Combination HIV prevention among MSM in South Africa: results from agent-based modeling

Abstract

HIV prevention trials have demonstrated the effectiveness of a number of behavioral and biomedical interventions. HIV prevention packages are combinations of interventions and offer potential to significantly increase the effectiveness of any single intervention. Estimates of the effectiveness of prevention packages are important for guiding the development of prevention strategies and for characterizing effect sizes before embarking on large scale trials. Unfortunately, most research to date has focused on testing single interventions rather than HIV prevention packages. Here we report the results from agent-based modeling of the effectiveness of HIV prevention packages for men who have sex with men (MSM) in South Africa. We consider packages consisting of four components: antiretroviral therapy for HIV infected persons with CD4 count <350; PrEP for high risk uninfected persons; behavioral interventions to reduce rates of unprotected anal intercourse (UAI); and campaigns to increase HIV testing. We considered 163 HIV prevention packages corresponding to different intensity levels of the four components. We performed 2252 simulation runs of our agent-based model to evaluate those packages. We found that a four component package consisting of a 15% reduction in the rate of UAI, 50% PrEP coverage of high risk uninfected persons, 50% reduction in persons who never test for HIV, and 50% ART coverage over and above persons already receiving ART at baseline, could prevent 33.9% of infections over 5 years (95% confidence interval, 31.5, 36.3). The package components with the largest incremental prevention effects were UAI reduction and PrEP coverage. The impact of increased HIV testing was magnified in the presence of PrEP. We find that HIV prevention packages that include both behavioral and biomedical components can in combination prevent significant numbers of infections with levels of coverage, acceptance and adherence that are potentially achievable among MSM in South Africa.

Figure 1. Results from 2252 simulations of agent-based model of HIV spread among MSM in South Africa corresponding to 163 distinct combinations of HIV prevention interventions.

Each point represents replicates for a particular combination of HIV prevention interventions. Plotted are the mean percentages infected over 5 years for each intervention (averaged over replicates) versus the standard deviations of those percentages. Combination prevention interventions which included a ≥25% reduction in UAIs are indicated in dark blue, all others are indicated in light blue. The data point in red corresponds to the 60 simulation runs for the control setting of no intervention.

Figure 2. HIV infections prevented over 5 years from combination prevention interventions with four components.

ART coverage of eligible persons who were not already receiving ART at baseline, PREP with 50% acceptance (dotted lines), 15% UAI reduction (blue lines; no UAI change are in red) and increase in HIV testing (black triangles). See Table 1 for further details about the components of the prevention interventions.

Figure 3. HIV infections prevented over 5 years from combination prevention interventions with four components.

ART coverage of eligible persons who were not already receiving ART at baseline, PREP with 25% acceptance (dotted lines), 25% UAI reduction (blue lines; no UAI change are in red) and increase in HIV testing (black triangles). See Table 1 for further details about the components of the prevention interventions.