Time Course of the Interaction Between Oral Short-Term Ritonavir Therapy with Three Factor Xa Inhibitors and the Activity of CYP2D6, CYP2C19, and CYP3A4 in Healthy Volunteers

This phase I trial was conducted as a single-centre, open-label, fixed sequence clinical trial. Ritonavir was administered orally as 100-mg tablets (Norvir^® 100-mg film-coated tablets, Abbvie, Ludwigshafen, Germany) bid for 5 days. Before ritonavir administration (baseline day) and on the last day of ritonavir intake, the participants received a microdosed cocktail of FXaI (Fig. 1) consisting of apixaban 25 µg (Eliquis^®, Bristol-Myers Squibb, Munich, Germany), edoxaban 50 µg (Lixiana^®, Daiichi Sankyo Europe, Munich, Germany), and rivaroxaban 25 µg (Xarelto^®, Bayer, Leverkusen, Germany) [25]. The oral stock solutions of edoxaban (30 µg/mL), rivaroxaban (25 µg/mL), and apixaban (25 µg/mL) were prepared by the Pharmacy of Heidelberg University Hospital in compliance with GMP guidelines according to the pharmaceutical development protocols approved by the competent authority (BfArM, Bonn, Germany).

To monitor the different CYP isozyme activities, midazolam (Dormicum^®; 5 mg/5 mL solutions for infusion; Roche, Grenzach-Wyhlen, Germany; 30 µg at baseline and 10 µg at all other administration times), yohimbine 50 µg (Yohimbinum hydrochloricum D4^® tablets, DHU-Arzneimittel GmbH, Karlsruhe, Germany), and omeprazole 100 µg (OMEP^® 40 mg power for solution for infusion; Hexal, Holzkirchen, Germany) were administered as oral solutions together with the FXaI cocktail before ritonavir was initiated, every day with the intake of ritonavir, and 2 and 4 days after stopping ritonavir. To buffer gastric acid and prevent degradation of the uncoated omeprazole, sodium hydrogen carbonate (4.2% w/v) was administered 10 min before (50 mL equivalent to 25 mmol) and simultaneously (100 mL equivalent to 50 mmol) with the 100-µg uncoated omeprazole dose as used previously [28]. Microdoses were always prepared immediately before administration and were administered with ritonavir always in the same sequence (first the FXaI, followed by yohimbine, midazolam, and omeprazole) within 1 min.

We enrolled volunteers who were in good physical and mental health and had no clinically relevant findings in medical history, physical examination, electrocardiogram, and routine laboratory analyses including renal and liver function. Not enrolled were pregnant or lactating women, participants with known intolerance or contraindication to any component of the trial medication, a history of anaphylactic shock, or participants of an interventional clinical trial within 30 days before inclusion. The use of medicines (except iodine and levothyroxine), cannabis, or citrus fruits (including grapefruit) was not allowed during the trial, and enrolment in the study was allowed at the earliest 5 half-lives (for inhibitors) or 2 weeks (for inductors) after discontinuation of the previously taken medicines. Alcohol and smoking were not allowed on study days. All participants used highly effective contraceptive methods (intrauterine devices, intrauterine hormone-releasing systems, vasectomized partner, or sexual abstinence for female volunteers and condom or sexual abstinence for men) during the trial and for a further 2 weeks after the last medication intake.

The trial was conducted at the Clinical Research Centre of the Department of Clinical Pharmacology and Pharmacoepidemiology (KliPS), Heidelberg University Hospital, which is certified according to DIN EN ISO 9001:2015. Participants arrived fasting (for at least 8 h) and remained fasting for a further 4 h after taking all microdoses simultaneously. Blood samples for the analyses of FXaI concentration were drawn before and 0.5, 1, 1.5, 2, 2.5, 3, 4, 6, 8, 23, and 24 h after drug administration on the first day (baseline) and at the fifth day of ritonavir intake (Fig. 1). On days with FXaI intake, the entire 24-h urine was collected. Blood samples for analyses of yohimbine, omeprazole, and ritonavir were taken before and 0.5, 1, 1.5, 2, 2.5, 3, and 4 h on every day of ritonavir intake and on days 2 and 4 after discontinuation of ritonavir. Additional ritonavir plasma samples were taken 6, 8, 23, and 24 h after dosing on baseline and on day 5. Midazolam plasma samples were taken before and 2, 2.5, 3, and 4 h after drug administration on each study day.

The concentrations of FXaI, midazolam, yohimbine, and omeprazole with 5’-hydroxy omeprazole were quantified by ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC–MS/MS) according to published assays [29‐32] with a lower limit of quantification (LLOQ) of 1 pg/mL for midazolam and yohimbine and 5 pg/mL for omeprazole, 5′-hydroxy omeprazole, and the FXaI. Ritonavir plasma concentrations were quantified by high-performance LC–MS/MS with a LLOQ of 10 ng/mL [2]. Urine samples were diluted 1:10 with blank human plasma for analyses of FXaI concentrations [33]. Accuracy and precision values were always ≤ ± 15% and all assays complied with the relevant guidelines for the validation of bioanalytical methods of the European Medicines Agency and the US Food and Drug Administration [34, 35].

The sample size was determined using the interaction data from a previous trial (EudraCT 2017-004453) [36] investigating the effect of ketoconazole on FXaI, which was assumed to be comparable to ritonavir. A sample size of 3 (apixaban and rivaroxaban) and 5 (edoxaban) was calculated to detect the observed area under the plasma concentration–time curve (AUC) extrapolated to infinity (AUC_inf) ratio of 1.68 ± 0.36 for apixaban, 2.28 ± 1.10 for edoxaban, and 2.37 ± 0.53 for rivaroxaban with a power of 0.8 and an α error of 5%. Using these data, a two-sided paired t-test with a sample size of N = 8 (log-transformed AUC_inf) would allow to detect a 20% change of exposure (i.e., an AUC_inf ratio of 1.20) for apixaban, rivaroxaban, and edoxaban with a power of 80% and an α error of 5%.

The standard non-compartmental pharmacokinetic parameters of each FXaI were calculated using Phoenix™ WinNonlin Software 8.3.5 (Certara, Princeton, NJ, USA). The calculated parameters included maximum plasma concentration (C_max), time to C_max (t_max), AUC_inf, AUC from 0 to 24 h (AUC_0–24), 0 to 4 h (AUC_0–4, omeprazole, yohimbine), and 2 to 4 h (AUC_2–4, midazolam), apparent volume of distribution at steady-state (V_z/F), apparent oral clearance (CL/F), terminal elimination half-life (t_½), renal clearance (CL_renal), expressed as the amount excreted unchanged in urine over 24 h divided by AUC_0–24, and non-renal clearance (CL_nonrenal) as the difference between CL/F and renal clearance. The AUC was calculated using a mixed log-linear trapezoidal model.

To determine the effect of ritonavir on CYP3A activity, the estimated midazolam partial metabolic clearance (eCL_met) was calculated using the data of a limited sampling methodology [8, 37, 38]. Due to the prolonged midazolam elimination half-life with coadministration of ritonavir, midazolam concentrations above the LLOQ were observed in predose samples from the second day of ritonavir intake until 4 days after ritonavir discontinuation, which was taken into account as follows: the residual AUC was calculated using the predose concentration and the current half-life in the interval 2–4 h of the individual visit and then subtracted accordingly. The CYP3A4 recovery half-life was calculated as described previously [39].

The effect of ritonavir on CYP2C19 and CYP2D6 activity was determined using AUC_0–4 and C_max of yohimbine and omeprazole. The exposure of ritonavir was determined using AUC_0–4 after dose 1, 3, 5, 7, and 9, and after discontinuation.

Data are reported as geometric means with 95% confidence intervals (CIs) except t_max, which is reported as median with range. To test for differences of each FXaI separately between baseline and ritonavir, the geometric mean ratios (GMRs) and their 90% CIs were calculated. The default no-effect boundary of point estimates is 0.8–1.25 and when 90% CIs for systemic exposure (AUC) ratios are entirely within the equivalence range of 0.85–1.25, according to the standard practice of the FDA, it is concluded that no clinically relevant interaction is present. For other FXaI PK parameters, a ratio paired t-test (two-tailed) was performed to compare day 5 of ritonavir intake with baseline.

To account for the multiple occasions of assessing midazolam, yohimbine, and omeprazole exposure, repeated-measure one-way ANOVA of logarithmic transformed data with correction for multiple testing (Dunnett’s multiple comparisons test) was performed to test for differences to baseline.

Over the observation period, all plasma concentrations were above the LLOQ, except for the 23- and 24-h edoxaban concentrations in two and the 24-h edoxaban concentration in three participants. The 24-h urine collection from one participant was excluded from analyses due to suspected incomplete urine collection on day 5.

Coadministration of ritonavir 100 mg bid for 5 days increased plasma concentrations of all three FXaI (Fig. 2; Tables 1 and 2). The individual increase in AUC_0–24 for apixaban and edoxaban was heterogeneous: three participants had an AUC_0–24 ratio of 0.8–1.2 versus two and three participants, respectively, with AUC_0–24 increases of 2–2.3 (Fig. 3). The largest and consistent change was observed for rivaroxaban. A slightly but significantly prolonged t_½ (from 5.14 h to 6.57 h) and a significantly reduced CL_renal (from 49.4 to 34.8 mL/min) were observed for rivaroxaban but not for the other two FXaI. The volume of distribution was significantly reduced for edoxaban and rivaroxaban but not for apixaban.

The AUC_0–4 of ritonavir was 906 ng/mL⋅h (95% CI 543–1512) and C_max was 373 ng/mL (95% CI 244–572) after intake of the first dose. Exposure and C_max were similar on the remaining 4 days, ranging between 4593–5331 ng/mL⋅h and 1776–2107 ng/mL, respectively (Fig. 7). Two days after the last ritonavir dose, concentrations were close to or below the LLOQ.

Overall, the study medication was well tolerated. Twenty adverse events (5 considered related to study drug) were observed in six participants with most adverse events occurring on baseline day, that is, before ritonavir administration (Supplementary Table S1, see electronic supplementary material [ESM]), and all were transient. One adverse event (presyncope) was moderate and one (syncope) was severe and both occurred as an immediate consequence of unsuccessful venepunctures; all other adverse events were mild. The most common adverse event was headache on the first study day without ritonavir therapy (5 out of 8 participants affected). A relationship was considered unlikely, as only microdosed substances were administered on this day and headaches only occurred in one participant after repeated administration on the fifth day of ritonavir intake.

Treatment with ritonavir 100 mg bid for 5 days increased the AUC_0–24 of all currently marketed FXaI but with different magnitude and likely little clinical relevance in the absence of other clearance impairment (e.g., renal dysfunction) [40‐42]. The largest increase was observed with rivaroxaban but this change was below the order of magnitude that usually prompts dose reductions (factor 2). Unless very high doses are administered, as required in the initial treatment of major embolic events [43], dose reductions should be individually planned and it should also be considered that COVID-19 can trigger a prothrombotic state and at least prophylactic anticoagulation is generally recommended [44].

The differences in the magnitude of interaction between the different FXaI are not fully elucidated. They are certainly influenced by the different effects of ritonavir on the diverse clearance mechanisms involved, and similar differences have also been observed in a series of patients receiving these FXaI together with protease inhibitors boosted with ritonavir 100 mg bid [18‐23]. In these earlier case series and in our study, the interaction was most pronounced with rivaroxaban, the FXaI with the largest bioavailability (and complete bioavailability at low doses [45]), indicating that the exposure change was not caused by a change in bioavailability but rather a reduction of systemic drug clearance, which is also reflected in a prolongation of the elimination t_½. Other than with other FXaI, these clearance changes were partly due to reduced renal excretion, likely reflecting combined inhibition of efflux transporters (P-gp and BCRP) and of the organic anion transporter 3 (OAT3, SLC22A8), which plays a major role in the renal elimination of rivaroxaban [46, 47]. In our study with low ritonavir doses, the effect was smaller than the interaction observed with very high doses of ritonavir (600 mg bid), where a 2.5-fold increase in AUC and an even more extensive reduction of CL_renal were observed [17, 25, 36]. These data therefore suggest that boosting doses of ritonavir do not completely inhibit renal efflux transporters while they are likely high enough to inhibit CYP3A4 (and possibly CYP2J2 [7, 17, 48]) as efficiently as higher doses [2].

In contrast, there was only a minor DDI with apixaban, which is in agreement with the minimal interaction with potent CYP3A inhibitors such as voriconazole [25, 36] and the limited suspected contribution of efflux transporters [49] or OAT3 [47].

Ritonavir also increased edoxaban exposure but, in contrast to rivaroxaban, CL_renal and t_½ were not altered, suggesting an increase in bioavailability as a result of intestinal inhibition of efflux transporters. This is consistent with other DDI studies and physiology-based pharmacokinetic modelling, which suggested a role of intestinal P-gp (and possibly BCRP) in edoxaban bioavailability [50‐52]. It also agrees well with the effect size observed with the P-gp substrate dabigatran etexilate when it was administered with the same dose of ritonavir [53]. In addition, ritonavir also inhibits carboxylesterase 1 (CES1) in vitro [54], suggesting that concurrent inhibition of this minor pathway [10] could additionally be involved. The different influence of ritonavir on the urinary excretion of FXaI could be due to the fact that edoxaban and apixaban are not substrates of OAT3 and therefore the activity of the downstream renal P-gp is less important for their urinary excretion [47, 51].

Our trial generated activity data of relevant drug metabolising CYPs over the time course of ritonavir treatment (5 days) and after its discontinuation. As expected [2], already after the first ritonavir dose, the CYP3A4 clearance profoundly decreased, indicating strong inhibition. This is in line with the dose–CYP3A4 inhibition relationship established in vivo, which estimated that the ritonavir dose inhibiting CYP3A4 by 50% (ED₅₀) was very low (3.4 mg) [2]. After ritonavir discontinuation, the CYP3A activity recovered slowly with a half-life of 2.3 days. On day 4 after ritonavir discontinuation, CYP3A activity was still reduced while no ritonavir was detectable in plasma, which suggests prolonged mechanism-based inactivation of CYP3A4. Based on these data, the full recovery of CYP3A4 activity is estimated to take 10–12 days after a 5-day treatment course with ritonavir 100 mg bid.

Regarding the effect of ritonavir on CYP2D6 activity, no significant changes of the probe drug yohimbine were observed. In contrast, using desipramine as an index substrate for CYP2D6, ritonavir (100 mg bid) caused a modest inhibitory effect on CYP2D6 activity in extensive metabolizers [55]. However, the dextromethorphan/dextrorphan metabolic ratio as another indicator for CYP2D6 activity was not altered. Thus, these findings with microdosed yohimbine as a sensitive CYP2D6 probe drug [27] confirm that no clinically relevant alteration of CYP2D6 enzyme activity during the course of a 5-day treatment with ritonavir 100 mg bid occurs.

Ritonavir had a small induction effect on CYP2C19 activity leading to slightly decreased omeprazole exposure on day 5 of ritonavir treatment, which was also reflected in corresponding changes in the molar AUC ratio of omeprazole and 5-hydroxy omeprazole. This inducing effect is time-dependent as it is more pronounced after a 10-day ritonavir (100 mg bid) treatment [56]. CYP2C19 induction is also dose-dependent as reported with other CYP2C19 substrates; as an example, after 20 days of ritonavir (100 mg bid) treatment, voriconazole exposure was decreased by 33% [57] or 39% [58], which is only a small effect, probably attenuated by the autoinhibition of voriconazole [59, 60]. However, higher doses of ritonavir 400 mg bid for 20 days decreased voriconazole AUC by 83% [58]. Although short-term ritonavir treatment with low doses induces CYP2C19 activity, this effect is short lasting and likely does not require any dose adaptation of CYP2C19 substrates.

The following limitations merit discussion. First, we only assessed the effect on the last day of the 5-day ritonavir treatment, which was expected to be the day with the maximum possible change in short-course low-dose therapies with ritonavir. Due to its short duration, this trial was not suitable to assess the potential inducing properties of ritonavir and its results should only be extrapolated to longer treatment regimens with caution. In addition, no transporter probe drugs were tested, so the time course of transporter inhibition is still not well characterized. Second, patients may differ from healthy volunteers in many ways. In patients, comorbidities such as renal impairment can potentiate apparently minor DDI [41, 42] and it will always be critical to consider all relevant clearance pathways when determining appropriate individualized treatment plans for FXaI. Third, coadministration of the probe drugs with sodium hydrogen carbonate to buffer gastric acid could potentially influence their pharmacokinetics. However, the baseline pharmacokinetics agree well with prior studies using microdosed FXaI [25, 36], suggesting that buffering did not alter the pharmacokinetics of the three microdosed FXaI. This is in line with the absence of clinically relevant DDI of therapeutic FXaI doses with acid lowering drugs such as famotidine, ranitidine, or omeprazole [61‐64]. In addition, the clearance values and relative contribution of renal excretion compare well with those reported for regular doses of FXaI in healthy volunteers [17, 18, 65], suggesting that microdoses adequately reflect the pharmacokinetics of regular doses and are therefore suitable for DDI studies.