What impact will reducing concurrency have on decreasing the incidence of HIV in heterosexual populations?

The importance of concurrency (i.e., multiple overlapping sexual partnerships) in driving HIV epidemics has been hotly debated in the past 20 years. Some have hypothesized that concurrency is an extremely important driver of HIV epidemics and may explain the high prevalence of HIV in the general population of many sub-Saharan African countries. Notably, proponents of this hypothesis acknowledge a host of other factors, such as multiple sex partners, core groups, and the presence of other sexually transmitted infections, are also important. A substantial amount of data have been collected on sexual behavior (e.g., the number of sex partners individuals’ acquire each year), but there are only limited data on the level and/or degree of concurrency in any population.1 Consequently, it has been impossible to use empirical data to evaluate whether concurrency is a major driver of HIV epidemics. This lack of empirical evidence has led others to dispute the idea that concurrency is an important driver. Mathematical modelers, unconstrained by empirical data, have been investigating the potential impact of concurrency on HIV epidemics since the early 1990s.2 Modeling by Morris and Kretzschmar3–4 has shown that if levels of concurrency in heterosexual populations are high (i.e., 50%), the prevalence of HIV could be approximately 10 times higher than if individuals were serially monogamous. In an article in this issue, McCreesh et al5 present a new model representing heterosexual transmission of HIV. They use their model to evaluate the potential epidemiologic impact of hypothetical public health interventions that focus on reducing concurrency in heterosexual populations. McCreesh et al have developed a stochastic network model of sexual partnerships, some of which can be concurrent, in a heterosexual population where HIV is being transmitted. They used data on demography, sexual behavior, and HIV prevalence collected in 2008 in Uganda to parameterize their model. The sexual behavior data they used contained information on both the number of sex partners and the level of concurrency. These data were from the Masaka General Population Cohort study, which is a longitudinal study that has been running since 1989. Study participants are approximately 7000 residents of 25 villages in the rural south-west of Uganda. Each year face-to-face interviews are conducted; blood samples are collected and tested for HIV. Concurrency data were collected from 1214 men and 1470 women, a subset of the cohort study. Concurrency was defined as “overlapping sexual partnerships where sexual intercourse with 1 partner occurs between 2 acts of intercourse with another partner.” Concurrency was further classified into 2 types: (i) sex with spouses and ex-spouses, defined as concurrent partnership of long duration, and (ii) sex with others, defined as concurrent partnership of short duration. The survey data, used to parameterize the model, show that 9.6% of men in these rural villages in Uganda reported having engaged in a concurrent partnership of either type, but only 0.2% of women reported doing so. McCreesh et al assumed that the concurrency level reported by women was significantly lower than the actual value. Consequently, in their modeling, they increased the level of concurrency for women from the reported level of 0.2% to 2.4%, a 12-fold increase. They then used their stochastic network model to evaluate the potential impact of hypothetical public health interventions that focus on reducing concurrency. The authors found that when concurrency was reduced by 20%, the incidence of HIV (during a 10-year period) decreased by 9.2% in women, but only by 4.1% in men. When concurrency was reduced by 50%, the incidence of HIV (during a 10-year period) decreased by 16.2% in women, but only by 6.0% in men. Notably, if the authors had not inflated the level of concurrency in women, the reductions in incidence would have been