Background:  HIV-1 prevalence is declining in Uganda, a finding often ascribed to increased rates of safe behaviors, especially abstinence and monogamy (be faithful), and to a lesser degree condom use (A, B, C).  We propose to extend the acronym by considering the contributions of HIV-related mortality (death/D), viral epidemiology/early infection (E), and potential future (F) effects of ART in Rakai, Uganda.

Methods:  To assess determinants of declines in HIV prevalence in Rakai, we used longitudinal population-based open cohort studies in 50 villages.

Results:  Between 1994/1995 and 2002/2003, age-standardized HIV prevalence declined significantly in all adults (17.6% to 11.4% [p = 0.007]) and in young adults aged 20 to 24 (16.9 to 7.7%), but not in adolescents aged 15 to 19 (3.6 to 3.2%).  Incidence did not decline overall (1.3/100 person-years in 1994/1995 and 1.7/100 person-years in 2002/2003), in young adults (1.1/100 person-years and 1.5/100 person-years, respectively), or in adolescents (1.3/100 person-years and 1.2/100 person-years). In the same period, age of sexual debut declined in both sexes, and the proportions of young adults reporting sexual activity, non-marital relationships, and multiple partnerships increased.  However, condom use with casual partners rose significantly, particularly at younger ages (for example, from 19% to 38% in males aged 15 to 19).  Mortality among HIV⁺ persons was 13.3/100 person-years.  Approximately half of all HIV transmissions in this population occurred from index HIV⁺ partners who had themselves recently seroconverted and were in early stage HIV infection.     

Conclusions:  We observed no increase in abstinence or monogamy (no evidence for A or B), but condom use increased in casual relationships (evidence for C). Mortality (death/D) removed ~70 more HIV⁺ persons per year than were added through new seroconversions, accounting for much of the observed decline in prevalence. HIV incidence and thus the number of HIV transmissions contributed by persons in early stage HIV infection (E) remained relatively stable in this period. ART is unlikely to reduce HIV incidence, since most transmissions occur prior to index partner ART eligibility. Moreover, ART availability may result in behavioral disinhibition, increased risk behavior, and higher HIV incidence. In summary, declines in Rakai HIV prevalence in the past decade are associated primarily with C and D.  Prevention of ART-related behavioral disinhibition is crucial to contain the future course of the epidemic.   

Keywords: declines in HIV prevalence; Uganda; biological and behavioral determinants