Portal de Periódicos da CAPES

Population-level risk factors for HIV transmission and ‘the 4 Cities Study'

2002; Lippincott Williams & Wilkins; Volume: 16; Issue: 15 Linguagem: Inglês

10.1097/00002030-200210180-00025

1473-5571

Marie‐Claude Boily, Catherine M Lowndes, Simon Gregson,

HIV/AIDS Research and Interventions

The authors of the carefully designed 4 Cities Study [1] reported that male non-circumcision and herpes simplex virus type 2 (HSV-2) were strong correlates of individual HIV infection in all cities, and were more common in high (Kisumu, Ndola) than low (Cotonou, Yaounde) HIV prevalence cities. Surprisingly, behavioural risk factors were not more common in Kisumu and Ndola [2,3]. We wish to elaborate on the potential significance of sexual mixing patterns and AIDS differential mortality. Mathematical models have shown the importance of sexual mixing patterns on HIV prevalence [4]. The limited data on mixing patterns in the 4 Cities Study suggest that significant differences may exist between cities [2,3,5]. For example, compared with Cotonou, the population of Yaounde might be more sexually homogeneous and the pattern of commercial sex less assortative. Indeed, both sexes in the general population reported more lifetime partners and greater proportions reported non-spousal partners irrespective of marital status; men reported more contacts with female sex workers (FSW), who were less likely to be brothel-based [3] and had a smaller average number of clients; the degree of concurrency among the most sexually active unmarried men was greater. As expected given these differences, HIV prevalence was lower in FSW but higher in the general population in Yaounde than in Cotonou, and the female-to-male HIV prevalence ratio was also greater [6,7]. The broader mixing patterns between FSW, their clients and the remaining population are more difficult to ascertain/compare because details on the sexual characteristics of the male clients, on how many and how frequently high-activity individuals have sex with lower activity individuals are not reported [1,3]. Findings from an independent study in Cotonou [8] (estimating that 13% of the adult male population were clients of FSW compared with < 4% in the 4 Cities Study [3]), reinforced the likelihood of the under-reporting of male contact with FSW [2,3,5,9], suggesting a strong social desirability bias, which may itself vary between cities according to the prevailing social norms and sexual structures of their populations. In Yaounde, where sexual promiscuity seems most common and where FSW are less sexually active and are rarely brothel based [3,9], less stigma may be attached to commercial sex. The impact of AIDS differential mortality on the distribution in sexual activity in Kisumu and Ndola may be underestimated, given the simulation procedure used (which compares age cohorts by adjusting the sexual activity levels of older cohorts using comparative data from younger cohorts [9]). This procedure does not capture either the impact of AIDS mortality on younger age groups, or the individual and population-level behaviour adjustments resulting from the re-organization of the population's sexual structure that must occur as current and potential partners die [6] (e.g. widowed women may remarry or may engage in commercial sex work to support their families, etc. [3]). AIDS differential mortality can dilute the differences in sexual activity and sexually transmitted disease (STD) levels between high and low HIV prevalence cities over time, and may partly explain the lack of apparent association with HIV. Crudely, one would expect to find a positive (negative/weaker) population-level correlation between a risk factor and HIV prevalence in cities with small (large) or young (old) epidemics in which AIDS differential mortality is less (more) pronounced [6,7]. Evidence for this does indeed appear to exist [7,10,11]. In the 4 Cities Study paper [11] on concurrency, the fraction and the number of non-spousal partners, the fraction with more than one sexual partner, the number of partners, and the three concurrency measures, increase (decrease) with epidemic sizes among the low – which includes Dakar, Senegal – (high) HIV prevalence cities. Regarding STD co-factors, AIDS differential mortality would be expected to have a larger impact for ‘fragile’ (STD with low transmission probability, a short duration of infection or for which effective treatments exist, e.g. Neisseria gonorrhea and Chlamydia trachomatis) infections (especially at low prevalence) in which persistence relies primarily on the presence of high-risk individuals. Temporal changes in prevalence within a city can then exceed any differences that existed between cities at the beginning of the HIV epidemic, and can therefore reverse the original population-level correlation. For higher prevalence (trichomoniasis, syphilis) and ‘non-fragile’ STD (HSV-2 [12]) temporal changes may never exceed the initial difference in prevalence between cities, which may emphasize the relative role of ‘non-fragile’ STD over time as natural selection eliminates the ‘fragile’ STD. Among the low prevalence cities, N. gonorrhea, C. trachomatis, syphilis, trichomoniasis and HIV were more common in Yaounde than in Cotonou. However, overall, C. trachomatis and N. gonorrhea were less common in the high HIV prevalence cities whereas the opposite was true for HSV-2 (both sexes), syphilis and trichomoniasis (women). Therefore differences in sexual behaviours cannot be excluded yet as potential determinants of heterogeneity in epidemic sizes between cities. Further analyses based on mathematical models should help to evaluate the independent contribution of individual co-factors [9,13], and also to understand the issues discussed above.