The risk of HIV associated with hormonal contraceptives is controversial. We assessed hormonal contraceptive use and HIV incidence in HIV-discordant couples in Rakai, Uganda.
HIV-discordant couples were retrospectively identified from a cohort between 1999 and 2009. Hormonal contraception included oral contraception, depomedroxyprogesterone acetate (DMPA), and implants (Norplant). Poisson regression estimated adjusted incidence rate ratios (adjIRRs) associated with hormonal contraceptive methods. A case–control subanalysis estimated odds ratios (ORs) of HIV associated with hormonal contraceptive, adjusted for viral load and age.
We identified 190 male HIV-positive/female HIV-negative (M+F−) and 159 male HIV- negative/female HIV-positive (M−F+) couples not using antiretroviral therapy or condoms. Female HIV incidence was 5.8/100 person-years (py) among nonhormonal contraceptive users, 12.0/100 py among oral contraceptive users [adjIRR 2.65, 95% confidence interval (CI) 0.82–8.60], 4.5 among Norplant users (adjIRR: 0.89, 95% CI 0.11–7.10), and 7.5/100 py among DMPA users (adjIRR 1.42, 95% CI 0.60–3.36). Male HIV incidence was 7.4/100 py during nonhormonal contraceptive use, 16.5/100 py during female oral contraceptive use (adjIRR 2.52, 95% CI 0.49–12.95), and 4.9/100 py with DMPA use (adjIRR 0.57, 95% CI 0.19–1.70). The number of female seroconverters was three among oral contraceptive users, one among Norplant users, and seven among DMPA users. Male seroconverters were two during female oral contraceptive use, none with Norplant use, and three with DMPA use. In a nested case–control analysis after adjustment for HIV viral load, the adjOR associated with oral contraceptive use was 1.59 (95% CI 0.32–97.85) for M+F− and 2.11 (95% CI 0.18–25.26) for M−F+ couples. For DMPA use, the adjOR was 1.44 (95% CI 0.46–4.51) for M+F− and 1.40 (95% CI 0.30–6.49) for M−F+ couples.
We did not observe significant risk of HIV acquisition or transmission with oral contraceptives or DMPA use in HIV discordant couples, but several point estimates were above 1.0 and statistical power was limited.