2.5 Data analysis
Stata 15.0 was applied to conduct meta-analysis, and the research results are presented as forest plots with 95% confidence intervals. The combined correlation between the rate of hearing recovery and incidence of vertigo was assessed by means of the Mantel–Haenszel method as dichotomous data, combined with the estimation of the ratio of the two groups of patients. The data from all the eligible studies in this research were quantitatively combined. The studies’ statistical variability was analyzed by the P value (P ≤ .05 implying obvious heterogeneity) and the I2 statistic as a derived Cochran Q statistic (≥50%, high level of heterogeneity; 25% to <50%, moderate level of heterogeneity; and <25% means slight level of heterogeneity). When I2 < 50% and/or P ≤0.05, a fixed-effects model was applied; otherwise, a random-effects model was applied. The funnel plots, which do not have sufficient power with fewer than 10 researches involved, were applied to visually evaluate the publication bias and potential asymmetry.[9]