Volume 12, Issue 4 p. 532-544
ARTICLE
Open Access

Tiered approach to evaluate the CYP3A victim and perpetrator drug–drug interaction potential for vonoprazan using PBPK modeling and clinical data to inform labeling

Darcy J. Mulford

Corresponding Author

Darcy J. Mulford

Phathom Pharmaceuticals, Inc., Buffalo Grove, Illinois, USA

Correspondence

Darcy J. Mulford, Phathom Pharmaceuticals, Inc., 2150 E. Lake Cook Road, Suite 800, Buffalo Grove, IL 60089, USA.

Email: [email protected]

Search for more papers by this author
Diane Ramsden

Diane Ramsden

Takeda Pharmaceuticals, Cambridge, Massachusetts, USA

Search for more papers by this author
Liming Zhang

Liming Zhang

Takeda Pharmaceuticals, Cambridge, Massachusetts, USA

Search for more papers by this author
Ingrid Michon

Ingrid Michon

Certara, Sheffield, UK

Search for more papers by this author
Eckhard Leifke

Eckhard Leifke

Phathom Pharmaceuticals, Inc., Buffalo Grove, Illinois, USA

Search for more papers by this author
Neila Smith

Neila Smith

Phathom Pharmaceuticals, Inc., Buffalo Grove, Illinois, USA

Search for more papers by this author
Hannah M. Jones

Hannah M. Jones

Certara, Sheffield, UK

Search for more papers by this author
Carmelo Scarpignato

Carmelo Scarpignato

United Campus of Malta, Msida, Malta

Search for more papers by this author
First published: 10 March 2023
Citations: 1

Darcy J. Mulford and Diane Ramsden contributed equally to this work.

Abstract

Vonoprazan is metabolized extensively through CYP3A and is an in vitro time-dependent inhibitor of CYP3A. A tiered approach was applied to understand the CYP3A victim and perpetrator drug–drug interaction (DDI) potential for vonoprazan. Mechanistic static modeling suggested vonoprazan is a potential clinically relevant CYP3A inhibitor. Thus, a clinical study was conducted to evaluate the impact of vonoprazan on the exposure of oral midazolam, an index substrate for CYP3A. A physiologically-based pharmacokinetic (PBPK) model for vonoprazan was also developed using in vitro data, drug- and system-specific parameters, and clinical data and observations from a [14C] human absorption, distribution, metabolism, and excretion study. The PBPK model was refined and verified using data from a clinical DDI study with the strong CYP3A inhibitor, clarithromycin, to confirm the fraction metabolized by CYP3A, and the oral midazolam clinical DDI data assessing vonoprazan as a time-dependent inhibitor of CYP3A. The verified PBPK model was applied to simulate the anticipated changes in vonoprazan exposure due to moderate and strong CYP3A inducers (efavirenz and rifampin, respectively). The clinical midazolam DDI study indicated weak inhibition of CYP3A, with a less than twofold increase in midazolam exposure. PBPK simulations projected a 50% to 80% reduction in vonoprazan exposure when administered concomitantly with moderate or strong CYP3A inducers. Based on these results, the vonoprazan label was revised and states that lower doses of sensitive CYP3A substrates with a narrow therapeutic index should be used when administered concomitantly with vonoprazan, and co-administration with moderate and strong CYP3A inducers should be avoided.

Study Highlights

  • WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

In vitro studies identified vonoprazan as a substrate and time-dependent inhibitor of CYP3A. Vonoprazan may alter the exposure of CYP3A substrates and similarly, modulators of CYP3A could impact vonoprazan exposure.

  • WHAT QUESTION DID THIS STUDY ADDRESS?

A clinical study determined the potential for vonoprazan to inhibit a sensitive CYP3A substrate (oral midazolam). The data were used to refine and validate a physiologically-based pharmacokinetic (PBPK) model in Simcyp to simulate the victim drug–drug interaction (DDI) potential for vonoprazan with moderate and strong inducers.

  • WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

The PBPK model was applied to simulate scenarios where clinical data are lacking, including the impact of CYP induction on vonoprazan exposure. The analysis supported the vonoprazan drug label, including to avoid co-administration of vonoprazan with moderate and strong inducers.

  • HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?

The PBPK modeling offers a viable approach to evaluate the victim and perpetrator DDI risks of CYP3A substrates, including vonoprazan. This study highlights the opportunity to apply PBPK modeling in lieu of conducting a dedicated DDI study with inducers.

INTRODUCTION

Vonoprazan, a potassium-competitive acid blocker, suppresses gastric acid secretion by competitively inhibiting the gastric enzyme hydrogen potassium adenosine triphosphatase (H+, K+–ATPase). Vonoprazan is being investigated in the United States and Europe for clinical use in patients with erosive esophagitis and nonerosive reflux disease. Vonoprazan is approved by the US Food and Drug Administration (FDA) for the treatment of Helicobacter pylori infection in combination with either amoxicillin or amoxicillin and clarithromycin.1-7 Orally administered vonoprazan is rapidly absorbed, displays time- and dose-independent pharmacokinetics (PKs), and its exposure is not significantly altered by food.8 Vonoprazan provides rapid, potent, and durable gastric acid suppression following once-daily dosing.8

Vonoprazan is highly metabolized through multiple pathways, including a combination of cytochrome P450 (CYP) isoforms, and conjugation pathways such as sulfo- and glucuronosyl-transferases. In vitro studies with recombinant CYP enzymes determined that CYP3A4 is the primary enzyme responsible for CYP-mediated metabolism of vonoprazan, with CYP2D6 and CYP2C19 providing minor contribution.9 Vonoprazan is also metabolized by sulfotransferase SULT2A1 in vitro. Clinical data confirm that polymorphisms of CYP2C19 do not impact vonoprazan PKs and together with the phenotyping data suggest that CYP3A plays a larger role in metabolic clearance of vonoprazan.2, 4 In vitro, vonoprazan provided reversible and time-dependent inhibition of CYP3A (half maximal inhibitory concentration [IC50] = 10 μM, inhibitory constant [Ki] = 3.0 μM, concentration of mechanism-based inhibition associated with half maximal activation rate [Kapp] = 1.22 μM, and inactivation rate of the enzyme [Kinact] = 0.0161 min−1).10 Given that CYP3A contributes greater than 25% to the overall metabolic clearance of vonoprazan and vonoprazan inhibits CYP3A in vitro, vonoprazan is potentially involved in clinical CYP3A drug–drug interactions (DDIs), both as a victim and a perpetrator.

Clinical evaluation of the impact of a strong inhibitor and inducer is indicated when CYP3A is a predominant clearance pathway.11-13 In a clinical DDI study, co-administration of vonoprazan with a strong CYP3A4 inhibitor (clarithromycin) resulted in a 1.58-fold increase in vonoprazan exposure.14 As vonoprazan is approved for the treatment of H. pylori infection when used in combination with amoxicillin or amoxicillin and clarithromycin, potential drug interactions were evaluated in a clinical study with the combined regimen of vonoprazan, clarithromycin, and amoxicillin.15 Co-administration with clarithromycin and amoxicillin increased exposure of vonoprazan by 1.8-fold. No difference was observed in the PKs of amoxicillin but the clarithromycin exposure increased by 1.5-fold when administered as triple therapy. A clinical study evaluating the impact of a strong CYP3A inducer, rifampin, on vonoprazan exposure was paused due to potential safety issues related to possible nitrosamine contaminants in rifampin clinical supplies.16, 17 Although other strong CYP3A inducers, including phenytoin or carbamazepine, were considered as an alternative to rifampin for the clinical DDI study, a modeling approach was instead chosen to avert potential safety concerns when these agents are administered to healthy participants.18-20

The goals of this study were to: (a) apply mechanistic static modeling to determine clinical relevance of the in vitro CYP3A time-dependent inhibition; (b) evaluate the CYP3A inhibition potential of vonoprazan in a clinical DDI study using oral midazolam as an index substrate; (c) develop and verify a physiologically-based PK (PBPK) model for vonoprazan using in vitro and clinical PK data, including DDI data with clarithromycin and midazolam; (d) apply the PBPK model to simulate the impact of moderate and strong inducers on vonoprazan exposure; and (e) inform the DDI portions of the labeling for vonoprazan.

METHODS

Mechanistic static modeling

The DDI risk potential for vonoprazan with sensitive CYP3A substrates was evaluated using basic and mechanistic static modeling as proposed by regulatory guidance documents.11-13 The mechanistic static model has been described elsewhere.21, 22 Several input parameters for vonoprazan were investigated, including using the maximum and average unbound plasma concentrations at steady-state (Cmax,ss, Cav,ss,u) observed following 20-mg once-daily (q.d.) or twice-daily (b.i.d.) vonoprazan and also following 20-mg b.i.d. vonoprazan co-administered with clarithromycin (Table S1). Vonoprazan plasma concentrations following b.i.d. dosing and in the presence of clarithromycin are higher than those observed following q.d. dosing.14, 15 Thus, these higher perpetrator concentrations were included to provide the worst case DDI estimates for vonoprazan when co-administered with a selective CYP3A substrate such as midazolam.

Clinical DDI study with midazolam

Participants

Participants were healthy adults aged 18–45 years who were nonsmokers without clinically significant medical conditions by history, physical examination, laboratory testing, or electrocardiography, with a body mass index (BMI) of 18–30 kg/m2. Of 32 participants screened, 20 were enrolled and completed the study (Table S2). Eight (40%) were women; mean age was 32 years; 10 (50%) were White; seven (35%) were Black; eight (40%) were Latinx; and mean BMI was 25.5 kg/m2. Participants were housed in the clinic from the day before the first dose (day −1) until the end of all study procedures (day 11). Participants received standardized meals and snacks; foods and beverages known to inhibit or induce CYP enzymes were prohibited throughout the study. The trial protocol was approved by the appropriate institutional review board.

Study medications and design

Participants received a single oral dose of midazolam 2 mg (syrup) on days 1 and 9, and vonoprazan 20-mg b.i.d. (oral tablets) on days 2 through 10 (Figure S1). Participants fasted overnight for at least 8 h before the morning dose of study medication. On day 9, midazolam was administered 1 h after the morning dose of vonoprazan. Participants continued fasting for 4 h after dosing on days 1 and 9. On vonoprazan-only days, food was allowed 1 h after vonoprazan dosing. Evening vonoprazan dosing occurred more than 1 h before or after a meal.

Pharmacokinetic and statistical analyses

Blood samples to measure midazolam and 1-hydroxymidazolam plasma concentrations were collected predose and at various time intervals through 24 h following midazolam dosing on days 1 and 9. Two additional blood samples were collected after dosing on day 9: at 36 and 48 h. Blood samples to measure vonoprazan plasma concentrations were collected predose on days 2–11. Samples were analyzed using validated liquid chromatography coupled with tandem mass spectrometry assays for midazolam, 1-hydroxymidazolam, and vonoprazan in human plasma. Calibration curves for midazolam and 1-hydroxymidazolam included eight concentrations within the validation range of 0.1 to 100 ng/mL. The accuracy and precision of the method was assessed with quality control (QC) samples at concentrations of 0.3, 40, and 75 ng/mL. Precision values, based upon coefficient of variation (%CV) of QC samples, were ≤5.66% for midazolam and ≤4.94% for 1-hydroxymidazolam. Accuracy values, based upon the calibration standards across the range, were between 98.1% and 104.19% for midazolam and between 98.53% and 104.27% for 1-hydroxymidazolam. Overall, the method was determined to be reliable based on validation results for accuracy, precision, specificity, linearity, and reproducibility. PK parameters for midazolam and 1-hydroxymidazolam calculated using noncompartmental methods included the area under the curve (AUC) from time 0 to the last quantifiable concentration (AUC0-t) and from time 0 to infinity (AUC0-), Cmax, time to reach Cmax (Tmax), apparent first-order terminal elimination-rate constant (Kel), and first-order terminal elimination half-life (t1/2). The effect of vonoprazan on the PKs of midazolam was assessed using an analysis of covariance model with treatment as a fixed effect and subject as a random effect performed on the natural log-transformed values of Cmax, AUC0-t, and AUC0- for midazolam and 1-hydroxymidazolam. Treatment differences were expressed using point estimates and 90% confidence intervals.

Safety and tolerability assessments

Participants were monitored throughout the study for adverse events (AEs). Safety and tolerability end points included monitoring and recording of AEs, clinical laboratory test results (hematology, serum chemistry, and urinalysis), vital-sign measurements, 12-lead electrocardiogram results, and physical examination findings.

PBPK model development, verification, and application

Input parameters for vonoprazan are specified in Table 1. Simulations were conducted using Simcyp version 19 (www.certara.com/software/simcyp-pbpk). The PBPK modeling strategy is outlined in Figure 1; a “bottom-up” approach was used by incorporating in vitro and physiochemical properties for vonoprazan. The fraction absorbed (fa) and the first-order absorption parameters were derived from in vitro permeability and mass balance data from the [14C] human absorption, distribution, metabolism, and excretion (ADME) study.1 In vitro supersome data were used to assign the relative contribution of CYP3A4 to the clearance of vonoprazan, Ki, and Kinact values for time-dependent inhibition of CYP3A4 by vonoprazan were incorporated within the PBPK model. Thus, auto-inhibition was simulated within the model (Figure S2). Observed oral clearance (CL/F) data obtained from healthy participants were used to optimize intrinsic clearance values to recover the observed vonoprazan plasma concentration–time profiles. Steady-state volume of distribution was predicted using the Rodgers and Rowlands method.23 Observed renal clearance (CLR) was input directly. A “middle-out” approach was used to optimize parameters including lag time (tlag) and first-order absorption rate constant (Ka) by manual adjustment during the model development phase.

TABLE 1. Final input parameters for vonoprazan in the PBPK model.
Parameter Vonoprazan Source
Physicochemical and blood binding
MW (g/mol) 345.39 Experimental data
log P 2.74 Experimental data
Compound type Monoprotic base
pKa 9.3 Experimental data
B:P 0.93 Echizen et al. 20161
f u 0.135 Experimental data
Distribution Full PBPK model
VSS (L/kg) 3.82 Method 2 predicted
Absorption First order
f a 0.99 User defined based on hADME
ka (1/h) 0.5 Optimized
tlag (h) 0.4 Optimized
Papp (x10−6 cm/s) 17.8 Experimental data
Calibrator Papp (x10−6 cm/s) 32.8 Experimental data
Peff,man (x10−4 cm/s) 2.55 Predicted
Qgut (user) (L/h) 13.5 Retrograde calculator
f u,gut 1 Default
Elimination Enzyme kinetics
CLR (L/h) 4

Clinical study

Retrograde model. CL/F obtained from clinical study (115 L/h); fa 0.99; FG 0.86; CLR 4 L/h; fmCYP3A4 45.4%

CYP3A4 CLint (μl/min/pmol) 0.601
Additional HLM CLint (μL/min/mg) 90.9
Interaction
CYP3A4 Ki (μM) 3.0 From experimental-derived IC50 (10 μM) using Cheng–Prusoff equation
f u,mic 0.94 Calculated at 0.2 mg/mL
CYP3A4 Kapp (μM) 1.22 Experimental data
CYP3A4 Kinact (1/h) 0.966 Experimental data
  • Abbreviations: B:P, blood-to-plasma ratio; CL/F, oral clearance; CLint, intrinsic clearance; CLR, renal clearance; CYP, cytochrome P450; fa, fraction absorbed; FG, intestinal availability; fmCYP3A4, fraction of vonoprazan metabolized by CYP3A; fu, fraction unbound in plasma; fu,gut, fraction of drug unbound in the gut; fu,mic, free fraction of drug in an in vitro microsomal preparation; hADME, human absorption, distribution, metabolism, and excretion; HLM, human liver microsomes; IC50, half maximal inhibitory concentration; ka, absorption rate constant; Kapp, concentration of mechanism-based inhibition associated with half maximal activation rate; Ki, inhibitory constant; Kinact, inactivation rate of the enzyme; MW, molecular weight; P, partition coefficient; Papp, apparent permeability coefficient; PBPK, physiologically-based pharmacokinetic; Peff,man, effective human jejunum permeability; pKa, acid dissociation constant; Qgut, flow rate for overall delivery of drug to the gut (drug dependent); tlag, lag time; Vss, volume of distribution at steady-state.
Details are in the caption following the image
PBPK modeling scheme. ADME, absorption, distribution, metabolism, and excretion; BID, twice-daily; CL/F; oral clearance; CLR, renal clearance; CYP, cytochrome P450; DDI, drug–drug interaction; fa, fraction absorbed; fmCYP3A4, fraction of vonoprazan metabolized by CYP3A; fumic, free fraction of drug in an in vitro microsomal preparation; HV, healthy volunteers; Kapp, concentration of mechanism-based inhibition associated with half maximal activation rate; Kinact, inactivation rate of the enzyme; Ki, concentration of inhibitor that supports half maximal inhibition; MBI, mechanism-based inhibition; PBPK, physiologically-based pharmacokinetic; QD, once-daily; SD, single dose; tlag, lag time.

Model development

Predictions of plasma drug concentration–time profiles, clearance, and DDIs were performed in the Simcyp Simulator using a population of healthy virtual individuals and the strategy outlined in Figure 1. Default Simcyp parameter values were used to create a virtual North-European White population, as described previously.24, 25 For model development, simulations were conducted for 20-mg single-dose (SD) and 20-mg q.d. vonoprazan with fasting. The trial design used for simulation of the plasma concentration–time profiles of vonoprazan in healthy participants was based on the clinical study design. In total, nine participants aged 21–36 years (0% women) were recruited and received either a 20-mg SD of vonoprazan or seven daily 20-mg doses. For the simulations, 10 virtual trials were generated to assess variability across groups, and simulated profiles of vonoprazan were compared with observed data. The model was considered to be performing well when simulated values were within 1.25-fold of the observed values.

Model verification

The model was verified by simulating data from additional studies (not used during model development) across dose levels and regimens. All simulations replicated the designs of the noted clinical studies.

Vonoprazan as a victim drug: simulating interaction with clarithromycin

The fraction of vonoprazan metabolized by CYP3A (fmCYP3A) was verified using a clinical DDI study with clarithromycin as a strong CYP3A inhibitor. In total, 16 healthy participants aged 19–44 years (0% women) were recruited. In treatment period 1, participants received a 40-mg vonoprazan SD after breakfast. In treatment period 2, participants received 500-mg clarithromycin b.i.d. alone on days 1–5 and day 7. On day 6, participants received 500-mg clarithromycin b.i.d., co-administered with a 40-mg vonoprazan SD 1 h after clarithromycin administration. In simulations, 10 virtual trials of 16 participants aged 19–44 years (0% women) were generated to assess variability across groups. Simulated profiles of vonoprazan were compared with observed data. The default compound library file for clarithromycin was used (refer to Table S3).

Vonoprazan as a perpetrator drug: simulating interaction with midazolam

The study design was as described herein. The PBPK model was used to verify the CYP3A time-dependent and direct-inhibition parameters (Ki, Kapp, and Kinact). In simulations, 10 virtual trials of 20 participants aged 21–43 years (40% women) were generated to assess variability across groups. Manual sensitivity analysis was performed to evaluate the impact of Ki on the estimated DDI magnitude by repeating DDI simulations after reducing the in vitro Ki value by 10-fold. Simulated midazolam profiles were compared with observed data. The default compound library file for midazolam was used (refer to Table S3).

PBPK model application

Vonoprazan as a victim drug: simulating interactions with vonoprazan in combination with efavirenz or rifampin

The final verified PBPK model was utilized to simulate changes in the PK profile for vonoprazan in the presence of moderate and strong CYP3A inducers (Figure 1). For each simulation, 10 virtual trials of 20 healthy participants aged 18–50 years (50% women) were used. Vonoprazan conditions included 10-, 20-, and 40-mg SD PK simulation at 1 h postdose on the 16th day of 18 days of rifampin or efavirenz (600-mg q.d.). In addition, PK data were simulated at 10- and 20-mg q.d., and 20-mg b.i.d. vonoprazan for 7 days followed by 16 days of rifampin or efavirenz. The default compound library files for rifampin and efavirenz were used (refer to Table S3).

Vonoprazan as a victim drug: simulating interactions with vonoprazan in combination with clarithromycin

The suggested treatment regimen for vonoprazan in patients with H. pylori infection is 20-mg b.i.d. for 14 days in combination with either 1000-mg amoxicillin three times a day or 1000-mg b.i.d. amoxicillin, and 500-mg b.i.d. clarithromycin. To explore the combination of vonoprazan with clarithromycin for the treatment duration of 14 days, 10 virtual trials of 20 participants aged 18–50 years (50% women) receiving multiple oral doses of vonoprazan (20-mg b.i.d. for 14 days) in the absence of rifampin/efavirenz or clarithromycin, and following 14 days with dosing of rifampin/efavirenz (600-mg q.d.) and/or clarithromycin (500-mg b.i.d.), were generated. The default compound library files for clarithromycin, efavirenz, and rifampin were used (refer to Table S3).

RESULTS

Mechanistic static modeling

The basic model could not exclude the potential for relevant clinical inhibition of CYP3A substrates by vonoprazan (kobs + kdeg/kdeg ≥ 1.25). The potential for vonoprazan, ± clarithromycin, to act as a clinically relevant inhibitor towards substrates of CYP3A was further investigated using mechanistic static modeling. The mechanistic static model using oral midazolam as the substrate of CYP3A, and the inhibition parameters derived in vitro, predicted that vonoprazan has the potential to be a clinically relevant, albeit weak, inhibitor of CYP3A (AUC ratio [AUCR] 1.82–2.49; Table S1).

Clinical DDI study with midazolam

PK and statistical results

Mean plasma concentrations of midazolam and 1-hydroxymidazolam were greater at all sampling times when midazolam was co-administered with vonoprazan compared with midazolam alone (Figure S3). Midazolam Cmax and AUC values increased 1.9-fold when midazolam was administered after repeated vonoprazan doses versus administration alone (Table 2). Plasma exposure of 1-hydroxymidazolam also increased following vonoprazan co-administration, but to a lesser extent. The 1-hydroxymidazolam Cmax increased 1.3-fold and AUC increased from 1.3- to 1.4-fold when midazolam was administered after repeated vonoprazan doses compared with administration alone (Table 2). Elimination of midazolam and 1-hydroxymidazolam was similar when midazolam was administered alone or concomitantly with vonoprazan, as reflected by no meaningful change in t1/2 for either and suggestive of intestinal inhibition of CYP3A.

TABLE 2. Plasma PK parameter estimates determined for midazolam and 1-hydroxymidazolam.
Parameter (units) Treatment Mean ratio (90% CI)
Midazolam alone (N = 20) Midazolam + vonoprazan (N = 20)
Mean (SD) LS mean Mean (SD) LS mean
Midazolam
Cmax (ng/mL) 10.3 (3.61) 9.74 20.3 (9.55) 18.8 1.93 (1.61, 2.33)
AUC0-t (ng*h/mL) 24.2 (8.76) 22.7 50.7 (39.8) 43.7 1.92 (1.53, 2.42)
AUC0- (ng*h/mL) 25.5 (9.0) 24.0 52.3 (39.8) 45.4 1.89 (1.51, 2.37)
Tmax (h) median (min, max) 0.67 (0.25, 1.00) 0.63 (0.50, 1.00)
t1/2 (h) 6.21 (1.95) 6.58 (1.45)
1-hydroxymidazolam
Cmax (ng/mL) 4.76 (2.42) 4.20 5.59 (2.00) 5.24 1.25 (0.98, 1.59)
AUC0-t (ng*h/mL) 9.67 (4.22) 8.87 13.1 (5.92) 12.2 1.37 (1.11, 1.70)
AUC0- (ng*h/mL) 10.5 (4.53) 9.63 13.8 (7.08) 12.6 1.31 (1.31, 1.67)
Tmax (h) median (min, max) 0.75 (0.25, 1.00) 0.75 (0.50, 1.00)
t1/2 (h) 4.54 (1.96) 5.30 (3.16)
  • Note: Data are presented as mean (standard deviation) unless otherwise stated.
  • Abbreviations: AUC0-, AUC from time 0 to infinity; AUC0-t, AUC from time 0 to the last quantifiable concentration; CI, confidence interval; Cmax, maximum observed concentration; LS, least square; max, maximum; min, minimum; PK, pharmacokinetic; t1/2, first-order terminal elimination half-life; Tmax, time to reach Cmax.

Steady-state plasma concentrations of vonoprazan were achieved after 3 days following 20-mg b.i.d. vonoprazan, with mean trough concentration (Ctrough) values of 15.0, 14.3, and 15.7 ng/mL on days 4, 5, and 6, respectively. Ctrough values of vonoprazan were similar whether administered alone (days 3–8) or following midazolam co-administration (days 9–11).

Safety

Overall, six AEs were reported in four participants after receiving vonoprazan alone; three AEs were reported in two participants after vonoprazan and midazolam co-administration (Table S4). All AEs were mild in severity and resolved by the end of the study. Two AEs (somnolence and feeling drunk) in the same participant were reported to be related to midazolam after vonoprazan and midazolam co-administration; no AEs were considered related to vonoprazan. No deaths or serious AEs were reported. No participants discontinued the study due to an AE.

PBPK model

A full PBPK model with first-order absorption described the disposition of vonoprazan with reasonable accuracy when compared with observed clinical data following dosing of vonoprazan in healthy participants (Figure 2a; Table S5).

Details are in the caption following the image
Comparison of simulated versus observed secondary PK parameters for vonoprazan PBPK model development and verification. AUC0–24h, AUC from time 0 to 24 h; AUC, AUC from time 0 to infinity; Cmax, maximum observed concentration; Ctrough, trough concentration; DDI, drug–drug interaction; PBPK, physiologically based pharmacokinetic; PK, pharmacokinetic; tmax, time to reach Cmax; Vono, vonoprazan.

Clearance

Cytochrome P450 correlation analysis in human hepatic microsomes showed vonoprazan metabolization to M-I, M-II, and other unidentified metabolites. The formation of metabolite M-I is most strongly correlated with CYP3A4/5 activity, and to a lesser extent with CYP2C19. Using the reported mean intrinsic clearance (CLint) of vonoprazan in human microsomes (2.73 μL/min/mg at 1 mg/mL protein) to scale in vivo clearance underpredicted the observed CL/F (CL predicted of 11.6 L/h vs. observed of 90–179 L/h). Therefore, retrograde calculations for CLint were made using the observed CL/F, CLR, and fmCYP3A4 values (refer to Appendix S1: Equation S1). The results of the human ADME with a 20-mg vonoprazan dose in healthy volunteers are summarized in Figure 2b. Unchanged drug accounted for 4.4% of 31% total feces radioactivity = 1.4% (Table S6). Based on these data, the fa was assumed to be 0.99. Considering the percentage of radioactivity excreted into feces and urine, the maximum percentage of CYP3A4-related metabolism (total of M-I, M-I-G, and M-III exposure) was 18%. CYP phenotyping studies were performed using BD Supersomes.9 Accounting for inbuilt intersystem extrapolation factors within the Simcyp Simulator and the abundance of each individual CYP in the human liver, CYP3A4 contribution was calculated to be 45.4% of total metabolism, with the other CYP-mediated metabolism assigned to CYP2D6 and CYP2C19 (refer to Appendix S1: Supplemental Methods, “Scaling in vitro clearance to in vivo clearance using the retrograde approach”). Next to CYP-mediated metabolism, sulfation of vonoprazan to the M-IV-sulf metabolite also plays a role (Figure S4). The total fraction of sulfation = 13% (=% M-IV), leaving 59% of metabolism unaccounted for. When applying the in vitro fmCYP3A4 of 45.4% to this 59% and adding it to the 18% CYP3A4-related metabolism, 45% of the total clearance was calculated to be CYP3A4-mediated (Figure 3).

Details are in the caption following the image
Mass balance and overall disposition for vonoprazan and metabolites. The disposition profile following a single dose of [14C]-vonoprazan (20-mg) during the human absorption, distribution, metabolism, and excretion study. Data shown are the geometric mean ±95% confidence intervals. F, fraction bioavailability; fa, fraction absorbed; Vono, vonoprazan.

CLR of vonoprazan represented a minor elimination pathway, with 8% of the dose excreted unchanged into urine and an estimated CLR of 4 L/h (Table 1; Figure 3).

Model development

After manual adjustment of tlag and ka, the simulated profile of vonoprazan was comparable to clinical data. The predicted median Tmax, geometric mean Cmax, and AUC from time 0 to 24 h (AUC0-24) values for vonoprazan were within 1.25-fold of the observed geometric mean values (Figure 2; Table S5). As the dose range of interest was 10–40 mg and the variability in clearance at the lower doses was high, doses below 10 mg were not considered in the PBPK model-building process. Instead, linear clearance was assumed based on the SD observed data in the 10–40-mg dose range from all available studies.

Model verification

The simulated profiles of vonoprazan were comparable to observed clinical data. The predicted median Tmax, geometric mean Cmax, and AUC0–24 values for vonoprazan were within 1.25-fold of the observed geometric mean values (Figure S2; Table S5).

Simulation of plasma concentration–time profiles of vonoprazan in healthy participants after SD (40-mg) oral vonoprazan co-administered with clarithromycin – verification of fmCYP3A4

Comparisons of simulated and observed plasma concentrations of vonoprazan following a vonoprazan 40-mg SD in the absence of clarithromycin and co-administered with clarithromycin (500-mg b.i.d.) on the sixth of 7 days of dosing are shown in Figure 2b. Predicted and observed mean Cmax and AUC values and corresponding geometric mean ratios for vonoprazan in the absence and presence of clarithromycin are shown in Table S5. The simulated profiles of vonoprazan were comparable to clinical data. In addition, the predicted geometric mean Cmax and AUC ratios for vonoprazan in the presence of clarithromycin were within 1.25-fold of observed values. Thus, the vonoprazan fmCYP3A4 of 0.45 was assumed in all further simulations.

Simulation of plasma concentration–time profiles of midazolam in healthy participants after SD (2-mg) oral midazolam co-administered with vonoprazan – verification of CYP3A4 Ki and mechanism-based inhibition parameters

Comparisons of simulated and observed plasma concentrations of midazolam following a midazolam 2-mg SD in the absence of vonoprazan and co-administered with vonoprazan on the eighth of 9 days of dosing (20-mg b.i.d.) are shown in Figure 2b. Predicted and observed mean Cmax and AUC values and corresponding geometric mean ratios for midazolam in the absence and presence of vonoprazan are shown in Table S5. The simulated profiles of midazolam were comparable to the clinical data. In addition, the predicted geometric mean Cmax and AUC ratios for midazolam in the presence of vonoprazan were within 1.25-fold of the observed values. A sensitivity analysis of the Ki and mechanism-based inhibition (MBI) parameters was performed. Based on this sensitivity analysis, the in vitro derived parameters, Ki value of 3 μM, Kapp of 1.22 μM, and Kinact of 0.966 h−1 were assumed in all further simulations.

Model application

Vonoprazan as a victim of induction DDI

The model was used prospectively to predict the outcome of DDI with rifampin or efavirenz following either single or repeat dosing of vonoprazan. The analysis indicates that the predicted changes in plasma exposure of vonoprazan during co-administration with rifampin, a strong inducer of CYP3A4, are borderline strong (AUCR reduced by ≥80%) for SDs, lower for multiple dosing of vonoprazan (10-mg q.d.), and moderate (AUCR reduced by ≥50% and <80%) for higher multiple doses (20-mg q.d. and 20-mg b.i.d.). Simulations with the moderate CYP3A4-inducer, efavirenz, predicted a moderate interaction (AUCR reduced by ≥50 and <80%) for all simulated vonoprazan doses (Table 3).

TABLE 3. Summary of predicted data for vonoprazan in the absence and presence of CYP3A4-inducers following single and multiple oral administration of vonoprazan.
Perpetrator Vonoprazan dosage Cmax (ng/mL) simulated (trial range) AUC0- (ng*h/mL) simulated (trial range) CmaxR AUCR
Control With inducer Control With inducer
Rifampin 600 mg q.d. 10 mg, SD 12.9 (12.0–15.2) 3.61 (3.16–4.43) 95.6 (84.3–114) 18.6 (15.5–24.1) 0.28 0.19
20 mg, SD 26.1 (24.2–30.8) 7.32 (6.40–8.99) 196 (173–234) 38.6 (32.1–50.1) 0.28 0.20
40 mg, SD 53.0 (49.2–62.5) 15.0 (13.1–18.4) 405 (357–484) 82.4 (68.2–107) 0.28 0.20
10 mg, q.d. 14.5 (13.5–17.1) 4.05 (3.54–5.02) 103 (90.8–123) 20.5 (16.9–26.8) 0.28 0.20
20 mg, q.d. 30.5 (28.2–35.9) 8.82 (7.68–10.9) 220 (193–264) 45.8 (37.6–59.9) 0.29 0.21
20 mg, b.i.d. 39.1 (35.8–45.5) 11.2 (9.72–13.9) 242 (210–293) 52.6 (42.8–69.0) 0.28 0.22
Efavirenz 600 mg q.d. 10 mg, SD 13.1 (12.0–15.6) 7.35 (6.18–9.31) 98.2 (89.1–111) 44.9 (39.0–52.3) 0.56 0.46
20 mg, SD 26.6 (24.2–31.6) 14.9 (12.5–18.8) 201 (182–229) 92.2 (80.0–108) 0.56 0.46
40 mg, SD 53.9 (49.1–64.2) 30.3 (25.4–38.3) 417 (377–474) 192 (167–224) 0.56 0.46
10 mg, q.d. 14.8 (13.4–17.6) 8.14 (6.83–10.3) 106 (95.5–121) 48.4 (41.9–56.6) 0.55 0.46
20 mg, q.d. 31.2 (28.0–37.0) 17.2 (14.4–21.8) 227 (204–259) 104 (89.8–122) 0.55 0.46
20 mg, b.i.d. 40.2 (35.9–47.1) 21.6 (18.1–27.1) 251 (224–288) 116 (100–138) 0.54 0.46
  • Abbreviations: AUC0-∞, AUC from time 0 to infinity; AUCR, AUC ratio; b.i.d., twice-daily; Cmax, maximum observed concentration; CmaxR, Cmax ratio; CYP, cytochrome P450; q.d., once-daily; SD, single dose.

Vonoprazan as a victim of induction DDI in the presence of clarithromycin

Simulations of 20-mg b.i.d. vonoprazan in combination with 500-mg b.i.d. clarithromycin showed that clarithromycin weakly inhibits steady-state vonoprazan (Cmax ratio [CmaxR] = 1.36 and AUC from time 0 over the dosing interval [AUCtau] ratio [AUCtauR] = 1.46; Table S7). Adding a CYP3A4-inducer to combined vonoprazan and clarithromycin resulted in projection of moderate induction of vonoprazan by rifampin (CmaxR = 0.58; AUCtauR = 0.48) but this was less than the projected induction of vonoprazan by rifampin when vonoprazan was administered alone (CmaxR = 0.28; AUCtauR = 0.22). As clarithromycin is an inhibitor of CYP3A it is likely that the combined effect of inhibition by clarithromycin and induction by rifampin resulted in a predicted lower net clinical induction for vonoprazan. Similar effects were predicted with efavirenz, a moderate CYP3A4-inducer. The combination of all three drugs showed only weak induction on vonoprazan exposure (CmaxR = 0.86; AUCtauR = 0.81) compared with moderate induction when vonoprazan was administered without clarithromycin (CmaxR = 0.54; AUCtauR = 0.46). As clarithromycin is metabolized by CYP3A, inhibition by vonoprazan or CYP3A4 induction by rifampin or efavirenz may affect clarithromycin exposure. Therefore, the effect of CYP3A4 inducers on clarithromycin PKs was also explored. Simulations of 20-mg b.i.d. vonoprazan in combination with 500-mg b.i.d. clarithromycin showed that clarithromycin exposure was not affected by vonoprazan (CmaxR = 1.03; AUCtauR = 1.05). The addition of CYP3A4-inducers resulted in strong induction by rifampin (CmaxR = 0.27; AUCtauR = 0.14) and moderate induction by efavirenz (CmaxR = 0.46; AUCtauR = 0.29).

DISCUSSION

In this study, a tiered approach was applied to understand the CYP3A victim and perpetrator DDI potential for vonoprazan utilizing both clinical DDI data as well as model-based simulations with the goal to inform labeling. A mechanistic static model was used to project the magnitude of change for substrates of CYP3A.13, 14 This model predicted weak inhibition of CYP3A by vonoprazan when co-administered with oral midazolam. A subsequent clinical interaction study, in which midazolam Cmax increased 1.93-fold and AUC increased 1.89-fold following a vonoprazan 20-mg b.i.d. regimen, confirmed vonoprazan as a weak inhibitor of CYP3A in vivo. In line with several publications, the observed interaction magnitude was well-predicted by the mechanistic static model, particularly when using average concentration as the input value (0.96-fold of observed).22, 26-28 The midazolam clinical data were then used to verify a PBPK model, which had been developed using drug- and system-specific parameters and study-specific data. The PBPK model was subsequently used to predict the outcome of co-administration of vonoprazan with the strong and moderate CYP3A inducers, rifampin and efavirenz, in lieu of conducting a clinical study.

Given the pleiotropic effects of rifampin on CYP3A and other inducible enzymes and transporters including P-glycoprotein (P-gp), there is a tendency for PBPK modeling to underpredict the magnitude of induction for compounds that are metabolized via multiple inducible pathways and those that are dual P-gp and CYP3A substrates.29, 30 This can be mitigated by incorporating an induction parameter for P-gp and the substrate kinetics toward P-gp into the PBPK model.29 In vitro studies confirm that vonoprazan is highly permeable and is not a substrate of P-gp (data on file). Comparisons of predicted versus observed exposure of drugs within twofold of observed data is considered to be “a primary metric for assessment of model fidelity.”31 When DDI data are available to optimize the model, this is often reduced to within 1.25- or 1.5-fold.32, 33 The final PBPK model was able to generate plasma concentration–time profiles and exposure levels of vonoprazan that were within 1.25-fold of observed data in healthy participants after an oral SD of 10, 20, 30, and 40 mg. Plasma concentration–time profiles and exposure levels of vonoprazan after an orally administered 20-, 30-, and 40-mg SD in healthy participants were predicted within 1.25-fold, and the lower doses of 10 and 15 mg within twofold and 1.5-fold, respectively, of the observed data. Failure to predict the plasma concentration-time profiles and exposure levels of vonoprazan within 1.25-fold compared with observed data at 10 and 15 mg may be attributed to the large variability in clearance values observed in low-dose groups.

The predicted geometric mean CmaxR and AUCR of vonoprazan 40-mg SD co-administered with clarithromycin in healthy participants were 1.51- and 1.78-fold, respectively, which were within 1.25-fold of the observed values of 1.35- and 1.58-fold, respectively. These data confirmed the estimated fmCYP3A4 of 45.4% used by the model, which was estimated from human ADME and in vitro metabolism studies. Analysis of observed data from the food-effect cohort showed no statistical effect of food on vonoprazan Cmax and AUC, but a delay in median Tmax from 2 to 4 h was observed.3 The Tmax in the clarithromycin DDI simulations was underpredicted, which could be explained by the fact that participants received vonoprazan in combination with food in this study. This resulted in a delay in Tmax similar to the delay seen in the fed cohort of the food-effect study, which was not accounted for by the model.

The predicted geometric mean CmaxR and AUCR of midazolam in healthy participants following co-administration of vonoprazan with a 2-mg SD of midazolam were 1.66-fold and 1.97-fold, respectively, which is within 1.25-fold of the observed values of 1.93- and 1.89-fold, respectively. These data confirmed the use of MBI of CYP3A4 by vonoprazan, as well as the use of the in vitro Ki for CYP3A4 of 3.0 μM. Sensitivity analysis showed that midazolam inhibition by vonoprazan could be simulated only when MBI of CYP3A4 by vonoprazan was considered, whereas 10-fold lowering of the CYP3A4 Ki for vonoprazan did not significantly improve the DDI simulations.

The model was used prospectively to predict the likely outcome with co-administration of rifampin and efavirenz following either single or repeat dosing of vonoprazan. The predictions demonstrated that impact on the plasma exposure of vonoprazan during co-administration with rifampin was borderline strong induction (AUCR reduced by ≥80%) for the SDs and the lower multiple doses of vonoprazan (10-mg q.d.), and moderate (AUCR reduced by ≥50% and <80%) for the higher multiple doses (20-mg q.d. and 20-mg b.i.d.). Simulations with the moderate CYP3A4 inducer, efavirenz, predicted a moderate interaction (AUCR reduced by ≥50% and <80%) for all simulated vonoprazan doses. These projections are in line with the magnitude of clinical induction observed for rifampin (strong) compared with efavirenz (moderate).27

An International Consortium for Innovation and Quality (IQ) PBPK-modeling induction working group (PBPK-IWG) publication highlights the potential for PBPK modeling of induction-based interactions of CYP3A substrates.34 It reports survey data indicating that there is greater confidence in using PBPK modeling to predict induction by rifampin when strong induction is indicated, highlighting a case study in which PBPK modeling was used in lieu of a rifampin DDI study to inform ivosidenib labeling.35, 36 The IQ PBPK-IWG proposed a strategy similar to the one taken here. In addition, the FDA has published findings from PBPK modeling submissions and noted situations where higher confidence in the model output is achievable.37

Considering vonoprazan increased oral midazolam exposure and was confirmed to be a weak CYP3A inhibitor, labeling indicating that lower doses of sensitive CYP3A4 substrates with a narrow therapeutic index should be used when administered concomitantly with vonoprazan alone or vonoprazan in combination with clarithromycin was implemented. Additional labeling to avoid concomitant administration of vonoprazan with moderate and strong CYP3A inducers was implemented based on the PBPK modeling. These studies highlight the potential for PBPK modeling to prioritize clinical studies, delineate complex interactions, and inform DDI labeling recommendations.

AUTHOR CONTRIBUTIONS

D.J.M. and D.R. wrote the manuscript. D.J.M., D.R., L.Z., I.M., and H.M.J. designed and performed the research. D.J.M., D.R., L.Z., I.M., E.L., N.S., H.M.J., and C.S. analyzed the data.

ACKNOWLEDGMENTS

The authors would like to thank Suresh Balani at Takeda for support and scientific advice. Medical writing and editorial assistance were provided by Abigail Killen-Devine and Kyle Lambe (Synergy, UK) and were supported by Phathom Pharmaceuticals.

    FUNDING INFORMATION

    Phathom Pharmaceuticals, Inc. funded the research presented in this publication.

    CONFLICT OF INTEREST STATEMENT

    D.J.M., E.L., and N.S. are employed by Phathom Pharmaceuticals. E.L. and N.S. also report stock/options in Phathom Pharmaceuticals. D.R. and L.Z. declared no competing interests for this work. I.M. and H.M.J. are employees of Certara, paid by Phathom Pharmaceuticals to consult and perform this work. C.S. discloses compensation from Phathom Pharmaceuticals as a consultant, advisory board member and speaker.

    DATA AVAILABILITY STATEMENT

    The data collected for this study will not be made available to others.