Prevalence of HIV Infection and Predictors of High-Transmission Sexual Risk Behaviors Among Men Who Have Sex With Men

Abstract

Objectives. We sought to determine the prevalence of HIV and novel cofactors of high-transmission-risk behavior in a probability sample of men who have sex with men (MSM).

Methods. We performed a cross-sectional telephone survey of 1976 adult MSM in San Francisco.

Results. We found an HIV prevalence of 25.2%. Predictors of unprotected insertive anal intercourse with a serodiscordant (not having the same HIV/AIDS serostatus) partner among HIV-infected men included use of Viagra and a greater number of partners in the past 12 months. Unprotected receptive anal intercourse with a serodiscordant partner among men not known to be HIV infected was independently associated with having lived in San Francisco for less than 1 year, use of crystal methamphetamine and amyl nitrites, a greater number of partners, and agreement with the statement, “You are less careful about being safe with sex or drugs than you were several years ago because there are better treatments for HIV now.”

Conclusions. Strategies to prevent HIV for urban MSM should focus on new predictors of HIV transmission.

Twenty-five years after the start of the HIV epidemic in the United States, HIV infection rates remain high and the majority of newly diagnosed HIV infections in 2004 occurred among men who have sex with men (MSM).1 Several recent studies of MSM living in urban areas have reported increased levels of unsafe sexual behaviors and HIV infection2^–10 and have identified novel cofactors likely to be associated with increased HIV transmission, including use of methamphetamines11^–14 and Viagra,15 meeting sexual partners on the Internet,16^–18 and treatment optimism (feeling less concern about acquiring HIV because of medications that can reduce HIV-related morbidity and mortality).19^–22 However, these studies have not used probability designs, resulting in uncertainty in how to incorporate this information into programs to reduce HIV transmission. Therefore, we conducted a household telephone probability survey between June 2002 and January 2003 of 1976 MSM living in San Francisco, Calif. We collected information on sexual and drug-using behaviors and tested men for HIV, gonorrhea, and chlamydia infections.

METHODS

Survey Methods

The MSM were sampled with a random-digit-dial telephone survey of households in San Francisco. The sampling frame consisted of 50 telephone exchanges from a set of 198 exchanges that covered 13 zip codes where an estimated majority of MSM resided in 1996.23 An adaptive strategy was used to increase the efficiency of sampling.24

Once a telephone number was found to correspond to a household, the household was screened to identify eligible subjects. A household was eligible for sampling only once. Eligible households must have been in San Francisco and had at least 1 adult male who self-identified as homosexual or bisexual or had sex with another man at least once since age 14. Previously validated methods to increase the likelihood that a respondent would acknowledge same-gender sexual behaviors were employed.23^,25 The respondent was asked how many eligible MSM resided in the household. If there was only 1 eligible man in the household, he was selected. If there were 2 eligible MSM, then the computer randomly selected 1 of the men to participate. If more than 2 men were eligible, then the respondent was asked to identify the housemate with the most recent birthday (excluding himself) and this man was asked to participate. If the respondent did not know the roommates’ birthdays, then the respondent was asked to choose any 1 of the eligible roommates (excluding himself).

Once an eligible respondent was selected, the purpose and procedures of the study were explained to the respondent and verbal consent was obtained. Interviews were conducted in English and Spanish with computer-assisted telephone interview technology by the Westat Corporation (Rockville, Md). Respondents were provided with $25 as compensation for the approximately 45-minute interview. At the end of the interview, respondents were asked to participate in testing for gonorrhea and chlamydia. Men who did not self-report being HIV infected were also asked to participate in HIV testing. All testing was done with urine-based tests. Those who consented to testing were mailed a home specimen collection kit along with instructions for obtaining and returning the specimen by mail to the local public health laboratory for testing. Participants in the testing portion of the study were provided with a unique study number and a telephone number to call to receive the results of their tests. Respondents who returned the test kits were compensated $30. Respondents who did not call for their test results within 1 month of submitting their specimen were sent reminder letters with $5 cash enclosured as an incentive to call for test results. Treatments for gonorrhea and chlamydia infections were mailed to participants who tested positive for these agents and who called for test results. The HIV-infected participants were provided with referrals for HIV-related care when they called for their test results.

Interview Measures

The interview collected demographic and socioeconomic data, HIV-related sexual and drug-using risk behaviors data, history of HIV and other sexually transmitted diseases, and use of HIV-related health services (for those who self-reported being HIV infected). Participants were asked if they agreed with the statements, “HIV-positive men on HAART [highly active antiretroviral therapy] or combination drug therapy are less infectious than HIV-positive men not on such drugs” and “You are less careful about being safe with sex or drugs than you were several years ago because there are better treatments for HIV now.”

Laboratory Methods

Urine was tested for HIV antibody with the Calypte HIV-1 urine enzyme-linked immunoassay (Calypte Biomedical Corporation, Alameda, Calif) and confirmed with the Cambridge Biotech HIV-1 urine Western blot kit (Calypte Biomedical Corporation, Alameda, Calif). Testing for gonorrhea and chlamydia was done using the Gen-Probe APTIMA combo 2 assay (Gen-Probe Inc, San Diego, Calif).

Statistical Methods

Comparisons between HIV infection and demographic and risk characteristics were made using the Wilcoxon rank sum test for differences in medians and the χ 2 test for differences in proportions. To identify the cofactors independently associated with the behaviors likely to lead to HIV transmission, we used multiple logistic regression models in which we entered all variables that were associated with the outcome in the bivariate analysis at a P value less than 0.1. For HIV-infected men, HIV-transmission–risk behaviors were defined as unprotected insertive anal intercourse with serodiscordant (not having the same HIV/AIDS serostatus), non-primary sexual partners. For non–HIV-infected men, HIV-transmission–risk behaviors were defined as unprotected receptive anal intercourse with serodiscordant, nonprimary partners. Because we were assessing risk behaviors that were based upon the participants’ known serostatus, we used self-report of HIV infection for analysis of high-transmission-risk behaviors. The differences in point estimates and 95% confidence intervals (CI) were calculated with weights that were constructed from the sampling probabilities and the nonresponse proportions and standardized to the size of the completed sample. Sampling weights were developed by the Westat Corporation (Rockville, Md). Crude analyses were conducted with SAS version 8.2 (SAS Institute Inc, Cary, NC) and weighted analyses were conducted with Stata SE 8.2 (Stata Corp, College Station, Tex).

RESULTS

A total of 733787 telephone numbers were called to identify 15272 households, of which 2676 held eligible MSM. Of the eligible MSM, 1976 (74%) participated in the interview.

Ninety-four percent (n = 1862) of the participating MSM had been tested for HIV in the past, and 492 of these reported that they were HIV infected. Of the remaining 1484 respondents, 1049 (71%) completed the urine-based HIV and sexually transmitted disease testing, and of the 492 respondents who reported being HIV infected, 347 (71%) completed the sexually transmitted disease testing. Urine-based testing identified 8 previously unknown HIV infections, 1 case of gonorrhea, and 8 cases of chlamydia. None of the individuals with these infections had more than 1 infection identified.

The men ranged in age from 18 to 92 years with a median age of 42 years (interquartile range, 35 to 52 years). The sample was predominately White (75%) and well educated (87% had completed at least some college; Table 1 ▶). More than half of the men had lived in San Francisco for more than 10 years.

TABLE 1—

Sample Sociodemographic Characteristics and HIV Prevalence Among MSM: San Francisco, 2002–2003

Characteristic	Total	Crude Prevalence, No. (%)	Pb	Weighted Prevalence,a % (95% CI)	P
Total	1976	500 (25.3)		25.2 (23.3, 27.2)
Age, y			< .001		< .001
18–24	52	0		0
25–34	375	50 (13.3)		13.6 (10.3, 17.9)
35–44	706	209 (29.6)		30.0 (26.5, 33.7)
45–54	485	168 (34.6)		34.1 (29.8, 38.7)
≥ 55	358	73 (20.4)		20.5 (16.5, 25.2)
Race/ethnicity			.019		.036
White	1490	370 (24.8)		24.7 (22.5, 27.1)
African American	63	24 (38.1)		40.1 (28.2, 53.4)
Asian/Pacific Islander	67	11 (16.4)		17.1 (9.5, 28.9)
Latino	193	60 (31.1)		29.3 (23.0, 36.4)
Mixed race	53	13 (24.5)		21.6 (12.4, 35.1)
Other	120	22 (20.0)		21.9 (14.7, 31.4)
Education			< .001		< .001
Some high school	44	18 (40.9)		40.9 (26.7, 56.9)
High-school graduate	180	55 (30.6)		31.8 (24.9, 39.5)
Some college	486	175 (36.0)		35.6 (31.3, 40.3)
College graduate	720	147 (20.4)		20.1 (17.2, 23.4)
Graduate degree	508	95 (18.7)		18.0 (14.8, 21.7)
No. years living in San Francisco			< .001		< .001
< 1	60	7 (11.7)		13.5 (6.4, 26.2)
1–5	368	52 (14.1)		14.1 (10.6, 18.5)
6–10	421	89 (21.1)		20.9 (17.1, 25.4)
> 10	1095	342 (31.2)		31.4 (28.6, 34.4)
Personal annual income, $			< .001		< .001
< 24 001	470	183 (38.9)		38.6 (34.1, 43.3)
24 001–48 000	524	123 (23.5)		22.0 (18.5, 25.9)
48 001–72 000	389	82 (21.1)		21.1 (17.1, 25.8)
72 001–96 000	225	47 (20.9)		21.4 (16.2, 27.7)
≥ 96 001	299	48 (16.1)		15.9 (12.0, 20.9)

Notes. MSM = men who have sex with men; CI = confidence intervals.

^aPrevalence based upon self-report and results from HIV testing in this study.

^bχ² test for associations between all variable categories.

Weighted HIV prevalence (based upon both self-report of HIV infection and results from HIV testing in this study) was 25.2% (95% CI = 23.3, 27.2) and differed little from the crude prevalence (Table 1 ▶). Prevalence of HIV was highest among men aged 45 to 54 years. There were no HIV infections among men aged younger than 25 years. African Americans had the highest HIV prevalence (40%), whereas prevalence was lowest among Asians and Pacific Islanders (17%). Prevalence of HIV was higher among those with less education and lower income levels. Prevalence increased with the number of years the individual had lived in San Francisco.

Eighty-five percent of the men were sexually active in the previous 12 months, and 81% reported having sex with a man in that time period (Table 2 ▶). Unprotected anal intercourse with a partner of the same serostatus was more frequent among men with primary partners than among men with nonprimary partners, and unprotected anal intercourse with serodiscordant partners was more frequent among men with nonprimary partners than among men with primary partners. HIV-infected persons (based upon self-report of HIV infection as well as results from testing in this study) were more likely to report high-risk sexual behaviors, diagnosis of a sexually transmitted disease, and use of Viagra in the past 12 months than were non–HIV-infected men.

TABLE 2—

Prevalence of Sexual Risk Behaviors and Drug Use Among HIV-Infected and Non–HIV-infected MSM: San Francisco, 2002–2003

Sexual Risk Behaviors	Percentagea Reporting Behavior	Percentage HIV-postiveb (n = 500)	Percentage HIV-negative (n = 1476)	Bivariate P
Sexual behaviorsc
Sexually active	85.1	87.7	84.2	.065
Sexual intercourse with a man	81.0	87.2	78.9	.001
Primary male partner	44.5	46.3	43.9	.369
Nonprimary male partner	62.4	71.7	59.2	< .001
Sexual intercourse with a woman	7.8	2.3	9.6	< .001
Sexual intercourse with a partner from Internet	26.9	30.9	25.6	.030
No. male partners, median		5	3	< .001
Received money or drugs for sex	2.3	3.8	1.8	.011
Gave money or drugs for sex	5.1	6.2	4.7	.206
Sexual behaviors with primary male partners
Unprotected receptive anal intercourse	17.2	19.7	16.4	.114
Unprotected receptive anal intercourse with partner of unknown or different HIV serostatus	3.1	6.1	2.1	< .001
Unprotected receptive anal intercourse with partner of same HIV serostatus	14.2	14.0	14.2	.900
Unprotected insertive anal intercourse	18.9	18.0	19.1	.607
Unprotected insertive anal intercourse with partner of unknown or different HIV serostatus	3.5	3.1	3.6	.616
Unprotected insertive anal intercourse with partner of same HIV serostatus	15.6	14.8	15.9	.551
Sexual behaviors with nonprimary male partners
Unprotected receptive anal intercourse	14.0	30.5	8.4	< .001
Unprotected receptive anal intercourse with partner of unknown or different HIV serostatus	8.7	18.7	5.3	< .001
Unprotected receptive anal intercourse with partner of same HIV serostatus	5.5	11.6	3.3	< .001
Unprotected insertive anal intercourse	18.9	31.5	14.6	< .001
Unprotected insertive anal intercourse with partner of unknown or different HIV serostatus	12.2	16.5	10.8	.002
Unprotected insertive anal intercourse with partner of same HIV serostatus	6.9	15.2	4.1	< .001
Diagnosed with STD	10.9	17.9	8.6	< .001
Drug usec
Used Viagra	27.8	44.2	22.4	< .001
Amyl nitrite use				< .001
Did not use at all	74.5	63.4	78.4
Used ≤ 3 days per month	21.4	29.1	18.7
Used weekly or daily	4.1	7.4	3.0
Crystal methamphetamine use				< .001
Did not use at all	83.2	74.0	86.4
Used ≤ 3 days per month	13.5	18.9	11.6
Used weekly or daily	3.3	7.1	2.0
Cocaine use				.547
Did not use at all	85.2	83.5	85.8
Used ≤ 3 days per month	14.1	15.7	13.6
Used weekly or daily	0.7	0.8	0.6
Club drugd use				.725
Did not use at all	79.4	78.7	79.7
Used ≤ 3 days per month	20.0	20.5	19.9
Used weekly or daily	0.5	0.8	0.5

Notes. MSM = men who have sex with men; STD = sexually transmitted disease.

^aPercentages are weighted for survey design.

^bHIV status determined on the basis of self-report and results from HIV testing in this study.

^cIn past 12 months.

^dSuch as ketamine hydrochloride (K or special K), methylenedioxymethamphetamine (ecstasy), gamma butyrolactone (G or GBH), or gamma hydroxybutyrate (G or GBH).

Men who were infected with HIV were more likely to use amyl nitrites (poppers) and crystal methamphetamine than were non–HIV-infected men (Table 2 ▶). Amyl nitrites were the most frequently used illicit drug among the men who were HIV infected; 36.5% used amyl nitrites in the past 12 months, and 7.4% reported using then on a weekly or daily basis. Crystal methamphetamine was used by 26% of HIV-infected men compared with 13.6% of non–HIV-infected men, and 7% of the infected men reported daily or weekly crystal methamphetamine use compared with 2% of the uninfected men. Although substantial proportions of the men reported using cocaine and “club drugs” (such as ketamine), the proportion of men using these drugs did not differ by HIV serostatus.

Of those men who knew that they were HIV infected at the time of the survey (those who tested HIV positive during this study were considered HIV negative or unknown for this analysis), 16.5% reported unprotected insertive anal intercourse with a serodiscordant partner in the past 12 months. This behavior was associated with younger age, having lived in San Francisco for fewer years, use of illicit drugs and Viagra, having met sexual partners on the Internet, having received money or drugs in exchange for sex, a greater number of partners, and agreeing with statements that indicated that men who are HIV infected who are using highly active antiretroviral therapy (HAART) are less infectious than men not using such therapy and that the respondent is less careful with sex or drugs now because of improved therapy for HIV infection (Table 3 ▶).

TABLE 3—

Unprotected Insertive Anal Intercourse (UIAI) Among HIV-Positive MSM With Nonprimary Partners Whose HIV Serostatus Was Negative or Unknown (n = 492): San Francisco, 2002–2003

Characteristic	% Reporting UIAI (n = 81)a,b	Adjusted OR (95% CI)c	Bivariate P
Age, y		0.93 (0.9, 1.0)d	< .001
18–24	NA
25–34	29.7
35–44	21.7
45–54	11.8
≥ 55	2.8
Race/ethnicity			.699
White	16.9
African American	16.2
Asian/Pacific Islander	0
Latino	18.7
Mixed race	7.4
Other	17.6
Education			.460
Some high school	6.1
High school graduate	17.0
Some college	14.0
College graduate	20.6
Graduate degree	18.8
No. years living in San Francisco			.007
< 1	53.2	2.48 (0.6, 11.1)
1–5	27.7	1.93 (0.7, 5.3)
6–10	15.3	0.58 (0.2, 156)
> 10	14.0	Referent
Personal annual income, $			.286
< 24 001	19.8
24 001–48 000	16.8
48 001–72 000	9.6
72 001–96 000	20.8
≥ 96 001	11.8
Risk behaviorse
Crystal methamphetamine usef	29.9	0.89 (0.4, 1.9)	< .001
Amyl nitrite (popper) usef	28.1	1.22 (0.6, 2.5)	< .001
Club drug usef,g	33.6	1.14 (0.5, 2.5)	< .001
Cocaine usef	24.1	0.99 (0.5, 2.1)	.079
Viagra usef	27.5	2.34 (1.1, 5.0)	< .001
Sexual intercourse with partner from Internetf	29.3	0.82 (0.4, 1.6)	< .001
Received money/drugs for sexf	42.6	2.07 (0.7, 6.0)	.004
Gave money/drugs for sexf	28.3		.101
No. of male sexual partners in past 12 months			< .001
1	14.0	0.04 (0.01, 0.3)
2–5	62.8	0.14 (0.05, 0.4)
≥ 6	30.3	Referent
HIV transmission attitudes
Agrees with the statement “HIV-positive men on HAART or combination drug therapy are less infectious than HIV-positive men not on such drugs.” f	25.1	1.77 (0.9, 3.4)	.003
Agrees with the statement “You are less careful about being safe with sex or drugs than you were several years ago because there are better treatments for HIV now.”f	26.5	1.28 (0.6, 2.7)	.005
Medical care for HIV
Has received HIV-specific medical care	16.4		.776
Receiving HAART	16.2		.762
Duration known to be HIV postitive, y
< 1	32.4
1–2	26.5
3–4	21.6
5–6	17.1		.316
7–8	13.6
9–10	21.2
> 10	14.1

Notes. MSM = men who have sex with men; OR = odds ratio; CI = confidence intervals; NA = not applicable; HAART = highly active antiretroviral therapy.

^aOnly men who self-reported being HIV infected were included.

^bResults were weighted for survey design.

^cAll variables found to be statistically significant (P < .1) in bivariate analysis were entered into the logistic regression model.

^dOdds of infection per year of age (with 18 years as the reference).

^eRisk behaviors in the past 12 months.

^fReference group was not having participated in the risk behavior or not agreeing with the statements.

^gSuch as ketamine hydrochloride (K or special K), methylenedioxymethamphetamine (ecstasy), gamma butyrolactone (G or GBH), or gamma hydroxybutyrate (G or GBH).

In multivariate analysis, independent predictors of unprotected insertive intercourse with a serodiscordant partner among the HIV-infected men included Viagra use and having a greater number of partners (Table 3 ▶). Agreeing that HAART decreases infectivity was of borderline significance (odds ratio=1.77; 95% CI=0.9, 3.4; P=.09), but use of HAART itself was not predictive of this risk behavior.

Unprotected receptive anal intercourse with a serodiscordant partner occurred among 5.4% of the men who reported being HIV negative or having an unknown serostatus. This behavior was associated with younger age, fewer years living in San Francisco, use of illicit drugs and Viagra, having met their sex partner on the Internet, having received money or drugs in exchange for sex, a greater number of partners, and agreeing with the statement that they are less careful about being safe with sex or drugs because of the availability of HAART (Table 4 ▶).

TABLE 4—

Unprotected Receptive Anal Intercourse (URAI) Among HIV-Negative^a MSM With Nonprimary Partners Whose HIV Serostatus Was Positive or Unknown (n = 1485),^b San Francisco, 2002–2003

Characteristic	% Reporting URAI (n = 72)a,b	Adjusted OR (95% CI)c	Bivariate P
Age group, y		1.0 (0.9, 1.0)d	.020
18–24	6.8
25–34	6.2
35–44	7.4
45–54	4.7
≥ 55	1.0
Race/ethnicity			.632
White	5.1
African American	6.9
Asian/Pacific Islander	7.2
Latino	5.5
Mixed race	0
Other	7.9
Education			.799
Some high school	7.2
High school graduate	4.4
Some college	6.2
College graduate	5.9
Graduate degree	4.3
No. years living in San Francisco			.001
< 1	15.8	5.45 (1.6, 19.1)
1–5	6.0	1.37 (0.5, 3.4)
6–10	7.5	1.34 (0.6, 3.1)
> 10	3.5	Referent
Personal annual income, $			.321
< 24 001	7.7
24 001–48 000	6.0
48 001–72 000	4.6
72 001–96 000	3.7
≥ 96 001	4.1
Risk behaviorse
Crystal methamphetamine usef	36.60	2.76 (1.3, 5.7)	< .001
Amyl nitrite (popper) usef	16.0	2.63 (1.3, 5.16)	< .001
Club drug usef,g	11.0	1.03 (0.5, 2.1)	< .001
Cocaine usef	10.6	1.12 (0.5, 2.5)	.002
Viagra usef	8.9	0.84 (0.4, 1.6)	.003
Sexual intercourse with partner from Internetf	12.1	1.07 (0.6, 1.9)	< .001
Received money or drugs for sexf	18.2	1.21 (0.4, 3.8)	.007
Gave money or drugs for sexf	7.6		.418
No. of male sexual partners in past 12 months			< .001
1	0.4	0.05 (0.01, 0.4)
2–5	2.6	0.28 (0.1, 0.6)
≥ 6	13.6	Referent
HIV transmission attitudes
Agrees with the statement “HIV-positive men on HAART or combination drug therapy are less infectious than HIV-positive men not on such drugs.”f	5.6		.843
Agrees with the statement “You are less careful about being safe with sex or drugs than you were several years ago because there are better treatments for HIV now.”f	16.3	3.39 (1.8, 6.3)	< .001

Notes. MSM = men who have sex with men; OR = odds ratio; CI = confidence intervals; HAART = highly active antiretroviral therapy.

^aOnly men who self-reported being non–HIV-infected were included.

^bResults were weighted for survey design

^cAll variables found to be statistically significant (P < .1) in bivariate analysis were entered into the logistic regression model.

^dOdds of infection per year of age (with 18 years as the reference).

^eRisk behaviors in the past 12 months.

^fReference group was not having participated in the risk behavior or not agreeing with the statements.

^gSuch as ketamine hydrochloride (K or special K), methylenedioxymethamphetamine (ecstasy), gamma butyrolactone (G or GBH), or gamma hydroxybutyrate (G or GBH).

After we controlled for confounding, independent predictors that were statistically significant of unprotected receptive anal intercourse with a serodiscordant partner among HIV-negative men included living in San Francisco for less than 1 year, use of crystal methamphetamine and amyl nitrites, a greater number of partners, and agreeing with the statement of being less careful with sexual intercourse and drug use now than several years ago because of the availability of more effective therapies for HIV infection (Table 4 ▶).

DISCUSSION

The findings from this population-based study—a high prevalence of HIV infection and evidence of high-transmission–risk behaviors among both infected and uninfected MSM—demonstrate that conditions for continued HIV transmission persist among MSM in San Francisco. Our prevalence rate of 25% is similar to the prevalence among MSM in San Francisco surveyed 2 years later (24%).26

Prevalence of HIV was higher among African American and Latino men compared with White men. Increased prevalence among African American and Latino men was found in a large survey of young MSM conducted in 7 large metropolitan areas of the United States in the mid- to late 1990s27 and in a more recent multicity study.24 Rates of AIDS nationally are also higher among African American and Latino men; for example, in 2004, 50% of new HIV/AIDS diagnoses occurred among African Americans.1 These findings are consistent with a shift in the HIV epidemic to minority populations and highlight the need for culturally relevant and effective prevention and care programs.

We found evidence of high-transmission sexual risk behavior especially among the HIV-positive MSM, with lower rates among the HIV-negative men. Unprotected anal intercourse between serodiscordant partners has been occurring despite specific prevention efforts designed to encourage men to disclose their HIV status to partners28 and to use condoms during anal intercourse.

Crystal methamphetamine use was high overall and was independently and strongly predictive of high-transmission sexual risk behavior among the non–HIV-infected men. Although crystal methamphetamine use was significantly associated with unprotected insertive anal intercourse among HIV-infected men with their uninfected or unknown serostatus partners in the bivariate analysis, it was not retained in the final multivariate model. One reason that crystal methamphetamine may not have been retained in the final model is that there was significant colinearity between methamphetamine use and club drugs (r = 0.47), amyl nitrites (r = 0.39), and cocaine (r = 0.29) among men who were HIV infected.

Although on the basis of these data we cannot determine which illicit drug is actually resulting in high-transmission-risk sexual intercourse, we believe that it is likely that it is crystal methamphetamine for several reasons. First, we found a significant association between crystal methamphetamine use and high-transmission–risk sexual intercourse in the much larger group of non–HIV-infected men. Second, crystal methamphetamine use has been associated with high-risk sex in other studies of MSM.12^–14 In addition, we conducted an additional logistic regression analysis in which we excluded all drugs except crystal methamphetamine from the model, and we found that use of methamphetamine was independently predictive of high-transmission–risk sexual intercourse among the men who were HIV infected (odds ratio = 1.9; 95% CI = 1.1, 3.3.)

Anecdotal reports have suggested that crystal methamphetamine is used by MSM in conjunction with Viagra to have increased frequency and duration of intercourse and to overcome erectile dysfunction that may occur in the presence of substance use. Methamphetamine use was found to be an independent predictor of syphilis infection in MSM in San Francisco, and this risk was greater among the men who used methamphetamines with Viagra than among those who did not combine the 2 drugs.12 We found that of the 293 participants who used crystal methamphetamine in the past year, 55% had also used Viagra, whereas only 24% of those who did not use methamphetamines had used Viagra (χ² test for associations, P< .001).

We found that meeting sex partners on the Internet was associated with high-transmission–risk sexual behavior in the bivariate analysis but not in the multivariate analysis. Upon further examination we found that meeting sexual partners on the Internet was confounded by the number of male partners. Although not independently predictive of high-transmission–risk sexual intercourse, meeting sexual partners on the Internet appears to provide an outlet for expanding sexual networks. The Internet has recently been identified as a source of meeting one’s sexual partners and has been associated with an outbreak of syphilis among MSM16 and with high-risk behaviors.17^,18 The Internet can be used as a new avenue for presenting HIV and sexually transmitted disease prevention messages and sources of referrals for testing and care.

The availability of effective antiretroviral therapies appears to have had the unfortunate effect of increasing risk behavior among non–HIV-infected men. High-transmission–risk sexual intercourse among the uninfected men was more likely among those who agreed with a statement that they were less careful with sexual intercourse and drug use now than in the past because of the availability of HAART. This is consistent with findings from studies exploring attitudes and high-risk behavior shortly after protease inhibitors became widely available.3^,19^–22 Prevention efforts must take into account current attitudes because of the effect they have on behavior.

Using information about the serostatus of one’s sex partner has been reported among MSM and cited as 1 method of reducing HIV transmission.29^–32 We found that sexual intercourse between seroconcordant partners was more frequent among men with their primary partners and that sexual intercourse between serodiscordant partners was more frequent with nonprimary partners. It appears that the type of partner (primary vs nonprimary) may play a more important role in HIV transmission than the selection of a partner or type of sexual activity based upon serostatus.

In addition to the recently identified cofactors for HIV transmission, long-standing risk factors for high-risk behavior continue to be important. Use of amyl nitrite, long associated with high-risk sexual behavior among MSM, was associated with high-transmission–risk sexual intercourse. In addition, illicit drug use was a common finding among both infected and uninfected men. The association between the use of these drugs and HIV risk behaviors and infection among MSM identified in this and other studies11^–14 highlights the need to develop drug treatment programs specifically tailored for MSM who use these drugs.

Limitations

Although a key strength of our study is that we used a probability sample of MSM, there are limitations to consider. There has been an increase in the use of cell phones in recent years, and our sample only included residential phones. However, this is not likely to be a source of substantial bias. In a survey during the first 6 months of 2003 only 3% of households were found to rely exclusively on cell phones.33

We selected telephone exchanges that corresponded to the MSM-dense areas of the city, and thus we were less likely to have included households outside these areas; risk behaviors may be different outside the traditionally gay neighborhoods. However, our estimates of HIV and risk behaviors are more representative than traditional convenience samples. Sampling of the MSM-dense areas of San Francisco may have underrepresented MSM of color. There are no reliable data regarding the racial/ethnic distribution of MSM in San Francisco. Compared with the 2000 census, our sample had a substantially greater proportion of White men. This discrepancy is most likely explained by influx to urban areas of White MSM who have higher education and income levels.21 In our study, participants were administered questions about changes in residence (excluding moves within the same city). The mean number of places that the respondents had lived in was greater for Whites (4.7) than for African Americans (4.1), Latinos (3.6), and Asians (3.3; P< .001, analysis of variance test for differences in means across multiple groups).

In addition, HIV status may have been misclassified. A previous study among a probability sample of MSM in San Francisco demonstrated that participants’ self-report of being infected with HIV was accurate.34 In addition, we minimized misclassification of HIV-infected persons by including HIV testing in the study. Although 29% of men who did not report being HIV infected did not participate in HIV testing, thereby creating the possibility of having misclassified some men who were HIV infected as uninfected. This would not have affected our analysis of the predictors of high-transmission–risk sexual intercourse because self-report of HIV status was used for that analysis. In addition, between self-report of HIV positivity and results from HIV testing, we had valid information on HIV status in 78% of the sample, thereby minimizing the effect of possible misclassification.

In addition, because this is a cross-sectional survey, we were only able to measure correlates of HIV infection and not a causal link between the predictor variables and high-risk sexual behaviors. A final limitation to consider is whether high-risk sexual behaviors among MSM from San Francisco are representative of MSM in other urban areas. We believe that they are. In the past, we have found that risk behaviors and HIV trends among MSM in San Francisco preceded similar trends among MSM in other cities. Shortly after HAART became widely available, we noted an increase in sexual risk behavior followed by increases in HIV incidence.3 Similar trends were later documented in the United States and elsewhere.4^–10^,35^,36 In addition, in a multisite interview project of HIV-infected men conducted between 2000 and 2002, 6.3% of subjects reported high-transmission-risk sexual intercourse in the past 6 months, comparable to our finding of 5.4% reporting this behavior in the past 12 months.37

Conclusions

Our findings suggest several areas for novel and improved interventions to reduce HIV transmission among MSM in the face of a mature epidemic. Specific drug treatment and prevention programs aimed at use of amyl nitrite and crystal methamphetamine appear necessary. In addition, expanding the types of counseling services for MSM may also be helpful. For example, cognitive therapy has been found to be effective in risk reduction among both HIV-infected38 and non–HIV-infected men.38^,39 We found that men with primary partners had lower rates of HIV infection, and with the exception of unprotected receptive anal intercourse with a serodiscordant partner, sexual behaviors with primary partners were not associated with HIV infection. Methods to promote developing and sustaining respectful, committed relationships such as laws permitting civil unions and marriage between MSM may contribute to decreased HIV transmission.

In the third decade of the HIV epidemic, it is clear that MSM remain severely impacted by HIV/AIDS and that risk behaviors in this group are resulting in ongoing HIV transmission. Research to identify models that understand and predict high-risk behavior may help in developing effective prevention programs. Efforts to conduct high-quality studies of HIV incidence, prevalence, attitudes, and detailed sexual and drug use behaviors should be a priority to keep prevention efforts relevant.