You are here
Impact of Male Circumcision Scale-Up on Community-Level HIV Incidence in Rakai, Uganda
Xiangrong Kong1, Godfrey Kigozi2, Fred Nalugoda2, David Serwadda3, Maria J. Wawer1, Ronald H. Gray1
1 Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, United States. 2 Rakai Health Sciences Program, Entebbe, Uganda. 3 Makerere University College of Health Sciences, Kampala, Uganda.
Background: Randomized trials show male circumcision (MC) reduces individual level risk of HIV acquisition by 50-60% in men, and MC has been scaled up in sub-Saharan Africa. We assessed the impact of MC coverage on community level HIV incidence among non-Muslim men in Rakai, Uganda.
Methods: The Rakai Community Cohort Study (RCCS) conducts ~ annual surveillance of consenting residents aged 15-49 in 43 rural communities consistently since 1999. Before 2003, MC was largely confined to Muslim men. MC services to non-Muslims were provided through randomized trials during 2004-07 and were scaled up via a PEPFAR program since 2008. For each community, the non-Muslim male population prevalence of MC and HIV incidence per 100 person years (py) was estimated before the MC trials (period 1: 1999 -2003), during the trials (period 2: 2004-07), and since the MC scale up (period 3: 2008-11). Incidence rate ratio (IRR) of community level (CL) HIV incidence associated with CL MC prevalence was estimated by Poisson log-linear regression. Adjusted IRR was also estimated adjusting for potential confounding due to potential population composition change over time (measured by the surrogate of time periods) and CL ART coverage among HIV+s.
Results: In period 1, the mean CL HIV incidence among non-Muslim men was 1.4/100pys (range 0-4.1/100pys); and mean CL MC prevalence was 4.1% (range 1-8%). ART was not available during this period. In period 2, the mean CL HIV incidence was 0.9/100pys (range 0-3.7/100pys), mean CL MC prevalence was 10.0% (range 0-30%), and the mean CL ART prevalence was 5.2% (range 0-13%). In period 3, the mean CL HIV incidence was 1.1/100pys (range 0-3.6/100pys), mean CL MC prevalence was 23.5% (range 0-45%), and mean ART prevalence was 17.9% (range 3-42%). For every 10% increase in CL MC prevalence among non-Muslims, the associated IRR of CL HIV incidence was 0.85 (p=0.006, 95% Confidence Interval [CI] 0.75-0.95). After adjusting for time period and CL ART prevalence, CL MC prevalence had an adjusted IRR of 0.81 (p =0.076, 95% CI 0.65-1.02).
Conclusions: The MC program reached 23.5% coverage among non-Muslim men in Rakai by 2011, with large variations across communities. A 10% increase in community level MC coverage was associated with a 15% (95%CI 5-25%) reduction in CL HIV incidence among non-Muslim men, and this reduction of CL HIV risk was robust after adjusting for potential confounding due to time period and CL ART coverage.