The literature search, study selection, and data extraction were done between Jan 11, 2014, and Sept 9, 2017, and data analysis was done from Sept 10, 2017, to Feb 24, 2018. The study selection process is shown in figure 1; a list of excluded studies, with reasons for exclusion, and a list of retained studies is provided in the appendix (pp 26–272). 133 studies were retained for the network meta-analysis, 81 in children and adolescents, 51 in adults, and one including children, adolescents, and adults. In total, 14 346 children and adolescents and 10 296 adults were included. For 83% of studies, additional data and information not reported in the full-text paper were used. The appendix (pp 283–381) reports the main characteristics of included studies. The risk of bias was rated overall low in 23·5% of studies in children and adolescents, unclear in 65·4%, and high in 11·1%. The risk of bias was overall low in 27·5% of studies in adults, unclear in 56·8%, and high in 15·7% (appendix pp 382–458).
