3.5. Statistical analysis

Revman 5.3 software (Cochrane Collaboration) was used for bias evaluation. For continuous variables (time to first urination), the results will be reported as mean difference (MD) and 95% confidence interval (CI); count data (effective rate) will be compared by odds ratio (OR) and 95% CI are calculated. In the heterogeneity test, if I² < 50%, P > .10, there is no significant heterogeneity. We chose the fixed effect model to combine the effect quantity. If the combined data is I² > 50%, P < .10, it indicates high heterogeneity. We choose the random effect model to combine the effect quantity.^[15]

Using Gemtc 0.14.3 (Developed by van Valkenhoef G et al) and Stata14.2 (STATA Corp., College Station, TX, USA) for mesh meta-analysis.^[16,17] In Gemtc software, Bayesian mesh meta-analysis is realized by Markov Chain Monte Carlo (MCMC) method, through 4 chains. For simulation, the number of iterations is set to 50,000, and the step size is set to 10.^[18] At the same time, the potential scale reduction parameter (potential scale reduced factor, PSRF) is used to evaluate the convergence of the results. When the PSRF is close to 1, it indicates that the results have good convergence and the obtained results are highly reliable.^[19]