Statistical analysis

All statistical analyses were performed using Cochrane Review Manager Version 5.3 (Cochrane Library, UK). Fixed-effect model (Mantel-Haenszel method) was used to estimate the pooled statistic. The heterogeneity among the included studies was detected using I². On the other hand, when the P-value from the heterogeneity test was less than 0.1, a random effects model (DerSimonian-Laird method) was conducted. Odds ratios (OR) were used for results of dichotomous data, and 95% confidence intervals (CI) were also counted. A P < 0.05 was regarded as statistical significance for the pooled OR. Publication bias was graphically judged from the Begg adjusted rank correlation test [30] and the Egger regression asymmetry test [31], when the number of the included studies was more than six.