HIV and hepatitis B, C co-infection and correlates of HIV infection among men who have sex with men in Rwanda, 2021: a respondent-driven sampling, cross-sectional study

Data came from a cross-sectional, biobehavioral survey (BBS) of MSM in Rwanda. Respondent-driven sampling (RDS) was used as a sampling approach to recruit MSM between November and December 2021 [29]. Seeds were identified through community leaders and associations across the country and were screened by study staff to ensure they had attributes necessary to facilitate productive recruitment. Selection criteria included large social network sizes, respected members of their communities, and motivation to contribute to the achievement of the study objectives by encouraging their peers to participate.

Each seed was given three RDS coupons to recruit peers from their networks who were members of the target population and who they knew and who knew them. Recruitment continued by providing each recruit with three RDS coupons until the target sample size was reached. Eligibility criteria for participation included being a male at birth, at least 18 years of age, residing in Rwanda for at least 12 months, having reported having had anal or oral sex with a man in the last 12 months, and consenting to participate in the study.

There were eight study sites distributed according to the administrative provinces, and three seeds were selected at each study site, making 24 seeds in total. Figure 1 shows the recruitment chains generated by each of the 24 RDS seeds. Compensation of 3,000 Rwandan francs (RWF) corresponding to 3·6 United States dollars (USD) with 2021 as the base year to cover transport expenses and an additional 2,000 RWF (2·4 USD) to compensate for time spent was provided. Secondary compensation of 3,000 RWF (3·6 USD) for each eligible recruit was offered, and participants returned to the study office approximately two weeks after their study visit to collect their recruitment reimbursement.

The survey was interviewer-administered, and data were collected on Android tablets using a standardized, structured questionnaire. The questionnaire collected data on demographics, sexual risk behaviors, HIV and STI knowledge, attitudes and beliefs, and other health-related information. Following pretest counseling, HIV tests were performed on all consenting participants by following the national-approved HIV testing algorithm of two tests: Alere Combo as the first screening test and Stat Pak as the second screening test (Fig. 2). Using the Rapid Test for Recent Infection (RTRI) assay and viral load testing, all participants diagnosed as HIV-positive were classified as likely having acquired an HIV infection within the last 12 months (i.e., recent infection) or having acquired an HIV infection more than a year ago (i.e., long-term infection). All HIV-positive participants were referred to health facilities for treatment after posttest counseling. Rapid hepatitis B (Fig. 3) and C (Fig. 4) tests were conducted on all participants to detect hepatitis B surface antigen (HBsAg) and anti-HCV using rapid diagnostic tests (RDTs), respectively.

The individual-level survey data were extracted directly from the Open Data Kit (ODK) and imported into RDS-Analyst software to be cleaned and recoded prior to any analysis. Since crossover recruitment between study sites was not observed, site-level data were analyzed separately, which means that there were eight RDS surveys where weight calculations considered the site-corresponding MSM population size estimates (PSE) [30].

Study sites were distributed by province: North with MSM PSE of 2,375, South with PSE of 2,109, and East with PSE of 2,287 had each one study site, whereas the City of Kigali with PSE of 7,842 and West with PSE of 2,469 had each 3 and 2 study sites, respectively, based on the MSM PSE 2023 published results [30]. To obtain weights for each study site, the provincial-level PSEs for the city of Kigali and Western Province were split to obtain site-level PSEs to be used for site-level weight computation.

As a result, the PSEs for the three study sites from the city of Kigali were assigned as follows: Gatenga HC with a PSE of 2,300; Remera HC with a PSE of 2,300; and We-ACT with a PSE of 3,200. Site-level MSM PSEs for the two sites from the West Province were assigned to Gisenyi HC with a PSE of 1,200 and Gihundwe HC with a PSE of 1,300. All site-level PSEs within the city of Kigali and West Province were assigned, assuming a distribution of site-level MSM PSEs to follow urban/rural KP dynamics (higher in urban areas and less in rural areas) using urban/rural ratios from the 2012 population census.

Finally, to obtain the nationwide pooled estimates, all study sites were aggregated using the “Aggregate Estimate’ option available within RDS-Analyst using site-level PSEs for all 8 study sites across the country [30]. Analyses were conducted using RDS-Analyst, software developed for the analysis of complex RDS data [31]. The data were later exported into a standard statistical package (STATA version 17 (College Station, TX)) with individual weights for further analyses [32]. Analyses provide adjusted population point estimates and a 95% Confidence Intervals (95%CI) of key indicators of interest.

Data were summarized using descriptive statistics as either proportions or medians and associated interquartile ranges (IQRs) for continuous variables. All statistics were weighted according to RDS-generated weighting techniques using RDS-analyst software. The primary outcome variable was summarized as the proportion of the number of participants with a positive HIV diagnosis divided by the total number of participants tested for HIV. A review of the literature identified factors associated with HIV infection among MSM in the region and the strength of the association between each of these variables and the primary outcome variable at both the bivariable and multivariable levels (Fig. 5). We constructed a bivariate logistic regression model for each individual factor to observe individual factor associations with the outcome of interest. For all significant factors at the bivariate logistic regression level, they were considered one by one for the multivariable logistic regression model. An explanatory, multivariable logistic regression model was constructed using forward stepwise selection of variables that minimized the Akaike information criterion (AIC) and model testing using the likelihood ratio test (Fig. 5). Effect estimates are presented as odds ratios (ORs) and adjusted odds ratios (AORs) along with associated 95% confidence intervals (95% CIs). All statistical tests were two-sided with alpha set at 0.05.

The survey received ethical approval from the local Institutional Review Board and the Rwanda National Ethics Committee (RNEC). It was also reviewed in accordance with the US. Centers for Disease Control and Prevention (CDC) human research protection procedures and was determined to be research with CDC nonengaged. Since same-sex relationships remain stigmatized in Rwanda, we anticipated possible risks of physical and/or verbal violence in cases where a study participant was identified in the community. Therefore, the study investigator ensured that the study was conducted anonymously to protect the identity of participants and ensure confidentiality of the data collected; no names or any other personally identifiable information were recorded anywhere.

Completed questionnaires (identifying individuals by only identification numbers) were kept with the study coordinator during the fieldwork. All forms containing any study information were kept in a locked cabinet accessible to only authorized study personnel. Additionally, the electronic data set identified by the unique code was password-protected and accessed by authorized personnel using a computer backed up to a server located at the Rwanda Biomedical Center. Furthermore, for all study members participating in human subjects’ research and data collection, it was a requirement for the study team to have participated in training on human subjects’ research, confidentiality, and interviewing techniques before commencing study activities. A confidentiality agreement was signed by all study investigators, coinvestigators, and data collectors.

The funders of the study had no role in the study design, in the collection, analysis, and interpretation of data, in the writing of the report, or in the decision to submit the paper for publication.

In total, 3,094 coupons were distributed. Of these coupons, 2,304 were returned by potential MSMs to the study site and were valid, corresponding to a coupon return rate of 74·5%. Of the 2,304 potential participants, 66 did not have valid coupons or were ineligible to participate. Therefore, the total number eligible was 2,238, with 2,211 (98·8%) consenting to providing interviews and blood samples for all survey biomarkers.