Statistical analysis

Review Manager software version 5.3 was used with a fixed-effect model to pool data, and a forest plot was generated to calculate risk ratios (RRs) and their 95% confidence intervals (CIs). The chi-square test was performed to test the hypothesis (Z distribution, P < 0.05 was considered statistically significant). I² statistics were used to assess the heterogeneity among studies (values less than 50% indicate acceptable heterogeneity and values greater than 50% indicate substantial heterogeneity). When heterogeneity was unable to be readily explained, one analytical approach is to incorporate heterogeneity into a random-effects model. For any particular set of studies in which heterogeneity is present, a confidence interval including the random-effects pooled estimate is wider than a confidence interval including a fixed-effects pooled estimate. Therefore, a random-effects model and subgroup analysis were performed if the source of heterogeneity was unable to be determined. A sensitivity analysis was performed to predict stability and bias (including publication bias, selective reporting within studies, etc.). A meta-analysis was conducted if sources were available from a minimum of two studies.