Risk of bias of included studies
As shown in Table S3 andFigure S2 , there were no specific descriptions about random sequence generation in nine studies (31, 48, 50, 60, 65, 72, 74, 77) and they were determined to be “unclear”. There were 5 studies with both a sham group and a no treatment control (35, 40, 64, 74, 77) which were defined as unclear risk for blinding of participants. However, studies with no treatment control were scored as high risk of bias. In all included studies, there was no attrition bias from reviewing methods and results. Since results could be directly obtained, reporting bias was assessed as ‘low risk’ for all included studies. In addition, other sources of bias such as improper funding were not described and were at unclear risk of this bias.