Heterogeneity and publication bias
Inter-study heterogeneity was explored using the Chi2-statistic, but the I2 value was calculated to quantify the degree of heterogeneity across trials that could not be attributable to chance alone. If heterogeneity was significant (I2 > 75%), 3 strategies were used to assess data validity and heterogeneity: (1) a subgroup analysis of higher quality studies (quality score ≥10); (2) funnel plots to evaluate publication bias (i.e., funnel asymmetry) with Egger’s test; and (3) a meta-regression to assess the effects of covariates on the primary outcome of interest.
A domain-based evaluation of risk of bias was performed in accordance with the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.08 as previously described7. Three authors (MM, MG, GN) subjectively reviewed all studies included in this review and assigned a value of ‘critical’, ‘serious’, ‘moderate’ or ‘low’ to the following questions: (Domain 1) Was the allocation sequence adequately generated? (D2) Was allocation adequately concealed? (D3) Was the treatment adequately classified? (D4) Were data affected by deviation from intended intervention? (D5) Were incomplete outcome data sufficiently assessed? (D6) Are reports in the study free of the suggestion of selecting outcome measures or (D7) of selective outcome reporting?; ‘Risk of bias’ plots were performed using package ‘plotvis’ R-project, following the Review Manager® Version 5.3 layout (The Cochrane Collaboration, Software Update, Oxford, UK).