Abstract

WEAC0202 - Oral Abstract Session

Association of injectable contraception and risk of HIV-1 acquisition in women in HIV-1 serodiscordant partnerships: persistence of effect in multiple sensitivity analyses

Presented by Renee Heffron (United States).

R. Heffron¹, D. Donnell², H. Rees³, C. Celum¹, N. Mugo^1,4, E. Were⁵, G. de Bruyn³, E. Nakku-Joloba⁶, K. Ngure⁴, J. Kiarie^1,4, R. Coombs¹, J. Baeten¹, Partners in Prevention HSV/HIV Transmission Study Team

¹University of Washington, Seattle, United States, ²Fred Hutchinson Cancer Research Center, Seattle, United States, ³University of Witswatersrand, Johannesburg, South Africa, ⁴Kenyatta National Hospital, Nairobi, Kenya, ⁵Moi University, Eldoret, Kenya, ⁶Makerere College of Health Sciences, Kampala, Uganda

Background: We recently reported that women in HIV-1 serodiscordant partnerships using injectable contraceptives had a 2-fold increase in HIV-1 acquisition risk (aHR=2.05, p=0.04; Heffron et al., Lancet Infectious Diseases 2012). Here, we report the results from multiple additional sensitivity analyses examining the robustness of our findings.
Methods: We used Cox proportional hazards regression, adjusting for age, male partner's plasma viral load, and time-dependent unprotected sex and pregnancy, to compare HIV-1 incidence in women using injectable contraception to women reporting no hormonal method. Sensitivity analyses included: 1) additional or alternative covariates to adjust for sexual behavior, 2) restricting to follow-up periods when unprotected sex was reported, 3) censoring visits after a woman switched her contraceptive method (to minimize the effect of changing methods on risk), and 4) restricting to consistent injectable users outside of South Africa (with access to only injectable depot medroxyprogesterone acetate [DMPA]).
Results: In all models, the approximately 2-fold increase in HIV-1 acquisition risk persisted. Neither additional adjustment for coital frequency (aHR=2.06, p=0.04) nor adjustment for the male partner's report of unprotected sex (where accuracy is less likely influenced by the woman's contraceptive choice, aHR=2.03, p=0.07) influenced the results. When limited to periods when unprotected sex was reported or including only visits prior to the first contraceptive switch, the aHR were 2.29 (p=0.17) and 2.62 (p=0.07), respectively. Finally, when only consistent, presumed DMPA users were retained (i.e. South Africans excluded), injectable contraception was associated with a nearly 4-fold increased HIV-1 risk (aHR=3.93, p=0.01).
Conclusions: Injectable contraception continued to be associated with increased HIV-1 risk despite numerous methods to control for potential imprecision in contraceptive exposure measurement and/or key behavioral confounders. Some analyses had p-values >0.05 due to reduced statistical power but the magnitude of association continued to be as strong as that seen in our primary analytic model.