Contraceptive implant use duration is not associated with breakthrough pregnancy among women living with HIV and using efavirenz: a retrospective, longitudinal analysis

Abstract

Introduction: Contraceptive implants containing etonogestrel and levonorgestrel have emerged as popular contraceptive options among women in areas of high HIV burden in sub-Saharan Africa. However, recent pharmacokinetic data have shown drug–drug interactions between implants and efavirenz-containing antiretroviral therapy (ART), reducing the effectiveness of the implants. Here, we evaluated pregnancy incidence in 6-month intervals following implant initiation among women using efavirenz and contraceptive implants to assess whether the risk of breakthrough pregnancy is higher after specific periods of implant use. Methods: We used data from a retrospective longitudinal analysis of women living with HIV ages 18–45 years in western Kenya who attended HIV-care facilities between 2011 and 2015. We used Cox proportional hazard models to compute haz- ard ratios (HRs) for breakthrough pregnancy by implant type and ART regimen. Depending on the model, we adjusted for socio-demographic and clinical factors, programme, site and interaction between calendar time and ART regimen. We utilized inverse probability weights (IPWs) to account for three sampling phases (electronic medical record [EMR], chart review and phone interview) and calculated overall parameter estimates. Results: Women contributed 14,768 woman-years from the largest sampling phase (EMR). The median age was 31 years. Women used etonogestrel implants for 26–69% of the time and levonorgestrel implants for 7–31% of the time, depending on the sampling phase. Women used efavirenz, nevirapine or no ART for 27–33%, 40–46% and 15–26% of follow-ups, respec- tively. When combining sampling phases, there was little evidence to suggest that the relative hazard of pregnancy among efavirenz-containing ART users relative to nevirapine-containing ART changed with length of time on implants: IPW-adjusted HR of 3.1 (CI: [1.5; 6.4]) at 12 months, 3.4 (CI: [1.8; 6.3]) at 24 months, 3.8 (CI: [1.9; 7.7]) at 36 months and 4.2 (CI: [1.6; 11.1]) at 48 months (interaction p-value = 0.88). Similarly, no significant change in HRs over time was found when comparing women not using ART to nevirapine-containing ART users (interaction p-value = 0.49). Conclusions: We did not find evidence to suggest implants being more fallible from drug–drug interactions with efavirenz at later time intervals of implant use. Thus, we would not recommend shortening the duration of implant use or replacing implants sooner when concomitantly used with efavirenz