Discussion
This is a comprehensive systematic assessment of HSV-2 epidemiology in China. We included publications in Chinese and English major bibliography databases and involved over one million study participants. This study has added to the existing literature by providing a most up-to-date, comprehensive picture of HSV-2 epidemiology in China, mapping the Chinese provincial level of HSV-2 seroprevalence differences, and investigating language bias between publications identified from Chinese and English databases. Around 1 out of 12 people in the general population were infected by HSV-2. Stratified by study populations and geographic regions, we observed higher HSV-2 disease burden among higher-risk populations such as FSW and participants from northeastern China, calling for more preventive interventions for these populations. By comparing the HSV-2 prevalence synthesized from Chinese and English databases, we found a higher overall HSV-2 prevalence from English bibliographic databases than Chinese, indicating the existence of language bias and implying the importance of including both Chinese and English publications in the review.
The HSV-2 seroprevalence among the general population is 7.9% (6.9-8.8%), similar to the overall prevalence of HSV-2 in Asia,21 lower than in Europe, Sub-Saharan Africa, and Australia but higher than the Middle East and North Africa.8,10,12,13,30 With other regions whose prevalence was flat8 or increasing10,11, we observed the HSV-2 prevalence in China reduced by about 2% every five years, which could be attributed to better STI clinic access and improved STI education among the general population nationwide.31,32 Stratified by sex, we observed comparable HSV-2 prevalence between men and women. However, among other populations, we found a higher HSV-2 seroprevalence among women than men, especially higher-risk populations, with the seroprevalence significantly higher among FSW than MSM (52.1% vs . 12.2%), which revealed on the one hand, sustained HSV-2 transmission among this population33 and on the other hand, the importance of continuing healthcare and behavioral intervention to reduce HSV-2 infection among the populations vulnerable to HSV-2 infection. Geographically, we found that the Northeastern region (Liaoning, Jilin, and Heilongjiang) had the highest HSV-2 prevalence due to the high proportion of STI clinic attendees with suspected genital herpes symptoms and MSM living with HIV within this sample (10,202/15,904, 64.1%). Stratified by age group, we observed a steady increase of HSV-2 prevalence with age, consistent with increasing cumulative exposure risk to the virus over the sexual life span and incurability of HSV-2. The assessment of HSV-2 risk factors through meta-regression confirmed the observations in synthesized seroprevalences: populations of higher vulnerability to HSV-2 had higher HSV-2 prevalence as earlier studies had higher HSV-2 prevalence than later studies.
We found heterogeneities in HSV-2 seroprevalence estimation and study population composition between English and Chinese bibliographic databases, indicating language bias. First, from the estimation perspective, the overall HSV-2 seroprevalence was significantly higher from English databases than Chinese databases (20.5% vs . 13.6%, RR=1.10 (1.05-1.14)). Stratified by risk levels, although we found comparable HSV-2 seroprevalence estimation among the general population, the HSV-2 seroprevalence estimation was significantly higher among higher-risk populations such as FSW and MSM/MSM in English databases. Second, the composition of the study population significantly differed between the publications identified from English and Chinese databases. The sample size of general populations in publications from Chinese databases is more than 15-fold that from English bibliography databases, mainly attributable to the large-scale toxoplasmosis, rubella cytomegalovirus, herpes simplex, and HIV (TORCH) screening among pre-pregnancy and pregnancy women in clinical and community settings. Meanwhile, the sample size of the higher-risk population in English databases accounted for one-third of the overall identified sample size, compared to around 3% in Chinese databases. Third, Chinese publications were more geographically diverse, providing HSV-2 seroprevalence not just in southern China (e.g., Guangdong and Yunnan) but around the country. These observations might be due to several reasons. First, the regular TORCH screening results among the general population were harder to be published in journals published in English bibliographic databases than in Chinese databases.34 Second, higher-risk populations had higher HSV-2 vulnerabilities and potentially higher public health significance.35 Investigators promoted their results in journals published in English databases for wider attention and higher citation.36 Interestingly, on the other hand, we found lower HSV-2 prevalence estimation among intermediate-risk populations and drug users. The sample sizes and numbers of the study were much lower than those of higher-risk populations, which indicates an unmet study need for these populations. These observations indicate a potential existence of language bias by only including publications published in English bibliography databases in the review. Considering the huge quantity and varying quality, most Chinese journals have yet to be published in English bibliographic databases.37 The synthesized results might be distorted by restricting publications to English-language only in literature searching.38 Previous studies have shown that positive results in randomized controlled trials are more likely to be published and published in English bibliographic databases, potentially leading to overestimating the drugs’ effectiveness. The Cochrane Handbook for Systematic Reviews and the United States Institute of Medicine Guidelines for Systematic Reviews recommends including non-English-language literature published in English bibliographic databases in the review.17,39 Our study revealed that this might not be enough to eliminate the language bias if not including non-English bibliographic databases in the review. As a result, it could y lead to misinterpretation of the disease’s unmet needs and a suboptimal distribution of public health resources in HSV-2 prevention across different populations.
There are some limitations to be noted. First, we should have included all Chinese bibliographic databases in the literature search scope. There are other databases, such as Weipu Database. However, we have included the two most popular Chinese bibliographic databases in this review and have yet to observe differences regarding the impact of journals published in different Chinese databases. Thus, the potential publication bias of excluding other Chinese databases is minimal. Second, due to heterogeneities in variable categorization (such as age group), some studies’ subgroups cannot be extracted and are categorized into the mixed group or other populations. These studies’ information was not fully utilized. However, the synthesized results did not differ significantly from the categorized groups’ results. The overall estimation should be robust to these heterogeneities in variable categorization. Third, the current included studies, 29.4% of them had high ROB in both quality domains, and only 4.0% of the studies had low ROB in both domains, which implies a space for future studies to employ probability-based sampling methods to improve study quality. Fourth, we only presented the synthesized results for studies across different regions. However, there were heterogeneities regarding the study population. For example, in some regions, we identified more studies among higher-risk populations than in other regions, which could inflate the HSV-2, which calls for caution in interpreting the current results prevalence in this region.