Men Together Making a Difference: Reducing HIV/STD Risk Behavior Among South African Men

Sub-Saharan Africa has just over 10% of the world's population, but according to UNAIDS was home to more than 60% of all people living with HIV in 2003-some 25.4 million people. South Africa continues to have the largest number of people living with HIV in the world. As in other parts of sub-Saharan Africa, heterosexual exposure is the primary HIV transmission category. An estimated 5.3 million South Africans-2.9 million women and 2.4 million men-were living with HIV at the end of 2003. Worldwide, efforts to stem the spread of HIV among heterosexuals have stressed the impact of HIV on women. Oft-cited statistics indicate that about half of all people living with HIV are women. It is often suggested that women are seldom free to make empowered choices and face a range of HIV-related vulnerabilities that men do not face-many of which are embedded in the social relations and economic realities of their societies. Women's economic dependence on their male partners and the fact that women do not have the power to abstain from sex or to insist on condom use-even when they suspect that their man has other sexual partners and might have HIV-is emphasized. The hazards of young women's sexual relationships with older men and the high rate of rape and other forms of sexual coercion are cited. Men are typically mentioned as injection drug users, as having sex with other men-whether as MSMs or being on the "down low"-or as being the cause of the spread of HIV in women. The strategies typically offered to address the impact of HIV on women are interventions with women. For instance, it is recommended that women be taught the information and skills to make decisions about the terms of their sexual relationships, that methods of protection that women can control (e.g., microbicides) be developed, and that boosting women's economic opportunities and social power should be seen as part and parcel of potentially successful and sustainable HIV prevention strategies. To be sure, there is an alternative approach to addressing women's risk of heterosexual transmission of HIV, one that would be important complement to the predominant approach: namely, focusing on men. Although it is noted that HIV is affecting women most severely in places where heterosexual exposure is a dominant mode of transmission, this seldom leads to a recommendation that interventions be developed to change men's behavior. Yet, by reducing men's sexual risk behavior, it should be possible to reduce rates of HIV in both men and women. Men's rates would decline because they are the recipients of the intervention; women's rates would decline because they have sex with men. Interventions targeting men could take into account the power that men have in sexual decision-making and risk taking. However, whether one considers the US literature or the international literature, few randomized controlled trials (RCTs) of HIV/STD risk-reduction interventions have targeted heterosexual men. Elwy and colleagues 2002 review of HIV/STD prevention intervention studies revealed that only 12 of 1157 studies worldwide were conducted on males only. Most were not RCTs demonstrating intervention efficacy. Only 2 demonstrated significant effects on mediators and behaviors, and neither was a RCT. In addition, scant attention has focused on the general population of heterosexual men in any region. Indeed, 8 of the 12 studies focused on incarcerated men, STD patients, substance abusers, miners, or truck drivers. To address this gap in the literature, the proposed trial will focus on men who are more representative of the general population. In many developing countries, including South Africa, the HIV/AIDS epidemic is generalized, and there is a need to develop and test interventions for a broad range of the population, not just special high-risk sub-populations. Accordingly, the purpose of this research is to develop and test the efficacy of an intervention to reduce behaviors that create the risk for contracting and transmitting STD, including HIV, among South African men who have sex with women. A cluster-randomized control trial design will be used to reduce the potential for contamination between treatment arms that would be present if individuals were randomized. An attention control group will be used to control for Hawthorne effects, special attention, and group interaction. Men will be recruited from Black townships surrounding East London, including Mdantsane, Scenery Park, Duncan, Village, and Gompo Town, and the semi-rural area of Berlin in the Eastern Cape Province of South Africa. More than 98% of the residents of these areas are Black Africans, and isiXhosa is the first language for 98.8% of the population. There are 206 neighborhoods defined as geographical clusters tied to census data in this catchment area, allowing the creation of 103 matched pairs of neighborhoods similar on the percentage isiXhosa-speaking, percentage married, percentage male, percentage living in informal dwellings, percentage unemployed, and population size. From the 103 matched pairs, 22 pairs will be randomly selected for the trial. One neighborhood in each pair will be randomly assigned to each of the 2 study arms. Before recruiting from a neighborhood, meetings with community leaders (e.g., councilor, clergy) will be held to enlist their support. In addition, a meeting will be held to inform men in the neighborhood about the study. Recruiters will inform potential participants about the study, obtain consent to be screened, and conduct a brief screening interview to determine eligibility and willingness to participate. Eligible men will be invited to participate in the "Men, Together Making A Difference Project" designed to understand men's behaviors that may create health risks such as heart disease, cancer, and STDs, especially HIV, and to find ways to teach men how to reduce these risks. A common participant recruitment and enrollment protocol, including use of the same posters and other materials, will be followed in the neighborhoods in both conditions. Eligible men will be recruited in advance of randomization so that at the time they agree to participate they will be blind to the specific intervention they will receive. This procedure will reduce the probability of self-selection bias into the different conditions of the trial. All participants will complete self-report measures via audio computer-assisted self-interviewing before the intervention, immediately after, and 6 and 12 months after the intervention. Several steps will be taken to increase the validity of self-report measures. Participants will be given a calendar, with the dates clearly marked. This will make salient to respondents the dates that are included when they are asked to recall their behavior "in the past 3 months" and that they should be specific. The importance of responding honestly will be emphasized. They will be informed that their responses will be used to create programs for South African men like themselves and that this will be possible only if they answer the questions honestly. This pits the social responsibility motive against the social desirability motive. Participants will be assured that their responses will be kept confidential. Facilitators who lead the intervention groups will not be involved in any way in the data collection. The use of ACASI should also serve to increase participants' motivation to respond accurately. There are 2 Specific Aims. Aim 1 is to test the primary hypothesis that men who receive a culturally appropriate theory-based HIV/STD risk-reduction intervention will be more likely to report consistently using condoms during vaginal intercourse in the 12-month post intervention period than will men who receive an attention-control intervention. Aim 2 is to test the secondary hypothesis that outcome expectancies and self-efficacy to use condoms mediate the HIV/STD risk-reduction intervention's effect on condom use. In addition, an exploratory aim is to conduct hypothesis-generating analyses on whether the efficacy of the intervention varies depending on neighborhood characteristics or participants' baseline characteristics. The effects of HIV/STD risk-reduction interventions may differ as a function of the neighborhood's unemployment rate, percentage living in informal dwellings/shacks, percentage married, or sex ratio. Potential individual-level moderators include age, marital status, language use (English versus isiXhosa), and alcohol and drug use. The unit of inference in this trial is the individual. This is because the trial is designed to test the efficacy of a behavior-change intervention based on individual-level behavior change theory. As Donner and Klar in 2000 noted in their influential textbook on cluster-randomized trial, the unit of inference, not the unit of randomization, determines the unit of analysis. The data will be analyzed using an intention-to-treat mode, with participants analyzed based on their intervention assignment, regardless of the number of intervention or data-collection sessions attended. The primary aim focuses on testing for significant differences between two treatment conditions over the post intervention period. Major statistical challenges arise in the proper handling of repeated clustered outcomes. Each variable of interest is completed by each man, nested within a neighborhood, thus creating a correlated outcome. Moreover, the multiple assessments of each variable over the study period produce correlated repeated outcomes. The primary challenge in the analysis of such data is appropriate adjustments for the differential treatment means between clusters and the correlations among the observations within a cluster (cluster effects). Most statistical models assume stochastic independence among observations and thus are inappropriate for clustered data. In this trial, generalized estimating equations (GEE) modeling will be employed to handle the clustered data appropriately. GEE modeling avoids explicit modeling of the within-cluster correlations by basing statistical inferences of model parameters on marginalized likelihood or generalized estimating equations. Since GEE requires a relatively large sample size, it is not appropriate for small studies. Given the large sample size, GEE is appropriate for in this trial. The implementation for this trial is relatively straightforward. For instance, to determine whether a greater percentage of the men who receive the HIV/STD risk-reduction intervention report consistent condom use 6 and 12 months post intervention, as compared with those who received the health-promotion intervention, the model will include time-independent covariates, baseline measure of consistent condom use, intervention type, and time (2 categories representing 6- and 12-month follow-up), which will provide the effect of the intervention over the 2 follow-up visits. A sample size calculation was performed to detect an a priori effect size of a 10% increase in consistent condom use from 32% to 42% in the HIV/STD risk-reduction intervention condition, adjusting for the expected variance inflation due to clustering. A 10% increase was selected as a clinically and substantively important effect size. Based on pilot data, an intraclass correlation (ICC) of 0.01 was estimated. Assuming alpha = 0.05, a 2-tailed test, ICC = 0.01, 15% attrition at 12-month follow-up, and N = 1,152 men in the trial from 44 neighborhoods with an average of 26 men in each neighborhood, the trial was estimated to have 81% power to detect a 10% increase in consistent condom use from 32% to 42% in the HIV/STD intervention group. Assuming the same effect size, hypothesis tests on secondary sexual behavior outcomes and theoretical mediator variables will have similar statistical power.