A prospective cohort of HIV-infected Malawian women on Efavirenz initiating the Levonorgestrel Implant or the Depot Medroxyprogesterone Acetate Injectable

ABSTRACT Sub-Saharan Africa (SSA) has high rates of unintended pregnancy, maternal mortality, and perinatal HIV. Increased use of effective contraception could reduce all three. The Levonorgestrel (LNG) implant is a highly-effective and reversible contraceptive that is particularly well-suited to settings like Malawi, where our research will take place. Sometimes referred to as ?forgettable contraception,? the LNG implant provides up to 5 years of protection and is not dependent upon external factors such as regular clinic attendance or breaks in the health system supply chain. Thus, the LNG implant?s typical-use failure rate is very low at 0.1% in the first year. The Depot Medroxyprogesterone Acetate (DMPA) injectable is the most commonly-used contraceptive in SSA. However, DMPA requires repeat injections every 3 months, leading to a higher typical- use failure rate of 6% in the first year. Recent observational studies suggest that co-administration of the antiretroviral Efavirenz (EFV) may reduce the contraceptive efficacy of the LNG implant, possibly due to pharmacokinetic interaction between the two drugs. Citing these small studies, some countries in SSA are considering policy recommendations against use of implants for women on EFV. South Africa and the U.S. have already issued such guidance. We argue in this application that the data are not strong enough to support this policy action and propose a 2-part study designed to provide critical evidence to providers and policymakers. Aim 1 will compare the typical-use pregnancy rates of the LNG implant versus the DMPA injectable in a prospective cohort of 1,420 HIV+ women on EFV (710 initiating the LNG implant and 710 initiating DMPA). Women will be enrolled after receiving their chosen contraceptive at their local health facility and will be followed in the study after 1 month and then every 3 months for at least 2 years and up to 4 years. A Poisson model will be used to estimate pregnancy incidence among implant and DMPA users. A marginal structural Poisson model will then be used to estimate the effect of implant compared to DMPA upon pregnancy incidence. Aim 2 will study 240 women in a 2:1 nested case-control study of women from the cohort to determine if higher EFV concentrations in hair (a biomarker for EFV exposure) are associated LNG implant contraceptive failure. 80 women (cases) who become pregnant while using the LNG implant will be compared with 160 LNG implant users who were not pregnant after the same duration of implant use (controls). Conditional logistic regression will be used to assess the association between hair EFV concentrations and LNG implant failure. We will also evaluate the association between being a ?slow? EFV metabolizer with LNG implant failure and assess if there is a LNG concentration threshold at which LNG implant failure occurs among EFV users. The study will be a collaboration between investigators at the University of North Carolina (UNC) in Chapel Hill, UNC Project-Malawi, and the Lighthouse Trust. We will work closely with members of the Malawi College of Medicine and the Malawi Ministry of Health to guide study implementation and any resulting policy decisions regarding the use of LNG implant and EFV in Malawi.