Scale-Up of ART and VMMC Explain a Two-Fold Decline in HIV Incidence in Western Anna Bershteyn1, Adam N. Akullian1, Daniel J. Klein1, Britta L. Jewell2, Kennedy K. Mutai3, and Samuel M. Mwalili4 1. Institute for Disease Modeling, 3150 139th Ave. SE, Bellevue, WA, USA 2. Department of Medicine, University of California and San Francisco, Zuckerberg San Francisco General Hospital, 1001 Potrero Avenue San Francisco, CA 94110, USA 3. National Aids Control Council, P.O. Box 61307-00200, Argwings Kodhek Rd, , Kenya 4. Centers for Disease Control and Prevention, PO Box 606, Village Market, Nairobi 00621, Kenya

Background Methods Figure 3. Model fit to age/sex-specific HIV prevalence in the six counties. Figure 4. Model fit to age-specific HIV prevalence (15-49) in County. Western Kenya has among the world’s highest prevalence of HIV. In Siaya EMOD-HIV, an individual-based HIV transmission and care and Counties, approximately one in four adults is living with HIV. continuum model [2], was used to simulate the HIV pandemic in Longitudinal surveillance of a community in Siaya [1] found that incidence Western Kenya. The model was calibrated to age-, sex-, and county- had fallen by two-fold between 2011 and 2016. We used mathematical specific HIV prevalence estimates from four national surveys, as modeling to estimate the relative contribution of antiretroviral therapy well as estimates of population size and structure, number on ART, (ART) and voluntary male medical circumcision (VMMC) to the declines in number receiving VMMC, and national targets for VMMC coverage HIV incidence in the Western Kenya region, including the Siaya County. [3]. No incidence data were available for model calibration. Conservatively, future projections assumed a sustained ART coverage of approximately 60%, lack of large-scale PrEP scale-up, and maintenance of 80% VMMC coverage. The baseline model trajectories were modified to simulate what would have happened in the absence of ART and/or VMMC. Results Calibration yielded 250 best-fitting model trajectories for each of six counties comprising the Nyanza region of Western Kenya. EMOD-HIV recapitulated the halving of HIV incidence over 2011-2016 in Siaya (county-wide), despite the model fitting process not directly utilizing Conclusions References incidence estimates from the longitudinal surveillance site in this county. Epidemiological modeling conducted prior to, and without use of, [1] Borgdorff MW, Kwaro D, Obor D, et al. HIV incidence in western Kenya during incidence estimates in Gem, Siaya – but using HIV prevalence and scale-up of antiretroviral therapy and voluntary medical male circumcision: a Estimated HIV incidence declined drastically in Siaya due to scale-up of intervention data from the same time period – agrees with the population-based cohort analysis. Lancet HIV. 2018;5(5):e241-e249. ART and VMMC, without which incidence would have remained stable at dramatic incidence declines observed in longitudinal surveillance. 1.7 new infections per 100 person-years among adults age 15-49. [2] Bershteyn A, Gerardin J, Bridenbecker D, et al. Implementation and Modeling suggests that observed incidence declines can be fully applications of EMOD, an individual-based multi-disease modeling platform. Incidence peaked in 2002, fell to half of its peak by 2018, and continued attributed to scale-up of ART and VMMC, without which incidence Pathogens and Disease 2018; 76 (5). to decline to one-third of peak levels by 2028. ART is the predominant Figure 1. Map showing the six would have remained stable at high levels. Incidence is expected to cause of incidence declines up until 2025, after which VMMC is expected [3] Bershteyn A, Mutai KK, Akullian AN, et al. The influence of mobility among counties of the former Nyanza continue to decline if intervention coverage is maintained, but to surpass ART as a driver of incidence decline, provided Siaya achieves high-risk populations on HIV transmission in Western Kenya. Infectious Disease province of Western Kenya. Siaya enhanced efforts to prevent HIV infections will be required to reduce and maintains a target of 80% VMMC coverage. Similar trends were Modelling 2018;3, 97-106. County is highlighted. Blue star shows incidence to levels consistent with the goal of ending the HIV epidemic. the location of the Gem community. found in other high-prevalence counties in Western Kenya. Image credit: Omondi [CC BY-SA 3.0] Figure 5. Modeled HIV incidence declines in Siaya County are attributable to ART and VMMC. Figure 6. ART has been the main driver of incidence declines, but VMMC may surpass it by 2025. Figure 2. Observed HIV incidence decline in Gem, Siaya. Adapted from Borgdorff et al., Lancet HIV 2018. 100 PY 100 PY 100 PY 100 PY per 100 PY 100 PY per per Modeled Adult HIV Incidence Modeled Adult Modeled Adult HIV Incidence Modeled Adult Adult (15+) HIV Adult (15+) Incidence

www.idmod.org | github.com/InstituteForDiseaseModeling Presented at the 25th Conference on Retrovirology and Opportunistic Infections | March 4-7, 2019 | Seattle, WA, USA