Male–male sex may now be the predominant mode of HIV transmission in China.1 China is no exception to a pattern of high and recently rising HIV prevalence and incidence among men who have sex with men (MSM) throughout Asia.1–6 For example, the proportion of MSM HIV infected has been reported at 4.2% in Singapore, 13.0% in India, and as high as 30.7% in Thailand.2–5 In China, the estimated number of HIV new infections occurring in 2009 was 48,000, with one-third attributed to MSM, a substantial increase from the projected 12.7% in 2007.1 In urban areas, upward trends in HIV prevalence among MSM have been observed in Beijing, from 0.4 to 5.8% during 2004–2006; in Shenzhen, from 1.7 to 3.8% during 2005–2007; in Nanjing, from 0 to 5.8% during 2003–2007; and in Chongqing, from 10.4 to 12.5% during 2006–2007.6 In the last 2 years, direct HIV incidence measures have been reported at 2.6 per 100 person-years in Beijing, 5.12 in Nanjing, 5.4 in Shenyang, and 7.8 in Chongqing.7–10 Such levels are comparable with those witnessed in the early explosive phase of the HIV epidemic among MSM in San Francisco in the early to mid-1980s.11
Faced with an escalating HIV epidemic, there is a great need for longitudinal studies to measure HIV incidence, to determine causes of HIV acquisition, and to test culturally appropriate prevention interventions through randomized controlled trials. However, stigmatization and discrimination against MSM make it difficult to identify, recruit, establish, and maintain cohorts of this population.12 Currently, only a few short-term cohort studies have been implemented in China.7–9
Typically, cohort participants are recruited by convenience sampling methods, taking eligible subjects who are most visible and willing to participate, although missing the less accessible or more hidden ones. Indeed, the less accessible parts of a population may actually be at higher risk for HIV infection, thereby producing underestimates of HIV incidence.7,9,13,14 Respondent-driven sampling (RDS) was developed to increase inclusion of hidden populations in research through peer referral and has been widely used as a recruitment strategy for cross-sectional surveys including many of MSM around the world.15–17 RDS incorporates several processes to enhance inclusion and representation of the target population.18 First, it enlists initial “seeds” purposely chosen to be well connected and influential within the population to begin recruitment of others. Second, recruitment is limited to a few others, typically 3–5, to force the chains of referral to permeate diverse social networks, including subpopulations who might not be directly accessible to researchers. Third, a system of incentives is used to motivate participation and recruitment of others. A primary incentive is given for eligible recruits to enroll; a secondary incentive is given for each person the individual successfully recruits for the study. Meanwhile, peer referral has the advantages of persons in the hidden population being more able to identify others in the population and can vouch for the legitimacy of the study and researchers. The process fosters long chains of peer referrals, which can grow exponentially and eventually reach “equilibrium” or the point at which the composition of the sample does not change substantially with further waves of recruitment.
We sought to use some features of RDS to create and maintain a longitudinal study of a hidden population. Specifically, we developed several hypotheses on how RDS methods could enhance inclusion and retention of MSM in an HIV incidence cohort in Nanjing, China. First, we believed that long chains of peer referrals would be efficient in accruing eligible subjects to enroll in the cohort. Second, we assumed that the composition of the cohort would diversify from the initial enrollees with subsequent recruitment waves and reach stability. For the establishment of the initial cohort enrollment from the community at large, we call our recruitment method “long chain-peer referral” (LCPR). Third, we further posited that peers who were successful in completing follow-up would be the most able to recruit other MSM likely to complete follow-up, thus enhancing cohort retention. For example, they would be best able to explain the expectations and benefits of follow-up and influence their peers to follow their example. We therefore incorporated a peer referral system whereby each man successfully completing follow-up would be encouraged to recruit other MSM into the longitudinal study. We called this method of recruitment “open cohort peer referral” (OCPR). Thus 2 parallel but linked cohorts are created as follows: one from the community at large using LCPR, the other an open cohort phase using OCPR. The present report presents the LCPR and OPCR methods, recruitment and retention results, and rates of HIV incidence by recruitment method and by calendar time.
Overall Study Design
We constructed 2 parallel longitudinal cohort studies by peer referral to measure HIV incidence over 2 years through follow-up visits at 6-month intervals.
Study Subjects and Initial Cohort Recruitment
Using LCPR, the initial cohort [Cohort1 (C1)] of MSM was established from May to July 2008 in Nanjing, Jiangsu Province, China. Initial seeds were screened and referred by local MSM community-based organizations, sexually transmitted infection (STI) clinics, and by key informants from the gay community. A total of 9 seeds were enlisted who were diverse in terms of age, education, and venues where they found male partners. Each seed was given 3 coupons to recruit up to 3 MSM peers. Each coupon had a validity period (2 weeks), a unique study ID number, information about the testing site location and hours, and a contact phone number for enquires. Those willing to further refer up to 3 of their peers were trained in recruitment procedures and were given recruitment coupons. Eligible participants were 18 years or older and having engaged in anal or oral sex with a man in the past 12 months. As C1 approached the projected 400 participants [an a priori calculation made to have 80% power to detect a significant relative risk of 2.0 where the risk factor was projected to be 50% unprotected anal intercourse (UAI)], recruitment was tapered off by reducing the recruitment coupons from 3 to 2 to 1. Enrollees were given 30 Renmenbi (∼4.5 USD) prepaid phone cards as an incentive for completing the survey and HIV testing. Unlike many RDS surveys, we did not give a secondary incentive for the recruitment of peers. This decision was based on input from key informants from the MSM population in Nanjing who wished to minimize demands to return for secondary incentives or put undue pressure on peers. The clinic at the Jiangsu Provincial Center for Diseases Control and Prevention was the project site.
Cohort Maintenance and Open Recruitment
All HIV-seronegative participants at baseline in C1 were followed up every 6 months. To maximize retention, participants' contact information was collected, including mobile phone numbers and Internet-based chat services (QQ and MSN), and used for multiple reminders for up-coming and missed appointments. Reminders could also be given to the peer acquaintances who recruited them and to the peer acquaintances they recruited. Two dedicated research staff were responsible for ensuring follow-up in the cohort. Reminder phone calls were made 2 to 3 weeks before scheduled visits. When a participant missed an appointment, the staff would call the participant at the beginning of the following week to reschedule. When participants could not be reached by phone, QQ and MSM messages were sent or peer contact was used to try to notify the participants. For participants who moved out of the city, an online questionnaire was implemented and blood test results were obtained by the national network of HIV counseling and testing clinics. A system of progressive incentives was implemented to further enhance follow-up. Men who completed their second visit were given 50 Renmenbi (∼7.5 USD) prepaid phone cards; those who completed the third visit and beyond were given 100 Renmenbi (∼15 USD) prepaid phone cards.
For the recruitment process of the open cohort phase [Cohort 2 (C2)] and to test our hypothesis that participants with successful follow-up would successfully recruit others who complete follow-up, we implemented OCPR. At each 6-month follow-up visit in C1, MSM were given 3–5 referral cards to recruit new eligible MSM within a validity period (1 week). Participants in C2 were also offered the chance to recruit additional participants at their 6-month 12-month, and 18-month follow-ups.
A brief standardized questionnaire was administered face-to-face to collect sociodemographic characteristics including age, marital status, Nanjing official residential status (“Nanjing hukou”), education attainment, monthly income, and risk-related variables including number of male sex partners, UAI with men by regular and casual relationship type, whether they exchanged sex for money, sex and unprotected sex with women, and history of sexually transmitted disease (STD). Behaviors were asked for the preceding 6-month time frame; history of STD was asked in the last year. We also collected information on where they found sex partners, self-described sexual orientation, and basic HIV/AIDS knowledge on modes of transmission and prevention. An educational booklet, 2 condoms, water-based lubricant, and HIV pretest counseling were given after the interview.
A venous blood sample (5 mL) was obtained for HIV rapid test and syphilis screening after pretest counseling. Rapid tests (Acon Biotech, Hangzhou, China) were used for HIV screening with HIV-positive results being confirmed with an additional rapid test (Serodia HIV, Fujirebio, Japan) and a Western Blot test (HIVBLOT 2.2, Genelabs, Singapore) according to standard laboratory procedures. Positive syphilis results (RPR, Beijing WanTai Biological Pharmacy Enterprise, Beijing, China) were confirmed by Treponema Pallidum Particle Agglutination assay (Livzon Group, Zhuhai, China). HIV-positive participants were referred to the HIV counseling center immediately for care and treatment counseling. Syphilis results and treatment referral services were provided within 2 weeks of the blood draw as results were available.
Written informed consent, questionnaires, educational package, laboratory tests, and counseling were implemented in the same manner as the baseline survey at each semiannual follow-up.
Data Collection and Statistical Analysis
Questionnaire-based data and laboratory testing results were recorded, double entered, and compared with Epi Data software (The Epi Data Association, Odense, Denmark). Respondent-driven sampling analysis tool software was used to determine if equilibrium on the LCPR sample was achieved. SPSS 16.0 (SPSS Inc) was used for statistical analysis. Demographic and risk behaviors were examined as proportions for each recruitment wave for the LCPR (ie, C1) portion of the data and cumulatively proportions for the OCPR (ie, C2) portion of the data. Retention rates were calculated as cumulative proportions remaining from those enrolled at each 6-month follow-up. To assess whether men lost to follow-up were likely to be at higher or lower risk compared with those retained in the cohort, we used logistic regression analysis with 5 variables selected based on the literature and our previous studies of MSM in Nanjing,8,9,19,20 including HIV knowledge, multiple male partners, any UAI with male partners, UAI with casual male partners, and history of STD. HIV incidence was calculated as the number of HIV seroconversions observed, divided by the person-years at risk during the period.21 For the present report, we assigned 6 months for each full follow-up visit achieved by those remaining HIV negative and 3 months for those found to have seroconverted to HIV positive between visits. Confidence intervals were created assuming a Poisson distribution. HIV incidence was examined within each group according to the method of recruitment and by calendar time from the initiation of the cohort.
The study protocol was reviewed and approved by the Ethics Committee of the Chinese Centre for Disease Control and Prevention. Written informed consent was obtained from all eligible participants.
The structure of the cohorts recruited through LCPR and 4 rounds of OCPR are shown in Figure 1. Of the 9 initial seeds selected to launch LCPR, 8 successfully recruited peers and began the long chains of peer referrals to the initial C1 phase of the cohort. By the eighth to ninth waves, equilibria were achieved on all 4 tracking variables (ie, education level, Nanjing hukou, marital status, and syphilis antibody prevalence). Equilibria for 6 variables in the LCPR recruitment are also shown as the line graphs in Figure 2. LCPR accrued 417 MSM, of whom, 397 were HIV negative, at which point the LCPR mode of recruitment was stopped. OCPR for the open cohort C2 phase was initiated and repeated at every subsequent follow-up visit for the continuing subjects. Recruitment through OCPR enrolled a total of 477 additional MSM. Of these, 460 were HIV negative at enrollment. Overall, the largest chain included 87 MSM in 14 waves; no individual recruited more than 3 other participants although more were possible in our design.
Figure 2 describes the accrual of cohort enrollment by 6 characteristics for each method by waves for C1 and by 6-month rounds for C2. The lines show the proportion of subjects by characteristics cumulatively by wave of recruitment (up to 14 and then aggregated thereafter) which comprises the LCPR portion of the cohort. Bars show the same variable by round of OCPR recruitment. For demographic characteristics, the figure shows a decrease in Nanjing hukou, a decrease in the proportion of men with low income (<1000 Renmenbi per month), whereas single marital status remained stable throughout the waves of LCPR recruitment. For the OCPR recruitment, Nanjing hukou was lower at each OCPR round and single marital status was higher. Two markers of HIV risk, UAI and meeting partners over the internet, showed some increase over LCPR recruitment waves, although syphilis seropositivity decreased. OCPR rounds showed lower UAI and syphilis seropositivity, whereas meeting partners on the Internet was higher than the LCPR waves.
Additional variables corresponding to the cumulative recruitment by LCPR and each round of recruitment by OCPR are shown in Table 1. OCPR tended to recruit younger MSM (younger than 30 years), more educated, and self-reported homosexual orientation (corroborated by less sex with women) compared with LCPR. OCPR-recruited MSM reported less history of STD, which tended to be corroborated by lower syphilis seropositivity. OCPR-recruited men also reported less commercial sex (buying or selling) than LCPR-recruited men. Few MSM reported drug use with either method of recruitment. Thus, the overall pattern of risk behavior was not strikingly or consistently different by LCPR versus OCPR method of recruitment.
Retention rates illustrated in Figure 1 were calculated as proportions in Table 2. For both, the LCPR portion and for each recruitment round of the OCPR portion, the largest lost to follow-up occurred within the first follow-up period. Of 397 MSM enrolled through LCPR, 286 (72.0%) were seen at their first follow-up visit at 6 months. The first 3 rounds of OCPR recruitment each retained 69.2%, 51.4%, and 59.1%, respectively, of their enrollees to their first follow-up visit. By time on study, OCPR subject retention tended to fare worse than the initial LCPR subject retention. By 2 years, 178 (44.8%) LCPR participants were retained, still higher than OCPR retention for shorter total follow-up periods (eg, only 36.2% of subjects were retained for 1.5 years after enrollment from the first round of OCPR; 43.0% were retained to 1 year from the second round).
Overall, there was a pattern of higher risk among MSM lost to follow-up compared with those retained. On 4 of 5 factors associated with HIV infection at enrollment, MSM lost to follow-up exhibited significantly higher risk compared with those retained from baseline to the first 6-month follow-up, including low baseline HIV knowledge [odds ratio (OR): 10.3, 95% confidence interval (CI): 2.2 to 47.8], having more male sex partners (OR: 1.3 per partner, 95% CI: 1.1 to 1.6), having UAI with casual male sex partners in the last 6 months (OR: 4.6, 95% CI: 3.0 to 7.1) and history of STD in the year before baseline (OR: 2.8, 95% CI: 2.0 to 3.9). Due to small numbers, subsequent follow-up periods did not show statistically significant differences among defaulters compared with MSM retained, with the exceptions of having UAI with casual male sex partners (OR: 1.7, 95% CI: 1.4 to 2.2) for those lost between 6 and 12 months and more male sex partners (OR: 2.0, 95% CI: 1.7 to 2.4) for those lost between 18 and at 24 months. Nevertheless, the direction of association was consistent for higher risk among defaulters compared with men retained for all follow-up periods for all variables for both recruitment methods.
Table 2 also provides HIV incidence rates by recruitment method, round, and follow-up period, tracking seroconversions rates forward from participants' recruitment group. For the study as a whole, combining the LCPR and OCPR recruitment, HIV incidence was 3.36 per 100 person-years (95% CI: 2.13 to 5.04). By recruitment method, the overall HIV incidence rates were comparable. For the LCPR recruitment portion (C1), incidence was 3.45 per 100 person-years (95% CI: 1.97 to 5.61); for the OCPR method overall (C2) incidence was 3.16 per 100 person-years (95% CI: 1.27 to 6.52). Thus, there was broadly no evidence of higher or lower risk of HIV acquisition by recruitment method. Given the distribution of the relatively rare events of HIV acquisition, no follow-up period showed a significantly higher or lower rate of seroconversion with all confidence intervals having substantial overlap.
Figure 3 presents HIV incidence by calendar period for all subjects present in the study for the 6-month interval. Thus, the trend line shows HIV incidence rates over the 2-year period. Despite the apparently higher incidence in the first 6-month interval, the overall trend in incidence was stable statistically (P > 0.05) and for the remaining 1.5 years visually flat.
We tested the ability of a novel method to establish a cohort of MSM at risk for HIV infection in Nanjing based on LCPR in an effort to capitalize on several advantages purported by RDS. Namely, that the method can reach social networks and parts of the population not accessible to researchers or nongovernmental organization, peers can vouch for the legitimacy of the researchers, MSM are best able to identify and recruit other eligible MSM, and ultimately that the sample accrued would be more inclusive of the diversity of the population. In this first attempt, we believe that we met with modest success in establishing a cohort. LCPR successfully recruited the targeted HIV-negative MSM within a few months. Progressive waves of LCPR did not substantially change the risk profile of MSM recruited in the long term; rather, there was a picture of general stability. Some indicators of risk tended to increase from the initial seeds and early waves (eg, any UAI, meeting partners on the Internet); others decreased (eg, syphilis seropositivity); yet many remained generally stable. The decrease in the proportion of MSM reactive to syphilis antibody over recruitment waves bears notice. The decline may have demonstrated the initial seeds and early waves from their social networks being associated with use of STI services. Also, the proportion of MSM with Nanjing hukou decreased over recruitment wave. We interpret this to signify that LCPR recruitment did succeed in moving away from MSM with greater access to services (public services are in principle only free to official residents). Thus, LCPR recruitment may have served to diversify the cohort beyond MSM accessible through existing services.
We further expanded on the LCPR method to sustain recruitment over time by encouraging those who successfully completed follow-up to recruit additional MSM in our OCPR method. Our assumptions were that MSM committed to the cohort would select and recruit similarly committed MSM and influence their peers to remain in the study. Our data can at best claim only partial support for these assumptions. OCPR did recruit an additional 460 eligible MSM over the subsequent year and a half. The composition was largely comparable with the aggregate LCPR portion of the cohort, but we did not observe better retention. On the contrary, OCPR-recruited MSM tended to drop out at a higher rate than LCPR-recruited subjects. Even more disappointing, neither method was able to retain a majority of the eligible MSM enrolled past the first follow-up visit.
We believe that our data produced a plausible but low estimate of HIV incidence for the population of MSM in Nanjing. The overall rate of 3.36 per 100 person-years is not good news. The level is very likely to increase HIV prevalence among MSM for some years to come unless effective prevention programs are increased. The figure falls above a recent synthesized estimate for MSM in the United States of 2.39 per 100 person-years in the last decade and typical figures for Europe and Australia,22 but below other recent estimates of MSM elsewhere in China and Asia.6,9,10,23 However, we believe the true incidence of HIV in Nanjing to be higher than observed in our cohort given that the loss to follow-up differentially occurred among the MSM reporting higher baseline risk. A plausible explanation is those more likely to be retained in the cohort were at lower risk or more prevention minded before recruitment or reduced their risk behaviors through repeated contact with prevention messages and HIV counseling and testing. This result is consistent with other cohorts conducted in the United States, the Netherlands, and Argentina.13,19,20 For these several potential biases, we suspect our estimate to be under the true level of our city.
When HIV incidence was calculated by calendar time, the trend was stable after the first 6-month interval (ie, the point at which OCPR recruitment or C2 began). We hypothesize that the open cohort process achieved a dynamic stability; that is, new subjects followed for at least 6 months replaced subjects of similar risk profile lost to follow-up after 6 months. Such an effect has been proposed to typically occur within open cohorts.21 The effect may also be analogous to the equilibrium achieved by RDS or LCPR; that is, a point is reach after which subsequent chains of peer referral do not substantially alter the composition of the sample. If true, our OCPR process may be an effective means of sustaining a cohort over long periods with stable composition and enabling the tracking of HIV incidence trends in the population. The differential loss to follow-up may still result in a bias in the absolute measure HIV incidence, but that level may track with real changes in incidence in the population over time. However, we do not wish to over interpret our estimate of HIV incidence given the high level of lost to follow-up overall and acknowledge that biases may result in underestimation or overestimation of HIV incidence in unpredictable ways.
We recognize other limitations of our data and methods. First, we wish to distinguish our study from conventional RDS used for conducting cross-sectional surveys. Researchers familiar with RDS will note that our analysis does not take into account statistical adjustments for recruitment patterns (ie, network size, homophily) to strengthen external validity. We point to the primary aims of the study being to capitalize on logistical strengths of RDS in efficiently recruiting and establishing a diverse cohort and subsequently focus on the internal validity of the longitudinal study. The present study is not a cross-sectional survey for which conventional RDS adjustments apply. As one example, because only HIV-negative subjects are enrolled in the longitudinal study, the recruiter-recruit linkage used for adjustment is disrupted with the exclusion of HIV-positive recruits. We recognize that the lack of RDS adjustments, such as those done in the respondent-driven sampling analysis tool (www.respondentdrivensampling.com), weakens inference that the sample enrolled is representative of the MSM population of Nanjing and new methods to adjust are needed. Although having a population-representative cohort is desirable, it is seldom achieved, especially in hidden and hard to reach populations. Cohorts recruited by other means (eg, by advertisement, self-referral, outreach, or clinic referral) seldom take into account how subjects are recruited nor make claims to be representative of all the target population. Moreover, the methods to statistically adjust longitudinal data are likely to be complex and as yet uncertain. We note that although our methods allowed for more recruitment than the 3 per subject typically used for RDS, in practice we had no more than 3. Large numbers of recruits per participant may cause the study population to resemble a “snowball” or convenience sample. Therefore, caution must be exercised in determining the number of recruits per participant. Just as RDS is held to be a step in the right direction to enhance inclusion and representation, with caveats, we believe LCPR and OCPR have potential to improve the rigor of longitudinal studies of hidden and hard to reach populations.
We identify potential ways that LCPR and OCPR can be improved, but first recognize that these novel methods in themselves provide no guarantee against loss to follow-up. Conventional means for ensuring retention need to be in place and actively pursued, including multiple contact information, multiple mechanisms to contact, multiple attempts to contact, and multiple types of incentives. Newer technologies such as mobile phones, text messaging, instant messaging as QQ, MSN, email, and diverse social network sites on the Internet are innovations that can be increasingly used for longitudinal studies. Additional medical care incentives could be added in our context, including free treatment for multiple STI, vaccinations such as for hepatitis B and A, and a more comprehensive men's health package. Specific to LCPR and OCPR, we point to ample room for incentives for retention that capitalize on social networks. As mentioned above, we did not provide secondary incentives (eg, phone cards for successful recruitment of peers) as is typical of RDS surveys worldwide although substantial numbers of studies have been done successfully without them and in some settings they are not permitted.24,25 Although not giving secondary incentives was a recommendation of our key informants in our preliminary phase, in hindsight, we believe they would have enhanced cohort retention. We recommend future cohorts using LCPR and OCPR to include progressive secondary incentives for peers who successfully bring their acquaintances back to the cohort when they miss their 6-month appointment. The incentives could be applied both to the person who initially recruited them and to the persons they recruited. We project that secondary incentives will enhance follow-up, reducing the bias caused by loss to follow-up, which in the present study would likely have increased the measure of HIV incidence.
Ultimately, however, success in ensuring the participation and retention of marginalized populations in research may not depend on innovative recruitment strategies, new technologies, health or material benefits, or other individual incentives so much as basic trust and respect. Those factors may best be bolstered by the uses of the research data from the population for improving the social, mental, and physical welfare of the community. In our setting in China, we are only just beginning the proper programatic responses to address the HIV epidemic among MSM. Our data suggest that the epidemic is still progressing rapidly. We hope that our effort is one step in strengthening methodological approaches that can be applied to prevention intervention trials and ultimately scaling up those programs with evidence of efficacy.
The authors thank the working group members from HIV/AIDS/STD prevention and control institute in Jiangsu provincial Centers for Disease Control and Prevention and Nanjing municipal Centers for Disease Control and Prevention, and nongovernmental organization members from Jiangsu Tongtian, for program implementation and assistance. The authors also would like to thank faculty and staff of University of California, San Francisco for peer reviews and technical support. Without them, this article would not have happened.
2. Toh P, Lim S, Chio M. Men having sex with men HIV sero-prevalence study in Singapore, a community based project to protect the MSM community. Presented at: XVII International AIDS Conference, 2008, Mexico City, Mexico [THPE0354].
3. Ko NY, Lee HC, Chang JL, et al.. Prevalence of human immunodeficiency virus and sexually transmitted infections and risky sexual behaviors among men visiting gay bathhouses in Taiwan. Sex Transm Dis. 2006;33:467–473.
4. UNAIDS/WHO. AIDS Epidemic Update: December 2007. Geneva, Switzerland; UNAIDS: 2007. UNAIDS/07.27E. ISBN 978 92 9 173 621 8.-INDIA.
5. Van Griensven F, Varangrat A, Wimonsate W, et al.. Trends in HIV prevalence, estimated HIV incidence and risk behavior among men who have sex with men in Bangkok, Thailand, 2003 to 2007. J Acquir Immune Defic Syndr. 2010;5:234–239.
6. van Griensven F, de Lind van Wijngaarden JW. A review of the epidemiology of HIV infection and prevention responses among MSM in Asia. AIDS. 2010;24(suppl 3):S30–S40.
7. Ruan Y, Jia Y, Zhang X, et al.. Incidence of HIV-1, syphilis, hepatitis B, and hepatitis C virus infections and predictors associated with retention in a 12-month follow-up study among men who have sex with men in Beijing, China. J Acquir Immune Defic Syndr. 2009;52:604–610.
8. Yang H, Hao C, Huan X, et al.. HIV incidence and associated factors in a cohort of men who have sex with men in Nanjing, China. Sex Transm Dis. 2010;37:208–213.
9. Xu JJ, Zhang M, Brown K, et al.. Syphilis and HIV seroconversion among a 12-month prospective cohort of men who have sex with men in Shenyang, China. Sex Transm Dis. 2010;37:432–439.
10. Han M, Feng LG, Jiang Y, et al.. Surveillance on HIV-1 incidence among men who have sex with men in Chongqing, China, 2006–2008. Zhonghua Liu Xing Bing Xue Za Zhi. 2009;30:878–881.
11. Katz MH. AIDS epidemic in San Francisco among men who report sex with men: successes and challenges of HIV prevention. J Acquir Immune Defic Syndr. 1997;14(suppl 2):S38–S46.
12. Bridge J, Lazarus JV, Atun R. HIV epidemics and prevention responses in Asia and Eastern Europe: lessons to be learned? AIDS. 2010;24(suppl 3):S86–S94.
13. Dukers NH, Fennema HS, van der Snoek EM, et al.. HIV incidence and HIV testing behavior in men who have sex with men: using three incidence sources, The Netherlands, 1984-2005. AIDS. 2007;21:491–499.
14. He Q, Wang Y, Li Y, et al.. Accessing men who have sex with men through long-chain referral recruitment, Guangzhou, China. AIDS Behav. 2008;12:93–96.
15. Ma X, Zhang Q, He X, et al.. Trends in prevalence of HIV, syphilis, hepatitis C, hepatitis B, and sexual risk behavior among men who have sex with men. Results of 3 consecutive respondent-driven sampling surveys in Beijing, 2004 through 2006. J Acquir Immune Defic Syndr. 2007;45:581–587.
16. Ramirez VJ, Heckathorn DD, Vázquez R, et al.. From networks to populations: the development and application of respondent-driven sampling among IDUs and Latino gay men. AIDS Behav. 2005;9:387–402.
17. Simic M, Johnston LG, Platt L, et al.. Exploring barriers to 'respondent driven sampling' in sex worker and drug-injecting sex worker populations in Eastern Europe. J Urban Health. 2006;83(suppl):S6–S15.
18. Heckatuorn DD. Respondent-driven sampling: a new approach to the study of hidden population. Soc Probl. 1997;44:174–199.
19. Buchbinder SP, Douglas JM Jr, McKirnan DJ, et al.. Feasibility of Human Immunodeficiency Virus vaccine trials in homosexual men in the United States: risk behavior, seroincidence, and willingness to participate. J Infect Dis. 1996;174:954–961.
20. Sosa Estani S, Marone R, Bautista CT, et al.. Buenos Aires cohort of men who have sex with men:prevalence, incidence, risk factors, and molecular genotyping of HIV type 1. AIDS Res Hum Retroviruses. 2007;23:1322–1329.
21. Heyward WL, Osmanov S, Saba J, et al.. Preparation for phase III HIV vaccine efficacy trials: methods for the determination of HIV incidence. AIDS. 1994;8:1285–1291.
22. Stall R, Duran L, Wisniewski SR, et al.. Running in place: implications of HIV incidence estimates among urban men who have sex with men in the United States and other industrialized countries. AIDS Behav. 2009;13:615–629.
23. van Griensven F, Varangrat A, Wimonsate W, et al.. Trends in HIV prevalence, estimated hiv incidence, and risk behavior among men who have sex with men in Bangkok, Thailand, 2003-2007. J Acquir Immune Defic Syndr. 2009;53:234–239.
24. Johnston LG, Malekinejad M, Kendall C, et al.. Implementation challenges to using respondent-driven sampling methodology for HIV biological and behavioral surveillance: field experiences in international settings. AIDS Behav. 2008;12(4 suppl):S131–S141. E-pub June 6, 2008.
25. Malekinejiad M, Johnston LG, Kendall C, et al.. Using respondent-driven sampling methodology for HIV biological and behavioral surveillance in international settings: a systematic review. AIDS Behav. 2008;12:S105–S130.
Keywords:© 2012 Lippincott Williams & Wilkins, Inc.
China; HIV incidence; longitudinal study; men who have sex with men