Understanding the differences between contrasting HIV epidemics in east and west Africa: results from a simulation model of the Four Cities Study.

OBJECTIVE To determine if the differences in risk behaviours, the proportions of males circumcised and prevalences of sexually transmitted infections (STIs) observed in two African cities with low prevalence of HIV (Cotonou, Benin, and Yaoundé, Cameroon) and two cities with high prevalence (Kisumu, Kenya, and Ndola, Zambia) could explain the contrasting HIV epidemics in the four cities. METHODS An individual-based stochastic model, STDSIM, was fitted to the demographic, behavioural and epidemiological characteristics of the four urban study populations based on data from the Four Cities Study and other relevant sources. Model parameters pertaining to STI and HIV natural history and transmission were held constant across the four populations. The probabilities of HIV, syphilis and chancroid acquisition were assumed to be doubled among uncircumcised males. A priori plausible ranges for model inputs and outputs were defined and sexual behaviour characteristics, including those pertaining to commercial sex workers (CSWs) and their clients, which were allowed to vary across the sites, were identified based on comparisons of the empirical data from the four sites. The proportions of males circumcised in the model, 100% in Cotonou and Yaoundé, 25% in Kisumu and 10% in Ndola, were similar to those observed. A sensitivity analysis was conducted to assess how changes in critical parameters may affect the model fit. RESULTS Population characteristics observed from the study that were replicated in the model included younger ages at sexual debut and marriage in east Africa compared with west Africa and higher numbers of casual partners in the past 12 months in Yaoundé than in the other three sites. The patterns in prevalence of STIs in females in the general population and CSWs were well fitted. HIV prevalence by age and sex and time trends in prevalence in the model were consistent with study data with the highest simulated prevalences in Kisumu and Ndola, intermediate in Yaoundé and lowest in Cotonou. The sensitivity analysis suggested that the effect of circumcision on the development of the HIV epidemics may have been mediated indirectly by its effect on ulcerative STI. CONCLUSIONS The contrasting HIV epidemics in east and west Africa could be replicated in our model by assuming that male circumcision reduced susceptibility to HIV, syphilis and chancroid. Varying rates of male circumcision may have played an important role in explaining the strikingly different HIV epidemics observed in different parts of sub-Saharan Africa.