Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda

OBJECTIVE: To monitor the HIV-1 epidemic in Western Uganda and the possible impact of interventions. DESIGN: Results from sentinel surveillance of HIV-1 seroprevalence were compared with cross-sectional serosurvey data and model simulations. METHODS: Age-specific trends in HIV-1 prevalence between 1991 and 1997 amongst antenatal clinic (ANC) attenders in the town of Fort Portal, where a comprehensive AIDS control programme has been implemented since 1991, were analysed. Results were compared with outputs from a mathematical model simulating the HIV-1 epidemic in Uganda. Two scenarios were modelled: one without and one with behaviour change. Sentinel surveillance data were compared with data from a population-based HIV-1 serosurvey at the study site, which was carried out in early 1995. RESULTS: Data from 3271 ANC attenders identified greater education and being single as risk factors for HIV-1 infection. A significant decrease of risk for women with secondary school education over time was observed, whereas the risk for illiterate women remained high. Among women aged 15-19 years (n = 1045) education and marital status-adjusted HIV-1 prevalence declined steadily from 32.2% in 1991 to 10.3% in 1997. For 20-24-year-old women (n = 1010) HIV-1 prevalence increased until 1993 from 19.9% to 31.7% and decreased thereafter (21.7% in 1997). These trends closely follow the prediction of the model simulation assuming behaviour change, and for 1995-1997, confidence intervals of the HIV-1 prevalence estimate exclude the model output for an uninfluenced epidemic. No clear trends of HIV-1 prevalence were found in older women (n = 1216) and comparisons with the model were ambiguous. Sentinel surveillance data at the time of the population survey closely reflected results for the female general population sample for the two younger age-groups (15-19 and 20-24 years). In contrast, pregnant women aged 25-29 years showed significantly lower rates than the population sample (20.8% versus 45.1%). CONCLUSION: HIV-1 prevalence amongst ANC attenders aged 15-24 years can be used to monitor the HIV-1 epidemic in the given setting. Declining trends of HIV-1 prevalence in women aged 15-19 and 20-24 years most likely correspond to a reduced HIV-1 incidence attributable to changes in behaviour. Our data also show that sentinel surveillance data need to be age-stratified to give useful information.