Country setting
GBMSM are disproportionately represented in prevalent cases of HIV in Singapore. As of 2019, a total of 8,295 incident HIV infections among Singapore residents have been notified to the Singapore ministry of health (MOH). In 2011, yearly incident cases of HIV transmitted through male ‘homosexual or bisexual’ modes exceeded that of ‘heterosexual’ modes for the first time, and that trend has persisted since (26). Recent studies have found that a majority of Singapore residents hold negative perceptions of, and attitudes towards lesbian, gay, bisexual, and transgender (LGBT) individuals. Specifically, most perceived same-sex relationships as being wrong, and are also not in favor of the repeal of Section 377A, the law that criminalizes sexual relations between men (27, 28). Past studies have also established the negative impact and trickle-down effects that such stigma has on HIV prevention efforts among GBMSM in Singapore (29-31).
Participants and recruitment
The PCYCS is a prospective cohort study exploring the syndemic risks associated with HIV and other sexually transmitted infections (STI) acquisition among YMSM in Singapore. This study was a partnership between Action for AIDS Singapore (AFA), one of Singapore’s longest-running community-based organizations serving the health of GBMSM, and the National University of Singapore (NUS). To be eligible for this cohort, participants had to be HIV-negative or unsure of their HIV status, between the ages of 18 to 25 years old, Singapore citizens or permanent residents, and identify as gay, bisexual, or queer men at the point of recruitment, which spanned across May to September 2019. Participants were asked to self-report these attributes. Assuming a population of 210,000 GBMSM in Singapore (32), we targeted recruitment of at least 384 participants to achieve a 95% confidence level and 5% margin of error. However, 600 participants were targeted for recruitment to account for potential attrition at each follow-up for this cohort study.
Participants were invited to participate in this study through a recruitment flyer that was disseminated through both online (e.g. social media) and offline (e.g. at the organization’s office or outreach activities) channels by a network of community-based organizations in Singapore who are engaged in health advocacy-related activities for GBMSM. Participants who were interested in participating and were eligible for the study signed up through an enrolment link with their self-reported alias, contact details, date of birth, gender, HIV status, sexual orientation, and their residence status. An AFA staff member subsequently verified the eligibility of participants who had signed up prior to sending them a unique identifier, and a link for the baseline survey.
It was imperative for the team to ensure that participants’ identities their data would remain confidential, as drug use and sexual relations between men are criminalized in Singapore. To do so, the researchers ensured that no staff member from AFA or NUS had full access to either the enrolment details held by AFA which contained aliases and contact details of participants, and the baseline survey results held by NUS. Both sets of data were only linked by the unique identifier which participants entered at the beginning of the survey. Upon completion of the survey, a NUS staff member provided AFA with the unique identifiers who had completed the baseline survey, and an SGD20.00 (approximately USD15.00) cash reimbursement was given to the participant. A total of 570 participants were recruited at the baseline of the cohort; the response rate could not be established as it was not possible to ascertain the total number of eligible participants that the recruitment flyers had reached. Participants could also refer their friends to participate in the survey and be reimbursed SGD5.00 (approximately USD3.75) for each eligible individual successfully referred and who had completed the baseline survey; a total of 171 (30.0%) of participants were recruited through referrals.
Ethics declaration
Ethics approval was obtained from the institutional review board at the National University of Singapore (NUS-IRB Reference Code S-19-007) prior to data collection.
Variable measures
The survey collected sociodemographic information from respondents, including age (in years), ethnicity (Chinese vs non-Chinese), gender (cisgender vs non-cisgender), sexual orientation (gay vs non-gay), and monthly household income (SGD5000 and above vs below SGD5000; SGD5000 is approximately USD3668.94). As the YMSM in our sample included respondents who were still schooling, educational attainment and gross monthly personal income were omitted as variables, though they were collected in the baseline survey. Household income was thus collected and utilized as a proxy variable for socioeconomic status among participants.
Participants were asked if they had ever used a series of substances in sexual contexts, including alcohol, poppers, meth, GHB/GBL as well as other ED medication or drugs (e.g. Viagra, Cialis, ‘black ants’). For sexual health outcomes, participants were asked about their patterns of unprotected anal sex, as well as STI diagnoses in the last six months. Unprotected anal sex with casual partners in the last six months was coded as a binary variable (yes vs no), and was derived from a series of questions that solicited frequency of self-reported condom use through a five-point Likert scale from 1 to 5, with 1 being that they did not use condoms and 5 being that they had always used condoms; this question was repeated for permutations of oral and anal sex with regular, casual, and sex worker partners in the last six months. Participants who had not used condoms all the time with casual and sex worker partners in the last six months were coded as ‘yes’ under this variable. Participants who also reported being tested positive for either gonorrhoea, syphilis, chlamydia, genital herpes, genital warts or hepatitis C were assigned as having been diagnosed with an STI in the past six months through a binary (yes vs no) variable.
For mental health outcomes, both depression severity and past suicide ideation were measured. Depression severity, which was measured through the well-established, nine-item patient health questionnaire-9 (PHQ-9) validated by Kroenke and colleagues (33, 34). Participants were asked “over the last 2 weeks, how often have you been bothered by any of the following problems?” to a total of nine statements, to which they could respond to four possible answers on a Likert scale; 1 being not at all and 4 being nearly every day. Depression severity was measured as an index that was the sum score of all nine items, with a minimum score of 0 and a maximum score of 27. Cronbach’s alpha of the scale was reported as 0.92. Participants were also asked about their suicide-related behaviors, including if they had ever contemplated suicide by responding to three possible answers: yes, no, or prefer not to say.
Statistical analysis
Statistical analysis was carried out using the statistical software STATA version 15 (Stata Corp, College Station, TX, USA). We employed descriptive statistics to identify trends in sample characteristics. Latent class analysis was performed with STATA’s gsem function to delineate classes of drug use in sexualized contexts. The chosen variables included a history of using alcohol, poppers, meth, GHB/GBL as well as other ED medication or drugs (e.g. Viagra, Cialis, ‘black ants’) in sexualized contexts. A latent class model was employed, whereby the conditional item probabilities for each class and class probabilities were estimated through maximum likelihood procedures. A posteriori probabilities were calculated using the predict post-estimation command. In order to determine the number of clusters, models with consecutively increasing were estimated and compared using both Akaike and Bayesian information criteria (AIC and BIC), but BIC was favored given its reliance on both the log-likelihood and the adjusted sample size (35). Given that the three-class model had the lowest AIC and BIC values, it was thus reported in this study. A summary of goodness-of-fit statistics are provided in Supplementary Table S1. Entropy, which provides a measure of how well individuals fit into each class, was also taken into account to determine the fit of the model. Following identification of latent classes, we sought to determine the association between varying classes with outcome variables of unprotected anal sex, STI diagnoses, depression severity and past suicide ideation while adjusting for key sociodemographic covariates. We employed multivariable Poisson regression models with robust sandwich variances to compute the crude prevalence ratio (PR) and adjusted prevalence ratio (aPR) estimating these outcome variables. Statistical significance was set at p<0.05. Analysis for this study was not pre-registered and the results reported here should be considered exploratory.