Features of replications.

Percentages refer to the total number of highly cited studies in each category. For studies with a replication, the table describes the type of study; most studies, however, did not have a valid replication. A network meta-analysis including a direct comparison for the groups of interest [44] was classified as a regular meta-analysis. Two meta-analyses [42, 43] replicating two highly cited studies are counted twice in their respective categories.