Stata 12.0 statistical software package

The Stata 12.0 statistical software package (Stata Corp, College Station, TX, USA) was used for all data analyses. We calculated the OR/RR/HR with a 95% CI. The Q-test was used to assess the statistical heterogeneity and judge the p-value qualitatively. The I² value in the I² test describes the proportion of the total variation due to heterogeneity rather than sampling errors, and I²> 50% or P<0.05 indicates a high degree of heterogeneity. The random-effects model was used; otherwise, there was no heterogeneity, and the fixed-effects model was used. The potential bias was assessed by a funnel plot and Egger’s test, and when the funnel plot was symmetrically distributed, there was no significant bias. Conversely, if the funnel plot exhibits skewness and asymmetry, this indicated bias.

The Stata 12.0 statistical software package (http://www.stata.com/) and the Review Manager 5.3 software (The Cochrane Collaboration) were applied to conduct publication bias analysis, meta-analysis and sensitivity analysis. Potential publication bias was determined by both the Egger regression test for a funnel plot [13 (link)] and the Begg–Mazumdar test, which is based on Kendall’s-τ [14 (link)]. For dichotomous variables, the odds ratio (OR) with 95% confidence interval (95% CI) was derived for each outcome; for continuous variables, we calculated the standardized mean difference and 95% CI. We performed the meta-analysis using a fixed-effect model if no significant heterogeneity was present; otherwise, a random-effect model was applied. To assess heterogeneity between studies, we performed a chi-square test and estimated the I² statistic. For all analyses, P < 0.05 was considered statistically significant. The study was approved by the ethics committee of Lishui People’s Hospital of Wenzhou Medical University. All the protocols and experimental procedures were in accordance with the Declaration of Helsinki and other international rules (S1 Table).

This study followed the recommended PRISMA statement. STATA 12.0 statistical software package (Stata Corporation, College Station, TX) was used for statistical analysis. All median with range or interquartile range were converted to the form mean with standard deviation^{20 (link)}. The weighted mean difference (WMD) and standard mean difference (SMD) with 95% confidence intervals (CIs) were used for appropriate continuous variables. According to previous studies, we used the method below to choose the effect model. The effects of the outcomes were pooled using a fixed-effect model, while a random model was employed when significant heterogeneity was detected. Heterogeneity was assessed by I² and considered significant when I² > 50%^{21 (link)–23 (link)}. Potential publication bias was assessed via Funnel plots, as well as the Bagger’s and Egger’s tests. P < 0.05 (two-sided) was considered statistically significant in our meta-analysis.