Drug interactions between hormonal contraceptives and antiretrovirals

Objective: To summarize published evidence on drug interactions between hormonal contraceptives and antiretrovirals. Design: Systematic review of the published literature. Methods: We searched PubMed, POPLINE, and EMBASE for peer-reviewed publications of studies (in any language) from inception to 21 September 2015. We included studies of women using hormonal contraceptives and antiretrovirals concurrently. Outcomes of interest were effectiveness of either therapy, toxicity, or pharmacokinetics. We used standard abstraction forms to summarize and assess strengths and weaknesses. Results: Fifty reports from 46 studies were included. Most antiretrovirals whether used for therapy or prevention, have limited interactions with hormonal contraceptive methods, with the exception of efavirenz. Although depot medroxyprogesterone acetate is not affected, limited data on implants and combined oral contraceptive pills suggest that efavirenz-containing combination antiretroviral therapy may compromise contraceptive effectiveness of these methods. However, implants remain very effective despite such drug interactions. Antiretroviral plasma concentrations and effectiveness are generally not affected by hormonal contraceptives. Conclusion: Women taking antiretrovirals, for treatment or prevention, should not be denied access to the full range of hormonal contraceptive options, but should be counseled on the expected rates of unplanned pregnancy associated with all contraceptive methods, in order to make their own informed choices.


Introduction
Women living with HIV will likely take combination antiretroviral therapy (cART) for much of their lives [1]. Those at high risk for HIV may also use antiretrovirals for preexposure prophylaxis (PrEP). Contraceptive use among women living with HIV or using antiretrovirals for PrEP is critical, as unintended pregnancy and short interpregnancy intervals can be associated with negative health consequences for both mother and infant [2][3][4].
Decreasing unintended pregnancies also reduces vertical HIV transmission [5]. Hormonal contraceptives are highly used worldwide, including in areas of high HIV prevalence; they are also among the most effective contraceptive methods [6,7]. Evidence-based guidance for hormonal contraceptives use among women using cARTor PrEP is needed to ensure access to a full range of the best contraceptive methods, and therefore increase the likelihood of achieving their reproductive life planning goals.
Concurrent use of hormonal contraceptives and antiretrovirals can lead to drug interactions, predominantly due to effects on liver metabolism (Tables 1 and 2). In the liver, cytochrome P450 (CYP) enzymes catalyze many important reactions, with the most significant for contraceptive metabolism being CYP3A4, which is also expressed in the intestines [8,9]. Antiretrovirals include different classes of drug (Table 2), including nonnucleoside reverse transcriptase inhibitors (NNRTIs), nucleoside analogue reverse transcriptase inhibitors or nucleotide analogue reverse transcriptase inhibitor (NRTIs), protease inhibitors, fusion inhibitors, and integrase inhibitors. The NNRTIs and integrase inhibitors are generally not substrates, inhibitors, nor inducers of cytochrome P450 enzymes [10]. In contrast, both protease inhibitors and NNRTIs are metabolized by CYP3A4 and also inhibit or induce this enzyme, resulting in increases or decreases in the concentration of concomitantly administered drugs [10].
Such interactions could lead to decreased contraceptive effectiveness (increasing risk of unintended pregnancy), decreased cART effectiveness (associated with resistance and/or HIV disease progression), decreased efficacy of PrEP (increasing risk of HIV acquisition), or increased antiretroviral or contraceptive toxicity. Based on theoretical concerns and limited data, women using cART are sometimes offered fewer contraceptive choices than their HIV-negative peers [11]. The objective of this review was to systematically examine published evidence on drug interactions between hormonal contraceptives and antiretrovirals, in order to contribute to improved clinical and policy decision-making.

Methods
We followed the PRISMA and MOOSE guidelines for conducting the review and reporting the results [12,13].
We searched PubMed, POPLINE, and EMBASE from database inception to 21 September 2015 for studies of hormonal contraceptive and antiretroviral drug interactions (Supplement 1, http://links.lww.com/QAD/ B36). We also hand-searched reference lists of published studies, and contacted topic experts.

Study selection
We included published studies of women using hormonal contraceptives (Table 1), including combined oral contraceptives (COCs), progestin-only pills (POPs), emergency contraceptive pills (ECPs), injectables, vaginal rings, patches, or implants. Studies included women who were either HIV-positive, HIV-negative but at risk of HIV, or healthy, who concurrently used cART, PrEP, or single antiretrovirals and hormonal contraceptives. We included studies reporting on women taking oral contraceptives where the type of oral contraceptive was not specified. We excluded studies evaluating women on cART without comparisons by contraceptive use, those evaluating only genital HIV viral load, and those 918 AIDS 2017, Vol 31 No 7

Data abstraction and management
After screening and removal of duplicates, we abstracted relevant data from each included report using a predesigned form. Two authors independently reviewed selected manuscripts, with differences resolved by consensus.
We described strengths, weaknesses, and funding source for each included study (Tables 3 and 4) , but did not do formal quality assessment because no formal evidence grading system exists for pharmacokinetic studies.

Results
Our search identified 1570 records. Fifty published reports from 46 individual studies met the inclusion criteria ( Fig. 1, Tables 3 and 4). Four reports were secondary analyses or subsets of the primary studies and are included with the primary study in the tables [14][15][16][17]. The results are presented by outcome assessed, focusing first on the most important clinical outcomes (contraceptive effectiveness, antiretroviral effectiveness, toxicity associated with combined administration), then the pharmacokinetic data (for contraceptives and antiretrovirals), in each case by antiretroviral class and by contraceptive method.

Contraceptive effectiveness
Although pregnancy is the most relevant outcome, few large studies were designed to investigate contraceptive effectiveness. Several secondary analyses helped fill this gap, particularly for women using nevirapine-containing or efavirenz-containing cART. Although some small pharmacokinetic studies of healthy women report on pregnancy, women were generally required to use additional contraception; these studies are included in Table 3 but not summarized here.

Nonnucleoside reverse transcriptase inhibitors
Fourteen reports from clinical trials and six secondary analyses described contraceptive effectiveness measures among women using NNRTIs and hormonal contraceptives (Table 3).
Pregnancy rates and ovulation rates did not differ between HIV-positive women taking COCs and nevirapinecontaining cART and those not yet taking cART [18]. In a small trial of women using COCs with efavirenzcontaining cART, three women ovulated (out of 25) but no pregnancies were reported [19]. Five small pharmacokinetic trials of NNRTIs and COCs also demonstrated no ovulation among study participants [26][27][28][29][30].
In large cohort studies, pregnancy rates were slightly higher among women taking efavirenz-containing cART (11-15/100 woman-years) compared with women taking oral contraceptive and nevirapine-containing cART or no cART (pregnancy rates 6-11/100 woman-years) [24,25]. Notably the reported pregnancy rates in the large cohort studies are still lower than an expected pregnancy rate of 40 per 100 woman-years among women not using any modern contraceptive and trying to prevent pregnancy.
Other retrospective cohort studies reported pregnancy rates among oral contraceptive users ranging from 2.6 to 5.8 per 100 woman-years (most, but not all, women were using nevirapine) [20,21].

Depot medroxyprogesterone acetate
In two pharmacokinetic studies, women using depot medroxyprogesterone acetate (DMPA) remained anovulatory when using cART containing either efavirenz or nevirapine [14,31,32]. Five cohort studies also presented pregnancy rates with injectables among cART users. In the largest, pregnancy rates ranged from 8 to 10 per 100 woman-years for injectable users, with higher rates in efavirenz users and those not on cART compared with pregnancy rates in those using nevirapine [24]. Another found women using DMPA had pregnancy rates from 3 to 5 per 100 woman-years, with lower rates among users of nevirapine and efavirenz compared with no cART [25]. Two additional studies reported pregnancy rates from 1.8 to 4.2 per 100 woman-years among DMPA users taking various antiretrovirals (primarily nevirapine) [20,21].

Implants
Pregnancy rates among users of cART and contraceptive implants differed by whether the implant contained levonorgestrel or etonogestrel, and whether women were taking efavirenz or nevirapine. Pregnancy rates were higher among women using the levonorgestrel implant concomitantly with efavirenz. Two prospective studies (N ¼ 79 and N ¼ 45) reported no pregnancies through 3 years and 6 months, respectively, among women using etonogestrel implants and NNRTI-containing cART [33,34], although the second study found that women taking efavirenz-containing cART had a 2.8% presumed ovulation rate over 6 months. In a small pharmacokinetic study of women using levonorgestrel implants and cART containing efavirenz or nevirapine, three pregnancies (3/20; 15%) occurred within 48 weeks, all in women taking efavirenz [35]. Similar findings were seen in a large retrospective study, where 15 of 121 (12.4%) women using levonorgestrel implants and efavirenz-containing cART became pregnant, at a mean duration of 16.4 months; no pregnancies occurred among women taking nevirapine-containing cART [36].
Two secondary analyses described pregnancy rates with implant use in women using efavirenz-containing or nevirapine-containing cART [24,25]. In the first, pregnancy rates for users of efavirenz, nevirapine, and no cARTwere 5.5, 2.3, and 3.4 per 100 woman-years for etonogestrel implant users, and 7.1, 1.9, and 3.3 per 100 woman-years for levonorgestrel implant users, respectively [24]. In the second, pregnancy rates per 100 womanyears were 1.4 for unspecified implant users not taking cART, 0 for nevirapine users, and 6 for women taking efavirenz [25].
Combined oral contraceptives/patches Co-administration of darunavir/ritonavir with COCs resulted in no ovulation [38]. Similarly, coadministration of lopinavir/ritonavir containing cART in women using a contraceptive patch also found no ovulations [37].
Progestin-only pills One report showed that oral norethindrone thickened cervical mucus and led to similar mucus scores in women using protease inhibitor-containing cART compared with those taking NRTIs alone [16].
Depot medroxyprogesterone acetate Reports of women using lopinavir/ritonavir-containing or nelfinavir-containing cART with DMPA found that no women ovulated [14,31,39].

Implants
No ovulations nor pregnancies were reported in a pharmacokinetic study of etonogestrel implant users taking lopinavir/ritonavir-containing cART [34].
Nucleoside/nucleotide reverse transcriptase inhibitors Analyses of two large trials of hormonal contraceptives and NRTIs used for PrEP found that use of tenofovir/ emtricitabine did not affect pregnancy rates among users of COCs, injectables, or implants [40,41].
Integrase inhibitors A small pharmacokinetic study of dolutegravir with COCs resulted in no ovulations [42].

Antiretroviral effectiveness
Eight reports evaluated the effects of hormonal contraceptive use on the effectiveness of NNRTI-containing or protease inhibitor-containing cART, and found no effects on death, CD4 þ cell count, or plasma viral load with concurrent use of DMPA, levonorgestrel implants, or oral contraceptives [14,[43][44][45][46][47][48][49]. Use of DMPA also did not affect the efficacy of PrEP [50].

Toxicity of combined administration
Studies among healthy women using hormonal contraceptives concurrently with single antiretrovirals, or HIVpositive women using cART, generally reported no difference in adverse events of concurrent treatment compared with use of either hormonal contraceptives or antiretrovirals alone (Table 3). One HIV prevention trial evaluated pharmacodynamic interactions between tenofovir-containing PrEP with oral contraceptives or DMPA, and found that bone mineral density was not significantly decreased [51].

Contraceptive pharmacokinetics
Thirty-two reports include contraceptive pharmacokinetic measures among women using antiretrovirals and hormonal contraceptives (Table 4).

Combined oral contraceptives
Two studies evaluated efavirenz with COCs: one in women taking only efavirenz, and one in women taking efavirenz-containing cART [15,26]. Ethinyl estradiol concentrations were not significantly changed, but progestin levels decreased by approximately 60%.
Three studies reported the effect of nevirapine on COC pharmacokinetics. Ethinyl estradiol levels varied from being unchanged to being approximately 30-60% lower [15,30,53]. Progestin levels were not significantly affected.
Studies of etravirine and rilpivirine in women taking COCs found minimal effects [27,28]. Similarly, a study of fosdevirine (the development of which was discontinued due to toxicity) found no effect on hormone levels in COC users [29].
Depot medroxyprogesterone acetate Two studies evaluated the effect of NNRTIs on DMPA, and found no difference in medroxyprogesterone acetate pharmacokinetics through 12 weeks with concurrent use of either efavirenz-containing or nevirapine-containing cART [31,32].

Implants
A study of women using etonogestrel implants found 54-70% lower etonogestrel levels among women taking efavirenz-containing cART compared with women taking no cART [34].
Emergency contraceptive pills One study in healthy women showed that levonorgestrel levels were 56% lower in ECP users after use of efavirenz [52].
Combined oral contraceptives/Patch Two studies examining concurrent use of ritonavir and COCs found decreased ethinyl estradiol levels, whereas progestin levels were unaffected [54,57]. Another study showed decreased ethinyl estradiol levels, but increased progestin levels, when COCs were used with atazanavir/ ritonavir [55]. Similar findings were reported with lopinavir/ritonavir-containing cART and the contraceptive patch [37]. This study also reported lower ethinyl estradiol levels with concurrent use of a single COC pill, but progestin levels were not evaluated. Only darunavir/ ritonavir was associated with significantly lower ethinyl estradiol levels as well as slightly lower norethindrone levels [38].

Progestin-only pills
In women receiving protease inhibitor-containing cART, norethindrone levels were higher compared with controls [56]. A subanalysis restricted to women using ritonavirboosted atazanavir confirmed this finding [17].
Depot medroxyprogesterone acetate One study showed significantly increased medroxyprogesterone acetate concentrations, compared with historical controls, in women using DMPA and lopinavir/ ritonavir-containing cART [39].

Implants
A study evaluating the effect of cART on the pharmacokinetics of the etonogestrel implant found women using lopinavir/ritonavir-containing cART had etonogestrel levels approximately 50% higher than women not taking cART [34].

Nucleoside/nucleotide reverse transcriptase inhibitors
Two studies evaluating NRTIs used for PrEP with COCs or levonorgestrel implants showed no change in hormone levels [58,59].

Chemokine receptor 5 antagonists
Two studies showed that neither maraviroc nor vicriviroc impacted hormone levels when used concurrently with COCs [57,60].

Integrase inhibitors
In two studies, concurrent use of COCs and raltegravir led to small increases in progestin exposure [61], but dolutegravir had no impact on hormone levels [42].

Nonnucleoside reverse transcriptase inhibitors
Three studies evaluated the impact of COCs on the pharmacokinetics of efavirenz [19,26,62]. Among women using COCs and efavirenz alone, concentrations were similar to historical controls [26]. However, in a trial of women on efavirenz-containing cART, use of COCs led to efavirenz concentrations lower than historical controls [19]. Another analysis of women taking efavirenz-containing cART found no difference in efavirenz concentrations between hormonal contraceptive users and nonusers [62].
Three studies evaluated the impact of COCs on nevirapine levels. In two reports of women using nevirapine-containing cART, nevirapine levels were not significantly different in women using COCs [19,53]. Time to undetectable nevirapine levels was longer in women receiving single-dose nevirapine and using COCs [63]. Another study found rilpivirine levels in COC users to be similar to historical controls [27].
Among women on various cART regimens, nevirapine levels were slightly higher after administration of DMPA, but no changes in efavirenz levels were noted [31].

Protease inhibitors
Four studies investigated the effects of COCs on the pharmacokinetics of protease inhibitors [38,54,55,64]. Levels of saquinavir were not affected by COC use, but atazanavir levels were slightly increased [55,64]. Coadministration with COCs resulted in darunavir and ritonavir levels comparable to those in historical controls [38,54].
Co-administration of the contraceptive patch with lopinavir/ritonavir-containing cART resulted in slightly decreased levels of both protease inhibitors compared with historical controls [37], whereas DMPA had no effect on protease inhibitor levels in women taking such regimens [39].
Nucleoside/nucleotide reverse transcriptase inhibitors One study found no effect of hormonal contraceptives (COCs and DMPA) on zidovudine plasma or intracellular pharmacokinetics [65].

Chemokine receptor 5 antagonists
When maraviroc was taken with COCs, levels were similar to those seen in historical controls [60].

Discussion
Few of the 50 reports included in this review provided relevant data that can be applied to clinical practice with certainty ( Table 5). The most significant interactions with hormonal contraceptives occurred in women using cART-containing NNRTIs, particularly efavirenz. However, even in these studies, the outcomes reported were often pharmacokinetic rather than clinical, involved small populations, which limited study power, or were derived retrospectively from secondary analyses of existing cohorts.
The most important outcome for contraceptive drug interactions is method failure resulting in pregnancy, but few studies reported this outcome (Table 5). Changes in contraceptive hormone levels do not necessarily translate into reduced efficacy or increased toxicity, as levels vary greatly within and between individuals and populations [66,67]. Further, the contraceptive threshold, or minimum steroid hormone level required to maintain contraceptive effectiveness, is difficult to determine [68,69]. However, when pharmacokinetic data show no or minimal changes, clinical effects are unlikely. Ovulation is used in many drug-drug interaction studies to indicate risk of pregnancy, but ovulation is also a surrogate marker. The occurrence of ovulation does not always result in pregnancy. For example, many women ovulate during levonorgestrel implant use, yet contraceptive effectiveness remains high [70]. Additionally, no included studies evaluated true ovulation; rather, ovulation was presumed based on serum progesterone measurements alone, which can be inaccurate [70].
The most clinically significant drug-drug interactions identified in our systematic review were reported in women using efavirenz-containing cART and COCs or progestin-containing subdermal implants. Although studies show DMPA is not impacted by efavirenz use, studies of women using efavirenz and contraceptive implants reported pregnancy rates ranging from 5 to 15 per 100 woman-years, and COC users taking efavirenz had pregnancy rates ranging from 13 to 15 per 100 woman-years [24,25,35]. For COCs, because contraceptive effectiveness relies on user adherence, potential additional reductions in effectiveness from a drug interaction, if confirmed, are concerning. Conversely, studies that reported on women using contraceptives with nevirapine-containing cART were generally reassuring. None of the studies that enrolled women using a number of different hormonal contraceptives with nevirapine   reported increases in pregnancy or ovulation rates [18,19,21,24,25,30,35,36,45].
The many other studies included in this review that evaluated hormonal contraceptives and antiretrovirals other than efavirenz reported no results that should change clinical practice. Antiretrovirals used in PrEP do not affect hormonal contraceptive effectiveness. Concurrent use of protease inhibitors with COCs does not alter contraceptive effectiveness despite the observed decreased ethinyl estradiol plasma levels found, as the progestin component is primarily responsible for contraceptive effectiveness. Minimal to no changes in progestin levels were reported in multiple studies of concurrent protease inhibitor and hormonal contraceptive use.
Although concomitant use of a few protease inhibitors led to increased progestin levels in some studies, these changes are unlikely to impact safety given the variable doses and wide safety margin of contraceptive progestins.
Despite the small number of reports in our review, studies were also reassuring with regard to the effect of hormonal contraceptives on cART or PrEP effectiveness, or antiretroviral pharmacokinetics. Pharmacokinetic studies were limited because they either only reported antiretroviral pharmacokinetics compared with historical controls or presented data from healthy women taking single antiretrovirals.
Concurrent antiretroviral and hormonal contraceptive use also does not appear to lead to increased toxicity, though most studies were of short duration (1-28 days). Short-term pharmacokinetic studies may not accurately reflect adverse effects that may occur during use of longterm cART or PrEP with hormonal contraceptives. The only study to evaluate long-term toxicity showed little impact of concurrent use of DMPA and tenofovir on bone mineral density over 1 year [51]. Pharmacokinetic effects may also be time-dependent, further limiting the utility of short-term evaluations.
Strengths of our review included a comprehensive search strategy, systematic review of study inclusion by all authors, and dual data abstraction. Limitations are generally due to lack of studies evaluating relevant clinical outcomes. In addition, few studies are available regarding whether implants containing levonorgestrel or etonogestrel have different contraceptive effectiveness when used with efavirenz. Furthermore, with both implants, it is possible that interactions with enzyme inducers such as efavirenz are time-dependent, because hormone levels decrease over time after implantation [71]. Other data gaps are whether injectable contraceptives other than intramuscular DMPA, such as the lower-dose subcutaneous DMPA or injectable norethisterone enanthate, might be susceptible to drug interactions. Questions also remain regarding any impact of lower-dose efavirenz regimens [72] on hormonal contraceptive effectiveness, which cannot be predicted. Another limitation is that intracellular antiretroviral concentrations are likely a better predictor of clinical effectiveness than plasma levels, but pharmacokinetic studies only reported the latter. Finally, women on cART may also take other drugs that can alter liver metabolism, such as rifampin, and the combined effect of multiple enzyme-inducing medications on contraceptive hormone levels remains poorly characterized.
Our review highlights the dearth of studies designed to provide meaningful clinical data to guide contraceptive choices for HIV-positive women taking cART. Studies should be designed to report clinical outcomes such as pregnancy and HIV disease progression during long-term administration. Currently, incomplete data are being used to limit contraceptive choices for HIV-positive women.
In the absence of well conducted prospective clinical trials, data from pharmacokinetic studies and secondary analyses have been used to make clinical judgments on medication effectiveness and inform contraceptive policy. For example, in October 2014 the South African authorities recommended that women using efavirenz or other enzyme-inducing drugs should not use etonogestrel implants [11]. In May 2016, the European Medicines Agency recommended that women taking hepatic enzyme inducing drugs, including efavirenz, be offered double doses of oral levonorgestrel for postcoital emergency contraception [73]. When such guidance is developed, the absolute risk of pregnancy should be considered and addressed in the guidance publications, as well as other considerations such as availability and contraceptive effectiveness of the alternatives proposed. Even if a particular contraceptive method is potentially less effective than usual in a woman using a concomitant antiretroviral, it may still be more effective than many alternative contraceptive methods [74]. Although nonhormonal methods such as copper IUDs are not affected by drug interactions, their use remains very low in many settings worldwide, and efforts to increase IUD use have had limited success [7]. If access to implants is restricted, in many settings DMPA would be the primary option available to women, virtually eliminating womancentered decision making.
In summary, current published data do not support limiting women's access to any hormonal contraceptives. Women taking antiretrovirals for HIV treatment (in the form of cART) or prevention (in the form of PrEP) should have access to the full range of hormonal contraceptive options, and be enabled to make informed decisions about their options. Contraceptive efficacy is only one of many factors that an individual may consider when choosing a contraceptive method, and some women who are motivated to use the etonogestrel implant may wish to do so even if there is concern for decreased efficacy when used with efavirenz. National or regional restrictions on contraceptive method access, while well intentioned, supersede women's personal decisions, which may actually increase risk of unintended pregnancy if remaining contraceptive options are unacceptable or inaccessible. More well designed prospective studies are needed to examine potential drug interactions between antiretrovirals and all contraceptive methods, to better inform guidelines and counseling for the more than 16 million women living with HIV.