Data analyses
Two authors separately collected the required, relevant data with any
discrepancies between them being resolved by group consultation. The two
authors used the Cochrane collaborative bias risk tool to assess the
risk of bias independently in five aspects and used the Grading of
Recommendations, Assessment, Development and Evaluation (GRADE) to
estimate the quality of evidence for each
outcome.[15] The results of each RCT were assigned
as dichotomous data, analysed by the Mantel Haenszel method, and
presented as relative risk (RR). The summary RR and 95% confidence
interval (CI) of the survey results were calculated using a random
effect model.[16] P ≤ 0.5 was considered to be
statistically significant, and heterogeneity was assessed through
I‐square (I2) and Q statistics; I2> 50% is considered
substantial[17,18]. Using funnel
plots[19], Egger and Begg tests were conducted to
evaluate the publication bias of studies with reported results ≥
10.[20]