Meta-analyses
We used the random-effects model to conduct the meta-analyses and to pool the proportions of HSV-2 positive cases.22 We used the Freeman-Tukey double arcsine transformation by specifying the command sm=“PFT” in the analysis.23 We presented Cochran’s Q statistic to assess the effect size heterogeneity.24 I2 was used to measure the between-study heterogeneity. We also presented prediction intervals, defined as the 95% confidence interval of the distribution of true HSV-2 seroprevalence around the estimation.
The pool results were stratified by different characteristics, including age groups, gender, year of data collection, and study sites. We mapped the pooled seroprevalence in the unit of provinces in China to demonstrate the geographic difference in the HSV-2 disease burden.