The 2008 UNAIDS report on the global epidemic shows a diverse range of HIV epidemics unfolding in all corners of the globe . Encouragingly, this includes substantial declines in some of largest epidemics, in parts of Western, Southern and Eastern Africa (Fig. 1). However, interpreting trends in prevalence requires substantial care, as the long survival time with HIV means that prevalence measurements record the historical, rather than the current, trajectory of the epidemic. Moreover, we expect that, even without changes in behaviour, there will be a natural evolution in HIV incidence early in the epidemic, as the focus of transmission shifts from those at higher risk of infection to those at lower risk of infection. Mathematical models indicate that this transition is expected to bring down prevalence in generalized epidemics, approximately 15–25 years after HIV spreads through the higher risk groups  – or, put another way: about now.
The substrate for these natural epidemiological dynamics is variation between individuals in the risk of acquiring and transmitting infection. Until now, models have concentrated on differences in sexual behaviour, including the number of sexual partners, the timing of partnerships, the chance that these partners are infected with HIV and, if they are, the stage of the infection [2,3]. It has also been hypothesized that subfertility associated with HIV could selectively remove infected women from the antenatal clinical data as the epidemic matures, or that asynchronous epidemics in subpopulations, could spuriously generate artificial trends in observed prevalence, but these are less likely to be significant factors [2,4,5].
The article by Nagelkerke et al., published today in AIDS, demonstrates the potential influence of another source of variation – differences in the biological susceptibility of individuals to HIV. The authors fitted a model, that allowed for heterogeneity in susceptibility, to observational data from a cohort of Kenyan sex workers showing that the average risk of infection per sex act declined by four-fold between 1985 and 2000. Extrapolating the inferred distribution of susceptibility to the general population, the authors show how this source of variation could cause epidemics to decline as they mature, even without behaviour change.
On one hand, the observations from the cohort of sex workers could underestimate susceptibility to HIV, as these women were presumably heavily exposed to HIV before recruitment to the cohort, removing those most at risk from the study. However, it seems more likely that the susceptibility is overestimated in the model (the best fit is for 40% of women to be almost resistant to infection). Reductions in the average transmissibility of HIV – either through control of bacterial sexually transmitted infection cofactors for HIV , or through the changing phase of the epidemic leading to fewer clients having highly infectious primary infection at the time of contact – could contribute to the observed reduction in risk for sex workers. Similarly, changes in the pattern of client–sex workers interactions (especially the degree of fidelity and the number of sex acts per unique client ) could also reduce risk. If these factors were included in the model, the independent effect of variation in biological susceptibility may be diminished. More simply, increasing social desirability bias for reporting more condom use (which is reported on a qualitative scale of ‘seldom’, ‘often’, etc. ), could lead to trends in the calculation of incidence per sex act when incidence per time remained constant.
Nevertheless, the authors demonstrate that, in principle, heterogeneity in susceptibility to HIV infection could lead to declines in prevalence, and the authors are right to contend that this confounds simplistic analyses of prevalence to infer changes in sexual risk behaviour. Such inferences, however, are possible if factors that could contribute to declines are appropriately represented in mathematical models. This is hampered by limited knowledge about the extent to which hypothetical factors come into play (through incomplete information about the variation in sexual behaviour and variation in biological susceptibility, etc.). One approach is to compare observed trends with an extreme model counterfactual, with the strongest possible ‘natural decline’ in prevalence. The prevalence declines in Uganda, for instance, cannot be replicated by models making even the most extreme assumptions about natural dynamics . Another approach would be to make a fairer representation of the uncertainty around each factor, and use model comparison techniques to draw conclusions that hold across the range of credible parameter values.
The future of HIV epidemic monitoring is likely to rely on HIV prevalence for many years to come. After years of intensive research, direct measurements of incidence in local cohort studies are becoming less and less representative of whole countries, and assays that discriminate recent infections in cross-sectional serosurveys have been shown to be unreliable in African countries without calibration [11–13]. Antiretroviral therapy will add a further layer of complexity, as longer survival times will tend to increase HIV prevalence; so that upturns in epidemics may not indicate increased risk behaviour, and stable prevalence rates could mask substantial reductions in incidence. It will, therefore, be essential to make maximum use of mathematical modelling in the interpretation of trends in HIV prevalence. To be conservative and defensible, these models must reasonably account for all other potential sources of natural changes in epidemics, so that the contribution of actual reductions in risk – if any – can be resolved. And only from that starting point, can the important investigations into the proximal and distal causes and reasons for the behaviour changes begin .
2. Hallett TB, Aberle-Grasse J, Bello G, Boulos LM, Cayemittes MPA, Cheluget B, et al
. Declines in HIV prevalence can be associated with changing sexual behaviour in Uganda, urban Kenya, Zimbabwe, and urban Haiti. Sex Transm Infect 2006; 82:i1–i8.
3. Kilian AH, Gregson S, Ndyanabangi B, Walusaga K, Kipp W, Sahlmuller G, et al
. Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda. AIDS 1999; 13:391–398.
4. Garnett GP, Gregson S. Monitoring the course of the HIV-1 epidemic: the influence of patterns of fertility on HIV-1 prevalence estimates. Math Popul Stud 2000; 8:251–277.
5. Walker P, Hallett TB, White PJ, Garnett GP. Interpreting declines in HIV prevalence: the impact of spatial aggregation and migration on expected declines in prevalence. Sex Transm Infect 2008; 84(Suppl 2):ii42–ii48.
6. Nagelkerke NJ, Vlas SJ, Jha P, Luo M, Plummer FA, Kaul R. Heterogeneity in host HIV susceptibility as a potential contributor to recent HIV prevalence declines in Africa. AIDS 2009; 23:123–128.
7. Rottingen JA, Cameron DW, Garnett GP. A systematic review of the epidemiologic interactions between classic sexually transmitted diseases and HIV: how much really is known? Sex Transm Dis 2001; 28:579–597.
8. Ghani AC, Aral SO. Patterns of sex worker-client contacts and their implications for the persistence of sexually transmitted infections. J Infect Dis 2005; 191(Suppl 1):S34–S41.
9. Kimani J, Kaul R, Nagelkerke NJ, Luo M, MacDonald KS, Ngugi E, et al
. Reduced rates of HIV acquisition during unprotected sex by Kenyan female sex workers predating population declines in HIV prevalence. AIDS 2008; 22:131–137.
10. UNAIDS. Evidence for HIV decline in Zimbabwe: a comprehensive review of the epidemiological data
. 2005. www.epidem.org
11. UNAIDS Reference Group on Estimates Modelling and Projections. Statement on the use of the BED-assay for the estimation of HIV-1 incidence for surveillance or epidemic monitoring.Wkly Epidemiol Rec
12. Karita E, Price M, Hunter E, Chomba E, Allen S, Fei L, et al
. Investigating the utility of the HIV-1 BED capture enzyme immunoassay using cross-sectional and longitudinal seroconverter specimens from Africa. AIDS 2007; 21:403–408.
13. Mermin J, Musinguzi J, Opio A, Kirungi W, Ekwaru JP, Hladik W, et al
. Risk factors for recent HIV infection in Uganda. JAMA 2008; 300:540–549.
14. Hallett TB, White PJ, Garnett GP. Appropriate evaluation of HIV prevention interventions: from experiment to full-scale implementation. Sex Transm Infect 2007; 83:i55–i60.