Evaluating HIV prevention efforts using semiparametric regression models: results from a large cohort of women participating in an HIV prevention trial from KwaZulu-Natal, South Africa

OBJECTIVE: To describe and quantify the differences in risk behaviours, HIV prevalence and incidence rates by birth cohorts among a group of women in Durban, South Africa. METHODS: Cross-sectional and prospective cohort analyses were conducted for women who consented to be screened and enrolled in an HIV prevention trial. Demographic and sexual behaviours were described by five-year birth cohorts. Semiparametric regression models were used to investigate the bivariate associations between these factors and the birth cohorts. HIV seroconversion rates were also estimated by birth cohorts. RESULTS: The prevalence of HIV-1 infection at the screening visit was lowest (20.0%) among the oldest (born before 1960) cohorts, while the highest prevalence was observed among those born between 1975 and 79. Level of education increased across the birth cohorts while the median age at first sexual experience declined among those born after 1975 compared to those born before 1975. Only 33.03% of the oldest group reported ever using a condom while engaging in vaginal sex compared to 73.68% in the youngest group; however, HIV and other sexually transmitted infection (STI) incidence rates were significantly higher among younger women compared to older women. CONCLUSIONS: These findings clearly suggest that demographic and sexual risk behaviours are differentially related to the birth cohorts. Significantly high HIV and STI incidence rates were observed among the younger group. Although the level of education increased, early age at sexual debut was more common among the younger group. The continuing increase in HIV and STI incidence rates among the later cohorts suggests that the future trajectory of the epidemic will be dependent on the infection patterns in younger birth cohorts.