Article Text

This article has a correction. Please see:

Download PDFPDF

High prevalence of risk behaviour concurrent with links to other high-risk populations: a potentially explosive HIV epidemic among men who have sex with men in Guangzhou, China
  1. Q He1,2,
  2. Y Wang1,
  3. P Lin1,
  4. H F Raymond3,
  5. Y Li1,
  6. F Yang1,
  7. J Zhao4,
  8. J Li1,
  9. L Ling2,
  10. W McFarland3
  1. 1
    Center for Disease Prevention and Control of Guangdong Province, Guangzhou, PR China
  2. 2
    School of Public Health, Sun Yatsen University, Guangzhou, PR China
  3. 3
    San Francisco Department of Public Health, and Center for AIDS Prevention Studies, University of California, San Francisco, California, USA
  4. 4
    Center for Disease Prevention and Control of Jiangsu Province, Nanjing, PR China
  1. Dr Q He, Center for Disease Prevention and Control of Guangdong Province, and School of Public Health, Sun Yatsen University, 176 Xin’gang Road West, Guangzhou, 510300, PR China; he-q{at}


Objectives: HIV disproportionately affects men who have sex with men (MSM), but HIV prevalence among MSM in Guangzhou has not shown the rapid increasing trend as it has elsewhere in China. The aim of this study is to detect the epidemic and to determine the characteristics of MSM in Guangzhou susceptible to HIV.

Methods: A cross-sectional survey with serological testing for HIV, syphilis, HBV and HCV through long-chain referral sampling strategy to help control the bias generated from non-statistic sampling.

Results: The most important features of MSM in Guangzhou are being young and mobile, and of comparable education and income level to that of the general population. The HIV prevalence was 1.3% (95% CI 0.3 to 2.7%) in 2006. Many HIV risk factors were identified: low awareness of HIV risk perception and prevention, high prevalence of diverse, multiple partners and versatile sexual role, more than half of them actively having sex with women, low persistent condom use with both male and female partners, commercial sex and one out of 27 practising needle or syringe sharing during illicit drug use.

Conclusion: HIV has been introduced into MSM in Guangzhou. Demographic and behavioural risk factors and overlapping risk populations contribute to a potentially rapidly rising epidemic among MSM and the potential for a bridge to female partners in Guangzhou if timely and effective interventions are not implemented.

View Full Text

Statistics from

In 2007, the total number of HIV/AIDS cases in China was estimated to be 700 000 but with the rate of new cases declining. Exceeding injection drug use as the predominant mode of transmission, sexually acquired HIV infection appears to be a factor in the majority of new infections. Men who have sex with men (MSM) represent 11% of sexually acquired infections, and these infections are centred in large and medium-sized cities and areas with large concentrations of migrants.1

Guangdong Province, one of the provinces earliest and most seriously affected by HIV, is experiencing a rapid increase in HIV infections. Guangzhou, the capital of Guangdong, reported the highest number of HIV/AIDS cases overall in 2006. MSM represented 2.14% of all cases in the province but only 0.6% in Guangzhou. Despite representing a small proportion of overall cases, HIV and syphilis prevalence from sentinel surveillance among MSM in 2006 were 1.33% and 12.6% respectively.2

Guangzhou is an attractive area for MSM in China because of the accepting and diverse culture of the numerous immigrants and the greater chances of employment than in smaller cities and rural areas. MSM in Guangzhou may be an important population affected by HIV because of their potential for high rates of infection and their potential role in bridging HIV infection to the general population.3 One of the most important tasks for HIV/AIDS prevention in Guangzhou is to ensure that the virus does not spread from high-risk groups to the general public.4 Over the last two decades, Chinese society has been increasingly accepting of discussion of sexual orientation. This period has also been hallmarked by the implementation of national opening-up policies and human-rights campaigns. Despite these improvements, stigma and discrimination towards homosexuality and bisexuality in China still remain.58 MSM in China are still a mostly hidden population and avoid disclosing their sexual orientation publically. This hidden nature has hampered HIV/AIDS prevention and intervention projects. However, recent improvements in sampling strategies for hidden populations have allowed a greater understanding of this vulnerable population. Recent studies have been able to overcome the limitations of early pilot studies which utilized only convenience sampling and, theoretically, allow generalisations to the population of interest to be made.3 9

HIV prevalence estimates among MSM across China show great variations from city to city. For example, HIV prevalence varies from 0% in Jiangsu to 10.4% in Chongqing.1015 Furthermore, a dramatic increase in infections among MSM in Hong Kong, a city within 175 km of Guangzhou and a city with which Guangzhou has strong trading and social connections, has been observed recently.16 However, discrepant results of studies of HIV prevalence have been reported elsewhere in China. It is thought that these differing prevalence estimates are mainly due to differences in sampling strategies. In order to clarify our previous pilot studies’ findings on HIV among MSM in Guangzhou, we conducted a survey in 2006 on the prevalence of HIV, syphilis, hepatitis B (HBV) and hepatitis C (HCV), demographics and sexual behaviours among MSM in Guangzhou using a long-chain referral sampling strategy.


Study design

We conducted a cross-sectional survey among MSM from May to August 2006 in Guangzhou, China. Subjects were recruited through a long-chain referral strategy known as respondent-driven sampling (RDS).17 Demographic characteristics and HIV-risk information were collected anonymously via face-to-face interviews. Blood was drawn using standard venipuncture for HIV, HBV, HCV and syphilis testing.

Study subjects and recruitment methods

The target population for this study were men who reported ever having had any sexual experience (anal or oral) with another male and who resided in Guangzhou at the time of the survey. Subjects were recruited through RDS, a modified chain-referral method designed to recruit hard-to-reach populations.18 Adjusting for recruitment patterns and the social network sizes of each participant, RDS can produce population-based estimates for demographic characteristics, risk behaviours and disease prevalence.19 We selected nine seed subjects who were diverse with respect to key demographic variables identified in our earlier pilot studies.3 9 Recruitment was tracked to monitor sampling efficiency. After completing the survey and testing, each seed subject was given three coded coupons and instructions on how to recruit up to three of their peers. Recruitment coupons had detailed information on how to contact the study site and how to participate in the study. Individuals who were given coupons by their peers could drop in, call or email for an appointment to participate in the study. After completing the survey and testing, subsequent participants were also given three coupons for further recruitment. The coupons expired 1 month after the issue date in an effort to stimulate rapid recruitment. We shortened the length of time the coupons were valid as equilibrium was approached in key demographic variables and stopped issuing coupons when we verified equilibrium in residence status (hukou), educational attainment and marital status. Every participant was paid 50 RMB (about $6.5) for his participation and for recruiting peers. In addition, each participant was provided testing for HIV, syphilis, HBV and HCV free of charge.

Experienced interviewers, from the Center for Disease Control and Prevention of Guangdong province (GDCDC) and the Public Health School of Sun Yatsen University, had been trained specifically on the study protocol, including RDS field procedures, coupon management, voluntary counselling, interview skills, building rapport, being sensitive to MSM sexual practices and cultural norms before the study began. These trainings were designed to help interviewers set a non-judgemental tone during interviews and thus to encourage honest responses and to minimise socially desirable answers. After providing informed consent, eligible respondents completed anonymous, structured, face-to-face interviews in a counselling room in one of the clinics at GDCDC to ensure confidentiality and quality. The completed questionnaires were cross-checked by the interviewers for missing or unclear answers before subjects left the study site.

We provided participants with tailored counselling on HIV and sexually transmitted diseases (STD). Blood was drawn for serological testing for HIV, HBV, HCV and syphilis infection. The participants were instructed to return or call in 1 month to obtain their results. Results were given anonymously: subjects were required to provide both the unique code given to them during the interview and their date of birth. All participants provided informed consent for each part of the study. The study protocol was reviewed and approved by the Internal Review Board (IRB) of GDCDC.


The questionnaire was developed and modified from a previous study of MSM conducted in Guangzhou.3 The instrument included measures of demographic characteristics, HIV/AIDS knowledge, perception of risk of HIV infection, self-identified sexual orientation, sexual behaviour with different types of partners and related condom use, and illicit drug use. Subjects were given a “Refuse to Answer” option for all questions. Persons testing positive for any infection were given individually tailored counselling including referrals and information regarding care and treatment.

All testing followed the same protocol as in our previous study,3 except for syphilis antibody testing. Syphilis antibody was screened using ELISA (Double Antigen Sandwich ELISA, Livzon Group Reagent Factory, Zhuhai, PR China) and then confirmed by Syphilis Rapid Plasma Reagin test (RPR, Shanghai Kehua Bioengineering, Shanghai, PR China).


Data were entered into an EpiData (v 3.1) database with logic check programs predesigned according to the questionnaire. Analyses were performed with SAS (v 8.2) and RDSAT (v 5.6.0). All missing data in the database were coded as “−1” for RDSAT analysis. Participants with missing or duplicate recruitment information were recoded as seeds for the purposes of RDS analyses. Key demographic variables were used to examine whether the sample achieved equilibrium, a measure of the robustness of the crude sample. RDSAT also provides a function where one can model how long it takes to achieve equilibrium in a given variable given a particular type of initial seed. We used this function to cross-check that we indeed achieved equilibrium in our key variables of interest.

We present adjusted point estimates, with their corresponding 95% CIs, of demographics, knowledge and risk behaviour for the population of MSM in Guangzhou. Additionally, crude frequencies and point estimates are given. Awareness of HIV/AIDS was defined as correctly knowing the blood and sexual transmission routes of HIV (five true-or-false questions) and giving more than 60% correct answers to eight true-or-false questions about non-transmission events or preventive methods. We measured unprotected anal intercourse (UAI) as any UAI with any type of male partner in the past 6 months. Furthermore, we defined unprotected intercourse (UI) as having any unprotected anal or vaginal intercourse with either female and/or male sex partners during the previous 6 months. Then, we additionally analysed the correlates of HIV and syphilis infection, by selecting a set of variables showing association with the outcome by bivariate analysis as well as a priori knowledge, using unweighted stepwise logistic regression and weighted stepwise logistic regression with weights generated by RDSAT. Weights were generated using HIV and syphilis infections as the variables of interest for their respective models. Both models were constructed using backward stepwise elimination retaining variables with an alpha of 0.1



Of the nine seeds initially selected to participate, six produced recruitment chains of MSM peers. The maximum recruitment waves among all productive seeds ranged from four to 21. Using the estimate waves function in RDSAT, we confirmed that the number of waves required by all key variables to achieve equilibrium was from three to six waves, far fewer than that achieved in practice. In terms of official Guangzhou hukou, never being married and obtaining only a high school education, equilibrium was obtained in approximately 2 months of recruitment (fig 1) or by the 18th wave (fig 2). Of the three recruitment coupons given to each participant, an average of one coupon was returned for an overall coupon return rate of 37.3%.

Figure 1

Equilibrium in main demographic features by week of recruitment among men who have sex with men in Guangzhou, 2006.

Figure 2

Equilibrium in main demographic features by wave of recruitment among men who have sex with men in Guangzhou, 2006.


Almost all MSM found in Guangzhou (94.6%, 400/423) lived in Guangzhou, but only a few of them had Guangzhou hukou, and less than 10% had participated in any one of our three previous biannual surveys or intervention activities between 2003 and 2005. Nearly 60% of MSM were aged 20 to 29. The majority were single, and most have a medium income of 1000 to 2000 yuan (RMB), whereas a few were unemployed. About two-thirds of MSM identified themselves as homosexual (table 1).

Table 1 Demographics of men who have sex with men in Guangzhou, 2006 (N = 423)

Sexual and other risk behaviour

About one in five MSM had their first sexual intercourse under 18 years old, and more than half of them had their first sex with another male. Three-quarters had only male partners whereas about one-fifth had both male and female partners within a recent 6-month period. The most common ways for MSM to find partners were through the internet, followed by contacts made through other sexual partners, cruising in parks and through friends or colleagues. MSM in Guangzhou had diverse partnerships including foreign partners, regular and casual partners and commercial partners of both genders. About half of MSM in Guangzhou had UAI with their male partners in the past 6 months. More than one-fifth were versatile (ie, taking both the receptive and insertive roles) during anal sex. About 7.5% of MSM had commercial UAI, which included 3.1% who bought and 2.8% who sold UAI.

A substantial proportion (14.3%) of MSM in Guangzhou had unprotected vaginal intercourse (UVI) with females. Together with UAI with male partners, unprotected intercourse (UI) with either male or female partners occurred in more than half of the population while UI with both male and female occurred in only 7.3% of the population. Furthermore, about 5.2% of MSM had UI with commercial female partners including 1.9% who sold unprotected sex to females. Finally, only 1.4% had UAI with foreigners (table 2). Notably, a very low proportion of MSM in Guangzhou used illicit drugs (5.4%, 95% CI 2.7 to 8.2%).

Table 2 Sexual partners and corresponding behaviour among men who have sex with men in Guangzhou, 2006 (N = 423)

HIV/AIDS awareness, risk perception and prevalence of HIV, syphilis, HCV and hepatitis B

Only 25.9% of MSM in Guangzhou correctly knew more than 60% of the non-transmission routes of HIV. About half of MSM perceived a possible personal risk of becoming infected with HIV.

The prevalences of HIV, syphilis, HBV and HCV among MSM in Guangzhou were about 1.3%, 14.8%, 10.9% and 2.6%, respectively. About 3.8% of MSM are currently infected with syphilis, and 0.5% are coinfected with both HIV and syphilis. About 16.8% of MSM have a history of symptoms of recent STDs (table 3).

Table 3 HIV/AIDS awareness, risk perception and infection prevalence among men who have sex with men in Guangzhou, 2006 (N = 423)

Associations in bivariate analysis

RDS stratification analysis showed that HIV infection was associated with participating in previous studies, knowing the HIV transmission routes and age of the sexual debut with either male or female partners. Syphilis infection was associated with unprotected sex with casual male partners, role in anal sex and number of female partners (table 4).

Table 4 Factors associated with HIV and syphilis infection by bivariate analysis among men who have sex with men in Guangzhou (α = 0.05), 2006 (N = 423)

Correlates of HIV and syphilis infection

Participating in previous studies, syphilis infection, age of the sexual debut, having foreign male partners and age were selected from demographics, types of sexual partners and corresponding sexual behaviour for multivariable analysis. Age was selected as a potential confounder of HIV infection, as age is a proxy for cumulative exposure. Unweighted logistic regression showed that HIV infection was positively associated with syphilis infection (OR = 13.34, 95% CI 2.53 to 70.40), whereas weighted analysis showed that HIV infection is associated with having foreign partners (OR = 12.177, 95% CI 2.089 to 70.971).

Modelling syphilis infection, we selected age, condom access, education attainment, having UAI with casual male partners, roles during anal sex and having female partners through bivariate analysis. In both the unweighted and weighted models, only UAI with casual male partners was negatively associated with syphilis infection (OR = 0.199, 95% CI 0.056 to 0.702) and (OR = 0.299, 95% CI 0.042 to 1.256) respectively.


First, we have confirmed earlier work that shows a rising HIV epidemic among MSM in Guangzhou but with an overall prevalence lower than that in other areas of China.8 11 13 2023 Second, levels of HBV, HCV and syphilis infection were not significant. Third, the majority of MSM in Guangzhou are immigrants from both inside and outside the province. MSM in Guangzhou were young but not highly educated, and monthly incomes were similar to those in our previous findings,3 9 This suggests that MSM in Guangzhou share many comparable features to the general population except for their sexual orientation. Fourth, risk behaviours were very prevalent. The majority of the respondents had multiple partners and of both genders. Anal intercourse with male partners was practised without the protection of condoms and was significantly associated with HIV infection among MSM,24 as reported in Hong Kong, the neighbouring city and elsewhere in the world.16 25 Fifth, the case for a potential bridge of HIV from MSM to female partners has been strengthened through the reported sexual networks and lower levels of condom use during heterosexual sex.3 Furthermore, we found an overlap between MSM, injecting drug users (IDUs) and those who have unprotected sex with commercial sex workers. This implies that the high prevalence of HIV among risk populations other than MSM in Guangzhou could spread into MSM and accelerate increases in prevalence.

Controlling for potential and known confounders, we found a positive association between HIV infection and having foreign male sexual partners. This suggests that HIV among MSM in Guangzhou may come from both internal and external sources, but with the extremely low prevalence of having foreign partners, this finding is inconclusive. The negative association between syphilis infection and UAI with casual male partners may suggest that those previously infected subsequently reduced more recent UAI with casual male partners because of their heightened perception of risk associated with UAI. This finding warrants further verification.

RDS is effective in reaching MSM in Guangzhou, even considering its relatively new use among MSM in China.26 We found that a robust sample can be obtained in as little as 2 months, while sometimes more waves of recruitment are needed than that of the number theoretically estimated through RDSAT. These additional waves were needed to reach equilibrium in key variables.

There are limitations to this study. First, there is no gold standard to sample hidden populations such as MSM, and while RDS may help to overcome some of the challenges in sampling hard-to-reach populations, it is unclear if RDS truly captures information on the entire population in question. Second, cross-sectional studies such as this are unable to establish causal inference. This limits our ability to discern what factors cause, or are in the causal chain leading to, HIV or syphilis infection. Third, participants were able to decline to respond to items freely. This leaves some of the most sensitive (and thus potentially most important) questions with small response rates, limiting the strength of our conclusions. Thus, these data need to be interpreted cautiously.

Despite these limitations, we feel that the data and conclusions presented here are helpful in furthering our knowledge of HIV risk among MSM in China. Furthermore, the low prevalence of HIV, the high prevalence of risk behaviour and complex social networks involving other risk groups suggest that a small window exists to prevent the rapid spread of HIV among MSM in Guangzhou. Timely prevention efforts will also be needed to prevent potential bridging of HIV from multiple high-risk groups to female partners of MSM. The relationship between risk behaviour and demographic characteristics should be explored in further studies.


We acknowledge the MSM NGOs and the interviewers, who devoted a large amount of time and effort in conducting and completing subjects’ recruitments, and W Wu, who cleaned and entered the data.


View Abstract


  • Funding: The study was partly supported by Guangdong provincial Medical Science Grant (A2006068), CDC-GAP China project and Guangzhou disease control project (2006-Zda-003).

  • Competing interests: None.

  • Ethics approval: Ethics approval was provided by the Internal Review Boards (IRB) of Center for Disease Prevention and Control of Guangdong province (GDCDC) in Guangzhou, China.

  • Patient consent: Obtained.

  • Contributors: QH was the PI for the study and lead author for the paper; QH, YW, PL, YL, JZ, LL and WMF contributed to the design of the study; QH, PL, JZ and WMF supervised the conduct of RDS; YW, FY and JL did field blood collection and laboratory tests; QH, LL and HFR performed all the statistical analyses; HFR reviewed the language; and all authors contributed to the write-up.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • PostScript
    BMJ Publishing Group Ltd