Sexual harm reduction practices of HIV-seropositive gay and bisexual men: serosorting, strategic positioning, and withdrawal before ejaculation

Transcription

1 Sexual harm reduction practices of HIV-seropositive gay and bisexual men: serosorting, strategic positioning, and withdrawal before ejaculation Jeffrey T. Parsons a,b,c, Eric W. Schrimshaw b, Richard J. Wolitski d, Perry N. Halkitis c,e, David W. Purcell d, Colleen C. Hoff f and Cynthia A. Gómez f Objective: This study assessed unprotected anal and oral sex behaviors of HIV-positive gay and bisexual men in New York City and San Francisco with their main and nonmain sexual partners. Here we focus on the use of three harm reduction strategies (serosorting, strategic positioning, and withdrawal before ejaculation) in order to decrease transmission risk. Method: The data from a baseline assessment of 1168 HIV-positive gay and bisexual men in the two cities were utilized. Men were recruited from a variety of communitybased venues, through advertising and other techniques. Results: City differences were identified, with more men in San Francisco reporting sexual risk behaviors across all partner types compared with men in New York City. Serosorting was identified, with men reporting significantly more oral and anal sex acts with other HIV-positive partners than with HIV-negative partners. However, men also reported more unprotected sex with partners of unknown status compared with their other partners. Some evidence of strategic positioning was identified, although differences were noted across cities and across different types of partners. Men in both cities reported more acts of oral sex without ejaculation than with ejaculation, but the use of withdrawal as a harm reduction strategy for anal sex was more common among men from San Francisco. Conclusion: Overall, evidence for harm reduction was identified; however, significant differences across the two cities were found. The complicated nature of the sexual practices of gay and bisexual men are discussed, and the findings have important implications for prevention efforts and future research studies. ß 2005 Lippincott Williams & Wilkins AIDS 2005, 19 (suppl 1):S13 S25 Keywords: HIV, seropositive, serosorting, harm reduction, gay, bisexual Introduction Recent research studies have documented increases in sexual risk behaviors among men who have sex with men (MSM) both in the United States and abroad [1 7]. Although survey data have shown that a minority (typically from 20 to 30%) of HIV-positive MSM engage in sexual practices that place seronegative partners at risk of HIV infection [8 11], recent reports of increases in HIV incidence [12,13] and sexually From the a Hunter College of the City University of New York, New York, NY, USA, the b Graduate Center of the City University of New York, New York, NY, USA, the c Center for HIV/AIDS Educational Studies and Training (CHEST), New York, NY, USA, the d Centers for Disease Control and Prevention, Atlanta, GA, USA, the e New York University, New York, NY, USA, and the f University of California, San Francisco, CA, USA. Correspondence to Jeffrey T. Parsons, Hunter College of the City University of New York, Department of Psychology, 695 Park Avenue, New York, NY 10021, USA. Tel: ; jeffrey.parsons@hunter.cuny.edu ISSN Q 2005 Lippincott Williams & Wilkins S13

2 S14 AIDS 2005, Vol 19 (suppl 1) transmitted infection (STI) rates [4,5,14,15] underscore the need to understand more fully the sexual behaviors of HIV-positive MSM. Clearly most HIV-positive gay and bisexual men neither want to nor intend to transmit HIV [16 18]; however, new infections continue to emerge among MSM, and in some areas of the United States diagnoses of HIV infection in this population have risen for three consecutive years from 2000 to 2003 [19]. Much of the discrepancy between the reported HIV risk reduction efforts of HIV-positive MSM and increasing rates of HIV infection may result from how sexual behaviors are assessed and what is perceived to constitute HIV transmission risk. For example, many of the initial studies to document the sexual risk behaviors of HIV-positive MSM did not assess the serostatus of sexual partners. Other studies assessed unprotected anal sex without consideration of whether the HIV-positive man was the insertive or the receptive partner, and few have assessed withdrawal before ejaculation. All of these factors can have an impact on the likelihood of HIV transmission [20,21]. Some have examined the strategies that gay and bisexual men use to enable them to obtain greater sexual satisfaction while reducing the risk of HIV transmission or infection. Many HIV-positive men choose to engage in unprotected sex with only, or primarily, other HIVpositive men [16 18]. The process of serosorting [18,22] is also used, by which individuals, regardless of their HIV status, engage in sexual risks only with those partners who they believe to be seroconcordant. Inherently problematic in this approach, however, is the potential for serosorting to fail. First, the approach assumes that HIV status disclosure has occurred; a recent study of HIV-positive gay and bisexual men found that 42% reported any sex (either protected or unprotected) without disclosing their status [23]. Second, the approach assumes that individuals are fully honest and accurate regarding disclosure of their status, which is not always the case [24,25], and that they actually know their HIV status in the first place [14]. Third, gay and bisexual men have been shown to make assumptions regarding the HIV status of their sexual partners, typically assuming that their partner s HIV status is concordant when engaging in unprotected sex [18,26,27]. These assumptions of seroconcordance are often made in sexual venues such as bathhouses and sex clubs, in which there is limited verbal communication, and MSM who frequent these venues have been found to be more likely to engage in unprotected sex [28,29]. As such, serosorting may reduce the risk of HIV transmission in some encounters, but questions remain about the effectiveness of this strategy in the light of serostatus disclosure and knowledge of serostatus among MSM. Van de Ven and colleagues in Australia [30,31] discussed strategic positioning to address patterns of sexual risktaking and risk management among gay and bisexual men. The underlying premise is that men have developed very clear understandings of HIV risks as well as differing levels of relative risk [21,22,32]. For example, despite the significant controversy and debate regarding the level of risk associated with oral sex resulting from conflicting empirical reports [33,34], gay and bisexual men understand the elevated risk of anal sex relative to oral sex [22,30,35,36]. Furthermore, these men tend to have a conceptualization of the relative risks of unprotected insertive versus receptive anal sex, in that they understand that HIV transmission and infection is more likely to occur if an HIV-positive man is the insertive partner and an HIV-negative man is the receptive partner during anal intercourse [22,37]. Such notions of strategic positioning to reduce the risk of HIV infection are supported by some epidemiological evidence [20,21,33,34]. Van de Ven et al. [30] found that of the 30% of gay men in serodiscordant relationships who reported unprotected anal sex, more than half engaged in strategic positioning, such that the HIV-negative men were insertive and the HIV-positive men were receptive. Furthermore, to ascertain whether or not this was indicative of intentional strategic positioning, rather than a preference for certain sexual positions, acts of protected anal sex in these serodiscordant relationships were examined. For protected sex acts, the majority of men, regardless of their HIV status, reported engaging in both insertive and receptive sex, lending support to the idea that such positioning for unprotected anal sex is strategic rather than preferential. Similar patterns of strategic positioning were identified for HIV-positive gay men with their serodiscordant casual sex partners. A third harm reduction technique used by some gay and bisexual men is the practice of withdrawal before ejaculation during anal sex. Withdrawal may serve as a compromise between using condoms or not having anal sex, and the practice is commonly cited as a justification for having engaged in unprotected anal sex [37]. Van de Ven et al. [30] found that 46% of the HIV-positive gay men reporting unprotected anal sex in a serodiscordant relationship engaged in consistent withdrawal. Although any unprotected anal sex, even that which includes withdrawal before ejaculation, represents a risk because of the presence of HIV in pre-ejaculatory fluid and the possibility that the insertive partner may not reliably withdraw [38,39], there is limited evidence from a study of heterosexual HIV transmission that such withdrawal does reduce the likelihood of HIV infection [40]. There is substantial evidence, however, that gay and bisexual men perceive withdrawal to be a safer option than ejaculation [37,41,42], and that such men are engaged in the use of withdrawal as a harm reduction strategy for HIV prevention [21,42 44]. Some men, however, appear to

3 Sexual harm reduction practices of HIV-seropositive gay and bisexual men Parsons et al. S15 use a resolve to withdraw as a justification for unprotected anal sex [45], and belief in the safety of withdrawal before ejaculation has been shown to predict HIV seroconversion [44]. Although condoms are recognized as providing the greatest protection in terms of HIV transmission, a minority of HIV-positive gay and bisexual men continue to engage in unprotected sex. However, HIV-positive men who engage in unprotected sex may attempt to minimize (but not eliminate) the risk of HIV transmission by having sex with other HIV-positive partners, taking the receptive role, or by withdrawal before ejaculation when taking the insertive role. In this paper, we focus on the sexual risk practices of HIV-positive gay and bisexual men in New York City and San Francisco, in order to understand these patterns of potential harm reduction strategies. Behaviors are examined separately for main and non-main partners, because studies have shown that gay and bisexual men within couples report higher frequencies of unprotected sex than single men [46,47]. Three specific hypotheses were tested: (i) HIV-positive gay and bisexual men would report behaviors consistent with serosorting, in terms of more sexual risk with other HIVpositive men than with HIV-negative or unknown-status partners; (ii) men would report strategic positioning, (more receptive rather than insertive sex with partners at risk); and (iii) men would report more acts of sex involving withdrawal before ejaculation than acts with ejaculation, with partners at risk. Methods Participants The data reported here are based on the 1168 participants who completed the Seropositive Urban Men s Intervention Trial (SUMIT) baseline assessment. A detailed description of the methods is provided elsewhere in this issue [48]. Measures Sociodemographics Participants were asked to indicate race/ethnicity, date of birth, gender identification (i.e. male, transgender), sexual orientation (i.e. gay, bisexual, heterosexual, none of the above/unsure), educational background, employment status, personal income, partnership status (i.e. whether or not participants have a primary partner), and city of residence. Sexual behaviors Sex behaviors were assessed by asking participants to indicate the frequency of four sexual behaviors (insertive oral, receptive oral, insertive anal, receptive anal) in the 3 months before completion of the survey, using a measure previously used with HIV-positive gay and bisexual men in the two cities [11]. Participants reported their frequency of engaging in each sexual behavior with and without the use of condoms, with and without ejaculation (when condoms were not used), and separate frequencies were obtained for main partners (defined as a partner you would call your boyfriend, spouse, significant other, or life partner ) and non-main (casual) partners. Furthermore, for non-main partners, participants reported sexual behavior frequencies for HIV-positive, HIV-negative, and unknown-status partners. For these frequencies, instructions were included such that participants responded to questions about HIV-positive and HIV-negative partners based on the number of nonmain partners who told the participant their HIV status, or in the case of unknown non-main partners, partners who did not tell their HIV status or said that they did not know. Open-ended frequency measures of sexual risk behaviors help to communicate expectations that such behaviors occur and are not abnormal, and as such have been recommended to improve the quality of self-report data [49,50]. Comparable measures of sexual risk practices used with gay men demonstrated that past 3 month assessment periods for sexual risk behaviors are reliable [51]. The measures used in this study included easy to understand terminology, which was developed in conjunction with a Community Advisory Board of HIV-positive men. Analytical approach The number of episodes for each sexual behavior were highly positively skewed, with the majority of men reporting no episodes. Because the non-normal distribution prohibits the use of typical parametric statistics (e.g. analysis of variance), we utilized non-parametric statistics, which do not require normally distributed data [52]. Comparisons between three independent groups (e.g. comparisons of behaviors based on whether participants had a main partner whowas HIV positive, HIV negative, or of unknown status) were conducted using the Kruskal Wallis test for continuous variables (the non-parametric equivalent of a one-way analysis of variance) and chisquare tests for categorical variables. A number of the reported comparisons, however, are based on nonindependent groups (e.g. a single participant could have both receptive and insertive sexual behaviors), and as such, require a separate set of non-parametric statistics. Comparisons of two non-independent groups were conducted using the Wilcoxon signed rank test for continuous variables and McNemar tests for categorical variables. For comparisons of three non-independent groups (e.g. comparison of men s behaviors with their HIV-positive, HIV-negative, and unknown-status nonmain partners), comparisons were made using Friedman tests for continuous variables and Cochran s Q for categorical variables. Post-hoc comparisons between three independent groups were made using Mann Whitney

4 S16 AIDS 2005, Vol 19 (suppl 1) tests (for continuous variables) and chi-square tests (for categorical variables). Post-hoc comparisons between three non-independent groups were made with Wilcoxon signed rank tests (for continuous variables) and McNemar tests (for categorical variables). Confidence intervals presented include the lowest possible real value. Results City comparisons Of the 1168 men, 590 (50.5%) were from the New York City metropolitan area and 578 (49.5%) were from the San Francisco Bay area. Differences in sociodemographics and health characteristics between men in the two cities are presented elsewhere in this issue [48]. Overall, across the two cities, 17.8% of men (n ¼ 151) reported having unprotected anal sex (either insertive or receptive) with a known HIV-negative non-main partner, but 34.0% (n ¼ 288) reported unprotected anal sex with a non-main partner of unknown HIV status. When HIVnegative and unknown-status non-main partners were combined, representing potential HIV transmission risk, 47.3% of men (n ¼ 390) reported any unprotected anal sex. When sex with main partners was examined across the two cities, 15.0% of men (n ¼ 65) reported having unprotected anal sex (either insertive or receptive) with a known HIV-negative main partner, and 6.3% (n ¼ 27) reported unprotected anal sex with a main partner of unknown HIV status. When HIV-negative and unknown-status main partners were combined, representing potential HIV transmission risk to main partners, 21.3% of men (n ¼ 92) reported any unprotected anal sex. When comparing unprotected sex among participants across the two cities, several significant differences were identified. Consistently, a greater percentage of men in San Francisco reported unprotected behaviors compared with men in New York, and men from New York typically reported lower frequencies of unprotected behaviors than men from San Francisco. Given the large number of significant differences between men from each city on sexual risk behaviors, men were analysed separately. Sexual behaviors with main partners Men from New York were significantly more likely to report a main sexual partner than men from San Francisco (42 versus 34%), x 2 (1) ¼ 7.07, P < In order to test the hypothesis that HIV-positive men with main partners at risk of HIV infection would report less sexual risk than men with HIV-positive main partners, comparisons were made across partner serostatus (see Table 1). Among men in both cities, those with HIV-negative or unknownstatus main partners reported less insertive oral and anal sexual risk behaviors than those with HIV-positive main partners. However, no significant differences were identified for receptive oral and anal sex risk behaviors by main partner serostatus. To examine strategic positioning, comparisons of insertive and receptive sex were made (see Table 2). The hypothesis was supported among men from San Francisco for both unprotected oral and anal sex behaviors, in that receptive sex was significantly more common than insertive sex. Among New York men, however, the hypothesis was supported for unprotected oral sex, but not for anal sex. Finally, the hypothesis that men attempt to reduce the risk of transmission to their main sexual partners by selectively engaging in sexual risk behaviors without ejaculation (withdrawal) rather than with ejaculation was examined (see Table 3). The hypothesis was supported for oral sex but not for anal sex among men from both cities. Sexual behaviors with non-main partners Men in both cities reported significantly more non-main sexual partners whose serostatus was unknown to them, than either HIV-negative or HIV-positive partners (see Table 4). Men also reported more HIV-positive than HIV-negative non-main partners. The hypothesis that men would report serosorting with non-main partners, such that they engage in fewer unprotected behaviors with partners at risk of HIV infection was partly supported (see Table 4). On most sexual risk behaviors, men reported a greater prevalence and frequency of sexual risk practices with their HIV-positive non-main partners, compared with their HIV-negative non-main partners. However, contrary to the hypothesis, post-hoc comparisons indicated that men in both cities reported a greater prevalence and frequency of most sexual risk behaviors with serostatus-unknown non-main partners than with either HIV-positive or HIV-negative non-main partners. In order to examine whether men attempted to reduce the sexual transmission risk to their non-main partners by using strategic positioning, comparisons in the reported prevalence and frequency of insertive behaviors relative to receptive behaviors were made (see Table 5). The hypothesis was supported among men from San Francisco, who were less likely to report and reported fewer unprotected insertive oral and anal sexual episodes (both in general and to ejaculation) with both their serostatus-unknown and HIV-negative non-main sexual partners (but no differences with their HIV-positive nonmain partners). The hypothesis was supported among men from New York, but only among their non-main partners of unknown serostatus. In particular, men from New York were less likely to report and reported fewer unprotected insertive than receptive oral and anal sexual episodes with their unknown-serostatus non-main

5 Sexual harm reduction practices of HIV-seropositive gay and bisexual men Parsons et al. S17 Table 1. Sexual behaviors with main partners: comparisons by serostatus of partner. New York (N ¼ 240) San Francisco (N ¼ 195) Variables Unknown status partner (n ¼ 50) Positive partner (n ¼ 100) Negative partner (n ¼ 90) x 2 or Kruskal Wallis test 1 Unknown status partner (n ¼ 19) Positive partner (n ¼ 94) Negative partner (n ¼ 82) x 2 or Kruskal Wallis test 1 % Any insertive oral 55% (41%, 69%) 73% (64%, 82%) 59% (48%, 69%) % (23%, 72%) 82% (75%, 90%) 72% (62%, 82%) % Unprotected insertive oral 45% (31%, 59%) 65% (56%, 75%) 49% (38%, 59%) % (23%, 72%) 76% (67%, 84%) 68% (58%, 79%) 6.08 Unprotected insertive oral a (2.16, 12.04) 8.68 b (5.70, 11.67) 5.96 a (3.43, 8.48) a (0.00, 10.53) 9.61 b (6.85, 12.36) 8.89 a,b (4.94, 12.84) 8.21 Unprotected insertive oral 0.47 a (0.00, 1.02) 2.85 b (0.72, 4.99) 1.42 a (0.00, 3.13) a,b (0.00, 1.59) 2.62 b (1.10, 4.13) 1.52 a (0.00, 3.49) to ejaculation 2 % Any receptive oral 84% (74%, 95%) 69% (60%, 78%) 83% (75%, 91%) % (66%, 102%) 75% (66%, 84%) 87% (79%, 94%) 3.98 % Unprotected receptive oral 73% (60%, 85%) 64% (54%, 74%) 73% (64%, 83%) % (59%, 99%) 71% (62%, 80%) 83% (75%, 91%) 3.68 Unprotected receptive oral (5.34, 14.70) 9.92 (6.73, 13.11) 9.76 (6.80, 12.71) (0.00, 56.95) (6.54, 15.27) (9.51, 18.36) 2.91 Unprotected receptive oral 2.18 (0.55, 3.80) 1.70 (0.55, 2.85) 2.59 (0.62, 4.55) (0.00, 50.62) 4.28 (0.32, 8.25) 6.34 (2.69, 9.99) 1.68 to ejaculation 2 % Any insertive anal 39% (25%, 53%) 61% (52%, 71%) 51% (41%, 62%) % (1%, 41%) 55% (45%, 65%) 31% (20%, 41%) % Unprotected insertive anal 20% (9%, 31%) 47% (37%, 56%) 11% (5%, 18%) % (0%, 16%) 47% (37%, 57%) 12% (5%, 19%) Unprotected insertive anal a (0.05, 2.87) 7.64 b (4.04, 11.25) 1.42 a (0.19, 2.65) a (0.00, 4.08) 6.35 b (3.73, 8.97) 1.09 a (0.15, 2.02) Unprotected insertive anal 1.10 a (0.00, 2.46) 3.82 b (1.11, 6.53) 0.64 a (0.00, 1.73) a (0.00, 4.08) 3.44 b (1.50, 5.37) 0.09 a (0.00, 0.21) to ejaculation 2 % Any receptive anal 69% (55%, 82%) 52% (42%, 62%) 59% (49%, 69%) % (33%, 82%) 54% (44%, 64%) 62% (51%, 73%) 1.17 % Unprotected receptive anal 26% (13%, 38%) 38% (28%, 48%) 24% (15%, 34%) % (23%, 72%) 45% (35%, 55%) 45% (34%, 56%) 0.03 Unprotected receptive anal (0.56, 5.95) 4.90 (2.53, 7.27) 3.78 (0.97, 6.58) (0.00, 7.43) 8.68 (3.53, 13.84) 4.33 (2.51, 6.15) 0.20 Unprotected receptive anal 2.04 (0.01, 4.07) 2.94 (0.87, 5.01) 1.62 (0.00, 3.34) (0.00, 7.25) 5.94 (0.87, 11.00) 2.98 (1.40, 4.55) 1.31 to ejaculation 2 CI, Confidence interval. 1 Analyses were conducted using Kruskal Wallis tests for the continuous variables, and chi-square tests for the dichotomous variables. Post-hoc comparisons using Mann Whitney tests (for continuous variables) revealed that means with differing superscripts (denoted with a and b ) were significantly different (P < 0.05). P < P < P < Mean number of sexual episodes in the past 90 days.

6 S18 AIDS 2005, Vol 19 (suppl 1) Table 2. Sexual behaviors with main partners: comparisons by insertive or receptive role. Variables Insertive New York (N ¼ 240) San Francisco (N ¼ 195) Receptive Wilcoxon signed rank test or McNemar test a Insertive Receptive Wilcoxon signed rank test or McNemar test a % Any oral 64% (58%, 70%) 77% (72%, 83%) % (68%, 81%) 81% (75%, 86%) 2.09 % Unprotected oral 55% (49%, 61%) 69% (63%, 75%) % (63%, 76%) 77% (71%, 83%) 2.82 Unprotected oral b 7.33 (5.48, 9.19) 9.88 (7.92, 11.84) (6.66, 11.01) (9.39, 17.50) 3.40 Unprotected oral 1.83 (0.73, 2.92) 2.13 (1.20, 3.06) (0.87, 3.07) 6.44 (2.60, 10.28) 3.57 % Any anal 53% (47%, 59%) 58% (52%, 64%) % (35%, 49%) 58% (51%, 65%) % Unprotected anal 28% (22%, 34%) 30% (24%, 36%) % (22%, 35%) 45% (38%, 52%) Unprotected anal b 4.06 (2.41, 5.71) 4.13 (2.61, 5.66) (2.26, 5.03) 6.37 (3.74, 9.00) 3.08 Unprotected anal 2.07 (0.83, 3.31) 2.26 (1.11, 3.41) (0.84, 2.80) 4.45 (1.91, 6.98) 3.11 CI, Confidence interval. a Analyses were conducted using Wilcoxon signed rank tests (for continuous variables) or McNemar tests (for dichotomous variables). P < P < b Mean number of sexual episodes in the past 90 days. partners. Few differences were found between unprotected insertive and receptive sexual behaviors with their HIV-negative non-main partners, and no differences were found with HIV-positive non-main partners. Finally, the hypothesis that HIV-positive men attempt to reduce the risk of transmission to their non-main sexual partners by selectively engaging in sexual risk behaviors without ejaculation (withdrawal) rather than with ejaculation was examined (see Table 6). The hypothesis was generally supported among both New York and San Francisco men. In particular, men from both cities reported fewer unprotected insertive and receptive oral episodes with ejaculation than without ejaculation with all partner types (HIV-positive, HIV-negative, and unknown), and reported fewer unprotected insertive anal episodes with ejaculation than without ejaculation with HIV-positive and unknown-status partners. Men in New York, however, did not demonstrate a significant difference in the number of acts of unprotected insertive anal sex with and without ejaculation with their HIVnegative non-main partners. Furthermore, no significant differences were noted in the number of unprotected receptive anal episodes that men reported with or without ejaculation, except that men from New York were more likely to report withdrawal during this behavior with their unknown-status partners. Discussion We sought to obtain a better understanding of the patterns of sexual risk behaviors and harm reduction practices among HIV-positive gay and bisexual men in two HIV epicenters in the United States. The first finding of note concerns the fact that a number of HIVpositive men did report unprotected anal sex acts with partners at risk. Across the two cities, a total of 17.6% of the men in SUMIT reported any unprotected anal sex (either insertive or receptive) with known HIVnegative non-main partners and 15.0% with a known Table 3. Sexual behaviors with main partners: comparisons of unprotected sexual behaviors with or without ejaculation. Variables With ejaculation New York (N ¼ 240) San Francisco (N ¼ 195) Without ejaculation Wilcoxon signed rank test With ejaculation Without ejaculation Wilcoxon signed rank test Unprotected 1.83 (0.73, 2.92) 5.48 (3.97, 6.99) (0.87, 3.07) 6.80 (4.88, 8.72) 6.56 insertive oral a Unprotected 2.13 (1.20, 3.06) 7.78 (6.05, 9.50) (2.60, 10.28) 7.02 (5.28, 8.76) 4.03 receptive oral a Unprotected 2.07 (0.83, 3.31) 1.99 (0.90, 3.07) (0.84, 2.80) 1.81 (0.99, 2.64) 1.23 insertive anal a Unprotected receptive anal a 2.26 (1.11, 3.41) 1.86 (0.89, 2.84) (1.91, 6.98) 1.94 (1.08, 2.80) 1.60 CI, Confidence interval. a Mean number of sexual episodes in the past 90 days. P <

7 Sexual harm reduction practices of HIV-seropositive gay and bisexual men Parsons et al. S19 Table 4. Sexual behaviors with non-main partners: comparisons by serostatus of partner. New York (N ¼ 400) San Francisco (N ¼ 440) Variables Unknown status partner (n ¼ 300) Positive partner (n ¼ 224) Negative partner (n ¼ 168) Friedman test or Cochrans Q a Unknown status partner (n ¼ 343) Positive partner (n ¼ 268) Negative partner (n ¼ 211) Friedman test or Cochrans Q a % Had non-main partner 73% 1 (69%, 77%) 54% 2 (49%, 59%) 41% 3 (36%, 46%) % 1 (72%, 80%) 60% 2 (56%, 65%) 47% 3 (43%, 52%) Non-main partners b (5.04, 7.84) (1.55, 2.44) (1.28, 2.76) (5.97, 10.92) (2.47, 3.99) (1.15, 2.26) % Any insertive oral 57% 1 (52%, 62%) 44% 2 (39%, 49%) 33% 3 (29%, 38%) % 1 (59%, 68%) 54% 2 (49%, 58%) 35% 3 (31%, 39%) % Unprotected insertive oral 53% 1 (48%, 58%) 40% 2 (35%, 45%) 29% 3 (24%, 33%) % 1 (57%, 66%) 51% 2 (47%, 56%) 33% 3 (28%, 37%) Unprotected insertive oral b (3.53, 5.72) (2.15, 3.75) (1.35, 2.54) (3.62, 7.86) (3.03, 6.56) (1.19, 2.03) Unprotected insertive (0.43, 1.19) (0.38, 1.12) (0.12, 0.65) (0.29, 3.48) (0.47, 3.34) (0.13, 0.78) oral % Any receptive oral 64% 1 (59%, 68%) 41% 2 (36%, 46%) 36% 2 (31%, 41%) % 1 (63%, 72%) 50% 2 (45%, 54%) 42% 3 (38%, 47%) % Unprotected receptive oral 60% 1 (55%, 65%) 38% 2 (33%, 42%) 33% 2 (28%, 37%) % 1 (61%, 70%) 49% 2 (43%, 53%) 41% 3 (36%, 45%) Unprotected receptive oral b (4.77, 7.63) (1.92, 3.21) (1.41, 2.77) (5.02, 9.70) (2.77, 5.05) (1.65, 4.00) Unprotected receptive (0.96, 2.25) (0.34, 0.84) (0.27, 0.94) (1.44, 3.28) (0.79, 2.31) (0.42, 2.33) oral % Any insertive anal 38% 1 (33%, 43%) 32% 2 (27%, 36%) 19% 3 (15%, 22%) % 1 (28%, 37%) 33% 1 (29%, 37%) 20% 2 (16%, 23%) % Unprotected insertive anal 19% 1 (15%, 22%) 21% 1 (17%, 25%) 7% 2 (4%, 9%) % 1 (18%, 26%) 27% 1 (23%, 31%) 10% 2 (7%, 12%) Unprotected insertive anal b (0.72, 1.48) (0.86, 1.84) (0.12, 0.63) (0.64, 1.32) (1.16, 2.29) (0.11, 0.45) Unprotected insertive (0.14, 0.37) (0.25, 0.86) ((0.01, 0.39) (0.19, 0.65) (0.37, 0.80) (0.01, 0.13) anal % Any receptive anal 42% 1 (37%, 47%) 30% 2 (26%, 35%) 22% 3 (18%, 26%) % 1 (34%, 43%) 35% 1 (30%, 39%) 26% 2 (22%, 30%) % Unprotected receptive anal 21% 1 (17%, 25%) 21% 1 (18%, 26%) 11% 2 (8%, 14%) % 1 (23%, 32%) 28% 1 (23%, 32%) 17% 2 (13%, 20%) Unprotected receptive anal b (0.90, 2.00) (0.74, 1.95) (0.29, 1.14) (1.05, 2.96) (1.31, 2.93) (0.26, 0.90) Unprotected receptive (0.34, 1.02) (0.32, 0.95) (0.14, 0.65) (0.44, 2.10) (0.69, 2.15) (0.11, 0.73) anal CI, Confidence interval. The sum of ns within each partner category exceed the N for each city because participants could have had sexual partners of varied HIV status. a Analyses were conducted using Friedman tests for the continuous variables, and Cochran s Q for the dichotomous variables. Post-hoc comparisons using Wilcoxon signed rank tests (for continuous variables) and McNemar tests (for dichotomous variables) revealed that means with differing superscripts (denoted with 1,2,3 ) were significantly different (P < 0.05). P < P < b Mean number of sexual episodes in the past 90 days.

8 S20 AIDS 2005, Vol 19 (suppl 1) Table 5. Sexual behaviors with non-main partners: comparisons by insertive or receptive role. New York (N ¼ 400) San Francisco (N ¼ 440) Variables Insertive Receptive Wilcoxon signed rank test or McNemar test a Insertive Receptive Wilcoxon signed rank test or McNemar test a With unknown-status partner % Any oral 57% (52%, 62%) 64% (59%, 68%) % (59%, 68%) 67% (63%, 72%) 5.14 % Unprotected oral 53% (48%, 58%) 60% (55%, 65%) % (57%, 66%) 66% (61%, 70%) 4.20 Unprotected oral b 4.63 (3.53, 5.72) 6.20 (4.77, 7.63) (3.62, 7.86) 7.36 (5.02, 9.70) 5.03 Unprotected oral 0.81 (0.43, 1.19) 1.60 (0.96, 2.25) (0.29, 3.48) 2.36 (1.44, 3.28) 4.40 % Any anal 38% (33%, 43%) 42% (37%, 47%) % (28%, 37%) 39% (34%, 43%) 4.93 % Unprotected anal 19% (15%, 22%) 21% (17%, 25%) % (18%, 26%) 27% (23%, 32%) 4.28 Unprotected anal b 1.10 (0.72, 1.48) 1.45 (0.90, 2.00) (0.64, 1.32) 2.01 (1.05, 2.96) 2.22 Unprotected anal 0.25 (0.14, 0.37) 0.68 (0.34, 1.02) (0.19, 0.65) 1.27 (0.44, 2.10) 3.23 With HIV-positive partner % Any oral 44% (39%, 49%) 41% (36%, 46%) % (49%, 58%) 50% (45%, 54%) 4.33 % Unprotected oral 40% (35%, 45%) 38% (33%, 42%) % (47%, 56%) 48% (43%, 53%) 2.94 Unprotected oral b 2.56 (2.15, 3.75) 2.95 (1.92, 3.21) (3.03, 6.56) 3.91 (2.77, 5.05) 1.39 Unprotected oral 0.75 (0.38, 1.12) 0.59 (0.34, 0.84) (0.47, 3.34) 1.55 (0.79, 2.31) 0.61 % Any anal 32% (27%, 36%) 30% (26%, 35%) % (29%, 37%) 35% (30%, 39%) 0.42 % Unprotected anal 21% (17%, 25%) 22% (18%, 26%) % (23%, 31%) 28% (23%, 32%) 0.09 Unprotected anal b 1.35 (0.86, 1.84) 1.34 (0.74, 1.95) (1.16, 2.29) 2.12 (1.31, 2.93) 0.21 Unprotected anal 0.56 (0.25, 0.86) 0.64 (0.32, 0.95) (0.37, 0.80) 1.42 (0.69, 2.15) 2.07 With HIV-negative partner % Any oral 33% (29%, 38%) 36% (31%, 41%) % (31%, 39%) 42% (38%, 47%) % Unprotected oral 29% (24%, 33%) 33% (28%, 37%) % (28%, 37%) 41% (36%, 45%) Unprotected oral b 1.94 (1.35, 2.54) 2.09 (1.41, 2.77) (1.19, 2.03) 2.82 (1.65, 4.00) 4.99 Unprotected oral 0.38 (0.12, 0.65) 0.60 (0.27, 0.94) (0.13, 0.78) 1.38 (0.42, 2.33) 4.83 % Any anal 19% (15%, 22%) 22% (18%, 26%) % (16%, 23%) 26% (22%, 30%) 8.08 % Unprotected anal 6% (4%, 9%) 11% (8%, 14%) % (7%, 12%) 17% (13%, 20%) Unprotected anal b 0.38 (0.12, 0.63) 0.71 (0.29, 1.14) (0.11, 0.45) 0.58 (0.26, 0.90) 3.77 Unprotected anal 0.19 (0.00, 0.39) 0.39 (0.14, 0.65) (0.01, 0.13) 0.42 (0.11, 0.73) 4.57 CI, Confidence interval. a Analyses were conducted using Wilcoxon signed rank tests (for continuous variables) or McNemar tests (for dichotomous variables). P < P < P < b Mean number of sexual episodes in the past 90 days. HIV-negative main partner. In addition, 33.6% of the men reported any unprotected anal sex with HIV statusunknown non-main partners, suggesting that the potential risk for HIV transmission among men in this sample (assuming some of these unknown-status partners are, in fact, HIV negative) is comparable to that found in other studies of HIV-positive gay and bisexual men [8 11,30]. It is important to note, however, that men in San Francisco were consistently more likely than men in New York to report having engaged in unprotected sexual activities. Overall, the data provided support for the use of serosorting, strategic positioning, and withdrawal before ejaculation as harm reduction techniques. Across the two cities, men reported limited unprotected sex with HIV-negative partners compared with other partner types, more acts of sex as the receptive partner than the insertive partner, and more acts of unprotected anal and oral sex with withdrawal, rather than with ejaculation. The data thus generally support these patterns of harm reduction, and demonstrate that HIVpositive gay and bisexual men engage in sexual behaviors that are perceived to be lower risk for the transmission of HIV. However, many of the patterns of HIV harm reduction were inconsistent across specific sexual behaviors, across partner types, as well as across the two cities. In both cities, men with HIV-positive main partners reported more unprotected sexual behaviors than men with HIV-negative or unknown-status main partners. In

9 Sexual harm reduction practices of HIV-seropositive gay and bisexual men Parsons et al. S21 Table 6. Sexual behaviors with non-main partners: comparisons of unprotected sexual behaviors with or without ejaculation. Variables With ejaculation New York (N ¼ 400) San Francisco (N ¼ 440) Without ejaculation Wilcoxon signed rank test With ejaculation Without ejaculation Wilcoxon signed rank test With unknown-status partners Unprotected insertive oral a 0.81 (0.43, 1.19) 3.82 (2.94, 4.70) (0.29, 3.48) 3.85 (3.09, 4.61) Unprotected receptive oral a 1.60 (0.96, 2.25) 4.59 (3.53, 5.65) (1.44, 3.28) 5.02 (3.10, 6.95) Unprotected insertive anal a 0.25 (0.14, 0.37) 0.82 (0.49, 1.16) (0.19, 0.65) 0.57 (0.37, 0.76) 3.24 Unprotected receptive anal a 0.68 (0.34, 1.02) 0.77 (0.47, 1.08) (0.44, 2.10) 0.74 (0.37, 1.10) 0.13 With HIV-positive partners Unprotected insertive oral a 0.75 (0.38, 1.12) 2.20 (1.52, 2.88) (0.47, 3.34) 2.89 (2.26, 3.51) 8.48 Unprotected receptive oral a 0.59 (0.34, 0.84) 1.98 (1.45, 2.51) (0.79, 2.31) 2.36 (1.80, 2.92) 6.43 Unprotected insertive anal a 0.56 (0.25, 0.86) 0.79 (0.45, 1.14) (0.37, 0.80) 1.14 (0.70, 1.58) 3.41 Unprotected receptive anal a 0.64 (0.32, 0.95) 0.71 (0.32, 1.09) (0.69, 2.15) 0.71 (0.43, 0.98) 0.13 With HIV-negative partners Unprotected insertive oral a 0.38 (0.12, 0.65) 1.56 (1.06, 2.07) (0.13, 0.78) 1.12 (0.86, 1.37) 6.98 Unprotected receptive oral a 0.60 (0.27, 0.94) 1.49 (0.97, 2.01) (0.42, 2.33) 1.43 (1.06, 1.81) 4.87 Unprotected insertive anal a 0.19 (0.00, 0.39) 0.19 (0.05, 0.33) (0.01, 0.13) 0.21 (0.09, 0.33) 4.18 Unprotected receptive anal a 0.39 (0.14, 0.65) 0.32 (0.09, 0.55) (0.11, 0.73) 0.16 (0.09, 0.22) 1.57 CI, Confidence interval. a Mean number of sexual episodes in the past 90 days. P < P < P < particular, the most risky behavior for HIV transmission, unprotected anal insertive sex with and without ejaculation, was less common with partners at risk of seroconversion, including partners of unknown status. However, these differences were not identified with regard to receptive anal sex (nor for receptive oral sex, a significantly less risky behavior). Therefore, for main partners, it appears that strategic positioning may function as a back-up strategy to serosorting in terms of reducing the likelihood of HIV transmission to partners at risk. For men in New York, however, strategic positioning was identified only with oral sex behaviors, and not for anal sex behaviors. It is unclear why HIV-positive men in San Francisco would be more likely to utilize strategic positioning with their main partners, whereas men in New York would not. It is possible that this finding emerged as a result of sampling differences, and the fact that non-representative samples of HIV-positive gay and bisexual men were enrolled. It is possible, however, that this difference is related to differential HIV risk reduction messages delivered in the two cities. It is also possible that the increased number of participants in New York who reported having a main partner is having an impact on this finding, or that the higher percentage of men with unknown-status main partners in New York resulted in these men making the assumption that their unknown-status main partner was actually HIV positive, decreasing their perceived need for strategic positioning. Finally, in terms of withdrawal before ejaculation with their main partners, men from both cities were significantly less likely to engage in oral sex with ejaculation than oral sex without ejaculation. However, there were no significant differences between the rates of anal sex with and without ejaculation. In both cities, men reported more unprotected sex with their HIV-positive non-main partners than with their HIV-negative non-main partners, indicating some efforts at harm reduction via serosorting. However, men in both cities reported more unknown-status non-main partners than HIV-positive partners, and more sexual risk with these unknown-status partners than either their HIVnegative or their HIV-positive casual partners. This was true for both anal and oral sex behaviors, regardless of sexual positioning in terms of insertive and receptive, and regardless of whether or not withdrawal was practised. There was thus support for serosorting when the partner s serostatus was known, but this does not appear to be the case for partners of unknown status. This may be because of the common assumption that gay and bisexual men believe that partners with an unknown serostatus actually have a concordant serostatus [26]. In addition, serosorting may be less common with unknown partners because of feelings of responsibility that emerge among some HIVpositive men, such that they believe that partners of unknown status who are willing to take risks must know the potential consequences and are thus making an informed decision to engage in sexual risk, which helps the HIV-positive man to distance himself from responsibility [17]. The hypothesis for strategic positioning with non-main partners was supported for men from San Francisco for both anal or oral sexual behaviors, and for both their HIVnegative and serostatus-unknown casual partners. For men in New York, the hypothesis was supported, but primarily for unknown-status partners. With known HIV-negative partners, the evidence for strategic

10 S22 AIDS 2005, Vol 19 (suppl 1) positioning was less clear; although men reported more receptive anal and oral sex to ejaculation compared with insertive sex, there was no difference in general rates of anal and oral sex in terms of taking a receptive versus an insertive role with HIV-negative casual partners. The data from SUMIT corroborate work [21,32,37] show that HIV-positive men perceive a reduced risk from taking the receptive role during sexual risk behaviors. In both cities, there was no clear evidence of strategic positioning with HIV-positive casual partners, suggesting that men have fewer concerns about re-infection with a different strain of HIV than they do concerns about transmitting HIV to a casual partner. The differences in the patterns of sexual behavior with HIV-negative and HIV-positive partners lends additional support to the conclusion of Van de Ven et al. [30] that strategic positioning is an intentional and deliberate HIV-related harm reduction practice rather than merely a reflection of sexual position preferences. The hypothesis for withdrawal as a harm reduction strategy for reducing the risk of HIV transmission to casual nonmain partners was supported, overall, for both cities in terms of anal and oral sex behaviors, and was mostly true regardless of the HIV status of the partners. However, there was not support for this hypothesis among men in New York in terms of the numberof acts of unprotected insertive anal sex with known HIV-negative partners. That is, men in New York, reported equivalent numbers of acts with and without ejaculation for this, the most risky of HIV transmission-related behaviors. It may be that when decisions have been made to engage in the most risky behavior for HIV transmission with known HIV-negative casual partners, the HIV-positive men (or their partners) do not feel that the added effort of withdrawal will significantly reduce risk. This may also help to explain why there were no significant differences in the number of acts of anal sex with and without ejaculation. The fact that men in both cities reported more unprotected sex with unknown-status casual partners than with either their HIV-positive or their HIV-negative partners is a source of concern. Other researchers have found that HIV-positive gay and bisexual men reported increased sexual risk with other HIV-positive men rather than with men of unknown status [53,54]. It is likely that men in our sample are making assumptions about the serostatus of their unknown-status partners, probably assuming that they are HIV positive. These findings may also be indicative of the nature of sexual activities with partners of unknown status, in that these sex partners may be met in more anonymous environments, such as public or commercial sex environments. Our previous work has shown that HIV-positive gay and bisexual men who report frequenting such venues for sex are more likely to report unprotected sex acts [28,29]. One of the primary problems with the use of serosorting and strategic positioning is the very real potential for the practice to be misguided. There is evidence that those in the seroconverting process are more infectious than those HIV-positive individuals who are asymptomatic [55]. As such, for these men (many of whom will not know their HIV status), strategic positioning techniques may be both misguided (because the individual who perceives that he is HIV negative is actually HIV positive), and even behaviors designed to minimize risk (e.g. withdrawal) may still result in transmission as a result of increased infectiousness. The evidence for strategic positioning suggests that messages about the differential risk of insertive versus receptive sex have been received by HIV-positive gay and bisexual men. It is unclear, however, the degree to which these men have accurate perceptions regarding the potential risks associated with unprotected receptive anal sex. For example, do these men feel that receptive anal sex is a no risk behavior in terms of HIV transmission to sexual partners, or do they more accurately perceive it to be a risk behavior, but one less risky than unprotected insertive anal sex. Prevention messages and interventions need to highlight that receptive anal sex without the use of a condom can transmit HIV, but at the same time such messages may need to be tempered for fear of HIVpositive men increasing their unprotected insertive acts. Similar concerns can be expressed regarding intervention efforts that would be designed to address withdrawal as a risk reduction practice. Like strategic positioning, withdrawal before ejaculation does not eliminate the risk of HIV transmission; however, the evidence that is currently available suggests that it can reduce the risk. HIV-positive gay and bisexual men need to understand that the potential reduced risk associated with withdrawal does not mean that there is the absence of risk. However, messages to communicate this point should take care to highlight the risk of withdrawal and receptive anal sex in ways that do not push men further away from harm reduction or cause them to give up on risk reduction in frustration. For example, Richters et al. [42] found that men who used withdrawal as a strategy were unlikely to use condoms at other times, indicating a strong motivation to avoid condoms. Intervention messages that are too strong regarding the risk of withdrawal could result in an individual who does engage in withdrawal increasing their acts of unprotected sex with ejaculation. Similarly, messages need to be tailored such that they do not have the effect of encouraging those who use condoms to believe that unprotected sex with withdrawal is acceptable. It is important to recognize that in any one sexual relationship, or even in any single sexual encounter, men may employ multiple harm reduction strategies. This paper involved summing across all sexual behaviors, thereby missing potential situations in which more than one strategy is employed. Complicating this further,

11 Sexual harm reduction practices of HIV-seropositive gay and bisexual men Parsons et al. S23 contextual variables (e.g. the use of alcohol or drugs, the location of sexual activity, the nature of the relationship, partner characteristics, the emotional condition of both parties at the time of having sex, etc.) are likely to impact the particular harm reduction strategies used in individual encounters. In our paper, we examined relationships across sexual encounters. In the real world, sexual behaviors among men are incredibly complicated and multilayered, such that a man who is high on methamphetamine in a bathhouse, having sex with an anonymous partner who is perceived to be incredibly attractive, is quite likely to employ different strategies than when he is sober, having sex in his own home with a partner he has had sex with before. There are a number of important research and prevention implications. The days in which simple HIV prevention messages are widely accepted among gay and bisexual men appear to be over. Prevention efforts should address the complexities inherent in the sexual risk behaviors and harm reduction strategies of HIV-positive men, as well as the contextual factors that come into play. Men are clearly very advanced and sophisticated with regard to the complexityof their sexual activities and prevention behaviors, and our assessment tools need to mirror this. The complexity of the behavioral strategies that men have created to minimize HIV transmission suggests that it is simply insufficient to restrict the measurement of high-risk sex as condomless sex. Timeline followback interviews and other techniques that capture the detailed aspects of sexual activity are necessary to understand more comprehensively and differentiate the relative risk of varied sexual behaviors. Finally, asrandomized controlledtrials of theeffectiveness of harm reduction strategies compared with other strategies areneitherfeasiblenorethical, wecanonlyrelyonadditional epidemiological investigations to continue to investigate harm reduction strategies. Gayand bisexual men are making decisions to use these strategies in the absence of complete data, and are empirically testing serosorting, strategic positioning, and withdrawal with their own bodies. Cohort studieswould be helpful in examining, at least on some level, the efficacy of these strategies in order to obtain prevention data that can be used by HIV-positive and HIV-negative gay andbisexualmentominimizetheir risks. Inconsiderationof the burdens inherent in cohort studies, case control comparisons using serological tests for recent seroconversion could differentiate men who seroconvert despite having employed harm reduction strategies from men who think they might have seroconverted but did not. The results presented should be examined in light of the fact that the data are both self-reported and crosssectional. As a result, causal implications should not be inferred, and longitudinal studies are needed to determine causative relationships as well as the degree to which strategic positioning may change over time. Self-reported data are subject to response bias, and it is possible that some of the HIV-positive gay and bisexual men in our sample under-reported unprotected sex behaviors with known HIV-negative partners, because of social stigma issues. The use of Audio-CASI, however, tends to have a beneficial impact on the honesty of reporting sensitive and potentially stigmatized sexual behaviors [56,57]. It is also difficult to know how to interpret the higher rates of sexual risk and less utilization of harm reduction strategies with partners of unknown status, as these partners could have been HIV negative or HIV positive. It is also important to consider the broader generalizability of the data, as it was collected from two large urban HIV epicenters in the United States. It is possible that HIVpositive gay and bisexual men from other geographical areas may demonstrate different patterns of strategic positioning and harm reduction with regard to sexual risk taking. This is further evidenced by the city differences noted between New York and San Francisco in SUMIT. Interventions aimed at HIV-positive gay and bisexual men need to recognize that such men are often utilizing harm reduction strategies, such as serosorting, strategic positioning, and withdrawal before ejaculation. Messages need to be carefully constructed so that men receive accurate information regarding the relative risk of various sexual risk practices, while at the same time making clear that lower risk does not equal low risk or no risk. Particular efforts should target men who engage in sexual risk practices with partners of unknown status, and it is likely that an increased focus on HIV testing and disclosure of one s status to sexual partners would have a minimizing effect on the number of unknown status partners. HIV-positive gay and bisexual men should be cautioned regarding making assumptions about the serostatus of their casual partners, as they may engage in inappropriate risk strategies based on inaccurate assumptions, resulting in the potential seroconversion of those partners who are, in fact, HIV negative. Acknowledgements The authors would like to thank Ron Stall and two reviewers for their helpful comments. Sponsorship: This research was funded by the Centers for Disease Control and Prevention through cooperative agreements with New Jersey City University (UR3/CCU216471, J.T. Parsons, PI) and the University of California, San Francisco (UR3/ CCU916470, C.A. Gomez, PI). References 1. Chen SY, Gibson S, Weide D, McFarland W. Unprotected anal intercourse between potentially HIV-serodiscordant men who have sex with men, San Francisco. J Acquir Immune Defic Syndr 2003; 33: