Statistical analysis
Meta-analysis was conducted using Stata software 16.0 (Stata, College Station, TX, USA). The outcomes of the dichotomous variables, such as odds ratio (OR) and hazard risk ratio, were included in this study. The included effect estimates (OR, PR, RR and HR) were combined together in subgroup analysis (16, 17). The I2 statistical test was used to assess the heterogeneity of included studies. The findings with p < 0.05 were statistically significant. When the estimated effect was heterogeneous (I2> 50% or p <0.05), the random effect model was applied; otherwise, the fixed effect model was applied. The sensitivity of each study was assessed by omitting one study at one time to reduce heterogeneity of the included studies. We did not assess publication bias because less than ten studies were included (18).