Objectives Syphilis and HIV epidemics overlap, yet little is known about combined network and behavioural factors that drive syphilis-HIV coinfection. Our study objective was to assess network contexts and sexual behaviours associated with syphilis-HIV co-infection and monoinfection among a particularly vulnerable subgroup: young Black men who have sex with men (YBMSM). To achieve this objective, we examined factors associated with coinfection by each subgroup as classified by syphilis-HIV infection status: (A) HIV monoinfected, (B) syphilis monoinfected and (C) neither syphilis infected nor HIV infected. In addition, we further identified the factors that are associated with HIV infection or syphilis monoinfection.
Methods Data were collected from a sample of 365 YBMSM, aged 16–29 years, recruited through respondent-driven sampling between 2014 and 2016, in two cities with large HIV epidemics: Houston, TX, and Chicago, IL. We conducted a series of multinomial logistic regression models to predict coinfection, HIV monoinfection and syphilis monoinfection as a function of network and sexual behavioural factors.
Results Coinfection was associated with having network members who are coinfected or HIV infected within one’s social network. Syphilis monoinfection was associated with a higher number of social venues attended, and HIV monoinfection was associated with having more condomless top partners.
Conclusion Public health interventions that address the diagnosis and treatment of syphilis infection and ensure that those with syphilis are being tested for HIV may be promising in limiting the synergy of syphilis-HIV infections in onward transmission. Advancing HIV and syphilis prevention efforts in high-prevalence networks may allow prioritisation of limited resources.
- social network analysis
- sexual risk behavior
- multinomial logistic regression
- young Black men who have sex with men (MSM)
Statistics from Altmetric.com
- social network analysis
- sexual risk behavior
- multinomial logistic regression
- young Black men who have sex with men (MSM)
In the USA, primary and secondary (P&S) syphilis rates have doubled between 2005 and 2013. Between 2009 and 2013, 91.1% of new infections occurred among men, and the highest percentage increase was among men who have sex with men (MSM) aged 25–29.1 The increase in syphilis rates among Black MSM was eight times greater than those among White MSM during 2005 and 2008.2 As of 2013, the rate of increase among Black MSM has slowed; however, Black MSM still bear a disproportionate infection rate, with 27.9 cases per 100 000 Black MSM as compared with 5.4 cases per 100 000 White MSM.1 Among Black MSM, syphilis rates were highest among the younger age groups of 20–29, whereas among White MSM, the older age groups of 35–44 had the highest rates.2
The P&S stages of syphilis are the most infectious stages of the disease and increase the risk for both HIV acquisition3 4 and onward HIV transmission.5 6 The syphilis epidemic among MSM also has been marked by a high rate of HIV coinfection.7 8 In particular, young Black MSM have high rates of both HIV and P&S syphilis infections.9 Some reports suggest an antecedent increase in P&S syphilis diagnoses prior to being diagnosed with HIV.10
The syphilis and HIV epidemics overlap, in part because syphilis facilitates HIV infection by disrupting normally protective mucosal and epithelial barriers.4 5 11 12 Commonly described risk factors for coinfection are behavioural and epidemiological networks. Behaviourally, syphilis infection provides an objective indication of condomless sex for both HIV-infected and HIV-negative MSM.13 14 Syphilis transmission among MSM can be facilitated by seroadaptive or serosorting behaviours (choosing sex partners based on HIV status),9 15 16 the practice of oral sex to avoid anal intercourse,15 17 engagement in condomless anal intercourse18 and having multiple sex partners.19
Infections with syphilis and HIV occur most frequently in network contexts in which HIV-uninfected MSM belong to sexual networks with a high prevalence of HIV and syphilis.4 Such sexual networks are formed through their affiliation with ‘risk environments’ or venues20 21 where HIV may be coprevalent with syphilis infection.15 Venue affiliation can serve as a proxy for membership in a network of densely connected, high-risk sexual networks among MSM.22 23 Previous members who have acquired HIV may re-enter the same sexual networks, and this increases the likelihood of contracting episodic STIs.24 As repeated P&S syphilis infection is associated with elevated HIV viral load,19 HIV transmission may be potentiated within sexual networks of young Black men who have sex with men (YBMSM).24
The social venues that YBMSM may frequent provide clues to sexual network formation and downstream syphilis infections. In general, attendance at social venues is a proxy for the formation of sexual networks.22 In contrast, attendance at health-promoting venues (eg, a community-based organisation) may serve as a protective factor within the networks of YBMSM25 and, thus, may impede HIV transmission. Many venues now promote biomedical prevention that includes HIV pre-exposure prophylaxis (PrEP). Some individuals may choose to use PrEP without also using condoms, thereby preventing HIV infection but not preventing syphilis infection. Clear delineation of YBMSM networks provides contextual clarity to HIV-syphilis transmission and will provide important and as yet undescribed data needed to determine how these epidemics can be addressed.
There were two objectives in our study. Our first objective was to determine risk or protective factors in network contexts associated with syphilis-HIV coinfection among YBMSM in comparison to each subgroup as classified by syphilis-HIV infection status: (A) HIV monoinfected, (B) syphilis monoinfected and (C) neither syphilis nor HIV infected. From a clinical standpoint, the identification of factors that predict coinfection is important, given the synergy of the two infections in onward transmission. Our second objective was to further identify factors that are associated with HIV monoinfection or syphilis monoinfection. This analysis will help us further our understanding of factors that seemingly are related to one group or another and, if distinct groups are identified, could suggest distinct sexual or social networks. The current analysis allows us to identify potential risk factors that could be leveraged for future syphilis and HIV prevention efforts among YBMSM.
Data and methods
Study setting and data collection
Analyses for this study were conducted using data from a multisite longitudinal network study, the Young Men’s Affiliation Project (YMAP). YMAP examines HIV transmission among MSM between the ages of 16 and 29 by investigating affiliation with social venues and health-promoting venues and determining how affiliation affects HIV risk and prevention in Houston, TX (n=378), and Chicago, IL (n=377). YMAP participants were recruited using respondent-driven sampling (RDS)26 between 2014 and 2016. Eligibility criteria include male sex assigned at birth and current male identification, engaging in oral or anal sex with another man in the prior year, residing in and planning to remain in Houston or Chicago for the following year and English-speaking ability.
Survey data were collected via computer-assisted personal interviewing. Topics included sociodemographic characteristics, HIV/STI risk/protective behaviours, social and sexual networks and venue attendance. For biological data, blood samples for HIV testing used site-specific platforms and algorithms and included Oraquick ADVANCE Rapid HIV-1/2 antibody, Clearview Complete HIV-1/2 point of care test, or Alere Determine HIV-1/2 Combo antigen/antibody test. Those with reactive samples were confirmed with laboratory testing using HIV-1/HIV-2 multispot differentiation and HIV RNA (viral load) tests, and were referred to a clinic for care. For syphilis testing, we used a fluorescent treponemal antibody (FTA) absorption test and a rapid plasma reagin (RPR) test, including measurement of titres to determine active syphilis infection.
The sample for this study was limited to 365 YBMSM participants (n=468 self-reported Black race, among whom we collected biomarker data from 396, with 31 participants having other missing values). A more detailed description is available in the online supplementary materials. We obtained assent/consent from all participants; parental consent was waived for minors (under 18 years of age).
Supplementary file 1
Statuses of syphilis and HIV infections. We created a dichotomous variable for active syphilis infection, defined as the participants having both a positive FTA and an RPR titre greater than or equal to 1:4 and as 0 otherwise. HIV seropositivity was based on confirmation test results by creating a dichotomous variable, coding 1 as seropositivity and 0 otherwise. Our outcome variable of syphilis-HIV coinfected statuses included the following four subgroups of: (A) coinfection (both syphilis and HIV infection), (B) HIV monoinfection, (C) syphilis monoinfection and (D) neither infection, which are illustrated in figure 1.
The social network was inclusive of the peer-referral network generated by the RDS process as well as the proximal social and sexual networks reported by respondents, that is, up to five people with whom they share personal information to document their social network and up to five people with whom they had anal, oral or vaginal sex within the past 6 months. These three network data sources were then combined for each city, using a matching procedure that made use of participants and their partners’ sociodemographic information, such as name, age, gender and race, to match the correct pair of participant and partner.27 Based on this aggregated social network, we created three network exposure variables by counting the number of directly connected members who were infected with syphilis, HIV or both (coinfection). We coded 1 for having at least one network member who had a syphilis infection, HIV infection, or coinfection, and 0 otherwise. A more technical description of how we generated these social networks and network variables is provided in the online supplementary materials.
Venue affiliation measure
Respondents were provided with city-specific rosters for social venues, such as clubs, bars and other spaces where men congregate, as well as with lists of health venues, such as education or health centres where young men receive services or support or socialise, which were compiled from formative research (refer to w1 in the online supplementary materials). Then, participants were asked to select each venue that they had attended in the prior 12 months. Two types of venue affiliation variables were created, one for the number of social venues and the other for the number of health venues that the respondent attended.
Sex behaviour measures
We assessed sex behaviour relevant to syphilis transmission. Participants self-reported the total number of sex partners in the past 6 months, which was transformed by taking the square root. The other three measures were assessed by all responses per sex partner elicited in the name generator (a maximum of five partners) and included the number of sex partners with whom participants used drugs during sex and had receptive condomless (or ‘inconsistent’ condom use) or condomless insertive anal intercourse (including versatile).
Age (continuous), education coded as high school or less (dichotomous) and lifetime housing instability experience (dichotomous) were measured.
Social network data were visualised using igraph network visualisation software (refer to w2 in the Supplementary file 1) in R V.3.4.3 statistical environment, with nodal colours indicating disease status.
We conducted two steps in our statistical analysis to address the first objective of identifying factors associated with coinfection. First, we conducted logistic regression analysis to model the log odds of coinfection (A) versus no coinfection (by combining three subgroup categories of (B) HIV monoinfection, (C) syphilis monoinfection and (D) neither infection), as depicted in figure 1. Then, we conducted a multinomial logistic regression analysis by specifying three logit models to predict coinfection in reference to each outcome subgroup category of (B) (model 1a), (C) (model 1b) and (D) (model 1c). To address our second objective of identifying factors associated with syphilis monoinfection or HIV monoinfection, we conducted additional multinomial logistic regression analysis by specifying one model to predict syphilis monoinfection (C) in comparison to without syphilis infection (D) (model 2) and another model to predict HIV monoinfection (B) in comparison to without HIV infection (D) (model 3). Additional details of these multinomial logistic regression models can be found in the online supplementary materials.
Our models included covariates of network factors, risky sexual behaviours and other socioeconomic and demographic characteristics previously defined. We fit separate models for each network exposure variable, and all models controlled for city (Houston vs Chicago). We used the logit and mlogit commands in Stata V.14.0, and our estimation was based on the adjusted RDS sampling method28 that approximates the inclusion probability as inversely proportional to the personal network size (the number of YMSM aged 16–29 in their city with whom they have regular contact). The techniques available to perform multivariate analyses are not sufficiently developed, and RDS weights may introduce additional variance that causes a wider CI.29 Thus, we also ran multinomial regression models without the RDS weights, using the Huber-White robust sandwich variance estimator with equivalent results (see table 1(S) in the online supplementary materials).
Social network visualisation
Figure 2 presents the social networks among 396 YBMSM for Houston (n=211, left) and Chicago (n=185, right), with nodal colours indicating syphilis-HIV infection status and tie colours indicating types of relation.
There are a couple of large clusters observed in these networks, but individuals within each cluster appear to comprise subgroup members with diverse syphilis and/or HIV infection statuses.
Table 1 presents our frequency and sample proportion of syphilis-HIV infection in addition to the RDS-adjusted estimate (lower, upper 95% CIs) that uses Gile’s Sequential Sampling estimators30 and was computed by the RDS V.0.7-8 program (refer to w3 in the online supplementary materials) in the R statistical environment, as well as descriptive statistics for sociodemographic, behavioural and network characteristics.
The sample proportion of syphilis-HIV coinfection was different between Houston and Chicago (P<0.05). The RDS-adjusted estimate of syphilis monoinfection yields a higher prevalence for Chicago compared with the sample proportion. Regarding network factors, on average, 30% of our participants in both cities had at least one syphilis-infected or at least one HIV-infected network member, and 20% had at least one coinfected member in their personal network. Houston participants attended more social venues and fewer health venues in the prior 12 months compared with their Chicago counterparts.
Multinomial logistic regression models
The results of the logistic regression analysis indicate that there was a significant association between coinfection and having syphilis-infected network member(s) (adjusted OR (AOR)=3.02; P<0.05, 95% CI 1.12 to 8.15), HIV-infected member(s) (AOR=3.90; P<0.01, 95% CI 1.46 to 10.43), members who are coinfected (AOR=4.64; P<0.01, 95% CI 1.72 to 12.51) and having more than a high school education (AOR=0.35; P<0.05, 95% CI 0.15 to 0.84), given all other variables being held constant. The results of the subsequent multinomial logic regression analysis are shown in table 2.
As compared with the HIV monoinfection subgroup (model 1a), the coinfection subgroup is 4.38 (P<0.05, 95% CIs 1.25, 15.27) times as likely to have an HIV-infected network member and 3.70 (P<0.05, 95% CIs 1.01, 13.53) times as likely to have coinfected network member(s). As compared with the syphilis monoinfection subgroup (model 1b), the coinfection subgroup is 6.15 (P<0.05, 95% CIs 1.50, 25.13) times as likely to have coinfected network member(s). As compared with neither infection subgroup (model 1c), the coinfection subgroup is 2.97 (P<0.05, 95% CIs 1.03, 8.52) times as likely to have syphilis-infected network member(s), 2.73 (P<0.01, 95% CIs 1.58, 14.14) times as likely to have HIV-infected member(s) and 4.48 (P<0.01, 95% CIs 1.54, 13.04) times as likely to have coinfected member(s).
Syphilis monoinfection is 1.19 (P<0.05, 95% CIs 1.02, 1.40) times as likely relative to neither infection subgroup for each additional number of social venues attended (model 2). HIV monoinfection is 2.06 (P<0.05, 95% CIs 1.13, 3.74) times as likely relative to neither infection subgroup for each additional number of condomless top partners (model 3).
Our study examined network and sexual behavioural factors that may be associated with syphilis-HIV infection among YBMSM. Strengths of our approach include the inclusion of both social and sexual network members, their connection to social venues and the collection of objective biomarkers for both HIV and syphilis infection. Our findings suggest that, in general, syphilis-HIV coinfected YBMSM tend to have network members who also are coinfected. Coinfected YBMSM also tend to have HIV-infected network members, but this tendency was not supported among individuals infected with syphilis. Coinfected YBMSM, unlike syphilis or HIV monoinfected YBMSM, are more likely to have syphilis-infected network members in comparison to non-infected YBMSM. These results indicate that HIV-infected or coinfected YBMSM are likely to be connected to others with syphilis infection. Regular syphilis testing and treatment among HIV-infected men may be important in limiting ongoing syphilis transmission in the community. In addition, syphilis-monoinfected YBMSM tend to attend a number of social venues in comparison to non-infected counterparts, indicating that social spaces may be potential intervention points for YBMSM.
Together, these findings suggest that network factors at the human and venue levels may influence syphilis and HIV infection, despite the commonly assumed syphilis followed by HIV infection chronology. Further analysis is required to disentangle the distinct yet overlapping HIV and syphilis transmission networks that occur among YBMSM.
On the surface, it is not surprising that having infected individuals within one’s social network might be associated with one’s own disease infection, although existence of this proximity within a sexual network might be more intuitive. This suggests that the surrounding risk environment is important, even though directly observed sexual connections may not exist. Our findings suggest that network proximity to HIV-infected and coinfected persons was evident in general but that network proximity to syphilis-infected persons was observed only when the comparison group was individuals who were HIV and syphilis uninfected. In addition, the social network may be more salient to understanding transmission risk within this group, given the dynamic nature of partnerships and frequent status shifting between sex and social interaction over time among young people.
An additional finding of interest was that the syphilis-monoinfected group relative to the dually non-infected one did not differ on any individual-level factors (ie, sex behaviour) but tended to have higher venue attendance. In contrast, the HIV-monoinfected group relative to the dually non-infected one tended to have more condomless top partners. Because causality cannot be inferred from these data, it also may be that there are other explanations such as dually uninfected individuals having access to healthcare and preventive healthcare that may assist them in staying HIV and syphilis negative.
The increasing use of PrEP without concomitant condom use may lead to an increase in STIs, including syphilis. Data from our YBMSM participants demonstrate that only six (2.5%) reported PrEP use at the time of the interview, and all were dually non-infected. Although this is an interesting finding and warrants further investigation, the number is small, and we do not have information on PrEP adherence or duration of PrEP use for this group. In addition, individuals on PrEP may have been treated for syphilis previously or as part of their PrEP care.
There are some limitations to this study. First, our analyses included cross-sectional data, which leaves unaddressed whether syphilis or HIV infection comes first or whether both occur at the same time. In addition, we did not have longitudinal data on syphilis infection to allow clear interpretation of titres. We therefore used a non-treponemal titre cut-off of 1:4. A person with a peak non-treponemal titre less than 1:4 might have been miscategorised as syphilis uninfected. Second, we had a relatively small sample size in some subgroups, such as syphilis-HIV coinfection. Third, our observed network data may reflect only a partial network of the complete transmission network, as our data were collected by an egocentric design. Thus, there are concerns about missing data that are not identifiable in social networks. Finally, our results may have limited generalisability to the broader urban YBMSM population, although RDS chains that consist of YBMSM are sufficiently long.
In summary, our findings suggest that network factors are important in describing syphilis and HIV infection patterns among YBMSM. Interventions that do not account for network forces at the person-to-person level or the person-to-venue level may be limited in efficiency and potential effectiveness. Clinicians should be aware of the potential for coinfection and routinely test and treat syphilis infection to limit onward transmission. Further research is needed to clarify how and if education and housing status may interact with network factors to influence transmission. Future research in this area could explore the dynamics of syphilis-HIV coinfection, using longitudinal data, to identify priority areas for interventions among subcommunities of YBMSM that are at increased risk.
Little is known about combined network and behavioural drivers of syphilis and HIV epidemics among young Black men who have sex with men.
Syphilis-HIV coinfection is associated with having coinfected or HIV-infected network members and, in some cases, with having syphilis-infected network members.
Syphilis monoinfection, in comparison to dual non-infection, is associated with a higher number of social venues attended.
Clinical practice and policy should underscore the opportunity for counselling related to syphilis-HIV transmission risks associated with social networks and sex behaviours.
We acknowledge Britt Skaathun, Dennis Li, Yucheng Zhao, Angela Di Paola, and YMAP staff for the contribution to this study.
Handling editor Jackie A Cassell
Contributors KF conceptualised the study, led the project, designed the network analysis, specified statistical models, conducted the statistical analysis and led the writing of the article. CAF interpreted the results and organised the literature on the topic. LMK designed the data collection, collected the data and interpreted the results. JYK prepared the data, computed the network measures and conducted network analysis. JAS led the project, designed the data collection, collected the data, interpreted the results and conceptualised the study. All authors substantially contributed to the study, engaged in writing and approved the manuscript.
Funding This work was supported by the National Institutes of Health (grant numbers 1R01MH100021, 1R01DA039934, K23-MH109358-02 and 1R21GM113694).
Competing interests None declared.
Patient consent Obtained.
Ethics approval All three institutions that participated in YMAP (The University of Chicago (UC), Ann & Robert H Lurie Children’s Hospital of Chicago (Lurie) and The University of Texas Health Science Center at Houston School of Public Health (UTHealth)) received approval from the irinstitutional review boards (IRB #HSCSPH120830).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Data will be shared on a requested basis. No additional unpublished data from the study are available.
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.