing studies because they are entirely unpublished i.e. not known to exist.