Statistical analysis
Statistical meta-analysis was performed using the RevMan 5.3 software (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2011). Confidence intervals (CI) were set at 95%. Mean difference (MD) and odds ratios (OR) were used in the analysis. The results were calculated using the DerSimonian-Laird random effect model (REM) revealing significant heterogeneity in the methodological characteristics of the included studies19. The cut-off for statistical significance was set at p<0.05. Mean values and standard deviations were calculated according to the equations proposed by Hozo et al when not provided by the studies20. To assess the between-study heterogeneity more precisely, both the Chi-square based Q statistic test for heterogeneity and the I2 statistic to quantify the proportion of the total variation attributable to heterogeneity were calculated. When there was no statistically significant heterogeneity (p>0.05, I2<50%), a fixed effect model was adopted. Otherwise, a random effect model was chosen21. The heterogeneity was considered statistically significant if a p<0.05. Publication bias was not tested due to heterogeneity of the included studies, which is a confounder that may influence the methodological integrity of these tests.