e., three meeting abstracts
were included in the analysis for China and two in the analysis for Mexico). Seroprevalence using any type of HBsAg assay was allowed (complete criteria are described in Supporting Table 1). Fixed effect (FE) and random effects (RE) meta-analyses of HBsAg seroprevalence rates from studies that met the inclusion criteria were conducted to calculate country-specific pooled CHB prevalence rates. RE analysis, which assumes heterogeneity among surveys, was considered more appropriate based on the nature of the data: HBV was unevenly distributed and we expected different rates from different surveys carried out in different see more populations in different locations at different times. FE analysis was conducted for comparison. Between-study heterogeneity was assessed for each country dataset using Cochran’s Q test
and the I2 statistic.14, 15 For most countries, data were insufficient for exploration of heterogeneity. Separate pooled rates were calculated for emigrants and for in-country populations for countries for which data were available, and results were compared using a Z test.15 Subgroup analyses were also done by decade of survey and by sex. For the 17 countries with at least 25 surveys, meta-regression analyses, based on the RE models using survey date as the covariate, Ku-0059436 chemical structure were done using Comprehensive Meta-Analysis software (Biostat, Englewood, NJ). For a few countries with low HBsAg seroprevalence rates (e.g., 上海皓元医药股份有限公司 Japan, Australia, New Zealand, Canada, and northern and western European countries), rates from large, population-based studies were used instead of meta-analysis. Study-quality assessment was done for only a subset of the data (i.e.,
Bangladesh, China, India, Iran, Korea, Pakistan, Philippines, Thailand, and Vietnam) to determine whether weighting based on study quality made a difference in the pooled prevalence rates. We developed a three-category scale (Supporting Table 2), scored each study, and calculated the pooled prevalence rates with and without the additional weighting factor, as described by Sutton et al.16 Flow of the systematic review is summarized by world region in Table 1. Results for individual countries are in Supporting Table 3. More than 17,500 articles were identified in PubMed searches; full text of 2,859 articles was assessed and data from 3,276 articles were entered into country-specific databases. In all, we found 1,373 articles reporting data meeting criteria for use in the meta-analyses. Many articles report data for more than one survey (e.g., pregnant women and military recruits) and these were entered separately. A total of 2,053 HBsAg seroprevalence surveys involving 18.6 million subjects were used in the meta-analyses (Table 2; Supporting Table 4).