1. K. Jean, P. Lissouba, D. Taljaard, R. Rain-Taljaard, B. Singh, J. Bouscaillou, G. Peytavin, R. Sitta, S.G. Mahiane, D. Lewis, A. Puren, B. Auvert Inserm CESP U1018, Villejuif, France Université Versailles Saint-Quentin-en-Yvelines, Versailles, France CHAPS, Johannesburg, South Africa National Institute for Communicable Diseases, Johannesburg, South Africa HIV incidence among women is associated with their partners' circumcision status in the township of Orange Farm, South Africa (ANRS-12126) Melbourne – July, 25 th 2014

2. Introduction- The effect of VMMC roll-out on HIV among women Background – The effect of VMMC on HIV among men 3 RCTs documenting a protective effect of VMMC among men Recommended by WHO-UNAIDS since 2007 Effect of roll-out demonstrated among men (Auvert et al, Plos Med 2013) An effect on HIV among women? No direct effect observed in epidemiological studies (Weiss et al, Lancet ID 2009) An indirect effect through reduced exposure? Predicted by modelling studies (Williams et al, Plos Med 2006; Njeuhmeli et al, Plos Med 2011) VMMC: Voluntary Male Medical Circumcision RCT: Randomized Controlled Trial 2

3. To assess the association between their partner’s circumcision status and HIV incidence among women in Orange Farm (South Africa) Objective Context - The ANRS « Bophelo Pele » project Orange Farm: Township of ~110,000 adults Since 2007: roll-out of free VMMC VMMC: Voluntary Male Medical Circumcision 3 ORANGE FARM

4. Context (2) - The ANRS « Bophelo Pele » project Auvert et al, Plos Med 2013 MC prevalence increased from 11% in 2007 to 53% in 2011 4

5. Context (3) - The ANRS « Bophelo Pele » project HIV prevalence reduced by ~ 50% among circumcised men Auvert et al, Plos Med 2013 5

6. Data collected among women: 3 independant surveys (2007-2010-2012) Questionnaire: age, ethnic group, religion, occupation, age at first sexual intercourse, alcohol consumption, education, ever having been married, number of lifetime partners, consistent condom use Blood sample: HIV Methods (1) – Surveys 6 Pooled sample N= 4538 women (15-49 y) having ever had sexual intercourse

7. Methods (2) : Modelling HIV incidence from observed age-specific prevalence A classical mathematical deterministic compartmental modeling approach (Gregson et al, AIDS 1996; Williams et al, Stat Med 2001) Finding the age-incidence function that best fits the observed age-prevalence curve Use of propensity score weigthing to account for other covariates : survey, age, ethnic group, religion, occupation, age at first sexual intercourse, alcohol consumption, education, ever having been married, number of lifetime partners, consistent condom use 3 steps in the modelling process 1. Assuming a parametric age-incidence function 2. Calculating age-specific prevalence from parametric incidence 3. Fitting the predicted age-specific prevalence to the observed age-specific prevalence → Estimating the incidence parameters → Estimating incidence rate → Computing 95% confidence interval (Bootstrap) 7

8. N= 4538 women (15-49 y) having ever had sexual intercourse n=1363 (30.0%) women having had only circumcised partners (HHOCP) HIV prevalence rate : 22.4% [20.2% - 24.6%] n=3175 (70.0%) other women HIV prevalence rate : 36.6% [35.0% - 38.3%] aPRR: adjusted Prevalence Rate Ratio (computed with Poisson regression) aPRR: 0.85 [0.76-0.95] Reduction: 15% 2[5%-24%] 8 Results (1) – HIV prevalence among women

9. Results (2) – Fitting age-specific HIV prevalence 9 Women HHOCP: n=1363 Other women: n=3175 HHOCP: Having Had Only Circumcised Partners

10. Results (3) – Estimating HIV incidence 10 Women HHOCP: n=1363 Other women: n=3175 HHOCP: Having Had Only Circumcised Partners Normal line: Women HHOCP Bold line: Other women (95% bootstrapped confidence intervals)

11. Results (4) – Effect on HIV incidence 11 Women HHOCP: n=1363 Other women: n=3175 HHOCP: Having Had Only Circumcised Partners Normal line: Women HHOCP Bold line: Other women (95% bootstrapped confidence intervals) Incidence rates among 15-49 y Women HHOCP: 0.032 [0.027-0.037] /py Other Women: 0.039 [0.036-0.042] /py  Incidence Rate Ratio: 0.83 [0.69-0.99]  Reduction: 16.9% [1.1%-31.0%] Among 15-29 y  Reduction: 20.3% [5.8%-33.8%]

12. Discussion 12 Main finding Reduced HIV incidence among women having only circumcised partners Interpretation Lower exposure due to a lower level of HIV among circumcised men Limitations Direct or indirect effect? HIV incidence modelled, not measured

13. Overview of results on VMMC roll-out obtained in Orange Farm (2007-2014) Among men VMMC uptake can be rapid and large VMMC is not statistically associated with condom use Circumcised men have a lower risk of prevalent HIV infection Circumcised men have a lower risk of incident HIV infection (BED assay) Among women Most women are in favor of VMMC Most women prefer circumcised men Partner’s circumcision status is not statistically associated with condom use Women having only circumcised partners have a lower risk of lower risk of prevalent HIV infection Women having only circumcised partners have a lower risk of incident HIV infection (modelled) VMMC: Voluntary Male Medical Circumcision Auvert et al, Plos Med 2013 Auvert et al, CROI 2014; AIDS 2014 13

14. Aknowledgments We thank all participants who took part in this study and the whole community of Orange Farm Portia Ntshangase Josephine Otchere-Darko Dino Rech Thabile Sekhukhune Daniel Shabangu Gaph Sipho Phatedi Yvon de la Soudière Brian Williams Muhammad Barmania Scott Billy Alexandre Blake Male Alina Chakela Ewalde Cutler Sasha Frade Agenda Gumbo Mohamed Haffejee Fikile Kate Bongiwe Klaas Paul Loubet Venessa Maseko Audrey Mkhwanazi Bantu Mupompa Cynthia Nhlapo Grace Nomsa Nhlapho