Determinants of HIV Risk Among African American Men Who Have Sex with Men

Abstract

Objectives

This study examined factors associated with the sexual risk practices of African American men who have sex with men (MSM).

Methods

Social networks composed of 445 Black MSM were recruited from 2012–2014 in Milwaukee, Cleveland, and Miami Beach. Participants completed assessments of past 3-month sexual behavior, background characteristics, substance use, and scales measuring HIV risk-specific characteristics as well as psychosocial domains potentially related to risk. Regression analyses determined factors associated with high-risk sexual behavior.

Results

Over one-third of men reported condomless anal intercourse (CLAI) outside of main concordant partnerships, predicted by alcohol and marijuana use, weak risk reduction intentions, and high gay community participation. High frequency of CLAI acts with non-main partners was associated with weak risk reduction intentions, the perception that safer sex is not a peer norm, low condom use self-efficacy, and longer time since last HIV testing.

Conclusions

Elevated HIV risk among Black MSM was primarily associated with substance use, perceived peer norms regarding condom use, other HIV risk-specific cognitive characteristics. Social network interventions hold promise for reaching racial minority MSM and addressing risk-related cofactors.

HIV infection in the United States falls along sharp lines of disparity related to sexual orientation and race. Every year since HIV surveillance began, men who have sex with men (MSM) have accounted for the majority of the country’s HIV cases.¹ The disease also disproportionately affects African Americans who constitute 12% of the American population but carry 44% of its HIV infection burden.¹ Yet, the starkest disparity emerges from the combined impact of race and sexual orientation. Black MSM represent only a fraction of a percent of the American population but accounted for over 25% of the country’s new HIV infections in 2010,¹ with HIV rates much higher among Black than white MSM.² HIV incidence among racial minority MSM in some cities ranges from 24% to 29%.^3,4 Considerable attention is being given to biomedical strategies such as early initiation of antiretroviral therapy (ART) and pre-exposure ART prophylaxis (PrEP) for HIV prevention.^5,6 However, the impact of these promising strategies will depend upon ART coverage and adherence, neither of which is likely to be quickly attained nor complete. For this reason, integrated HIV prevention approaches are needed, including improved interventions to reduce risk behavior among racial minority MSM.

Prior research has examined but has generally failed to establish differences in individual-level risk practices between Black and white MSM.^7–10 However, sexual network characteristics, high rates of undiagnosed and untreated HIV infection, high sexually transmitted disease (STD) prevalence, and unsuppressed viral load among HIV-positive African American MSM are believed to contribute to HIV disparities.^9–16

Black MSM are not a monolithic population,^17–20 and multiple factors may influence extent of HIV vulnerability within the community of racial minority MSM. These include risk-related sexual behavior norms, attitudes, and intentions;^21–25 substance use;^{12, 26–30} poverty and disadvantage;^19,20 and psychosocial domains including internalized homonegativity or homophobia,^31–34 self-perceived masculinity,^35,36 HIV conspiracy beliefs or mistrust,^37,38 religiosity,³⁹ and resilience.^40,41 It is important to ascertain the relative importance of these and other factors in order to properly tailor HIV prevention interventions for racial minority MSM.

This study recruited social networks of African American MSM and sought to determine the relationships of four types of factors with the riskiness of men’s sexual behavior practices: (1) social, economic, and demographic background characteristics; (2) substance use; (3) HIV risk-specific knowledge, attitudes, beliefs, and intentions; and (4) psychosocial domains including internalized homonegativity, self-ascribed masculinity, AIDS conspiracy beliefs, resilience, religiosity, and gay community participation. HIV risk-specific characteristics were examined because they are proximal to adopting protective actions based on many behavioral science theories.^42–44 Psychosocial domains were examined because broader personal and contextual life experiences may also potentiate risk. We sought to identify characteristics related not only to some men’s high-risk behavior but also the adoption of very safe behavior by other African American MSM, a strengths-based question that has been insufficiently explored.

METHODS

Participant Recruitment

The study recruited social networks of racial minority MSM in Milwaukee, WI, Cleveland, OH, and Miami Beach, FL in 2012–2014. Network recruitment began by identifying initial “seeds” in community venues where Black MSM were known to meet including bars, clubs, pageants, house balls, and neighborhood hangout places. Field staff systematically observed “social circles”⁴⁵ of racial minority men in these settings and identified each circle’s center of social attention. All seeds were Black MSM. Field staff approached the potential seed and explained the study. If recruited, the seed was asked to identify—by first name only—his close MSM friends. Seeds were asked to invite into the study each named friend. These individuals constituted the network’s first “ring.” When members of the first ring entered the study, they also invited their own personal friends, the second network ring. They, in turn, invited their own friends. In this way, sociocentric networks reached out three waves from each initial seed. Participant inclusion criteria were being at least 18 years old, living in the study city, being named as a friend by an already-enrolled participant, and providing written informed consent. Networks were eligible for inclusion if at least 50% of the seed’s first-ring friends were enrolled.

Thirty-five networks consisting of 464 participants (230, 180, and 54 from Milwaukee, Cleveland, and Miami Beach, respectively) were recruited. Networks ranged in size from 3 to 47 (mean=13) enrolled members. Nineteen men who reported their race as white, reported no history of same-sex behavior, or declined to answer sexual behavior questions were excluded. The sample size for analysis was 445.

Interview Assessment Measures

Participants individually completed audio computer assisted self-interviews (A-CASI) and received risk reduction counseling. The study protocol was approved by IRBs of each participating institution.

Demographic characteristics

Participants responded to questions about their gender at birth, self-identified present gender (male, female, or transgender), age, race, ethnicity, employment, income, education, and housing stability. Participants used a 5-point scale to describe their sexual orientation (from exclusively gay to exclusively straight) and were asked whether they ever had an HIV test, whether it was in the past year, and the last test’s result.

Sexual risk practices during lifetime, the past year, and the past three months

Participants indicated their number of male and female sexual partners in the previous year. Participants then described their sexual behaviors over the past three months on a partner-by-partner basis for up to the most recent 5 male and 5 female sexual partners. Respondents indicated their relationship with each partner (main and committed, regular but not main and committed, casual, or commercial), how often in the past 3 months they had anal intercourse (AI) with each male partner, and how many of those acts were condomless anal intercourse (CLAI). For each partner, respondents reported whether they disclosed their HIV serostatus before sex and whether the partner did so. Participants with >5 partners of either gender in the past 3 months summarized their sexual practices with all additional partners.

The sample was divided into two groups based on their sexual behavior. Men were considered to be low risk in their behavior (n=292) if they reported no CLAI during the past 3 months (either because they did not engage in any anal intercourse or reported 100% condom use for all AI acts), or if they reported CLAI only with their single main male partner when: (1) both the respondent and that main partner had been tested; (2) the partner disclosed his HIV-positive or HIV-negative serostatus before sex; and (3) they were HIV seroconcordant. Men were considered to have engaged in high-risk acts (n=153) if they reported CLAI with partners other than their single seroconcordant main male partner in the past 3 months.

Substance use

Participants were asked on how many days they drank alcohol in the past month and their number of drinks in a typical drinking day. In addition, participants indicated whether, and on how many days in the past month, they used each of a series of listed drugs including heroin, other opiates, cocaine, crack, amphetamines or methamphetamines, marijuana, ecstasy, gamma hydroxybutyrate (GHB), ketamine, inhaled nitrites (“poppers”), non-prescribed medications for erectile dysfunction, illicit prescription drugs, and any injected drug.

HIV risk-specific scales

The assessment included five HIV risk-specific scales. A 9-item scale measured knowledge of HIV risk reduction steps (sample item: “If a man pulls out before orgasm, it protects from getting AIDS and venereal diseases,” scale range 0–9). An eight-item scale measured perceived condom use peer norms (sample item: “Condom use is well-accepted among my friends”) with 3-point scales for each statement (scale range 0–16 Cronbach’s alpha, current sample=0.75). The same response format was used to measure condom attitudes (8 items, sample item: “Using condoms interrupts the pleasure of sex,” range 0–16, Cronbach’s alpha=0.80); risk reduction behavioral intentions (8 items, sample item: “A condom will be used if I have sexual intercourse with a casual partner,” range 0–16, Cronbach’s alpha=0.75); and risk reduction self-effficacy (8 items, sample item: “I am sure that I can overcome my partner’s objections to condoms,” range 0–16, Cronbach’s alpha=0.66).

Psychosocial domains

Five scales measured broader psychosocial domains hypothesized in the literature to influence HIV risk among minority MSM. Self-ascribed masculinity was measured with a 4-item scale adapted from Garcia et al.⁴⁶ (sample item: “I can pass as a straight man,” 5-point Likert response options from strongly disagree to strongly agree, range 4–20, Cronbach’s alpha=0.83). Internalized homonegativity was assessed using a measure adapted from Herek et al.,³³ Myers,⁴⁷ and Wagner,⁴⁸ with 5-point Likert scales to indicate agreement with 9 statements (sample item: “I wish I were not sexually attracted to men,” range 9–45, Cronbach’s alpha=0.86). Resilience, the perceived internal capacity to handle challenging life situations, was assessed using 10 items from a 25-item scale originally developed by Wagnild and Young⁴⁹ (sample item: “My belief in myself gets me through hard times,” range 10–50, Cronbach’s alpha=0.88). AIDS conspiracy beliefs were measured with Bogart and Thorburn’s³⁷ 9-item scale (sample item: “AIDS was created by the government to control the Black population,” 5-point Likert scales, range 9–45, Cronbach’s alpha=0.89). Religiosity and church involvement was measured with 6 items adapted from Forehand and Brady⁵⁰ (sample item: “How often do you attend religious services?” range 6–30, Cronbach’s alpha=0.80). Finally, a 7-item scale was developed specifically for this research to measure gay community participation, the extent to which one actively participates in gay-identified community activities (sample item “How often do you visit gay clubs or websites?,” range 7–35, Cronbach’s alpha=0.86).

Statistical Methods

Means and standard errors (SEs) were first calculated for continuous variables and relative frequencies for categorical variables to characterize the overall sample. Bivariate and multivariable statistical analyses were performed using random effects models to account for the potential effect of social network. Single predictor logistic regressions with a random network effect were first used to investigate the statistical significance of differences between men who did or did not engage in any high-risk acts on individual variables. Multiple logistic regressions with a random network effect were then performed to identify significant main effects using forward stepwise variable selection. All variables that had achieved p-values <0.2 in the bivariate analyses were investigated for statistical significance. All two-way interactions between significant main effects were also tested for statistical significance.

Finally, we examined variables associated with reporting three or more high-risk CLAI acts with non-main partners in the past 3 months. This cutoff was used because it categorizes participants as above the median (median=2) or at or below the median. Logistic regression analyses were used to compare the groups based on their frequency of high-risk CLAI acts. Differences were considered significant at the 5% significance level. Due to occasional small counts, some categories were aggregated into larger groups to secure at least 10 participants per cell after cross tabulation with the risk level indicator.

RESULTS

Sample Characteristics

Participants’ mean age was 27.3 years (SE=1.04, range 18–57). Although all reported being of male gender at birth and the large majority still identified as male (n=408, 92%), 6% of participants (n=28) now identified as transgender and 2% (n=8) as female. Seventy-five percent (n=335) described their sexual orientation as mainly or exclusively gay, 21% (n=94) as bisexual, and 3% (n=15) as mainly or exclusively straight. Seven percent (n=29) of participants reported being of Hispanic ethnicity. Nearly half (n=199) of participants were unemployed, 76% (n=334) had annual incomes below $20,000 per year with 56% (n=245) below $10,000, and 55% (n=246) had high school or less education. Nine percent (n=39) of participants reported unstable housing. Alcohol and marijuana were the substances most commonly used by study participants. Eighty-five percent (n=377) of participants drank alcohol and did so on a mean of 8 days in the past month, and 60% (n=266) smoked marijuana on an average of 17 days. Other substances used by over 2% of participants in the past month were crack/cocaine (9%, n=41), ecstasy (7%, n=31), illicit drugs (19%, n=83), opiates (4%, n=18), and inhaled nitrites (3%, n=14). Injection drug use in the past month was reported by only 2 participants. Because many individuals used illicit drugs but the number using a particular substance was often modest, substances other than alcohol and marijuana were combined into a category of any illicit drug use.

Participants reported a mean of 5.4 (SE=0.66) male partners in the past year and 2.4 (SE=0.23) in the past 3 months. Most men (93%, n=415) had no female partners during these periods. Most participants (96%, n=426) said they had been tested for HIV at some point, 66% (n=281) in the past year, and 25% (n=105) said their most recent HIV test was positive.

Factors Associated with Engaging or Not Engaging in High-Risk CLAI

Table 1 shows results comparing participants who did or did not engage in high-risk acts. Men who engaged in high-risk CLAI were significantly more likely to have incomes below $10,000, have had an STD in the past 6 months, to be HIV-positive, and tended to more often be unemployed. The groups significantly differed on all of substance use variables reported in Table 1, with substance use always greater among men who engaged in high-risk CLAI. There were consistent differences between groups on four of the five HIV risk-specific scales. Men reporting high-risk acts had weaker perceived peer norms for condom use, risk behavior reduction intentions, attitudes toward condoms, and self-efficacy for condom use. With respect to psychosocial domains, participants reporting high-risk CLAI scored higher in gay community participation and lower in resilience. Those who did not report high-risk CLAI tended to report greater religious and church involvement.

Table 1.

Unadjusted Comparisons of Participants Who Reported Any or No High-Risk Condomless Anal Intercourse (N=445)

Domain/Variable	High-Risk CLAI (N=153)	No High-Risk CLAI (N=292)	p
SOCIODEMOGRAPHIC CHARACTERISTICS:
Male gender—% (n)a	92.8 (142)	91.1 (266)	.45
Age in years—Mean (SE)a	27.0 (1.12)	27.5 (1.07)	.41
Hispanic ethnicity—% (n)	6.5 (10)	6.5 (19)	.95
Low education (high school or <) —% (n)	55.6 (85)	55.1 (161)	.97
Attends school-% (n)	21.6 (33)	23.6 (69)	.54
Currently employed—% (n)	49.7 (76)	58.2 (170)	.06
Low income (<$10K annually)b—% (n)a	62.5 (95)	52.6 (150)	.03
Unstable housing situation—% (n)	9.8 (15)	8.2 (24)	.65
SEXUAL, STD, AND HIV TESTING HISTORY
Completely gay orientation—% (n)a	49.0 (75)	42.1 (123)	.22
Had main male partner > 1 year—% (n)	21.6 (33)	25.3 (74)	.43
Had an STD in past 6 months—% (n)a	13.2 (20)	5.2 (15)	.01
Never tested for HIV or tested more than 1 year ago—% (n)	39.2 (60)	35.6 (104)	.46
Was HIV+ at most recent test—% (n)	30.7 (47)	19.9 (58)	.02
SUBSTANCE USE IN PAST 30 DAYS:
Drank alcohol—% (n)	92.8 (142)	80.5 (235)	<.01
Smoked marijuana—% (n)	71.9 (110)	53.4 (156)	<.01
Used crack/powder cocaine—% (n)	16.3 (25)	5.8 (17)	<.01
Used any illicit drug (excluding alcohol or marijuana) —% (n)	28.1 (43)	13.7 (40)	<.01
HIV RISK-SPECIFIC SCALES:
HIV risk knowledge—Mean (SE)	7.2 (0.14)	7.0 (0.11)	.17
Perceived condom use peer norms—Mean (SE)a	10.4 (0.34)	11.5 (0.29)	<.01
Risk reduction behavioral intentions—Mean (SE)a	10.1 (0.31)	12.8 (0.26)	<.01
Condom use attitudes—Mean (SE)	12.0 (0.31)	13.4 (0.26)	<.01
Condom use self-efficacy—Mean (SE)	13.0 (0.23)	13.7 (0.18)	<.01
PSYCHOSOCIAL DOMAIN SCALES:
Self-ascribed masculinity a	13.7 (0.36)	13.7 (0.29)	.99
Gay community participation—Mean (SE)	21.2 (0.55)	20.0 (0.45)	.05
Internalized homonegativity—Mean (SE)a	19.4 (0.67)	19.8 (0.51)	.56
Resilience—Mean (SE)a	43.6 (0.42)	44.7 (0.31)	.03
AIDS conspiracy beliefs—Mean (SE)a	18.9 (0.62)	18.1 (0.47)	.26
Religious and church involvement—Mean (SE)a	17.0 (0.48)	17.9 (0.39)	.06

^aUp to 4 participants were missing data on predictor variables.

^b8 individuals did not disclose their income.

Results of the multiple mixed logistic regression analysis predicting whether or not participants reported high-risk acts in the past 3 months are shown in Table 2, with reported odds ratios (ORs) adjusted for significant covariates. Four variables remained significant in the regression model: gay community participation, drinking alcohol in the past month, using marijuana in the past month, and risk reduction behavioral intentions. The odds of reporting only safe behavior decreased by 4% with a one-unit increase on the scale measuring gay community participation. A one-unit increase in strength of risk reduction behavioral intentions was associated with a 30% increase in the odds of reporting no high-risk CLAI. The odds of being safe in behavior among those who drank alcohol in the past 30 days were 37% of those participants who did not drink alcohol. The odds of being safe in behavior among those who smoked marijuana in the past 30 days were 51% of those who did not use marijuana.

Table 2.

Results of a Multiple Mixed Logistic Regression Predicting Participants’ Reports of No High-Risk Condomless Anal Intercourse (CLAI)^a

Covariate	Odds Ratio	p-value	95% CI
Gay community participation	0.96	.05	0.93, 1.00
Drank alcohol in the past 30 days	0.37	.01	0.18, 0.79
Used marijuana in the past 30 days	0.51	.01	0.32, 0.83
Risk reduction behavioral intentions scale	1.30	<.01	1.21, 1.40

^aArea under the Receiver Operating Characteristic (ROC) Curve=0.778, a measure of the model’s discrimination ability where random classification corresponds to 0.5 and perfect classification 1.0. All variables with p-value < 0.2 in Table 1 were tested for inclusion in the set of predictors.

Factors Associated with Frequency of High-Risk CLAI with Non-main Partners in the Past 3 Months

Analogous to Table 1, Table 3 compares sociodemographic characteristics, sexual and HIV history characteristics, substance use variables, HIV risk-specific scales, and scales measuring psychosocial domains between participants engaging in 3 or more CLAI acts with non-main partners in the past 3 months and participants who engaged in fewer than 3 CLAI acts. Longer time since one’s last HIV test, the perception that safer sex is not an accepted social norm, weaker risk reduction behavioral intentions, and lower condom use self-efficacy were each associated with more frequent high-risk CLAI, although HIV risk knowledge scores were positively associated with reporting 3 or more unprotected acts with non-main partners.

Table 3.

Unadjusted Comparisons of Higher-Risk Participants Who Had Condomless Anal Intercourse (CLAI) with Non-Main Partners Fewer Than Three Times Versus Three or More Times (N=153)

Domain/Variable	Less Than 3 High-Risk CLAI Acts (N=81)	3 or More High-Risk CLAI Acts (N=72)	p
SOCIODEMOGRAPHIC CHARACTERISTICS:
Male gender—% (n)	92.5 (75)	93.1 (67)	.91
Age in years—Mean (SE)a	26.7 (1.3)	27.5 (1.2)	.38
Hispanic ethnicity—% (n)b	3.8 (3)	9.9 (7)
Low education (high school or <) —% (n)	54.3 (44)	56.9 (41)	.75
Attends school —% (n)	18.5 (15)	25.0 (18)	.33
Currently employed—% (n)	50.6 (41)	48.6 (35)	.81
Low income (<$10K annually)a—% (n)	61.7 (50)	63.9 (46)	.79
Unstable housing situation—% (n)	8.6 (7)	11.1 (8)	.61
SEXUAL, STD, AND HIV TESTING HISTORY
Completely gay orientation—% (n)	54.3 (44)	43.1 (31)	.17
Had main male partner > 1 year—% (n)	24.7 (20)	18.1 (13)	.32
Had an STD in past 6 months—% (n)a	12.5 (10)	13.9 (10)	.80
Never tested for HIV or tested more than 1 year ago—% (n)	58.0 (47)	63.9 (46)	.46
Was HIV+ at most recent test—% (n)	29.6 (24)	31.9 (23)	.76
Time since last HIV test (years)—Mean (SE)b,c	0.9 (0.5)	2.3 (0.5)	.03
SUBSTANCE USE IN PAST 30 DAYS:
Drank alcohol—% (n)	92.6 (75)	93.1 (67)	.89
Smoked marijuana—% (n)	67.9 (55)	76.4 (55)	.25
Used crack/powder cocaine—% (n)	14.8 (12)	18.1 (13)	.59
Used any prescription illicit drug (excluding alcohol or marijuana)—% (n)	22.2 (18)	34.7 (25)	.09
HIV RISK-SPECIFIC SCALES:
HIV risk knowledge—Mean (SE)	7.2 (0.19)	7.2 (0.20)	.96
Perceived condom use peer norms—Mean (SE)	11.1 (0. 49)	9.5 (0.49)	<.01
Risk reduction behavioral intentions—Mean (SE)	10.7 (0.50)	9.6 (0.49)	.01
Condom use attitudes—Mean (SE)	12.6 (0.44)	11.6 (0.45)	.08
Condom use self-efficacy—Mean (SE)	13.5 (0. 39)	12.2 (0.39)	.01
PSYCHOSOCIAL DOMAIN SCALES:
Self-ascribed masculinity	13.5 (0.44)	13.7 (0.46)	.81
Gay community participation—Mean (SE)	21.4 (0.66)	21.4 (0.68)	.98
Internalized homonegativity—Mean (SE)	18.9 (0.87)	19.9 (0.92)	.40
Resilience—Mean (SE)	43.9 (0.55)	43.2 (0.58)	.38
AIDS conspiracy beliefs—Mean (SE)	18.4 (0.97)	19.6 (0.98)	.32
Religious and church involvement—Mean (SE)	17.1 (0.66)	16.9 (0.67)	.84

^aUp to 2 participants were missing data on predictor variables.

^bP-value was not reported due to a small cell count (< 5).

^c10 individuals did not report time since their last HIV test.

The multiple mixed logistic regression analysis showed that the sole significant predictor of having 3 or more CLAI acts with non-main partners in the last 3 months was perceived condom use peer norms (OR = 0.85, 95% CI = 0.77–0.94, p = 0.01). When the perceived condom use peer norm score increased by one unit, the odds of reporting 3 or more CLAI acts with non-main partners decreased by 15%.

DISCUSSION

The present study examined the influence of a diverse array of characteristics that have been hypothesized to be associated with HIV risk using analyses that took into account their intercorrelated nature. The picture that emerged underscores the combined associations of sexual risk practices with alcohol and illicit drug use; perceived safer sex peer norms, condom attitudes, and risk reduction intentions; recency of HIV testing; and indicators of socioeconomic distress. Many of these variables were associated not only with whether or not men engaged in CLAI outside of a seroconcordant main partner relationship but also how often they did so.

Psychosocial domains such as internalized homonegativity, self-ascribed masculinity, and HIV conspiracy beliefs were generally not associated with risk behavior. Although resilience distinguished between men who did or did not report high-risk condomless sex in unadjusted bivariate analyses, gay community participation was the only psychosocial domain that remained significant in the adjusted analyses and it was related to greater risk. This may be because an individual’s presence in gay-identified venues and online environments presents greater risk opportunities. There can be little doubt that factors such as resilience, masculinity, and internalized homonegativity play important roles in the lives of many racial minority MSM. However, and like other studies,^21,22,51 this research did not confirm the independent association of these domains with riskiness or safety in sexual behavior.

One in four participants reported that he was HIV-positive, an alarming level of disease prevalence comparable to that usually found in developing countries devastated by AIDS. Although most men in this sample reported having an HIV test at some point in their lives, one-third of men had not been tested in the past year. Efforts are needed to encourage more regular and frequent HIV testing among Black MSM.

Although the field often emphasizes the identification of factors associated with high-risk sexual behavior, the majority of participants in this sample either did not report CLAI in the past 3 months or did so only with their single HIV-concordant main partner. From a strengths-based perspective, these findings suggest that HIV prevention interventions should not only help persons develop protective HIV-related norms, attitudes, and intentions but also attempt to address socioeconomic disparities—including those related to income and employment—that contribute to risk. Integrated HIV and substance abuse prevention and treatment are also critical.

In contrast to most prior research, the present study recruited social networks of African American MSM based on their friendship interconnections rather than presence in gay-identified venues. Racial minority MSM do not always attend gay-identified venues, and network recruitment affords a strategy for reaching men who might otherwise be hidden in the community.⁵² It is noteworthy that, when asked to identify MSM friends in their social networks, the large majority of participants identified friends who were other African American men. Network enrollment methods may be useful not only for reaching racial minority MSM in the community—many of whom in this study were HIV-positive—but also for delivering HIV prevention interventions for risk behavior reduction, to promote regular HIV testing, or to encourage HIV medical care engagement.⁵² The current study’s findings that perceived social norms strongly influenced participants’ behavior practices provides support for the approach of strengthening norms for HIV prevention within naturally-existing social networks of African American MSM.

The study has several limitations. Because the study recruited social networks of Black MSM, rather than a true representative probability sample, analyses had to take into account potential dependence of participant responses within social networks. Although we determined associations between conceptually-defined characteristics with risk behavior, this methodology does not demonstrate causality, especially because predictor and outcome variables were assessed at a single point. Stigmatized activities may have been underreported, although A-CASI assessment reduces self-presentation bias.⁵³ The retrospective period for defining participants’ level of risk behavior was the past 3 months, short enough for reliable reports of sexual practices⁵⁴ but of a duration that could underestimate levels of longer-term risk behavior. It is possible that the risk levels of some participants were miscategorized. For example, an HIV-negative participant who engaged in CLAI with an HIV-infected main partner who is in medical care and virally suppressed is at lower risk for contracting HIV infection than if the partner was viremic. We did not assess whether HIV-positive partners were in medical care. HIV risk among African American MSM is influenced by the makeup of their sexual networks,^52,25 and this study did not assess sexual network characteristics. Finally, HIV concordance was defined based on what participants reported about themselves and about what their partners said to them. Some individuals may have misportrayed or not known their true HIV status.

Antiretroviral therapy can greatly reduce viral load among persons living with HIV infection and, in turn, reduce their likelihood of transmitting the disease to sexual partners.⁵ PrEP regimens diligently followed by high-risk but uninfected MSM also carry protective benefit.⁶ These biomedical developments are critical new tools for HIV prevention. However, their impact will be determined by ART coverage and adherence, and neither is likely to be quick or complete. There remains an urgent need for improved risk behavior reduction interventions for Black MSM and for integrated behavioral, social, and biomedical prevention.