Since the emergence of HIV as a global pandemic in the 1980s, men who have sex with men (MSM) have shared a disproportionately large burden of infection in many high-income countries in Western and Central Europe, Australia, and North America [1]. Due to this recognized high burden, MSM represent a large target population for resources on HIV/AIDS prevention, treatment, and research in these areas. By contrast, Africa's HIV/AIDS epidemic has long been understood primarily as a “heterosexual epidemic”, with an estimated 80% of HIV infections being tied to heterosexual transmission [2]. This focus has led to HIV/AIDS prevention efforts in Africa being targeted primarily to heterosexuals.

Recent epidemiological evidence has shown that MSM in Africa share a disproportionate burden of HIV infection [3]. Prevalence estimates of MSM in Africa range from 1–4% of the general population, but high levels of HIV infection and a high prevalence of MSM also engaging in sex with women has led MSM transmission to be linked to over 20% of all HIV cases in several countries of the Middle East, North Africa, and West Africa [4]–[6]. These data are at odds with the fact that most African countries have not dedicated any national HIV/AIDS funds to specifically target HIV/AIDS among MSM [6].

The 2009 UNAIDS report on universal access for MSM and transgender people highlights the global failure in addressing the needs of MSM regarding HIV/AIDS education, prevention, treatment, research, and care. One of the foci of this report is increasing access to HIV/AIDS prevention materials for MSM and transgender individuals [7]. Although increasing HIV/AIDS knowledge alone is not sufficient to promote sustainable behavior change, accurate knowledge of transmission and prevention of HIV is necessary if MSM are to adopt risk reduction strategies.

Globally, reporting on HIV knowledge among MSM is sparse. Only 33 out of 147 low and middle income countries (LMIC) reported knowledge data through the 2008 United Nations General Assembly Special Session (UNGASS) [8]. Only 2 of these 37 countries reporting UNGASS HIV knowledge data were in Africa, with Nigeria and Mauritius reporting that only 44% and 48% of MSM respectively could “correctly identify ways of preventing sexual transmission of HIV and could correctly reject major misconceptions about HIV transmission” [8]. Additionally, across all low and middle-income countries reporting knowledge scores, less than half of MSM held correct HIV knowledge.

Other studies from Sudan and Kenya indicate that MSM in Africa may have low knowledge regarding HIV prevention and transmission. More than half (55%) of a sample of MSM in Sudan and 35% of respondents in Mombasa, Kenya did not understand the link between anal sex and HIV infection [9]–[10]. By contrast, over 90% of samples of MSM in Malawi, Botswana, and Namibia understood that HIV can be transmitted through anal sex with a man [11]. However, of these samples, only 57%, 50%, and 85% respectively had ever received educational materials on preventing HIV transmission between men.

Data on levels and correlates of HIV/AIDS knowledge in Africa are essential to develop effective prevention and education strategies. Previous studies of HIV knowledge among MSM have focused on levels of HIV knowledge, but have not systematically examined factors associated with low knowledge. The present study aims to fill this gap by examining factors associated with low HIV/AIDS knowledge among MSM in South Africa and the United States using a validated HIV knowledge scale.

We chose to focus on South African and US MSM because: (1) the HIV-KQ-18 knowledge scale has only been previously validated and described among populations in the US, (2) South Africa has a large population of people living with HIV/AIDS, representing 25% of all those living in Sub-Saharan Africa, and (3) there is emerging evidence that MSM in South Africa share a disproportionate burden of HIV infection [1], [12]. This information will support the development and scale up of educational programs to address the specific needs of MSM, and may aid in optimizing existing prevention programs for MSM.

1,154 internet-using MSM in the United States and 439 in South Africa were recruited through banner advertisements on Facebook.com targeted to men who stated they were interested in men on their Facebook profiles. Facebook ads were displayed in South Africa from June 1 to June 30, 2010 and in all US states from October 1 to November 30, 2010. Banner ads included a range of images, including groups of men, individual men of a range of ethnic and racial backgrounds, and rainbow-themed images. Facebook users who clicked on the banner ads were taken to an internet-based survey. To be eligible to begin the survey men had to report male-to-male sex in the past year.

The survey collected information on participant demographics (age, race, education, and employment), sexual orientation, the number of friends, colleagues, or acquaintances they felt identified as gay or bisexual, HIV testing behavior, knowledge of HIV transmission, condom and water-based lubrication use, and questions on gay identity. HIV knowledge was quantified using the brief HIV knowledge scale (HIV-KQ-18), an internally consistent and stable HIV knowledge scale shown to be appropriate for low-literacy populations [13]. Questions in the HIV-KQ-18 focus on basic HIV transmission and prevention and are summed to form an index of overall HIV knowledge (0–18) with non-responses and “don't know” coded as incorrect.

Gay identity was quantified using a modified version of the US-validated “Gay Identity Questionnaire” consisting of 20 questions summed to form an index of decreasing pride and acceptance of homosexuality as one moves from 0 to 80 [14]. Race in the US was classified into white non-Hispanic, black non-Hispanic, Hispanic, and other (including Asian/Pacific Islander, Native American/Alaska Native, Multi-Racial, and other). Race in South Africa was classified into black African, white/European/African, and other (including Asian, colored, and other). For both groups of respondents, individuals responding “don't know” for previous HIV testing were coded as not being tested and individuals who indicated receiving “indeterminate” HIV results were coded as testing HIV negative.

Statistical analyses were conducted using SAS 9.2 (Cary, NC, USA). Statistical significance was assessed using an alpha value of .05 and two-tailed tests. Individuals were excluded from analyses if they did not report being male and having had male-to-male sex in the past year (n = 15 US; n = 80 South Africa) or if no knowledge questions were answered (n = 883 US; n = 85 South Africa). Individuals excluded for not answering any knowledge questions did not have significantly different age or gay/bisexual peer network distributions using t-tests.

Separate multivariable logistic regression models were built, one for the US and one for South Africa, using the dichotomous “low knowledge” variable as our outcome. Knowledge scores were dichotomized using a data-derived cutoff of greater than or less than and equal to 13/18 correct, with 19.9% and 15.3% scoring at this level or below in the US and South Africa respectively (Figure 1; Figure 2). We dichotomized knowledge this way because the proportional odds assumption was grossly violated for use of ordinal logistic regression with raw scores, and because there is no standard cutoff for a critical level of knowledge for the HIV-KQ-18. As a sensitivity analysis, changing this “critical knowledge” cutoff to other possible cut points of 14, 15, or 16 out of 18 correct on the knowledge scale did not meaningfully change factors associated or their magnitudes.

Variables considered for inclusion into each model were age, race, education, sexual orientation, HIV testing behavior, employment status, number of gay or bisexual friends known, having and/or had a female sex partner, relationship status (male or female), condom and lubrication use, and score on the gay identity scale. Backward selection procedures (α = .05) were used to arrive at the final models. Wald chi-square tests were used to establish significance of individual predictors, whereas likelihood ratio tests (LRT) were used to evaluate significance of groups of predictors (race, HIV-testing behavior). Hosmer and Lemeshow's goodness of fit test was used to determine if final models were adequate.

Basic demographics showed both groups of respondents to be primarily of white race, homosexual sexual orientation, and having had sex with only men (Table 1). Median knowledge scores were 16/18 correct for both cohorts, with 13.4% (59) and 17.1% (197) respondents in South Africa and the United States correctly responding to all 18 knowledge questions, respectively.

For South African MSM, less than 70% of respondents correctly answered that all pregnant women infected with HIV will not have babies born with AIDS, that a natural skin condom does not work better against HIV than a latex condom, and that a person can get HIV from oral sex (Table 2). For US MSM, less than 70% of respondents correctly answered that there is a female condom that can help decrease a woman's chance of getting HIV, and that a natural skin condom does not work better against HIV than a latex condom.

Hosmer and Lemeshow goodness-of-fit tests for each final model revealed no significant lack-of-fit (SA, p = .60; US, p = .95). Sexual orientation, having/and or had a female sex partner, sexual relationship status, and condom use were eliminated from both final models through backward elimination.

Among Facebook-using MSM in the US and South Africa, HIV/AIDS knowledge levels were high: the median respondent in both groups of respondents missed only 2 out of 18 questions. In both US and South African MSM, men with less than a high school education had significantly lower HIV/AIDS knowledge. This is in corroboration with a large body of research around the world showing that as general education level increases, so does knowledge of HIV/AIDS prevention and transmission [15]–[18]. Efforts to increase general educational opportunities for MSM in South Africa and the US, would be helpful inherently, and might also support higher HIV/AIDS knowledge.

Second, previous HIV testing was associated with having higher knowledge for MSM in the US and South Africa. Although causality is unclear from our study design, we hypothesize that MSM who have tested for HIV and received negative results gain some knowledge through contact with testing facilities, while individuals testing positive show a trend towards further increased knowledge due to continued contact with healthcare providers, educational materials, or peer groups. According to our data, traditional HIV/AIDS education and counseling associated with voluntary counseling and testing centers (VCT), hospitals, or other care centers is associated with higher levels of HIV knowledge among MSM.

Among South African MSM, having fewer acquaintances gay or bisexual was associated with lower HIV/AIDS knowledge, which could indicate that, in South Africa, peer networks are key avenues where HIV/AIDS information is shared. Additionally, after accounting for HIV testing behavior, not using condom-compatible lubrication during anal sex was associated with lower HIV knowledge in South Africa. South African MSM who receive condom-compatible lubrication from a community based organization or an NGO may also receive educational materials, making use of lubrication use a marker for access to outreach or comfort with service utilization.

For US MSM, there was a trend towards lower HIV knowledge among some racial and age groups. Hispanic MSM had significantly lower HIV/AIDS knowledge and black non-Hispanic MSM showed a trend towards lower knowledge, compared to White non-Hispanic MSM. Thirty years after HIV was first propelled into global consciousness, US MSM of color are at risk for gaps in HIV knowledge at a time when new HIV infections among these men are increasing.

US MSM age 18–24 or 50+ were at significantly increased risk for low HIV knowledge compared to men age 25–29. That these two age groups have the lowest HIV knowledge is troubling because sex with older partners has been shown to be associated with HIV infection, and US MSM age 50+ have an estimated HIV prevalence of 25% compared to 10.5% for men age 18–24 [19]–[20]. Tailored educational interventions, perhaps using social media for young MSM, and other creative methods to reach older MSM are urgently needed.

A novel finding of the current study is that low pride and acceptance of homosexuality was one of the factors most strongly associated with low HIV/AIDS knowledge among US MSM. Low pride and acceptance of homosexuality may correlate with willingness or ability of US MSM to access prevention services or with the extent of peer networks through which HIV education is shared. Additional studies are needed to characterize how gay identity formation in the United States may be related to other HIV-associated factors, especially in light of the recently questioned relationship between internalized homophobia and risky sexual behaviors [21].