Use of antiretroviral therapy (ART) regimens that are based on the drug efavirenz may compromise the effectiveness of the contraceptive implant, according to a retrospective longitudinal cohort study of HIV-positive Kenyan women.1 Among implant users, the adjusted incidence of pregnancy for those whose ART regimen was based on efavirenz was three times that for peers whose regimen was based on nevirapine, a drug from the same class that causes less activation of the enzyme that metabolizes hormonal contraceptives. In contrast, the pregnancy rate among users of the contraceptive injectable did not differ significantly by ART regimen. Implant users had a lower risk of pregnancy than users of other contraceptive methods except IUDs and permanent methods, both in the subsample of women on efavirenz-based regimens and in the entire study cohort.

Study participants were HIV-positive women aged 15–45 enrolled at HIV health facilities in western Kenya, where efavirenz-based therapy is now recommended as initial therapy. From electronic medical records, the investigators extracted data collected during clinic visits made by women between 2011 and 2013; typically, women visited the clinic every 1–6 months. For each visit, the researchers categorized the contraceptive method women were using and the ART regimen they were on at the time. Pregnancies were ascertained from women’s self-report or presenting as pregnant during a clinic visit. Poisson multivariate analyses were used to examine pregnancy rates for various combinations of contraceptive method and ART regimen.

Analyses were based on 24,560 women who contributed 94,162 observations. Women’s mean age at the start of the study was 31 years; 44% were married or cohabiting. Women were using the implant in 7% of observations, the injectable in 17%, the pill in 3%, the IUD or permanent methods in 3% and less effective methods (condoms or natural methods) in 37%; in 33% of observations, women were not practicing contraception. With respect to ART, women were on a nevirapine-based regimen in 49% of observations, an efavirenz-based regimen in 14%, and a lopinavir- and ritonavir-based regimen in 4%; in 32% of observations, women were not receiving ART. Overall, women had 3,337 pregnancies during follow-up, which corresponds to an unadjusted pregnancy rate of 8.9 per 100 person-years.

The pregnancy rate among all women using the implant (after adjustment for factors potentially affecting this outcome) was 1.4 per 100 person-years; however, the pregnancy rate was 3.3 among women on efavirenz-based ART regimens and 1.1 among those on nevirapine-based regimens—a difference that translated to an adjusted pregnancy rate ratio of 3.0. Findings were similar whether women were using the etonogestrel implant or the levonorgestrel implant. In contrast, the incidence of pregnancy among injectable users did not differ by ART regimen. The findings held up in sensitivity analyses that used more stringent definitions of contraceptive and ART exposure, and of pregnancy.

In the entire study cohort (irrespective of ART receipt and regimen), the adjusted pregnancy rate for users of other types of contraceptives—except the IUD and permanent methods—was 3–4 times that for implant users. Specifically, compared with a pregnancy rate of 1.4 per 100 person-years for implant users, the rates for injectable users, pill users and users of less effective methods were 4.3, 5.8 and 5.6, respectively; the rate for women not using any method was 4.8. Similarly, in the entire subset of women on an efavirenz-based ART regimen, the adjusted pregnancy rate for users of other types of contraceptives—except the IUD and permanent methods—was 2–3 times that for implant users. Specifically, compared with a pregnancy rate of 3.3 for implant users, the rates for injectable users, pill users and users of less effective methods were 5.4, 9.3 and 5.4, respectively; the rate for women not using any method was 4.0.

Study limitations included possible inaccurate documentation of contraceptive use and uncertainty about dates of initiation, higher than expected pregnancy rates among women using more effective methods, lack of information on pregnancy intentions, and inability to account for certain important covariates, such as sexual activity and ART adherence. Despite these, the authors comment that their findings add to evidence that efavirenz-based ART may reduce the effectiveness of the implant, but also suggest that the implant is still among the most effective reversible contraceptives, even among women on efavirenz-based ART. They suggest that prospective research is needed on interactions between hormonal contraceptives and ART regimens that assesses pharmacokinetics and more rigorously ascertains contraceptive and ART adherence and pregnancies. The authors conclude that “HIV programmes, providers, and ministries of health should continue to offer HIV-positive women the choice of selecting concomitant implants and efavirenz-based ART until better contraceptive and ART alternatives are shown to be more effective and become readily available.”

The authors of an accompanying commentary2 note that current guidance for South Africa recommends that women on efavirenz-based ART not use implants, and use the IUD or injectable instead. They caution that the newly reported pregnancy rates should be interpreted with caution because of the study’s reliance on medical records and the approach to analysis. Nonetheless, they conclude that “these data might encourage reconsideration of blanket guidance against concomitant use of implants and efavirenz, or emphasis on potentially less-effective alternative methods, and focus on appropriate counselling to help HIV-positive women choose the contraceptive method most suited to their situation.”—S. London

REFERENCES

1. Patel RC et al., Pregnancy rates in HIV-positive women using contraceptives and efavirenz-based or nevirapine-based antiretroviral therapy in Kenya: a retrospective cohort study, Lancet HIV, 2015, 2(11):e474–e482.

2. Morroni C, Bekker LG and Rees H, Contraceptive implants and efavirenz-based ART: friend or foe? Lancet HIV, 2015, 2(11):e454–455.