Polly Clayden, HIV i-Base
Three unintended pregnancies in women with levonorgestrel sub-dermal implants, receiving efavirenz (EFV)-based ART in a pharmacokinetic (PK) study, were reported in a late breaker presentation at CROI 2015. 
Preliminary data from this study, conducted in Uganda, were shown at the HIV Drug Therapy Glasgow Congress, 2014. [2,3] At 24 weeks EFV significantly decreased exposure to levonorgestrel – the active progesterone component of one commonly used contraceptive implant. Levonorgestral concentrations in women receiving EFV-based ART were approximately half of those in women not receiving ART, despite the EFV group having significantly lower body weight.
The final 48-week results were shown at CROI. Kimberly Scarsi presented the findings on behalf of investigators from University of Nebraska Medical Center, Makerere University, Uganda, and Liverpool University.
The study was a non-randomised, parallel group, PK evaluation comparing levonorgestrol concentrations in Ugandan women not yet eligible for ART (control group, n=17) and on stable EFV-based ART (n=20).
The women had a two-rod (75 mg/rod) levonorgestrol sub-dermal implant inserted at enrolment. Sampling for PK was performed before the implant and at 1, 4, 12, 24, 36 and 48 weeks after insertion. The investigators used a validated LC-MS/MS method, with an assay calibration range of 50-1500 pg/mL for the evaluation. PK data were reported as geometric means (GM) and GM ratio.
All participants were black Africans, with a mean age of 30 years at baseline. Women in the control group had a higher baseline body weight (73 kg) compared to those in the EFV group (59 kg), p<0.01. Women in the EFV group had undetectable viral load for a median of 10 months (range 5-66) before study entry.
Levonorgestrol PK data are presented in Table 1. These data show a 45-57% reduction in levonorgestrol concentrations, starting with a rapid reduction in the first week and persisting throughout the study period. Levonorgestrel AUC was reduced by 48%.
|Week||Control group (n=17)||EFV group (n=20)||EFV:control GM ratio|
|1||1070 (783 – 1356)||462 (370 – 553)||0.43 (0.41 – 0.47)|
|4||667 (541 – 792)||359 (280 – 473)||0.54 (0.52 – 0.55)|
|12||590 (475 – 704)||327 (268 – 385)||0.55 (0.55 – 0.56)|
|24||528 (423 – 633)||280 (212 – 348)||0.53 (0.50 – 0.55)|
|36||618 (520 – 716)||279 (149 – 409)||0.45 (0.29 – 0.57)|
|48||580 (477 – 684)||247 (209 – 285)*||0.43 (0.42 – 0.44)|
|AUC**||22.24 (18.55 – 25.92)||11.60 (9.38 – 13.83)||0.52 (0.51 – 0.53)|
* n=11, excludes 2 pregnant women and 6 who did not reach week 48 because arm closed; **AUC 0-36 wk (ng*wk/mL). AUC = area under the curve.
There was considerable interpatient and intrapatient variation in levonorgestrol concentration: coefficient of variation, respectively 43% and 26%.
Dr Scarsi noted that although PK was the primary objective of this evaluation, three (15%) unintended pregnancies occurred in the EFV arm during the study period (none in the controls). Two were identified at week 48 visit – after which the EFV arm was closed – and one at an early discontinuation visit.
|Week||Pregnancy 1||Pregnancy 2||Pregnancy 3|
|Pregnancy identified||Wk 48 (2 weeks post conception)||Wk 48 (10 weeks post conception)||Wk 43 (2 weeks post conception)|
The proposed levonorgestrol threshold for efficacy – based on historical data – is 180 pg/mL. Levonorgestrol concentrations for the three women who became pregnant during the study are shown in Table 2. The highest levonorgestrol concentration at which a woman became pregnant in the study was 303 pg/mL. Fifteen out of 20 participants (75%) in the EFV arm (none of the controls) had levonorgestrol concentrations below this threshold.
EFV induces cytochrome P450 3A4, giving the potential for drug-drug interactions with concomitant medications. Levonorgesterol is primarily metabolised by cytochrome P450 3A4, because of this, manufacturers and guidelines recommend avoiding co-administration with medications known to induce this pathway. So perhaps the results are not surprising.
That the proposed threshold for levonorgesterol efficacy (180 pg/mL) was inadequate in this study population is particularly worrying.
These data support a previous retrospective study of unintended pregnancies with levonorgesterol use in a low-income country: 12.4% (15/121) in women receiving EFV-based ART vs 0% (0/208) in those receiving NVP-based. 
Dr Scarsi rightly remarked: “Effective contraception for women on EFV-based ART is a public health priority.” She recommended that health workers discuss the risk of unintended pregnancy with women in this situation and other methods of contraception in the short term.
But she noted that sub-dermal implant use is continuing to rise: in sub Saharan Africa the procurement increased from 1 million to 5 million purchases in 2013.
It is vital that novel sub-dermal dosing strategies to use with EFV are explored as it is likely that similar effects on levonorgesterol levels might be observed with lower doses of EFV (e.g. 400 mg OD). However, more data to confirm this are warranted.
This serious interaction is also another reason to consider alternatives to EFV first line and appropriate trials must be designed and funded to generate sufficient data to support future guideline changes for low-income countries (particularly with dolutegravir-based regimens).
- Scarsi K et al. Levonorgestrel implant + EFV-based ART: unintended pregnancies and associated PK data. CROI 2015. Seattle, Washington. 23-26 February 2015. Oral abstract 85LB.
- Scarsi K et al. Efavirenz- but not nevirapine-based antiretroviral therapy decreases exposure to the levonorgestrel released from a sub-dermal contraceptive implant. HIV Drug Therapy Glasgow Congress, 2-6 November 2014. Oral abstract 0131. Journal of the International AIDS Society 2014, 17(Suppl 3):19484.
- Clayden P. Efavirenz decreases exposure to hormonal contraceptive implant. HTB. 1 December 2014.
- Perry SH et al. Implementing the Jadelle implant for women living with HIV in a resource-limited setting: concerns for drug interactions leading to unintended pregnancies. AIDS. 2014;28:791-793.