We conducted a prospective observational, parallel group, single-site PK study among women living with HIV on first-line ART, who were virally suppressed and between the ages of 18 and 45 years at the Wits RHI Research Centre, University of the Witwatersrand, in Johannesburg, South Africa.

The study was conducted in Johannesburg, Gauteng province, South Africa. The estimated HIV prevalence in Gauteng in 2022 was 15% and an estimated 67% of people living with HIV receiving care in the province are women [24]. At the time of the study design, efavirenz-containing ART was the recommended first-line ART in South Africa but by the time of study implementation, the South African guidelines recommended optimization of first-line ART from efavirenz- to dolutegravir-containing ART [25].

The study was reviewed and received ethical approval from the University of the Witwatersrand Human Research Ethics Committee, in South Africa (#201010), and the University of Washington Human Subjects Division, in the United States (#11487). The study also received approval from the South African Health Product Regulatory Agency, as at the time of initial review, doravirine was not yet a registered drug in South Africa. The study was registered with ClinicalTrials.gov (NCT04669678) and the South African National Clinical Trials Registry (DOH-27-062022-5454) [26, 27].

The 5 groups enrolled in EPIC were (1) doravirine + intramuscular (IM) depot medroxyprogesterone acetate (DMPA); (2) doravirine + subcutaneous (SC) MPA; (3) doravirine + subdermal ENG implant; (4) doravirine + nonhormonal intrauterine device (IUD); and (5) dolutegravir + IM DMPA. The sixth group, a historical comparator group, consisted of women living with HIV of similar age on dolutegravir + ENG implant from a Kenyan PK study called “PK of ARVs and Implant” (PARVI) [28]. Since the manufacturer of doravirine already conducted the industry standard PK study with doravirine and a combined oral contraceptive method, containing ethinyl estradiol and levonorgestrel [29], which did not demonstrate any bidirectional DDIs, we chose to exclude an oral contraceptive pill method in EPIC. We chose to not randomize assignments to contraceptive groups but to allow women to choose a method that best suited their preferences and allowed for enough persistence to the method for us to complete study follow-up, noting that women choosing one contraceptive method over another may have some differences [30].

Once enrolled, participants underwent a 6-week “lead-in” period with oral doravirine-containing ART to ensure that any hepatic induction with prior ARVs, especially efavirenz, was minimized before beginning contraception [26]. After the lead-in period, once participants initiated their chosen contraceptive method, they were followed every 2–4 weeks for an additional 12 or 24 weeks (primary endpoint), depending on the contraceptive method chosen (12 weeks for the IM DMPA or SC MPA groups and 24 weeks for the ENG implant and nonhormonal IUD groups). Additional details regarding eligibility, recruitment, retention, procedures, historical comparator group, and sample size calculations can be found in the Supplementary Text.

We considered the study groups the primary exposures. Groups 1–4 received doravirine in a single-pill combination of doravirine/lamivudine/tenofovir disoproxil fumarate (100 mg/300 mg/300 mg) taken orally once a day through the study. Groups 5 and 6 received dolutegravir/lamivudine/tenofovir disoproxil fumarate (50 mg/300 mg/300 mg) taken orally once a day through routine care, not from the study. At study exit, women in groups 1–4 were asked to resume their ART from routine care and their providers were notified of cessation of doravirine-containing ART.

Groups 1–4 received the IM DMPA, SC MPA, ENG implant, and nonhormonal IUD, respectively, while group 5 received IM DMPA and group 6 was on ENG implant. The study team administered IM DMPA (150 mg/mL) and SC MPA (104 mg/mL) every 3 months and placed the ENG implant (68 mg single rod) and nonhormonal IUD (T-shaped polyethylene frame wound with a copper wire) once. At study exit, IM DMPA and SC MPA group women were offered additional doses, while women on the ENG implant and nonhormonal IUD were offered removals if desired.

The co-primary outcomes were (1) the hormonal contraceptive concentrations and (2) doravirine concentrations, ideally timed at 24 hours since last dose to approximate concentration at 24 hours or minimum concentration (C_min), at each study visit. For the serum hormonal contraceptive concentrations, we utilized a validated, liquid chromatography–heated electrospray ionization–tandem triple quadrupole mass spectrometry (LC-MS/MS) technique, with the lower limit of quantification (LLOQ) of 10 pg/mL for ENG or MPA [27]. For the plasma doravirine concentrations, we also utilized LC-MS/MS techniques, with an LLOQ of 20 ng/mL.

Our secondary outcomes included (1) safety of ART (ascertained via solicited self-reports of frequency and severity of adverse events [AEs] at every study visit); (2) satisfaction of both ART and contraception (ascertained via self-reports at every study visit); and (3) doravirine-containing ART’s effectiveness (measured via HIV-1 RNA <40 copies/mL, via GeneXpert quantification) at study exit.

We present baseline characteristics by group descriptively. We calculated the geometric mean and ratios (GMR) for the log-transformed contraceptive hormone concentrations at each study visit by group, with 90% confidence intervals (CIs), comparing doravirine + IM DMPA (group 1) and doravirine + ENG implant (group 3) to its respective dolutegravir groups (ie, groups 5 and 6, respectively). Post hoc, in a sensitivity analysis, we censored visit concentrations for the contraceptive hormones when the ARV concentrations were below the LLOQ when expected (eg, doravirine concentration was <20 ng/mL when the participant was supposed to be taking doravirine-containing ART). Thus, for the contraceptive hormones, we presented both the primary outcomes as planned (regardless of detecting doravirine concentrations), termed “original sample,” and in the sensitivity analysis, termed “censored sample.” Similarly, when comparing each contraceptive method (ie, groups 1 and 3) to the IUD group (group 4) to examine differences in doravirine concentrations, post hoc, we censored visits when doravirine concentrations were below the LLOQ; this is the only doravirine concentration analysis presented. We calculated GMRs with linear mixed models per study visit. To account for any potential baseline differences by participant characteristics of women self-selecting which study group they would join, we utilized a priori multivariate models that helped account for potential confounding by age and body mass index (BMI), both measured at study start and treated as a continuous variable. In this analysis, we exclude planned comparisons between the IM DMPA and SC MPA groups.

This analysis includes 104 participants from the EPIC study and 24 from the PARVI study (Figure 1). The median age for all EPIC participants was 37 years (interquartile range [IQR], 29–40 years); among them, 8.4% were married, median parity was 2 (IQR, 2–3), median BMI was 24 (IQR, 20.7–26.9) kg/m², and median duration on ART prior to study start was 7 (IQR, 3–11) years (Table 1). Many of these characteristics were comparable between EPIC and PARVI participants, and across each group, though noting ethnic differences between South African and Kenyan participants.

There were a total of 6 participants who had initially enrolled but did not complete the study, 2 were terminated early from the study due to incident pregnancies prior to initiation of the contraceptive method, 2 discontinued their ENG implant, and 2 were considered lost to follow-up (Figure 1). Three participants, including the 2 already mentioned, became pregnant with varying amounts of exposure to doravirine; all 3 exited the study at pregnancy detection, stopped using doravirine, and were followed for pregnancy outcomes. One participant in group 4 was detected pregnant at study exit visit (with IUD strings visible on examination), estimated at 3 weeks of gestation, and was referred for IUD removal; her data are included in this analysis. This participant had twin gestation and went into labor preterm, with one fetus delivered stillbirth and the second infant born alive but deceased the next day; none of these adverse outcomes were deemed related to study or study product. The other 2 participants had uneventful pregnancies and normal vaginal deliveries, with infants well soon after birth.

During study follow-up, the total number of participants who did not have ARV concentrations detected above the LLOQ on at least one study visit were 12 in group 1 (doravirine + IM DMPA), 8 in group 2 (doravirine + SC MPA), 10 in group 3 (doravirine + ENG implant), 9 in group 4 (doravirine + nonhormonal IUD), and 3 in group 5 (dolutegravir + IM DMPA) (Table 1).

Comparator groups are delineated in Figure 2. Comparing IM DMPA hormonal contraceptive concentrations between doravirine- and dolutegravir-containing ART groups, the GMRs at each study visit ranged from 0.85 (90% CI, .65–1.11) to 1.04 (.91–1.19) in the original sample (most P values >.10; Table 2, Figure 3A). Comparing ENG implant hormonal contraceptive concentrations between doravirine- and dolutegravir-containing ART groups, the GMRs at each study visit ranged from 1.16 (95% CI, 1.07–1.26) to 1.22 (95% CI, 1.13–1.31) in the original sample (all P values <.10; Table 2, Figure 3C). For the above comparisons, no significant differences were detected between the original and censored samples (Table 2, Figure 3B and 3D).

Comparing doravirine concentrations between IM DMPA and nonhormonal IUD groups, the GMRs at each study visit ranged from 1.25 (95% CI, .57–2.74; P = .64) to 2.23 (95% CI, 1.23–4.06; P = .027) (Table 3, Figure 4A). Similarly, comparing doravirine concentrations between ENG and IUD groups, the GMRs at each study visit ranged from 0.78 (95% CI, .46–1.34; P = .46) to 2.61 (95% CI, 1.57–4.33; P = .002) (Table 3, Figure 4B).

A total of 264 solicited AEs were recorded from 104 EPIC participants, with 2 being at grade 3 or higher from 2 separate participants (Supplementary Tables 1 and 2). A total of 8 (3%) AEs grade 2 or higher were considered attributable to doravirine-containing ART and 22 (8%) to the contraceptive method.

Regarding satisfaction with their ART regimen, all participants indicated being satisfied or very satisfied with their ART during the study period (Supplementary Table 1). However, 22 (21%) participants indicated being somewhat or very dissatisfied with their contraceptive method ever during the study period.

Adherence to ART by pill count ≥80% per study visit was generally low (range, 35%–41%; Supplementary Table 1). ARV concentrations detected above the LLOQ per study visit were modest and variable (range, 66%–91%).

Viral suppression (<40 copies/mL) at study exit ranged from 86% to 96% between groups (Supplementary Table 1).

Despite being one of the most comprehensive PK studies to profile DDIs regarding concomitant use of doravirine-containing ART and hormonal contraceptives, our study faces several limitations. First, that not all participants adhered to their ART optimally limits exposure–outcome relationship ascertainment; nonetheless, our sensitivity analysis excluding ARV samples below limits of quantification did not alter the overall findings of no detrimental associations. Second, our historical comparator group was from a different country; our sociodemographic characteristics highlight some differences, and pharmacogenetic differences may play a role in PK associations for DDIs [25]. Third, our implementation of the timing of blood collection to truly generate ARV C_min outside of an intensive 24-hour sampling strategy were suboptimal. Thus, our analyses of doravirine concentrations resemble “random” time sampling; assuming this sampling is nondifferential between the groups, which we believe it is, our estimates will be biased toward the null. Therefore, future studies should better account for timing of blood collection, working carefully with study participants, to ensure the samples collected better represent the C_min time windows or conduct observed therapy with intensive 24-hour sampling.