As shown in Figure 1, a total of 3,368 relevant articles were identified, of which 760 were further screened, and 355 cross-sectional studies (51 published in English and 304 in Chinese, which included 571,328 MSM) were finally included in the systematic analysis [10-364]. The characteristics of the included studies are summarized in Table 1.
Characteristics of the selected studies
The 355 eligible articles included in this study covered 59 cities from 30 provinces and municipalities of China (no studies were from Tibet). The sample size of the selected studies ranged from 30 to 1,498,841 (mean 4373, 95% confidence interval [CI]: 455 – 8,290) [17, 18]. Among the 355 selected articles, 6 reported results for multiple study sites, and 79 reported HIV prevalence estimates for more than one time period, resulting in a total of 581 HIV prevalence estimates. The first included study was reported in 2001 [12], with an HIV prevalence of 1.31%, and the studies provided overall estimates of the HIV prevalence among MSM in China from 2001 to 2018.
The sampling methods varied, with 123 studies using snowball sampling, 39 studies using respondent-driven sampling (RDS), 82 studies sampling from time-venue, 39 studies using multiple recruitment methods, and 11 applying convenience sampling methods. The analysed data indicated that the largest proportion (35.6%, 95% CI: 32.3-39.9%) searched for sexual partners on the internet; among the other locations, 25.9% (95% CI: 22.4-27.0%) searched at bars/night clubs/tearooms, 13.2% (95% CI: 12.4-14.5%) searched at public bathhouses/saunas, and 12.4% (95% CI: 10.9-14.3%) searched at parks.
We found a publication bias, the problem that results from systematic differences between the results of all the completed studies on a topic and the results of the subset of those studies that are published, across the studies reporting the HIV prevalence (t=-4.12, p=0.0011) (Fig. 2), which must be taken especially seriously, as it presents perhaps the greatest threat to the validity of this method.
Study quality assessment
In the quality assessment, 214 of the included studies were considered “good” quality (values between 67% and 100%), 141 were considered “satisfactory” (values between 33% and 66%), and none were considered “poor” (values between 0% and 33%).
Demographic information
The demographic characteristics are detailed in Table 1. Most participants (68.7%) were single, although 22.3% had married a woman, 2.5% were cohabiting with their same-sex partner, and 5.0% were divorced or widowed. The sexual behaviour data indicated that 58.9% exclusively had sex with men, 33.8% had sex with both men and women, 2.8% were heterosexual, and 5.7% were undetermined.
Prevalence of HIV infections among MSM
Overall, the national HIV prevalence among MSM from 2001 to 2018 was 5.7% (95% CI: 5.4–6.1%, N=355), with study prevalence rates ranging from 0% (95% CI: 0.1-2.5%) to 22.91% (95% CI: 18.1-28.3%) [10, 334]. Substantial heterogeneity existed between the studies (p for Q test, p<0.0001; I2=98.0%).
Chronological prevalence of HIV infections among MSM
Based on the study years selected for our investigation, we estimated chronological prevalence of HIV in China by meta-regression analysis (I2= 95.9%, P < 0.001), demonstrating an increased tendency as time progressed (Fig. 3).
Relationships between HIV prevalence and age, education, marital status, occupation, ethnicity, and sexual debut partner
Table 2 shows the age-specific prevalence of HIV. For each 10-year increase in age, the prevalence of HIV also increased from 4.6% in people aged <20 years to 7.2% in people aged 20-29 years to the highest prevalence (19.3%) in those aged >50 years.
The prevalence of HIV decreased with increasing years of education, with a prevalence in the illiterate group of 16.8% (95% CI: 6.4-37.3%), which was higher than the prevalence rate among those who had received an education. Of the assessed occupations (including teacher, office staff, farmer, service business employee, unemployed and job-seeking, and worker), the prevalence of HIV was highest among farmers at 14.8% (95% CI: 3.8-43.2%).
The prevalence rates of HIV by sexual orientation were 7.0%, 6.1%, 7.5%, and 7.6% for the homosexual, bisexual, heterosexual, and undetermined groups, respectively. Although the internet was a major venue for seeking male sex partners among Chinese MSM (35.6%, 95% CI: 32.3-39.9%, N=101), MSM seeking male sex partners in bathhouses/saunas had the highest prevalence of HIV (13.4%, 95% CI: 10.3-17.1%, N=22).
The odds ratio (OR) of HIV for those whose first sexual encounter was with a male compared to those with a first sexual encounter with a female was 0.6 (95% CI: 0.5-0.7), suggesting that in China, MSM with a female sexual debut partner had a higher HIV prevalence than those with a male sexual debut partner. Drug use was not a significant contributor to HIV transmission among Chinese MSM (OR: 1.14, 95% CI: 0.31-4.21).
Geographical characteristics of the HIV prevalence
To determine the geographical characteristics of the HIV prevalence in China, we analysed the differences in prevalence by geographical divisions in China based on its provinces or municipalities. The number of studies, total HIV-positive population, and pooled sample size were summarized for the different geographical divisions (Fig. 4).
Overall, the highest pooled HIV prevalence among MSM was found in southwest China (10.7%, 95% CI: 9.3-12.2%, N=91), in which Chongqing city had the highest HIV prevalence (13.8%, 95% CI: 12.8-14.9%, N=24), followed by east China (6.3%, 95% CI: 5.6-7.0%, N=167), which had a prevalence higher than the national average. The pooled HIV prevalence among MSM in northeast China (4.2%, 95% CI: 3.4-5.0%, N=58), north China (5.2%, 95% CI: 4.4-6.1%, N=66), northwest China (3.7%, 95% CI: 3.0-4.5%, N=57), and south China (5.1%, 95% CI: 4.4-5.9%, N=121) were lower than the prevalence of the country as a whole.
Condom use information (in the last 6 months)
Participants who had engaged in unprotected sex in the past 6 months had a higher risk of HIV infection than those who reported protected sex (either sometimes or always using a condom). The ORs for participants who never used a condom during sex with men, during commercial sex with men, and during sex with a woman in the past 6 months were 0.11 (95% CI: 0.08-0.14), 0.11 (95% CI: 0.09-0.13), and 0.07 (95% CI: 0.03-0.14), respectively.