Gravar-mail: Relation between treatment benefit and underlying risk in meta-analysis. Standard of "label invariance" should not be abandoned.