Serosorting, a risk reduction strategy that involves preferentially seeking sex partners of the same HIV status or limiting unprotected anal sex to seroconcordant partners, has been described among men who have sex with men (MSM) in the United States, Europe, and Australia.1–8 Some research suggests that serosorting may provide modest protection against HIV if it replaces a higher risk behavior such as serodiscordant unprotected anal intercourse (UAI).2,9,10 However, serosorting carries a substantial risk for HIV-negative men owing to factors such as acute infection, infrequent testing, lack of HIV status disclosure, and alcohol and substance use, which can interfere with accurate determination of individual and partner serostatus2,11,12 and reduce the effectiveness of serosorting as a risk reduction strategy. Furthermore, owing to the epidemiologic synergy between sexually transmitted infections (STIs) and HIV,13 STI coinfection may further increase the risk of HIV transmission from patients unaware of their infection when serosorting occurs between presumed HIV-negative partners.
Some researchers distinguish serosorting from “seroguessing,” or serosorting based on the assumption of seroconcordance rather than on direct determination of serostatus.14 Zablotska and colleagues14 found that among HIV-negative MSM, the likelihood of UAI with a casual partner was greater both when the partner’s HIV status was known and assumed concordant compared with unknown. Whether seroconcordance is assumed or assessed directly may have different implications for disease transmission and prevention messaging. We examined the relationship between seroconcordant UAI and bacterial STIs according to the method by which partners’ HIV status was determined (direct vs. indirect methods) to provide preliminary information on the association between direct and assumed serosorting and bacterial STIs among HIV-negative MSM at an urban STI clinic.
MATERIALS AND METHODS
Subjects and Data collection
Data were collected from patients seeking clinical services for STI testing at Howard Brown Health Center, the Midwest’s largest lesbian, gay, bisexual, and transgender health center. The analysis was limited to men 18 years and older who reported at least 1 male sex partner in the previous 90 days. Sexual and risk behaviors were measured with a set of questions that were added to the standard risk assessment used in routine STI surveillance. The questionnaire and use of data in subsequent analyses were approved by the institutional review boards of Howard Brown Health Center and the University of Illinois at Chicago.
Patient characteristics, measured at each visit, included age, race/ethnicity, HIV status, and STI history. Individual sexual risk behaviors included total number of sex partners in the past 90 days and reported anonymous partners (any vs. none) and use of alcohol (any vs. none) or other substances (including ecstasy, methamphetamine, club drugs, cocaine, and poppers) in the past year. For the most recent sex partner, patients reported their partners HIV status (positive, negative, unknown), sex, partner type (main or casual), relationship type (monogamous or nonmonogamous), and whether they had unprotected insertive and/or receptive anal sex with the partner. All data were based on self-report. Classification of seroconcordance was based on perceived patient and partner serostatus and was not laboratory confirmed because anonymous HIV testing precluded linkage between STI and HIV results.
The analysis was restricted to MSM who self-reported as HIV negative (86% of visits) and provided information on the HIV status of their most recent male partner and sexual behavior with this partner. We excluded 97 (7%) visits by HIV-positive men because a small sample size limited subgroup comparisons, because of the different implications of serosorting for HIV-positive and HIV-negative men, and because we were specifically interested in examining the association between serosorting and STI among those at risk for HIV infection. Ninety-five (7%) visits where HIV status was reported as unknown were also excluded because of the role of knowing one’s own status in serosorting. Visits that occurred less than 30 days from the previous encounter were excluded because they were considered to represent the same clinical episode.
Age was collected and analyzed as a continuous variable in univariate analysis and dichotomized about the median for multivariable analysis. The total number of recent sex partners was collected as a continuous variable and dichotomized as 2 or more versus 1 for analysis because risk of STI increased among men reporting 2 or more partners versus 1 partner but remained stable with increasing partners more than 2. Alcohol use in the past year was collected and analyzed as a dichotomous variable: any versus none. Reported substances including ecstasy, methamphetamine, cocaine, poppers, and other club drugs were analyzed as a single measure of substance use (any vs. none) for multivariable analysis because of similarity in direction and magnitude of association when analyzed separately, positive correlation between the variables, and their presumed shared mechanism of influence on sexual risk taking. Unprotected anal intercourse was defined as any unprotected insertive and/or receptive anal intercourse with most recent sex partner.
Definition of STI Outcomes
The outcome for analysis was infection with early syphilis (primary, secondary, early latent) or urogenital or rectal gonorrhea or chlamydia. Syphilis infection was detected by enzyme immunoassay testing with confirmation by fluorescent treponemal antibody, and diagnoses were based on Centers for Disease Control and Prevention staging criteria.15 Urogenital gonorrhea and chlamydia were identified by nucleic acid amplification testing of urine specimens using Strand Displacement Amplification with the Becton-Dickinson ProbeTec assay (BD Diagnostic Systems, Sparks, MD). Screening for rectal gonorrhea and chlamydia was offered to men reporting receptive anal intercourse or rectal symptoms, and infections were identified using culture or the Becton-Dickinson ProbeTec assay.
Definitions of Serosorting Practices
Assessment of partner HIV status was based on categories that were adapted from qualitative research by Parsons and colleagues16 examining physical, behavioral, and circumstantial factors on which serosorting assumptions are based. Participants were given a list of items from which they could check all that applied, including a qualitative “Other” category, and mutually exclusive categories were created for analysis. Serostatus assessment was classified as known (i.e., “My partner told me and I had no reason to doubt them” or “We got tested together”) or assumed (i.e., based on partner’s appearance, education, occupation, etc.). Participants who did not report a means of assessing partner’s status were classified as using assumed assessment. The primary explanatory variable was based on the serostatus of the most recent partner, how the partner’s HIV status was determined, and whether UAI was reported. The 5 categories were as follows: known seroconcordance (UAI with a partner of the same serostatus in which partner’s serostatus was ascertained through direct assessment), assumed seroconcordance (UAI with a partner of assumed seroconcordance), UAI with a partner of unknown serostatus, serodiscordant UAI (UAI with an HIV-positive partner), and no UAI (condom-protected anal intercourse regardless of partner’s serostatus or no anal intercourse reported).
Definitions of serosorting vary in the literature but have often been based on sexual behavior—exclusively selecting sex partners of the same HIV status or limiting episodes of unprotected sex to partners of the same serostatus. Eaton and colleagues1,17 defined serosorting as “only having anal sex without a condom with partners of the same HIV status.” McFarland and colleagues18 recently distinguished between “condom serosorting” (using condoms with partners of discordant or unknown serostatus but having unprotected sex with partners of the same serostatus) and “pure serosorting” (having sex exclusively with partners of the same HIV status). We collected data only on behavior with the most recent sex partner because the setting and type of data collection (surveillance data from a busy STI clinic) did not allow us to collect detailed information on sexual behavior across multiple partners or sexual episodes. Because our data are cross sectional in nature, we contextualize UAI as either seroconcordant or serodiscordant because the term serosorting implies an a priori partner selection process. However, we attempted to identify potential serosorting by distinguishing between known and assumed seroconcordance to provide preliminary information on serosorting behavior in our population and prevalence estimates for future research.
This was a cross-sectional, visit-based analysis in which participants could contribute data at more than 1 visit, and exposure and outcome data were measured at the same point in time. Unadjusted and adjusted prevalence ratios (aPRs) were calculated using generalized estimating equations (GEEs) with an exchangeable correlation structure and log-binomial link. Generalized estimating equation methods allow for inclusion of multiple observations per subject and accommodate analysis of data from patients with varying number of visits because they account for correlation between responses from the same subjects.
The primary explanatory variable was UAI with the most recent sex partner (known seroconcordant UAI, assumed seroconcordant UAI, UAI with a partner of unknown status, serodiscordant UAI, or no UAI). Other explanatory variables included age, race/ethnicity, STI history, total sex partners in the past 90 days, and any anonymous sex partners and alcohol or substance use in the past year.
Exploratory analysis included Pearson χ2 tests for categorical variables and Wilcoxon tests to compare distributions of nonnormally distributed continuous variables by STI outcomes. Factors that were significant at P < 0.20 in univariate analysis were entered in multivariable regression models and retained in models with P < 0.05. Age and race were maintained in multivariable models for consistency with other published literature and to control for residual confounding. Analyses were conducted with SAS version 9.2.
From May 2010 through October 2011, 1046 HIVnegative MSM were screened for gonorrhea, chlamydia, and syphilis at 1208 visits. The analytic sample consisted of 1110 (92%) of 1208 visits among 961 men (844 with single visits, 117 with multiple visits) at which the most recent sex partner was identified as male and information was provided on the serostatus of the partner and sexual behavior. Men were 35 years (median age; interquartile range, 28–45 years), 75% white non-Hispanic, 8% black non-Hispanic, 11% Hispanic, and 6% other race/ethnicity (Table 1).
Participants reported their partner’s HIV status as concordant negative at 49.2% (546/1110), discordant at 5.1% (57/1110), and unknown at 45.6% (506/1110) of visits. Among 603 visits where the HIV status of the partner was known, 59 (9.8%) got tested together, 441 (73.1%) were told by the partner, and 103 (17.1%) used indirect (3.5%) or unknown (13.6%) methods for assessing the partner’s serostatus.
Known seroconcordant UAI was reported at 17.4% of visits; UAI with an assumed seroconcordant partner, 4.6% of visits; UAI with a partner of unknown serostatus, 15.0% of visits; serodiscordant UAI, 2.1% of visits; and no UAI, 61.0% of visits. During the study period, 133 gonorrhea infections, 106 chlamydia infections, and 16 early syphilis infections were identified. Overall STI prevalence was 20.1% (223/1110 visits): 20.2% (39/193) at visits with known seroconcordant UAI, 35.3% (18/51) at visits with assumed seroconcordant UAI, 29.5% (49/166) at visits with UAI with a partner of unknown serostatus, 34.8% (8/23) at visits with serodiscordant UAI, and 16.1% (109/677) at visits where no UAI was reported.
Association Between Seroconcordance and STI Outcomes
In univariate analysis, STI prevalence at visits where men reported known seroconcordant UAI was not significantly increased compared with visits where no UAI occurred (prevalence ratio [PR], 1.24; 95% confidence interval [CI], 0.85–1.79). However, compared with visits where no UAI was reported, STI prevalence was 2.16 (95% CI, 1.40–3.33) times greater at visits where assumed seroconcordant UAI occurred, 1.88 (95% CI, 1.39–2.54) at visits where UAI with a partner of unknown status occurred, and 2.07 (95% CI, 1.14–3.78) times greater at visits where serodiscordant UAI was reported. Other factors associated with STI were STI history, alcohol use, substance use, having 2 or more recent sex partners, younger age, and black race (Table 2).
In multivariable regression, assumed seroconcordant UAI (aPR, 2.51; 95% CI, 1.79–3.51), UAI with an unknown status partner (aPR, 1.76; 95% CI, 1.31–2.38), and serodiscordant UAI (aPR, 2.57; 95% CI, 1.76–3.75) remained significant predictors of STI after controlling for age and race/ethnicity, STI history (aPR, 1.44; 95% CI, 1.12–1.84), alcohol use (aPR, 1.54; 95% CI, 1.00–2.37), substance use (aPR, 1.45; 95% CI, 1.13–1.85), and multiple sex partners (aPR, 1.51; 95% CI, 1.04–2.20) (Table 2). Known seroconcordant UAI was not associated with increased STI risk (aPR, 1.20; 95% CI, 0.83–1.73).
Similar associations were observed when the analysis was restricted to visits where men reported casual partnerships (n = 722; 66.0% of visits). The adjusted STI prevalence was 2.47 (95% CI, 1.47–4.14) times greater at visits with assumed seroconcordant UAI, 1.75 (95% CI 1.23–2.49) times greater at visits where UAI with an unknown status partner occurred, and 2.65 (95% CI, 1.62–4.34) times greater at visits where serodiscordant UAI occurred compared with those where no UAI was reported (Table 3).
Unprotected anal intercourse with a seroconcordant partner in which serostatus was assessed directly was not associated with increased STI prevalence compared with no UAI. In contrast, assumed seroconcordant UAI was associated with significantly increased STI prevalence, an association that remained when the analysis was restricted to visits where men reported casual partners. Directly assessing a partner’s HIV status may reflect greater concern for one’s health in general or unmeasured relationship characteristics (higher trust and communication) that are associated with lowered STI risk. Among men reporting seroconcordant-negative partners, UAI was somewhat less common among men reporting direct versus assumed assessment (42.9% vs. 53.1%; P = 0.067), suggesting lower risk among those using direct assessment. There was no association between direct versus assumed assessment and age, race/ethnicity, number of sex partners, alcohol or substance use, or gay versus bisexual orientation. Although men were less likely to know the HIV status of casual partners compared with main partners (42.4% vs. 77.3%; P < 0.01), among those who reported a seroconcordant-negative partner, direct assessment was somewhat more common with casual partners than with main partners (86.1% vs. 79.2%; P = 0.032). This suggests that men may be more likely to use direct assessment with partners they know less well, although the magnitude of risk associated with assumed seroconcordant UAI was similar with main and casual partners.
Unprotected anal intercourse was also more common with concordant partners than with unknown status partners regardless of assessment method. Prevalence of UAI was 32.8% with unknown status partners, 53.1% with assumed concordant partners, and 42.9% with known concordant partners (P < 0.01). Unprotected anal intercourse with an assumed concordant partner was associated with higher STI prevalence (aPR, 2.51) than UAI with a partner of unknown serostatus (aPR, 1.76), suggesting that men who make assumptions about the status of their partners may have a unique risk profile that may reflect differences in sexual mixing patterns or partnership types, or other unmeasured confounders.
The increased STI prevalence associated with UAI in the context of indirect assessment suggests that the effectiveness of assumed serosorting as an HIV risk reduction strategy may also be limited, given the shared behavioral risks for HIV and STI transmission and the biologic mechanism by which undetected STI may increase HIV transmission. Furthermore, in settings with infrequent HIV testing and high partner turnover, accurately determining one’s own HIV status may also be difficult. In our study, multiple recent sex partners were reported at 81% of visits and 6 months or longer since the last HIV test at 59% of visits. Alcohol and substance use was also common and may further interfere with accurate determination of partner serostatus, particularly if used in the context of the sexual encounter.
Our results contrast other research that has examined the association between serosorting and STIs. Among MSM in Seattle, serosorting offered partial protection against HIV but did not protect against STI, and the magnitude of STI risk associated with serosorting was similar to that associated with discordant UAI.2 Distinguishing how and when men who practice serosorting establish seroconcordance in relation to the sexual encounter may thus be important when assessing the risk associated with serosorting.
Because the length and content of our survey were limited by the setting in which the data were collected, information on sex partner characteristics was available only for the most recent sex partner, and we did not quantify the number of episodes of UAI that occurred. Although this method limits difficulty with recall and maintains brevity of the survey, it precluded the examination of variation in behavior across sexual episodes according to the serostatus of the partner or the context of the sexual encounter. Because many patients reported multiple sex partners, behavior with last partner may not have been an accurate representation of global behavior patterns. However, the prevalence of known seroconcordant UAI in our study (17%) is fairly consistent with other studies reporting a serosorting prevalence of 20% to 33% in different populations of MSM.1,2,6,8
Because the study was cross sectional, we were unable to establish a temporal relationship between assumed seroconcordance and STI outcomes. Although we measured direct versus indirect ascertainment of partner serostatus to differentiate between known and assumed serosorting, the cross-sectional nature of the study limited the extent to which we were able establish the timing of discussion of serostatus. Furthermore, we were not able to distinguish whether the behaviors in our study represented intentional serosorting or post hoc rationalization of unsafe sex. A prospective measure of serosorting intentions, behaviors, and STI and HIV outcomes is needed to differentiate the temporality of these potential associations.
Self-report may have led to misclassification of participant and partner HIV status and seroconcordance. In general, risk behaviors may have been underreported because of the sensitive nature of the data, which is supported by the fact that STI prevalence was 16% even among patients reporting no UAI with their most recent sex partner. This may also be a limitation of using behavior only with last sex partner as a proxy for overall behavior patterns, particularly in a setting where most men reported multiple recent sex partners.
Despite these limitations, our study demonstrates the feasibility of assessing sexual behaviors by partner status and sexual decision making in an STI clinic setting and provides preliminary information on the association between serosorting and STIs in a high-risk population of MSM.
Among HIV-negative MSM in our study, engaging in UAI with a partner whose HIV status was assumed to be seroconcordant was associated with increased STI prevalence. The importance of establishing seroconcordance through testing and direct discussion should be emphasized for HIV-negative MSM who practice serosorting.
1. Eaton LA, Kalichman SC, Cain DN, et al.. Serosorting sexual partners and risk for HIV among men who have sex with men. Am J Prev Med 2007; 33 (6): 479–485.
2. Golden MR, Stekler J, Hughes JP, et al.. HIV serosorting in men who have sex with men: Is it safe? J Acquir Immune Defic Syndr 2008; 49 (2): 212–218.
3. Hart GJ, Elford J. Sexual risk behaviour of men who have sex with men: Emerging patterns and new challenges. Curr Opin Infect Dis 2010; 23 (1): 39–44.
4. Lattimore S, Thornton A, Delpech V, et al.. Changing patterns of sexual risk behavior among London gay men: 1998–2008. Sex Transm Dis 2011; 38 (3): 221–229.
5. Mao L, Crawford JM, Hospers HJ, et al.. “Serosorting” in casual anal sex of HIV-negative gay men is noteworthy and is increasing in Sydney, Australia. Aids 2006; 20 (8): 1204–1206.
6. Snowden JM, Raymond HF, McFarland W. Seroadaptive behaviours among men who have sex with men in San Francisco: The situation in 2008. Sex Transm Infect 2011; 87 (2): 162–164.
7. Truong HM, Kellogg T, Klausner JD, et al.. Increases in sexually transmitted infections and sexual risk behaviour without a concurrent increase in HIV incidence among men who have sex with men in San Francisco: A suggestion of HIV serosorting? Sex Transm Infect 2006; 82 (6): 461–466.
8. Xia Q, Molitor F, Osmond DH, et al.. Knowledge of sexual partner’s HIV serostatus and serosorting practices in a California population–based sample of men who have sex with men. Aids 2006; 20 (16): 2081–2089.
9. Jin F, Crawford J, Prestage GP, et al.. Unprotected anal intercourse, risk reduction behaviours, and subsequent HIV infection in a cohort of homosexual men. Aids 2009; 23 (2): 243–252.
10. Philip SS, Yu X, Donnell D, et al.. Serosorting is associated with a decreased risk of HIV seroconversion in the EXPLORE Study Cohort. PLoS One 2010; 5 (9): e1662.
11. Pinkerton SD. Acute HIV infection increases the dangers of serosorting. Am J Prev Med 2008; 35 (2): 184.
12. Wilson DP, Regan DG, Heymer KJ, et al.. Serosorting may increase the risk of HIV acquisition among men who have sex with men. Sex Transm Dis 2010; 37 (1): 13–17.
13. Fleming DT, Wasserheit JN. From epidemiological synergy to public health policy and practice: the contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect 1999; 75 (1): 3–17.
14. Zablotska IB, Imrie J, Prestage G, et al.. Gay men’s current practice of HIV seroconcordant unprotected anal intercourse: serosorting or seroguessing? AIDS Care 2009; 21 (4): 501–510.
15. Centers for Disease Control and Prevention. 2012 Case Definitions: Nationally Notifiable Conditions Infectious and Non-Infectious Case. Atlanta, GA: Centers for Disease Control and Prevention. 2012.
16. Parsons JT, Severino J, Nanin J, et al.. Positive, negative, unknown: Assumptions of HIV status among HIV-positive men who have sex with men. AIDS Educ Prev 2006; 18 (2): 139–149.
17. Eaton LA, West TV, Kenny DA, et al.. HIV transmission risk among HIV seroconcordant and serodiscordant couples: Dyadic processes of partner selection. AIDS Behav 2009; 13 (2): 185–195.
18. McFarland W, Chen YH, Nguyen B, et al.. Behavior, intention or chance? A longitudinal study of HIV seroadaptive behaviors, abstinence and condom use. AIDS Behav 2011; 16 (1): 121–131.