Study Population and Procedures
Between July 2008 and November 2010, 4747 heterosexual HIV-1-serodiscordant couples (1785 in which the HIV-1-uninfected partner was female) from Kenya and Uganda were enrolled in the Partners PrEP study, a randomized, placebo-controlled trial of PrEP. HIV-1-uninfected partners were randomized 1 : 1 : 1 to daily oral TDF, co-formulated FTC/TDF, or placebo, and followed monthly up to 36 months. Women who were pregnant or intending to become pregnant were not eligible for enrollment. Contraceptive use was not required for study participation, but monthly study visits included contraceptive counseling and provision of contraception on site free of charge, if desired. Direct monitoring of contraceptive use was not conducted. Sexual behavior in the prior month and contraceptive use were recorded monthly on standardized interviewer-administered questionnaires; condom use was recorded in the sexual behavior assessment, not as contraception. At each study visit, women were screened for incident pregnancy with urine testing and for HIV-1 seroconversion with rapid HIV-1 antibody tests. Women with positive pregnancy tests were referred to antenatal care and study drug was withheld until the end of pregnancy and breastfeeding. HIV-1-infected male partners were not eligible for antiretroviral therapy (ART) under the national guidelines of Kenya and Uganda at enrollment; they were followed quarterly, with clinical and CD4+ cell count monitoring, and were actively referred for ART initiation during follow-up.
To estimate the effectiveness of hormonal contraceptive methods for pregnancy prevention, we used the Andersen-Gill extension of Cox proportional-hazards regression to account for repeated events. Pregnancy start was estimated by the last menstrual period prior to the first positive pregnancy test. Follow-up time was measured by study visit months; therefore the first study visit during pregnancy was classified as the incident visit, and all subsequent visits during pregnancy were censored. Women re-entered the risk set when no longer pregnant. Contraceptive method was included as a time-dependent variable with no contraceptive use as the reference group. Whether PrEP modified the effect of each hormonal contraceptive method was tested with interaction terms on the log scale. Because both active PrEP arms contained TDF, they were analyzed as a single PrEP-exposed group. Covariates were selected a priori, including, at enrollment, age, number of children, partnership duration, and sexually transmitted infections, and, during follow-up, any unprotected sex, sexual frequency with the study partner, any sex with additional partners, and ART initiation by the HIV-1-infected partner. Women who acquired HIV-1 were censored at seroconversion.
Follow-up time when women reported nonhormonal contraceptive methods (tubal ligation, hysterectomy, IUD, diaphragm, or 'other') was excluded from the analysis. Because injectable contraceptive efficacy increases after initiation, we also estimated effectiveness for each of the first 3 months after initiation and for longer duration of use, and assessed whether PrEP modified these effects. As a sensitivity analysis, all analyses were repeated excluding visits when PrEP/placebo study drug was withheld according to protocol-defined indications and when PrEP adherence was estimated to be low, classified by pill counts of returned study medication below 80% or above 104%, as excess doses potentially represent nonuse (e.g. 'pill dumping'). Analyses were conducted in SAS 9.3.
AIDS. 2014;28(12):1825-1830. © 2014 Lippincott Williams & Wilkins
Lippincott Williams & Wilkins