Ekstrand, Maria L.; Stall, Ron D.; Paul, Jay P.; Osmond, Dennis H.; Coates, Thomas J.
Maintenance of lower levels of HIV risk-taking within populations characterized by high prevalence rates of HIV infection is one of the most important and daunting challenges in AIDS prevention. Even a relatively small increase in sexual risk-taking over time can yield a new wave of HIV infection in sexual networks with high background prevalence rates of HIV, even though such behavioral changes might not have equally damaging consequences in a low HIV seroprevalence population. New HIV infections within heavily impacted populations are both personal tragedies and a public health disaster, as they replenish the background HIV seroprevalence rate, and maintain the conditions for recurrent waves of HIV infection.
Preventing new infections within the population of young gay men is not only important in its own right, but such work also prevents the re-establishment of a high background rate of HIV infection within this vulnerable population . Currently, rates of new HIV infections within this age group in San Francisco tend to cluster at an annual rate of approximately 2% [2,3,4], with prevalence rates of HIV infection in a household-based sample reaching 30% by the age of 30years . Similar findings have been reported from a sample of young gay men in New York City . These data indicate that in the long-term, seemingly low rates of annual seroconversion can lead to very high prevalence rates, a conclusion that is also supported by epidemiological modeling exercises .
Recent increases in rates of rectal gonorrhea have been documented among male STD clinic patients . It has been suggested that the availability of new antiretroviral and combination therapies may have led to increasing levels of sexual risk-taking among some gay men . Since these reports are based on clinical samples of gay men, it is unclear whether they reflect broader trends within this population as a whole.
In order to address some of these issues, this paper describes cohort data on sexual risk-taking from the San Francisco Young Men‚s Health Study (SFYMHS). These data, taken over 4 years, offer a measure of the extent to which rates of UAI are changing over time in a non-clinical and more diverse sample of gay/bisexual men. The goals of this paper are: (1) to examine changes in rates of UAI over time; (2) to describe the conditions under which UAI occurs; and (3) to identify factors that correlate with UAI in 1996-1997 separately for men who perceived at least some of their sex partners to be of a discordant or unknown serostatus, and for the men whose partners were presumed to be HIV seroconcordant.
Participants and procedures
The data for this study were collected as part of waves two to five of the SFYMHS, which was originally designed as a survey of HIV infection and risk behavior devised after the San Francisco Men‚s Health Study [9,10,11]. The SFYMHS cohort was recruited through multistage probability sampling of English-speaking, unmarried men aged 18-29years residing in households from the 21 census tracts in San Francisco with the highest cumulative number of AIDS cases in 1992. All 19 census tracts surveyed in the SFMHS were included in the 21 census tracts sampled for the SFYMHS. Two additional census tracts were included to improve representation of areas of residence among men under age 30years. HIV seroprevalence rates and risk-taking behavior were previously reported by Osmond, et al. .
A multi-stage sample was drawn of 6671 addresses in the 21 census tracts. Ninety-three percent (n=6186) proved to be households with residential occupants. The sampling fractions were varied within blocks from one in four to one in eight households to oversample areas with higher concentrations of young men. Each selected household was approached to count the residents and determine their eligibility for study. Eligible subjects had to be unmarried, English-speaking men between 18 and 29years of age.
This process yielded 1387 eligible young single men of whom 1076 (78%) agreed to participate in a screening interview in their home. Subsequently 227 (16%) refused participation, 39 (3%) moved after enumeration, 31 (2%) were excluded due to language barriers, and 14 (1%) were never at home when contacted. Among those interviewed between March 1992 and April 1993, 408 (38%) either identified their sexual orientation as gay/bisexual or reported engaging in sexual behavior with a male partner in the previous 12months. The final probability sample in wave 1 thus comprised 408 men. In wave two, 330 of these men returned for follow-up assessment.
At wave 2 (April-November 1993), participants were asked to provide the names of male gay/bisexual friends who might be interested in participating in the study. Those friends who took part were then asked to provide additional names. This ‚snowball‚ (or referral) sample of 620 participants who reported having had sex with another man in the previous 12months was recruited between May 1993 and January 1994. The combined sample at wave 2 thus consisted of 950 men who reported having had sex with another man in the previous 12months.
The protocol assessed a wide variety of topics related to HIV and sexual behavior, using a self-administered questionnaire as follows.
(1) Demographic characteristics. Participants reported their age, ethnicity, length of time resident in San Francisco, level of education and income.
(2) HIV status. Both actual HIV antibody status and self-reported HIV antibody status were measured. Peripheral blood samples were obtained on filter paper by a finger stick. Samples were tested for HIV antibodies by the California State Department of Health Services‚ Viral and Rickettsial Disease Laboratory in Berkeley, California, USA, using enzyme-linked immunosorbent assay (ELISA; Organon Teknika Corporation, Durham, North Carolina, USA). Positive specimens were confirmed by Western blot (Organon Teknika). Self-reported HIV antibody status was assessed from questions in the annual survey.
(3) Sexual behaviors. Participants reported the numbers of male sex partners, relationship status, number of times they had practiced unprotected receptive and insertive anal intercourse and frequency of condom use during the previous 30days and number of partners with whom they had engaged in these behaviors in the previous 12months. Questions on sex in commercial sex environments (CSE), such as sex clubs, bathhouses and bookstores, as well as in public sex environments (e.g., parks and public restrooms) were also included. Participants also rated the frequency with which they engaged in sex while under the influence of drugs or alcohol during the previous 30days. In 1996-1997, we asked participants to indicate whether they knew the HIV status of all, some, or none or their UAI partners. Those who indicated that they knew the HIV status of ‚some‚ or ‚all‚ of their partners, were asked if those partners were HIV-positive, -negative or both.
(4) Alcohol and drug use. Frequency of alcohol consumption in the previous 30days was assessed using a 7-point scale, ranging from ‚never‚ to ‚at least once-a-day‚. Quantity of use during this time period was assessed using a 6-point scale, with response options ranging from ‚one or two drinks‚ to ‚12 or more drinks‚ consumed on typical days of alcohol use. Participants were classified as frequent/heavy alcohol users if they reported at least five drinks on one occasion, at least once-a-week. Use of various drug categories (e.g., marijuana/ hashish; nitrite inhalants; cocaine; amphetamines; ethyl chloride) was assessed retrospectively for the previous 12months. Participants who had indicated some use during this time, also reported whether they used substances on a daily, weekly, monthly, or less frequent basis. In addition, a measure was constructed of total number of drugs (other than alcohol) used in the previous 12months.
(5) Reasons for UAI during the previous 12months. Participants were presented with 13 reasons for UAI and asked how frequently they had applied during the previous 12months. (using a 5-point scale, ranging from ‚Never‚ to ‚Every time‚). These reasons had been frequently encountered in our previous studies of gay male sexual risk-taking (e.g., ‚we were in love‚, ‚I was tired of safe sex‚, ‚it felt better without a condom‚).
(6) Affective and cognitive consequences of last UAI. Participants rated how applicable they found each of 18 post-UAI reactions. This list of thoughts and feelings was compiled from previously reported accounts given by gay men after having UAI. Participants scored each item on a 4-point scale from ‚Not true‚ (0) to ‚Definitely true‚ (3). Items describing negative emotional consequences of the UAI episode (e.g., ‚I blamed myself‚) were combined into a 7-item subscale, ranging from 0 to 21 points (Cronbach‚s agr;=0.91 ). Items ascribing positive meaning to the UAI event (e.g., ‚I felt closer to my partner‚) were combined into a 5-item subscale, ranging from 0 to 15 points (Cronbach‚s agr;=0.81). Two items suggesting a diminished sense of self-efficacy with respect to condom use (e.g., ‚I felt discouraged about only being able to have anal sex with condoms‚) were also combined into a scale (Cronbach‚s agr;=0.77).
(7) Attitudes toward sex and HIV. Two 5-item scales were created from factor analyses of a series of items exploring attitudes toward sexual risk-taking, which were rated from ‚Strongly disagree‚ (1) to ‚Strongly agree‚ (4). These 5-item scales were: Difficulty Avoiding Risk (e.g., ‚I find it difficult to tell a sex partner not to do something I think is risky‚, Cronbach‚s agr;=0.79); and Commitment to Safer Sex (e.g., ‚If I‚m going to have anal sex, I‚m going to use a condom‚, Cronbach‚s agr;=0.80).
We examined sexual risk-taking in three different ways: (1) prevalence rates of behaviors among the cohort over several waves, (2) fluctuations of individual patterns of UAI over time, and (3) combining reported serostatus of partners with whom one had UAI to define a variable of risk for transmission of HIV. First, we examined the annual prevalence of any UAI, unprotected insertive anal, and receptive anal intercourse, among the cohort members. Changes in the proportion of men reporting UAI across the four waves were assessed using the method of Generalized Estimating Equations repeated measures logistic model [13,14], assuming an unstructured covariance matrix (PROC GENMOD, SAS Version 6-12) . Following examination of the overall linear trend, we examined changes between specific waves of data collection, using McNemar tests of significance.
The maintenance of protected and UAI practices over time was also examined to determine the stability of these behavioral patterns for individuals. If a respondent had engaged in any unprotected receptive or insertive anal intercourse during a given year, he was defined as ‚unprotected‚ during that time. Based upon these data, an individual was classified as falling into one of three categories: (1) ‚no UAI‚, (no UAI in any wave); (2) ‚UAI 1-3 waves‚ (UAI in 1, 2, or 3 annual waves); or (3) ‚UAI every wave‚ (UAI reported in all study waves).
During the most recent year of the study period (1996-1997), additional measures were added which allowed us to examine sexual risk-taking in this cohort more closely. Participants were asked to estimate whether they knew the HIV serostatus of ‚all,‚ ‚some‚ or ‚none‚ of the men with whom they had UAI. Participants who knew the HIV antibody status of at least some of these sex partners were then asked whether they were all HIV-negative, all HIV-positive, or a combination of both. Participants who reported practicing UAI were then classified into two groups: low-risk of anal transmission and high-risk of anal transmission of HIV. The first group consisted of those who knew that all their unprotected partners were seroconcordant. The second group included those who reported UAI with a partner of unknown or discordant serostatus.
Bivariate cross-sectional analyses were performed to examine correlates of ‚high transmission risk‚ behaviors, i.e. reporting UAI with a partner of unknown or discordant HIV antibody status. This group was compared both to the ‚no risk‚ (no UAI) and the ‚low transmission risk‚ groups. T-tests of significance were performed for those comparisons that involved continuous variables, such as age, number of partners, and perceived consequences of UAI. χ2tests of significance were conducted for correlates with dichotomous response options, such as the proportions of participants who had sex in commercial sex environments and who used inhalant drugs.
Finally, two multiple logistic regression analyses were conducted comparing ‚high transmission risk‚ men to ‚low transmission risk‚ men, and ‚high transmission risk‚ men to ‚no transmission risk‚ men. All correlates that had distinguished between the subject groups at P<0.10 in the bivariate analyses were included in the multivariate models. Each logistic analysis was then rerun to refine the set of variables in the final model. Odds ratios for the scaled continuous variables, such as ‚difficulty avoiding sexual risk‚, refer to the increase in sexual risk-taking associated with a 1-point increase in the scale. However, it should be noted that these are average increases only, which may not represent an equal increase in sexual risk-taking at every increment of the scale.
This analysis uses a subset of the ‚combined sample‚ of men who participated at every wave of data collection from 1993-1994 to 1996-1997 (n=510). It should be noted that the discrepancy between the sample size of this subset and the ‚combined sample‚ was due mainly to a loss of respondents between the 1993 and 1994 waves, when previously identified HIV seropositive men were not interviewed. Although no previously identified HIV-positive men were supposed to be seen in wave 3, two such men were seen unintentionally. In addition, 29 seroconverters were also seen in wave 3, bringing the total number of HIV-positive men to 31. Bias analyses indicate that the demographic characteristics of this subset of men in the present longitudinal cohort were similar to those of 518 men who did not participate at every wave. The mean age of the men who participated in all four study waves was 26.3years in 1993, compared to 26.4years for those who were absent during one or more waves of data collection. Retention rates for members of different ethnic groups were similar, ranging from a high of 63% for African American men to a low of 50% for Latino men. None of these differences were statistically significant. Similarly, in 1993 the difference in rates of UAI was not statistically significant (38% among men who participated in all subsequent waves, versus 35% among those who were absent during one or more waves).
Comparisons of members of the original household sample versus men recruited into the referral sample show that the two samples were very similar in terms of sexual risk-taking. In the household sample, 53% of the men reported ‚no transmission risk‚, 25% reported ‚low transmission risk‚, and 23% were classified as ‚high transmission risk‚. The proportion of men in the referral sample who were similarly classified were 50%, 28%, and 23%, respectively. None of these differences were statistically significant. Although the household sample was significantly younger than the referral sample (mean ages=26.4years versus 26.0years), this difference may not be socially relevant. None of the other demographic characteristics showed significant differences between the two subsets.
Rates of unprotected anal intercourse
Table 1 shows the proportion of participants who reported engaging in either unprotected receptive, unprotected insertive, or both forms of unprotected anal sex during the four time-periods studied. In 1993-1994, 37% of the cohort participants reported having engaged in either unprotected insertive and/or receptive anal intercourse. By 1996-1997, this number had increased to 50% of the sample. In the Generalized Estimating Equations analyses, the overall linear trend on time was significant (z=5.86, P<0.0001). The MacNemar tests showed that these rate increases occurred primarily between 1995 and 1997, indicating that the increases in UAI are a recent phenomena. The table identifies the years in which the rates were statistically significantly greater than in the previous year.
Individual behavior patterns
Although the rates reported in Table 1 are cause enough for concern, our analysis of individual behavior patterns over time suggests that these rates do not adequately reflect the tendency to engage in UAI over longer periods of time. Over a 4-year period, 68% of the participants reported at least some UAI. Most men (51%) practiced UAI in one, two or three out of the four years studied, whereas 17% reported practicing some form of UAI every year. Only 32% of the participants reported no UAI in any of the years studied. Occasional UAI has thus become the modal sexual behavior pattern in this population.
High-risk of transmission
It might be argued that the high rates of UAI observed in this cohort could represent carefully negotiated sex with a seroconcordant partner. To examine this possibility, we added a series of questions to the 1996-1997 questionnaire. All participants who reported having engaged in UAI were asked whether any of it had occurred with a partner of unknown or different HIV status. Men were coded as ‚high transmission risk‚ if they reported UAI with a partner of different or unknown HIV status, as ‚low transmission risk‚ if they reported all UAI partners to be seroconcordant, and as ‚no risk‚ if they reported no UAI at all. Twenty-two percent (n=111) of the participants were classified as ‚high transmission risk‚, 26% (n=131) as ‚low transmission risk‚ and 49% (n=250) as ‚no transmission risk‚. Reports of condom breakage or slippage were not considered for this variable. Eighteen of the participants could not be classified due to missing data on one or more of these variables. In 1996-1997, about half of all UAI occurred in situations that can only be described as of very high-risk for the transmission of HIV.
Correlates of risk of transmission
The next task in our analyses was to identify potential cross-sectional bivariate correlates of such ‚high transmission risk‚ sexual behavior. Table 2 identifies those variables that distinguish between ‚high transmission risk‚ men and ‚low transmission risk‚ men and between ‚high transmission risk‚ men and ‚no transmission risk‚ men. Table 3 compares the circumstances reported by ‚low transmission risk‚ versus ‚high transmission risk‚ participants, as well as their attitudes and reactions to the UAI episodes.
Cross-sectional correlates of ‚high transmission risk‚
There were no age or ethnic differences between ‚high transmission risk‚ men and the other two subgroups in this sample. The mean age in all three groups was between 25 and 26years, and the majority of participants (77-83%) were White. Men at ‚high transmission risk‚ reported having moved to San Francisco more recently (mean=38months of residence) than the ‚low transmission risk‚ men (mean=55months of residence). Despite these differences, it should be noted that the ‚high transmission risk‚ men were typically not recent immigrants to San Francisco.
The difference between reported HIV status and serological HIV test results in identified HIV-positive respondents among the ‚high transmission risk‚ group was notable. No statistically significant differences exist between the groups in terms of self-reported HIV-positive HIV status. However, serological findings indicate that ‚high transmission risk‚ men were significantly more likely to be HIV antibody positive (12%) than the ‚low transmission risk‚ group (2%) and the ‚no transmission risk‚ group (2%). The discrepancy between serological HIV test results and self-reported knowledge of HIV status is large (12% versus 5%) among these men, whom we have defined as engaging in UAI that is ‚high risk‚ for HIV. Among the 21 HIV-positive (established by serological testing) men in the present cohort, 57% (n=12) are classified as ‚high transmission risk‚, 14% as ‚low transmission risk‚ (n=3), and 28% (n=6) as ‚no transmission risk‚. Due to the small sample size of HIV-positive men, we lack the statistical power to conduct comparisons of risk behaviors or correlates by antibody status.
As shown in Table 2, men who practiced UAI with a partner of unknown or discordant serostatus in the 1996-1997 wave also reported greater numbers of sexual partners in the previous 30days than the ‚no transmission risk‚ or ‚low transmission risk men‚ (means of 4.1, 2.0 and 2.9 partners, respectively). They also reported more frequent sex in the previous 30days than did the ‚no transmission risk‚ men (respective means of 9.9 versus 4.9). During the previous year, ‚high transmission risk‚ men reported significantly higher numbers than did ‚low transmission risk‚ and ‚no transmission risk‚ participants for male sex partners (means of 27.9, 17.6, and 10.7, respectively); male sex partners under 30years (means of 13, 6.4, and 5.9, respectively); and male sex partners in sex clubs and bath-houses, (respective means: 9.3, 3.8, and 2.4).
‚High transmission risk‚ men were more likely to report use of nitrite inhalant drugs than either of the other two groups, with 52% stating that they had tried this drug in the previous 12months, and 14% reporting use at least once-a-month. They were also more likely to report having engaged in sex under the influence of alcohol and other drugs, than did the men in the ‚no transmission risk‚ group.
From Table 3, we note that only 32% of the ‚high transmission risk‚ men reported having disclosed their HIV status to all of their anal sex partners, compared to 91% of the ‚low transmission risk‚ men. Similarly, only 39% of the ‚high transmission risk‚ participants stated that at least some of their anal sex partners had reported that they were HIV-negative, compared with 83% of the ‚low transmission risk‚ men. ‚High transmission risk‚ men were also significantly more likely to experience difficulty avoiding UAI than were ‚low transmission risk‚ men. Men at ‚low transmission risk‚ were more likely than ‚high transmission risk‚ men to say that practicing UAI had positive meanings for them, e.g. by improving their relationships.
Several reasons, given retrospectively, for practicing UAI also distinguished between the two groups. ‚High transmission risk‚ men were significantly more likely to report that there were no condoms around at the time (43% versus 24%), that they were ‚really turned on‚ (77% versus 49%), and that they had been ‚drinking or using other drugs‚ (55% versus 22%). ‚Low transmission risk‚ men, on the other hand, were significantly more likely to report that UAI occurred because they ‚were in love‚ (69% versus 32%).
Multivariate correlates of sexual risk-taking
Table 4 identifies the variables that were independently correlated with reporting having UAI in the previous year with a partner of unknown or different HIV antibody status in two multiple logistic regressions.
Three variables, i.e. greater frequency of sex, any use of nitrite inhalant drugs in the previous 12months, and difficulty avoiding sexual risks, differentiated membership in the ‚high transmission risk‚ group from the ‚no transmission risk‚ men in multivariate analyses.
The comparison of ‚high transmission risk‚ and ‚low transmission risk‚ men included additional items that were applicable only to men who reported having UAI during the previous year. In a similar finding to the first logistic analysis, ‚high transmission risk‚ men were more likely than ‚low transmission risk‚ men to report nitrite inhalant use and difficulty in avoiding sexual risk- taking. ‚High transmission risk‚ men were also more likely to report heavy/frequent use of alcohol in the previous 30days and drug use in the previous 12months. They were less likely to report mutual disclosure of HIV status than were ‚low transmission risk‚ group.
Between 1995-1997, the annual prevalence rate of UAI rose to 50% in a community-based sample of gay/bisexual men in San Francisco. Over a 4-year period, more than two-thirds of the men in this cohort reported engaging in UAI, suggesting that occasional high-risk sex has now become pervasive among gay men in San Francisco. About half of the men reporting UAI in 1996-1997 did so with partners of unknown or discordant HIV status. This latter behavioral pattern can only be described as being of very high-risk for the transmission of HIV infection, especially given the high background rates of HIV infection found among San Francisco gay men. It also provides an explanation in behavioral terms for the increasing rates of rectal gonorrhea found among homosexually-active men in San Francisco. This suggests that the previously reported increases in UAI among gay men studied in clinical samples have also occurred in the broader population.
Given the significance of these findings, we must acknowledge the limitations of these data. First, although a substantial proportion of these men were selected using a household-based sampling frame, the majority of the sample was recruited through a snowball sampling technique. Although initial levels of risk-taking were similar in the two samples, the use of a snowball sampling strategy may have biased the results in other, unknown, ways. Secondly, we only have measures that allowed us to define UAI levels that were high-risk for HIV transmission during the last wave of data collection; thus we cannot determine whether rates of UAI with partners of unknown or discordant serostatus have been increasing in tandem with rising rates of UAI. Thirdly, the time-frame and measures used in the present analyses do not permit us to draw any conclusions regarding the potential influence of HAART (highly active antiretroviral therapy) on sexual risk-taking. Future studies need to examine this possibility more directly, now that these drugs have become more widely used. Finally, it should be noted that the vast majority of the participants in this study are HIV-negative. Findings of correlates of sexual risk-taking identified in these analyses may thus not be generally applicable to HIV-positive gay men.
We further suspect that the primary measure of high-risk sex for HIV transmission used here - UAI with partners of unknown or discordant serostatus - may be an overly restrictive as a measure of actual risk for HIV transmission. A sizable proportion of HIV antibody-positive gay men in this  and other samples  have been reported as unaware of their actual HIV status. Given the discrepancy between reported HIV antibody status and serological tests results for HIV antibodies among gay/bisexual men, some proportion of those men who perceived themselves as having sex with HIV-negative men or within HIV-concordant relationships may not actually be doing so.
Men who reported practicing UAI with a partner of unknown or different HIV antibody status also reported more sexual partners than other men in this sample. Although it is possible that having a greater number of partners automatically decreases the likelihood of serostatus knowledge, this finding still reinforces the importance of including the issue of multiple partners in future prevention messages. ‚High transmission risk‚ participants were also more likely to report use of nitrite inhalant drugs and having had UAI under the influence of alcohol and other drugs.
Men who reported having had UAI with a partner of unknown or discordant HIV antibody status were less likely to disclose their HIV status than were other men. They were more likely than ‚low transmission risk‚ men to agree with the items suggesting difficulties controlling sexual risk impulses with partners and greater distress/concern following UAI. These findings are consistent with previously reported correlates of high-risk sexual behavior [18,19]. Finally, it should be noted that although ‚high transmission risk‚ men were more likely to have had sex in sex clubs and bath-houses, we have no information on whether the sex that occurred there was unsafe. Current AIDS prevention efforts may not be effectively incorporating these variables into intervention activities or may not be reaching gay men who are characterized by these variables, or gay men frequenting commercial public sex environments. If so, this should be regarded as an ongoing failure of current AIDS prevention efforts.
We will soon be entering the third decade since recognizing the worldwide AIDS pandemic. During this period, gay/bisexual men have had to curtail highly valued sexual behaviors as a means of both individual and community survival. The initial risk reductions measured among gay men are among the most profound ever measured in response to a public health education campaign . Nonetheless, it is apparent that many gay men have found initial risk reduction easier than long-term risk reduction maintenance.
Clearly there is a need for proven models of AIDS prevention that support maintenance of long-term sexual risk reductions. The levels of sexual activity reported by the men in this sample, and in all other AIDS behavioral studies among gay men, strongly suggest that sexual expression is highly valued within this population. Accordingly, AIDS prevention models designed for gay men must begin with a recognition of the value of sexual expression in these men‚s lives, and they must be designed to respond to the variety of needs found within the gay male community in maintaining sexual safety. These data could be taken to suggest that some men are indeed attempting to seek out seroconcordant partners with whom they feel safer engaging in UAI, especially within relationships. However, given the discrepancies between serological data and self-presentation of HIV status, it appears that future campaigns need to send clear messages regarding the use of testing and disclosure as part of their risk reduction strategies.
Interventions designed to maintain sexual safety should incorporate characteristics found to predict effectiveness in meta-analyses of the AIDS intervention literature . These should include giving careful attention to the role of substance use, especially alcohol and nitrite inhalants, in conjunction with sexual activity among gay men. Prevention programs may also need to make use of a range of intervention modalities, specifically those designed to reach the individual, the sub-group, and the population at large. Recent research suggests that interventions may be more likely to reach their target audience if the program is designed to meet other perceived needs in addition to AIDS prevention . This may include incorporating AIDS prevention into fun social activities that are likely to be attended by a variety of gay men. We should, however, not forget that these are men in a large urban center, where the intervention activities have to compete with a wide variety of other social activities. Interventions that are designed to help gay men build strong friendship networks and connections within a community of choice may be especially effective in this and other large urban centers. Finally, it is also important to recognize that ‚declaring victory and leaving the field‚ in the fight against AIDS among gay men  is antithetical to supporting long-term behavioral risk reductions. This may be especially true in the era of protease inhibitors, with their impact on the well-being of many HIV-positive men and popular perceptions of the dangers of HIV disease. These beliefs contrast with the real dangers of the emergence and spread of resistant HIV strains. Unless we can find ways to help gay men devise practical, achievable, long-term HIV prevention strategies, ongoing HIV infection will remain a permanent problem among gay male populations.
This research project is made possible by the continued willingness of gay and bisexual men to reveal intimate details of their lives in the hope that it may help to bring the AIDS epidemic to an earlier end. We dedicate this paper to the memory of Aryae Levy, a joyful man who in living and in dying taught his friends a lot about courage. The authors also gratefully acknowledge Dr. Edwin Bryant, whose careful editing greatly improved the clarity of the manuscript.
1. De Wit, JB. The epidemic of HIV among young homosexual men. AIDS 1996, 10 (suppl 3):S21-S25.
2. Osmond DH, Page K, Wiley J, et al. HIV infection in homosexual and bisexual men 18 to 29years of age: the San Francisco Young Men‚s Health Study. Am J Publ Health, 1994, 12:1933-1937.
3. Dean L, Meyer, I. HIV prevalence and sexual behavior in a cohort of New York City gay men (aged 18-24). J Acqu Immune Defic Syndrome 1995, 8:208-211.
4. Osmond DH, Charlesbois E, Page-Shafer K, et al. HIV seroconversion in the San Francisco Young Men‚s Health Study: 1993-1996. XI International Conference on AIDS. Vancouver, July 1996 [abstract C562].
5. McFarland, W, Kellogg, T, Dilley, J, Katz, M. Estimation of human immunodeficiency virus (HIV) seroincidence among repeat anonymous testers in San Francisco. Am J Epidemiol 1997, 146:662-664.
6. Hoover, D, Muñoz A, Carey V, et al. Estimating the 1978-1990 and future spread of human immunodeficienvy virus type 1 in subgroups of homosexual men. Am J Epidemiol 1991, 134:1190-1205.
7. US Department of Health and Human Services. Gonorrhea among men who have sex with men - selected sexually transmitted disease clinics, 1993-1996. Morbid Mortal Weekly Rep 1997, 46:38.
8. Dilley JW, Woods WJ, McFarland W. Are advances in treatment changing views about high-risk sex? [letter]. New Engl J Med 1997, 337:501-502.
9. Ekstrand ML, Coates, TJ. Maintenance of safer sexual behaviors and predictors of risky sex: the San Francisco Men‚s Health Study. Am J Publ Health, 1990, 80:973-977.
10. Winkelstein W, Lyman D, Padian N, et al. Sexual practices and the risk of infection with the human immunodeficiency virus. J Am Med Assoc 1987, 257:321-325.
11. Piazza T. Sampling methods and wave 1 field results of the San Francisco Men‚s Health Study. Report to the National Institutes of Allergy and Infectious Disease. April 1986.
12. Cronbach, LJ. (1951). Coefficient alpha and the internal structure of tests. Psychometrika 16:297-334.
13. Liang K-Y, Zeger SL. (1986). Longitudinal data analysis using generalized linear models. Biometrika 73:13-22.
14. Zeger SL, Liang K-Y. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986, 42:121-130.
15. SAS Institute. SAS/STAT software: Changes and enhancements through Release 6-12. Cary, North Carolina: Author; 1997.
16. Hays RB, Paul J, Ekstrand M, Kegeles SM, Stall R, Coates TJ. Actual versus perceived HIV status, sexual behaviors and predictors of unprotected sex among young gay and bisexual men who identify as HIV-negative, HIV-positive and Untested. AIDS 1997, 11:1495-1502.
17. Lemp G, Hirozawa A, Givertz D, et al. Seroprevalence of HIV and risk behaviors among young homosexual and bisexual men; the San Francisco/Berkeley Young Men‚s Survey. J Am Med Assoc 1994, 272:449-454.
18. Ekstrand ML. Safer sex maintenance among gay men: are we making any progress? AIDS 1992, 6:875-877.
19. Stall, RD, Ekstrand, ML, McKusick L, Pollack L, Coates TJ. Relapse from safer sex: the next challenge for AIDS prevention efforts. J Acq Immune Defic Syndrome, 1990, 3:1181-1187.
20. Stall R, Coates TJ, Hoff C. Behavioral risk reduction for HIV infection among gay and bisexual men: a review of results from the United States. Am Psychologist 1988, 43:878-885.
21. Choi K, Coates T. The prevention of HIV infection. AIDS 1994, 8:1371-1389.
22. Kegeles SM, Hays RB, Coates TJ. The Empowerment Project: a community-level HIV prevention intervention for young gay men. Am J Publ Health 1996, 86:1129-1136.
23. Stall, R. How to lose the fight against AIDS among gay men. BMJ 1994, 309: 685-686.
© 1999 Lippincott Williams & Wilkins, Inc.