Physiologically-Based Pharmacokinetic Modeling of the Drug-Drug Interaction of the UGT Substrate Ertugliflozin Following Co-Administration with the UGT Inhibitor Mefenamic Acid.

The sodium-glucose cotransporter 2 inhibitor ertugliflozin is metabolized by the uridine 5'-diphospho-glucuronosyltransferase (UGT) isozymes UGT1A9 and UGT2B4/2B7. This analysis evaluated the drug-drug interaction (DDI) following co-administration of ertugliflozin with the UGT inhibitor mefenamic acid (MFA) using physiologically-based pharmacokinetic (PBPK) modeling. The ertugliflozin modeling assumptions and parameters were verified using clinical data from single-dose and multiple-dose studies of ertugliflozin in healthy volunteers, and the PBPK fraction metabolized assignments were consistent with human absorption, distribution, metabolism, and excretion results. The model for MFA was developed using clinical data, and in vivo UGT inhibitory constant values were estimated using the results from a clinical DDI study with MFA and dapagliflozin, a UGT1A9 and UGT2B4/2B7 substrate in the same chemical class as ertugliflozin. Using the verified compound files, PBPK modeling predicted an ertugliflozin ratio of area under the plasma concentration-time curves (AUCR ) of 1.51 when co-administered with MFA. ClinicalTrials.gov identifier: NCT00989079.

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Physiologically‐based pharmacokinetic (PBPK) prediction of enzymatic drug–drug interactions (DDIs), especially in relation to cytochrome P450 (CYP) inhibition, has been increasingly used during drug development and in support of regulatory submissions. However, there are few examples of the use of PBPK modeling and simulation for evaluating DDIs due to uridine 5’‐diphospho‐glucuronosyltransferase (UGT) inhibition.

WHAT QUESTION DID THIS STUDY ADDRESS?

What is the predicted DDI between ertugliflozin, UGT1A9 and UGT2B4/2B7 substrate, and the UGT inhibitor mefenamic acid (MFA) using PBPK modeling?

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

In this study, we utilized a novel approach by developing and verifying PBPK models for ertugliflozin and dapagliflozin, which have similar metabolic pathways, and leveraged the observed data from the MFA‐dapagliflozin DDI study in order to predict the magnitude of the DDI between MFA and ertugliflozin.

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

This approach would enable wider application of PBPK modeling of DDIs resulting from inhibition of non‐CYP enzymatic pathways.

Ertugliflozin, an oral sodium‐glucose cotransporter 2 (SGLT2) inhibitor, is approved in the United States and European Union for the treatment of adults with type 2 diabetes mellitus at daily doses of 5 or 15 mg, either as monotherapy or in combination with metformin or sitagliptin. In phase III, randomized, double‐blind trials, ertugliflozin (alone or with metformin or sitagliptin) was associated with significant reductions in fasting plasma glucose, glycated hemoglobin, body weight, and blood pressure, and was well tolerated. , , , , , , , The pharmacokinetics (PKs) of ertugliflozin are similar in healthy subjects and patients with type 2 diabetes mellitus. After oral administration, ertugliflozin is rapidly absorbed, with time to maximum concentration (Tmax) of 1 hour, and exposure increases proportionally with dose over the range of 0.5–300 mg. , The absolute bioavailability of ertugliflozin is ~ 100% and co‐administration with food does not have a clinically meaningful impact on ertugliflozin PKs.

The absorption, distribution, metabolism, and excretion (ADME) of ertugliflozin was evaluated after oral administration of 25‐mg 14[C]‐ertugliflozin. The primary clearance mechanism is metabolism, mainly by glucuronidation (~ 86%) with oxidation playing a minor role (~ 12%) ; renal excretion of unchanged ertugliflozin is low (~ 2%). The main enzyme responsible for the major glucuronidation pathway, as determined by in vitro reaction phenotyping studies, , , is the uridine 5’‐diphospho‐glucuronosyltransferase (UGT) [Correction added on 21 January 2021, after first online publication: the abbreviation for (UGT) has been corrected to uridine 5’‐diphospho‐glucuronosyltransferase throughout the article.] isozyme UGT1A9 (81%), with a lesser contribution from UGT2B4/2B7 (19%), leading to two pharmacologically inactive glucuronide metabolites. Metabolism of ertugliflozin via the minor oxidative pathway is mainly by the cytochrome P450 (CYP) isozyme CYP3A4 (85–100%), with minor contributions from CYP3A5 (0–10%) and CYP2C8 (0–4%). , As oxidation is a minor pathway, clinically significant drug–drug interactions (DDIs) are not expected following co‐administration of ertugliflozin with CYP inhibitors. However, given that metabolism by UGT1A9 is the major clearance pathway of ertugliflozin, the potential for DDIs after co‐administration with a UGT1A9 inhibitor needs to be evaluated.

Historically, drug interactions with UGT inhibitors have been evaluated in clinical studies and typically result in a less than two‐fold increase in substrate drug exposures. Two clinical studies have evaluated DDIs between UGT inhibitors and the UGT substrates canagliflozin and dapagliflozin, which are SGLT2 inhibitors in the same chemical class as ertugliflozin. Canagliflozin is mainly metabolized by glucuronidation via UGT1A9 and UGT2B4 with a minor contribution from oxidation. Co‐administration of canagliflozin with probenecid (UGT1A9 inhibitory constant (Ki) = 2,452 µM ; probenecid 500‐mg twice‐daily dose estimated unbound maximum steady‐state concentration (Cmax,ss,u) = 29 µM; inhibitor concentration/Ki ≈ 0.01) led to a 21% increase in canagliflozin exposure. Dapagliflozin has physicochemical and ADME properties similar to ertugliflozin, with the primary route of elimination of dapagliflozin being metabolism by glucuronidation (88%), predominantly by UGT1A9 with minor contributions from UGT2B4/2B7, whereas oxidative metabolism (10%) and renal elimination of unchanged drug (2%) are minor. Co‐administration of dapagliflozin with mefenamic acid (MFA; UGT1A9 and UGT2B4/2B7 inhibitor) led to a 51% increase in dapagliflozin exposure (ratio of the area under the plasma concentration–time curves (AUCR) = 1.51).

Recent publications from the US Food and Drug Administration (FDA), the European Medicines Agency (EMA), and a position paper from the International Consortium for Innovation and Quality in Pharmaceutical Development consisting of over 35 pharmaceutical companies highlight the utility of physiologically‐based pharmacokinetic (PBPK) modeling approaches and affirm the acceptability of PBPK simulations in regulatory reviews. , Consequently, the number of FDA New Drug Applications and EMA Marketing Application Authorisations containing PBPK analyses has increased by > 20‐fold since 2004, with > 50% of the models assessing enzyme‐based DDIs particularly CYP‐mediated DDIs. , However, there are few published examples of PBPK modeling approaches to simulate UGT‐mediated DDIs. A mirabegron PBPK model incorporated UGT2B7‐mediated metabolism (~ 30%) and was applied to simulate DDIs with fluconazole (CYP3A4 and UGT2B7 inhibitor) and probenecid. PBPK modeling has also been used to simulate DDIs with the UGT substrate and inhibitor valproic acid, and with a regimen of three antiviral drugs (ombitasvir, dasabuvir, and paritaprevir) that are known UGT1A1 inhibitors. An in vitro–in vivo scaling approach for UGT metabolism data combined with PBPK modeling was used to successfully simulate a DDI between the antiretroviral drug zidovudine and fluconazole (UGT2B7 inhibitor).

In the current study, the DDI between ertugliflozin and the UGT1A9 inhibitor MFA was evaluated using PBPK modeling by leveraging available clinical DDI results between dapagliflozin and MFA. The PBPK models for ertugliflozin and dapagliflozin were developed and verified based on results from clinical and in vitro studies using a similar modeling approach. The MFA model was developed based on available clinical data for MFA, and verified by simulating the DDI following co‐administration with dapagliflozin. The impact of MFA on the PK of ertugliflozin was then simulated at clinically relevant doses using the verified PBPK models within the commercial PBPK platform Simcyp.

METHODS

In vitro studies for dapagliflozin and MFA

In vitro dapagliflozin UGT reaction phenotyping studies, in vitro MFA plasma protein binding, and UGT inhibition studies were conducted to provide additional input parameters for the dapagliflozin and MFA PBPK models ( ).

PBPK modeling and simulation strategy

The modeling strategy involved the development of PBPK models for ertugliflozin, dapagliflozin, and MFA using available clinical PK (i.v./oral/ADME) and in vitro data. The models were then verified using observed results from the oral PK studies, and the in vivo MFA UGT K i values were estimated using results from the clinical DDI between MFA and dapagliflozin. Last, the verified PBPK models were applied to simulate the DDI following co‐administration of ertugliflozin and MFA.

PBPK modeling and simulation was conducted using Simcyp version 15, release 1 (Certara USA, Princeton, NJ). The simulations were performed in a virtual population library of healthy volunteers within Simcyp (Sim‐Healthy Volunteers) using the same age range (18–55 years) and gender distribution (0–50% women) as the clinical studies included in the model development and verification. A summary of the trial designs for all simulations is listed in Table .

Development of PBPK models for ertugliflozin and dapagliflozin

The ADME properties of ertugliflozin and dapagliflozin are similar (Figure ) and both PBPK models were developed using a middle‐out approach. The measured PK properties (i.v. clearance (CLiv) and volume of distribution at steady state (V ss)) after i.v. administration were used as input parameters during PBPK model development, and ADME results enabled the assignment of fraction metabolized (f m) values due to glucuronidation or oxidation. The specific enzymatic clearance rates for ertugliflozin and dapagliflozin were assigned using reaction phenotyping results and the approach defined in Figure . This approach was needed as the Retrograde Translation Tool in Simcyp version 15 did not incorporate metabolism by UGT enzymes.

Figure 1

Flow chart showing ertugliflozin and dapagliflozin metabolism and disposition. See Table and Table for additional information on f m assignments for ertugliflozin and dapagliflozin, respectively. Asterisk indicates values reported as % of dose; the identified parent, oxidative metabolites, and glucuronidation metabolites observed in feces and urine were scaled to 100% of the dose. ABA, absolute bioavailability; CLiv,p, intravenous plasma clearance; CLr, renal clearance; CYP, cytochrome P450; F, bioavailability; Fa, fraction absorbed; Fe, fraction excreted; fm, fraction metabolized; Gluc, glucuronidation; Oxid, oxidation; UGT, uridine 5’‐diphosphate‐glucuronosyltransferase.

Figure 2

Flowchart showing the ertugliflozin and dapagliflozin elimination model development strategy. Asterisk indicates ertugliflozin and dapagliflozin metabolism by UGT2B4/2B7 was assigned to UGT2B7 in the PBPK model. ADME, absorption, distribution, metabolism, and excretion; B/P, blood/plasma ratio; CLint, intrinsic clearance; CLint,u, unbound intrinsic clearance; CLiv,p, intravenous plasma clearance; CLiv,b,metab, intravenous metabolic blood clearance; CLiv,b, intravenous blood clearance; CLint,scaled,u, unbound intrinsic clearance scaled; fm, fraction metabolized; PBPK, physiologically‐based pharmacokinetic; UGT, uridine 5’‐diphospho‐glucuronosyltransferase.

Ertugliflozin model development

The ertugliflozin PBPK model was developed using clinical PK and in vitro data (Table ). A minimal PBPK distribution model with the observed V ss was used. Due to the biphasic i.v. PK profile, the volume of an additional single adjusting compartment (V sac) [Correction added on 21 January 2021, after first online publication: the abbreviation for volume of an additional single adjusting compartment (V sac) has been included.] and the first‐order rate constants K in and K out were estimated in Simcyp using the parameter estimation module. For simulations after an oral dose, the first‐order absorption model was used, with the fraction absorbed (F a) value set to 1 based on results from the absolute bioavailability study. The coefficient of variation for F a was set to 0% to capture the observed area under the plasma concentration–time curve (AUC). The human permeability coefficient was estimated from in vitro Caco‐2 results, and absorption rate constant (K a) values were estimated using sensitivity analysis to capture the maximum plasma concentration (C max). An absorption lag time of 0.5 hours was estimated to match the observed Tmax values.

Table 1

Simcyp input parameters for ertugliflozin, dapagliflozin, and MFA

Parameters	Ertugliflozin	Source	Dapagliflozin	Source	MFA	Source
Physicochemical properties
Structure		Ref. 48		Ref. 23		Ref. 49
Molecular weight, g/mol	436	Ref. 42	408	DrugBank 50	241	DrugBank 49
LogP	2.5	Ref. 42	2.52	DrugBank 50	5.4	DrugBank 49
Compound type	Neutral	–	Neutral	–	Acid	–
pKa	–	–	–	–	3.89	DrugBank 49
Fu,plasma	0.064	Ref. 45	0.09	Ref. 46	0.002	Measured (see Supplementary Information)
B/P ratio	0.66	Ref. 45	0.88	Ref. 46	1	Assumed
Absorption
Fa	1.0	Ref. 15	0.9	Ref. 38	1.0	Assumed
Ka, h−1	1.2	Estimated (Simcyp) a	1.2	Estimated (Simcyp) a	0.85	Estimated (Simcyp) a
Tlag, hours	0.5	Estimated (Simcyp) a	–	–	–	–
Caco‐2, 10–6 cm/s	4.1	Ref. 17	15.9	Ref. 46	–	–
Qgut, L/h	5.62	Predicted (Simcyp)	8.5	Predicted (Simcyp)	–	–
Distribution
Vss, L/kg	1.23	Ref. 15	1.19	Ref. 38	0.4	Estimated (Simcyp) a
Vsac, L/kg	1.12	Estimated (Simcyp) a	0.9	Estimated (Simcyp) a	–	–
Kin, h–1	3.34	Estimated (Simcyp) a	–	–	–	–
Kout, h–1	0.56	Estimated (Simcyp) a	–	–	–	–
Q, L/h	–	–	10	Estimated (Simcyp) a	–	–
Elimination
CLiv, L/h	11.2	Ref. 15	12.4	Ref. 38	–	–
CLr, L/h	0.1	Ref. 42	0.2	Ref. 23	–	–
CLint, CYP, µL/min/pmol	CYP3A4 = 0.041 CYP3A5 = 0.006 CYP2C8 = 0.011	Retrospectively calculated based on: CLiv (11.2 L/h) CLr (0.1 L/h); CYP fm (0.12) CYP3A4 (86%) CYP3A5 (10%) CYP2C8 (4%)	–	–	–	–
Additional HLM CLint, µL/min/mg	–	–	4	Retrospectively calculated based on: CLiv (12.4 L/h; Ref. 38) CLr (0.2 L/h; Ref. 23) CYP fm (0.10; Ref. 23)	–	–
CLint, UGT, µL/min/mg	UGT1A9 = 35 UGT2B7 = 8	Calculated based on: UGT1A9 (81%) UGT2B7 (19%)	UGT1A9 = 30 UGT2B7 = 3	Calculated based on: UGT1A9 (90%) UGT2B7 (10%)	–	–
CLpo, L/h	–	–	–	–	17	Estimated (Simcyp) a
Ki vs. UGT1A9, µM	–	–	–	–	0.038	Fitted based on clinical DDIb
Ki vs. UGT2B7, µM	–	–	–	–	0.051	Fitted based on clinical DDIb

B/P ratio, blood/plasma ratio; Caco‐2, permeability coefficient; CLint, intrinsic clearance; CLiv, clearance observed for intravenous administration; CLpo, clearance observed for oral administration; CLr, renal clearance; CYP, cytochrome P450; DDI, drug‐drug interaction; Fa, fraction absorbed; fm, fraction metabolized; Fu,plasma, fraction unbound in plasma; HLM, human liver microsomes; Ka, absorption rate constant; Ki, reversible inhibitory constant; Kin, first‐order rate constant in; Kout, first‐order rate constant out; LogP, partition coefficient; MFA, mefenamic acid; pKa, acid dissociation constant; Q, intercompartmental clearance; Qgut, hybrid term including both villous blood flow and permeability through the enterocyte membrane; Ref., reference citation; Tlag, lag time; UGT, uridine 5’‐diphospho‐glucuronosyltransferase; Vsac, volume of single adjusting compartment; Vss, volume of distribution at steady state; –, data not available or not applicable.

Simcyp parameter estimate.

Ki values were fitted to recover the observed clinical DDI between MFA and dapagliflozin. Relative in vitro potency between UGT1A9 Ki = 0.11 µM and UGT2B7 Ki = 0.15 µM was maintained.

The in vivo metabolism of ertugliflozin was characterized in the human ADME study following administration of a single oral dose of 25‐mg 14[C]‐ertugliflozin. In this study, the measured renal clearance (CLr) was low, with ~ 2% of the dose excreted unchanged (F e = 0.02), and ~ 12% of ertugliflozin was converted to oxidative metabolites and assigned to CYP metabolism (f m,CYP = 0.12). The remaining metabolism was assigned to glucuronidation (f m,UGT = 0.86; Table ).

The assignment of enzymatic clearance rates is defined in Figure . The Simcyp Retrograde Translation Tool was used to estimate intrinsic clearance (CLint) values for ertugliflozin based on measured CLiv and CLr values. The f m,CYP (0.12) from the ADME study and available ertugliflozin in vitro CYP reaction phenotyping results (CYP3A4 = 86%; CYP3A5 = 10%; and CYP2C8 = 4%) assigned f m values for specific CYP enzymes (f m,CYP × % contribution): f m,CYP3A4 = 0.103; f m,CYP3A5 = 0.012; and f m,CYP2C8 = 0.005. These f m values were used to calculate the ertugliflozin CLint,CYP values for each enzyme. The remaining CLint was assigned to glucuronidation (UGT CLint,u) and scaled to give a systemic enzymatic clearance (UGT CLint,scaled,u). In vitro studies in human liver and kidney microsomes showed that 10% of the systemic ertugliflozin UGT clearance occurs in the kidneys, whereas 90% occurs in the liver. Thus, 10% of the UGT CLint,scaled,u was subtracted and the remaining CLint,scaled,u was assigned to liver UGTs. In vitro ertugliflozin UGT reaction phenotyping studies , , showed that the relative contribution of UGT1A9 was 81%, and the contribution of the UGT2B4/2B7 isozymes was 19%. For modeling purposes, the minor UGT contribution was assigned to UGT2B7. Therefore, an ertugliflozin UGT1A9 CLint,u of 35 µL/min/mg and a UGT2B7 CLint,u of 8 µL/min/mg were calculated and used as input parameters in the ertugliflozin compound file.

Dapagliflozin model development

The dapagliflozin PBPK model was developed using clinical and in vitro results, similar to the approach used for the ertugliflozin model (Table ). A minimal PBPK distribution model was used. Due to the biphasic i.v. PK profile, parameters for V sac and intercompartmental clearance were determined by fitting the i.v. PK profile using the parameter estimation module in Simcyp. The first‐order absorption model was used for fitting the oral data, and the K a was estimated using a sensitivity analysis in Simcyp. The F a was estimated using the measured CLiv (12.4 L/hour) and bioavailability (78%) using the following relationship: F = F a × F g × F h. The dapagliflozin F a was 0.9, where the fraction of drug remaining after first‐pass through the intestinal wall (F g) was assumed to be 1 due to low oxidative metabolism (F a × F g = 0.91). In the human ADME study, 10% of dapagliflozin was converted to oxidative metabolites (f m,CYP = 0.10) and the measured CLr of dapagliflozin was low, with 2% of the dose excreted unchanged (F e = 0.02). The remainder of dapagliflozin metabolism was assigned to glucuronidation (f m,UGT = 0.88), as defined in Figure and Table ; assignment of dapagliflozin enzymatic clearance rates is defined in Figure . Using the Simcyp Retrograde Translation Tool and assigning 10% to CYP metabolism, the dapagliflozin oxidative CLint was estimated (CLint = 4 µL/min/mg) and listed as human liver microsomal clearance. The remaining CLint was assigned to glucuronidation (UGT CLint,u) and scaled to give a systemic enzymatic clearance (UGT CLint,scaled,u). Similar to ertugliflozin, 10% of the systemic dapagliflozin UGT clearance occurs in the kidneys, whereas 90% occurs in the liver. Thus, 10% of the UGT CLint,scaled,u was subtracted and the remaining CLint,scaled,u was assigned to liver UGTs. In vitro UGT reaction phenotyping studies indicated that 90% of the dapagliflozin UGT metabolism was due to UGT1A9, and the remaining 10% of UGT metabolism was assigned to UGT2B7 ( ; ). Thus, a UGT1A9 CLint,scaled,u of 30 µL/min/mg and a UGT2B7 CLint,scaled,u of 3 µL/min/mg were the input parameters for the dapagliflozin compound file.

Development of a PBPK model for MFA

The PBPK model for MFA was developed using a top‐down approach using clinical data ; the input parameters are listed in Table . A minimal PBPK distribution model with first‐order absorption was used and V ss, K a, and clearance values after oral administration were determined by fitting the clinical data to match the observed PK profile following a 500‐mg dose of MFA. In vitro K i values for MFA inhibition of UGT1A9 (0.11 µM) and UGT2B7 (0.15 µM) were the initial input parameters ( ; ) used to simulate the DDI following co‐administration of MFA (500‐mg loading dose then 250 mg every 6 hours for 4 days) with dapagliflozin (10‐mg single dose on day 2). The DDI was underpredicted, thus the MFA in vivo UGT K i values were optimized by fitting the clinical DDI data, while keeping the relative potencies of UGT1A9 and UGT2B7 inhibition. The estimated in vivo MFA UGT1A9 and UGT2B7 K i values were adjusted ~ 2.9‐fold to 0.038 µM and 0.051 µM, respectively, and were used as input parameters for verification of the MFA compound file.

PBPK model verification

Ertugliflozin and dapagliflozin model verification

The ertugliflozin model was verified by comparing simulated plasma concentration–time profiles and PK results from single‐dose (0.5–300 mg) and multiple‐dose (5 mg and 15 mg) studies, where steady‐state is achieved after 6 days of dosing, to observed clinical data. , , , The dapagliflozin model was verified by comparing simulated plasma concentration–time profiles and PK results from multiple‐dose studies (10 mg and 50 mg) to observed clinical data. The ertugliflozin and dapagliflozin Simcyp‐estimated f m and F e values were compared with observed results from the clinical ADME study and in vitro reaction phenotyping studies.

MFA model verification and in vivo UGT K i values

The MFA model was verified by comparison of the simulated plasma concentration–time profile and PK results following a single 500‐mg dose. The in vivo UGT K i values were estimated by simulating DDI following co‐administration of MFA (500‐mg loading dose then 250 mg every 6 hours for 4 days) with dapagliflozin (10‐mg single dose on day 2).

Simulation of co‐administration of ertugliflozin and MFA

The co‐administration of MFA (500‐mg loading dose then 250 mg every 6 hours for 4 days) and ertugliflozin (15‐mg single dose on day 2) was simulated to assess the impact of a UGT inhibitor on ertugliflozin PK.

RESULTS

From the in vitro studies (see ), UGT1A9 (90%) was the primary enzyme involved in dapagliflozin glucuronidation. MFA inhibited UGT1A9 and UGT2B7 with unbound half‐maximal inhibitory concentration/K i values of 0.22/0.11 and 0.30/0.15 µM, respectively, but did not inhibit UGT2B4 in vitro. The MFA plasma unbound fraction was 0.0022.

Ertugliflozin and dapagliflozin PBPK model verification

The predicted and observed PK parameters of ertugliflozin and dapagliflozin following single (ertugliflozin 0.5–300‐mg; dapagliflozin 10‐mg) and multiple (ertugliflozin 5‐mg and 15‐mg; dapagliflozin 10‐mg and 50‐mg) oral doses , , , , are in Table . The predicted plasma vs. time profiles of dapagliflozin and ertugliflozin are shown in Figure , respectively, for the single 10‐mg dose, and in Figures –S3 for the other dose regimens. The predicted plasma profiles after oral dosing are comparable to the observed clinical data for ertugliflozin and dapagliflozin. After single‐dose and multiple‐dose administration of ertugliflozin or dapagliflozin, the Cmax predicted/observed ratios were within 80–125% of observed values. The ertugliflozin AUC predicted/observed ratios were within 77–84% of observed after a single dose and within 80–125% after multiple doses across the dose range of 0.5–300 mg. The dapagliflozin AUC predicted/observed ratios were within 80–125% of observed values at all doses simulated.

Table 2

Predicted and observed geometric mean (%CV) PK parameters of ertugliflozin and dapagliflozin after single and multiple oral doses, and MFA after a single oral dose

Dose	Predicted Cmax, ng/mL	Predicted AUCinf, ng·h/mL	Observed Cmax, ng/mL	Observed AUCinf, ng·h/mL	Cmax predicted/ observed ratio	AUC predicted/ observed ratio
Ertugliflozin: single dose
0.1 mga, b	8.46 (63)	9.19 (35)	8.51 (32)	8.48 (15)	0.98	1.08
15 mga	249 (69)	1140 (50)	256 (14)	1400 (13)	0.97	0.81
0.5 mg	8.38 (64)	37.4 (44)	7.23 (11)	45.7 (10)	1.16	0.82
2.5 mg	41.9 (64)	187 (44)	42.8 (21)	231 (22)	0.98	0.81
10 mg	168 (64)	749 (44)	182 (22)	909 (15)	0.92	0.82
30 mg	503 (64)	2250 (44)	545 (24)	2810 (18)	0.92	0.80
100 mg	1680 (64)	7490 (44)	1620 (16)	9610 (16)	1.04	0.78
300 mg	4890 (66)	22000 (38)	4330 (20)	26400 (16)	1.13	0.83
Ertugliflozin: multiple dosec
5 mg	99.3 (61)	410 (47)d	81.3 (29)	398 (18)d	1.22	1.03
15 mg	279 (61)	1160 (47)d	268 (20)	1190 (22)d	1.04	0.97
Dapagliflozin: single dose
10 mga	124	580	143	628	0.87	0.92
Dapagliflozin: multiple dose
10 mg	123	537d	119	506d	1.03	1.06
50 mg	614	2690d	728	2540d	0.84	1.06
MFA: single dose
500 mge	6370	30400	6900	34200	0.92	0.89

%CV, geometric coefficient of variation; AUC, area under the plasma concentration–time curve; AUCinf, AUC from time 0 to infinity; Cmax, maximum plasma concentration; Ka, absorption rate constant; MFA, mefenamic acid; PK, pharmacokinetic; Vss, volume of distribution at steady‐state.

Model development.

Administered intravenously.

Exposures at steady‐state, following 6 days of single‐dose administration.

AUC from time 0 to 24 hours postdose.

PK profile was used to estimate Vss and Ka in Simcyp.

Figure 3

Clinically observed and PBPK model‐predicted plasma concentrations of ertugliflozin and dapagliflozin in healthy subjects following single‐dose oral administration with or without multiple‐dose administration of MFA. (a) Dapagliflozin 10 mg, (b) ertugliflozin 10 mg, (c) dapagliflozin 10 mg (dosed on day 2) following MFA (500‐mg loading dose and 250 mg every 6 hours for 4 days), and (d) ertugliflozin 15 mg (dosed on day 2) following MFA (500‐mg loading dose and 250 mg every 6 hours for 4 days). The observed and predicted plasma concentrations were expressed as mean (green or purple circles) and mean (green or purple lines), respectively, with 5th and 95th percentiles shown (gray lines), in the control treatment (green) and following co‐administration with MFA (purple). Where available, SD around the observed means are also shown (black whiskers). CSys, systemic concentration; DAPA, dapagliflozin; ERTU, ertugliflozin; MFA, mefenamic acid; PBPK, physiologically based pharmacokinetic; PO, prescribed orally.

The Simcyp‐estimated ertugliflozin and dapagliflozin f m and F e values based on the input parameters were similar to the f m and F e values derived from ADME and disposition studies. For ertugliflozin, the estimated values were f m,UGT1A9 = 0.71, f m,UGT2B7 = 0.17, f m,CYP3A4 = 0.10, f m,CYP3A5 = 0.002, f m,CYP2C8 = 0.01, and urine F e = 0.01; observed values were f m,UGT1A9 = 0.70, f m,UGT2B7 = 0.16, f m,CYP3A4 = 0.103, f m,CYP3A5 = 0.012, f m,CYP2C8 = 0.005, and urine F e = 0.02 (Figure ). For dapagliflozin, estimated values were f m,UGT1A9 = 0.81, f m,UGT2B7 = 0.076, f m,CYP = 0.094, and urine F e = 0.018; observed values were f m,UGT1A9 = 0.79, f m,UGT2B7 = 0.09, f m,CYP = 0.10, and urine F e = 0.02 (Figure ). For ertugliflozin, the Simcyp‐estimated kidney UGT clearance was 14%, which was similar to the initial estimation of 10%. These simulation results provided confirmation of the absorption, distribution, f m,UGT, f m,CYP, and F e assignments of the ertugliflozin and dapagliflozin PBPK models.

MFA PBPK model verification and in vivo UGT K i values

The predicted and observed PK parameters of MFA following a 500‐mg single oral dose are listed in Table , and the plasma PK profile is shown in Figure . The Cmax and AUC predicted/observed ratios were within 80–125% of observed values. The simulation results provided verification of the observed MFA PK profile.

The MFA PBPK model was evaluated by simulation of the DDI with dapagliflozin using the verified dapagliflozin PBPK model, and using the optimized K i values for UGT1A9 (0.038 µM) and UGT2B7 (0.051 µM) obtained by fitting the in vivo data. The predicted and observed PK parameters are in Table , and dapagliflozin plasma profiles in the absence or presence of MFA are shown in Figure . Using the optimized MFA PBPK model, the predicted PK impact of MFA on dapagliflozin (AUCR = 1.53; C maxR = 1.18) recapitulated the observed values (AUCR = 1.51; C maxR = 1.13).

Table 3

Predicted and observed geometric mean ratios (CI ) for AUC (AUC

	AUCR	CmaxR
Dapagliflozin
Predicted	1.53 (1.50–1.55)	1.18 (1.17–1.19)
Observedb	1.51 (1.44–1.58)	1.13 (1.03–1.24)
Predicted/observed ratio	1.0	1.0
Ertugliflozin
Predicted	1.51 (1.48–1.54)	1.19 (1.17–1.20)

AUCinf, area under the plasma concentration–time curve from time 0 to infinity; AUCR, ratio of AUCinf of substrate drug with co‐administration of the interacting drug to AUCinf of substrate drug alone; CI, confidence interval; Cmax, maximum plasma concentration; CmaxR, ratio of Cmax of substrate drug with co‐administration of the interacting drug to Cmax of the substrate alone; DDI, drug–drug interaction; Ki, inhibitory constant; MFA, mefenamic acid.

Predicted ratios show a 95% CI; observed ratios show a 90% CI.

Results from a clinical DDI study between dapagliflozin and MFA were used to fit the in vivo MFA Ki values.

Simulation of ertugliflozin PK following co‐administration with MFA

Using the verified ertugliflozin and MFA PBPK models, the DDI following co‐administration of ertugliflozin and MFA was simulated. Results are summarized in Table , and the simulated PK profiles in the absence or presence of MFA are shown in Figure . The predicted ertugliflozin AUCR value was 1.51 and the C maxR value was 1.19 in the presence of the UGT inhibitor MFA.

A sensitivity analysis was conducted to evaluate the increase in UGT f m (from 0.86 to 0.93) and decrease in CYP f m (from 0.12 to 0.06) by assuming that the secondary metabolites undergo glucuronidation as the initial step in metabolism, while maintaining the in vitro reaction phenotyping results. The simulated co‐administration with MFA predicted an ertugliflozin AUCR = 1.55 and C maxR = 1.20, which was consistent with results using the verified ertugliflozin model.

DISCUSSION

PBPK modeling and simulation have shown utility in drug development as they support the complexity required to evaluate mechanistic questions that require an in‐depth understanding of human physiology. This analysis uses PBPK modeling to simulate the DDI following co‐administration of ertugliflozin with the UGT inhibitor MFA. The key components in predicting UGT DDIs are the quantitative assessment of the UGT enzymes responsible for metabolism and the validation of in vivo UGT K i values for inhibitors.

There are several areas of uncertainty when assigning the contribution of UGT metabolism, including hydrolysis of glucuronide metabolites in the gastrointestinal tract after biliary excretion, and occurrence of UGT metabolism in hepatic and extrahepatic tissues due to the expression of UGTs in various organs. In the case of ertugliflozin, ~ 46% of the dose was recovered in excreta as primary glucuronides following an oral dose in the human ADME study (Table ), suggesting that the f m,UGT was at least 0.46. Additionally, ~ 40% of unchanged drug was recovered in the feces after an oral dose (Figure ; Table ). The unchanged drug in feces may be due to unabsorbed drug, biliary or gastrointestinal secretion of ertugliflozin, or hydrolysis of the primary glucuronides back to the parent compound. However, the absolute bioavailability study indicated complete absorption of ertugliflozin, and in vivo studies in animals and in vitro studies in human hepatocytes were indicative of minimal biliary excretion of ertugliflozin. , , Although gastrointestinal secretion of drugs cannot be ruled out, a large proportion of the unchanged ertugliflozin in feces is likely due to hydrolysis of primary glucuronides, which would indicate a higher f m,UGT value. As a worst‐case scenario for the estimation of the DDI due to UGT inhibition, all of the unchanged drug in feces was assumed to be formed due to hydrolysis of primary glucuronides. Therefore, an f m,UGT value of 0.86 was used for the ertugliflozin PBPK model.

UGTs are expressed in multiple organs—mainly the liver, kidneys, and gastrointestinal tract—and metabolism is dependent on the specific UGT isoform expressed in each tissue. Therefore, it is important to determine the UGT isoform involved in order to understand the relative contributions of hepatic vs. extrahepatic metabolism. In vitro reaction phenotyping studies showed that ertugliflozin is metabolized mainly by UGT1A9 (81%). , The minor UGT enzymes were assigned based on studies with the UGT inhibitor 16β‐phenyllongifolol, which was recently shown to be a nonselective inhibitor of both UGT2B4 and UGT2B7. As selective inhibitors of UGT2B4 and UGT2B7 are not available, the minor UGT enzymes are reported as UGT2B4/2B7 (19%). , , Because MFA inhibits UGT1A9 and UGT2B7, and not UGT2B4, the minor UGT responsible for metabolism of ertugliflozin was assigned to UGT2B7 in the PBPK model. This assignment provides a conservative assessment of the DDI following co‐administration of ertugliflozin with the UGT1A9/2B7 inhibitor MFA. These UGT enzymes are found in the liver and the kidneys; therefore, the compound is subject to metabolism in both organs. Differential metabolism in these two organs was accounted for in the PBPK models for ertugliflozin. UGT metabolism in the kidneys can be estimated by Simcyp, which accounts for differences in tissue blood flows and enzyme expression between the two organs, using liver UGT1A9 and UGT2B7 CLint values. For ertugliflozin, the liver and kidney UGT CLint values were estimated and incorporated into the PBPK model. The resulting f m assignments from the Simcyp output were similar to the f m assignments based on the ADME, bioavailability, and in vitro reaction phenotyping data. , Together, these results show that the liver is the main organ of clearance for ertugliflozin. The same approach was used for the assignment of f m and CLint values for dapagliflozin and the development of the dapagliflozin PBPK model was similar.

After derivation of the various input parameters, including assignment of enzymatic pathways and f m and CLint values for the ertugliflozin and dapagliflozin PBPK models, both models were verified through comparison of simulated results to available clinical data. The close concordance between the simulated outcome and observed data serves to confirm that the models adequately captured the metabolism and disposition characteristics of each compound.

The MFA compound file was developed using available clinical PK data, as a PBPK model was not available, and optimized with the results from a clinical DDI study with dapagliflozin. Given the similarities in ADME between ertugliflozin and dapagliflozin, the clinical DDI study results between MFA and dapagliflozin were critical to confirming the MFA UGT K i values. The initial MFA model predicted the PK of MFA, but it could not be further verified due to lack of published PK and clinical DDI data for MFA. The initial model underestimated the observed DDI following co‐administration of dapagliflozin and MFA when using the experimentally determined UGT1A9 and UGT2B7 K i values. Therefore, the UGT1A9 and UGT2B7 K i values were optimized (increased potency ~ 2.9‐fold) using the clinical DDI data such that the predicted AUCR and C maxR matched the observed values as reported in the literature.

Given the similarities in ADME and bioavailability between ertugliflozin and dapagliflozin, , , , , , the verified MFA PBPK model with optimized UGT K i values was used to simulate the DDI following co‐administration with the UGT substrate ertugliflozin. The PBPK model predicted an AUCR of 1.51 and a C maxR of 1.19 for the interaction between ertugliflozin and MFA. The magnitude of the predicted interaction is similar to the clinical DDI observed between MFA and dapagliflozin (AUCR = 1.51; C maxR = 1.13).

The safety and tolerability of ertugliflozin have been extensively studied throughout the drug’s clinical development program. In phase I and II studies, single doses of ertugliflozin of up to 300 mg, and multiple daily doses of up to 100 mg for ≤ 14 days and up to 25 mg for ≤ 12 weeks, had an acceptable safety profile. In phase III studies, ertugliflozin was safe and well‐tolerated at both the 5‐mg and 15‐mg approved doses, with a safety profile that is generally consistent with other members of the SGLT2‐inhibitor class. , Based on the overall safety profile of ertugliflozin across a wide dose range and variety of patient populations, combined with the AUCR of 1.51 for the predicted DDI between ertugliflozin and MFA from PBPK modeling, dose adjustment for safety reasons is not necessary when ertugliflozin is co‐administered with MFA or other UGT1A9 and UGT2B7 inhibitors of similar potency. The predicted magnitude of the DDI was within the ertugliflozin concentration range with established safety data and supports the justification that a clinical DDI study with a UGT inhibitor or ertugliflozin dose adjustment was not necessary. The use of PBPK modeling and simulation using well‐developed and verified compound models, as demonstrated in this study, contributes to improved efficiency and streamlining of drug development.

Funding

This study was sponsored by Pfizer Inc., New York, NY, USA in collaboration with Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA. Editorial support was provided by Shirley Smith, PhD, of Engage Scientific Solutions (Horsham, UK) and was funded by Pfizer Inc., New York, NY, USA in collaboration with Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Conflict of Interest

E.C., J.L., S.T., T.C.G., and V.S. are employees of Pfizer Inc., New York, NY, USA and may own shares/stock options in Pfizer Inc., New York, NY, USA.

Author Contributions

E.C., J.L., S.T., T.C.G., and V.S. wrote the manuscript. E.C., J.L., S.T., T.C.G., and V.S. designed the research. E.C. and J.L. performed the research. E.C., J.L., S.T., T.C.G., and V.S. analyzed the data.

Data‐sharing Statement

Upon request, and subject to review, Pfizer will provide the data that support the findings of this study. Subject to certain criteria, conditions, and exceptions, Pfizer may also provide access to the related individual anonymized participant data. See https://www.pfizer.com/science/clinical‐trials/trial‐data‐and‐results for more information.

Fig S1

Click here for additional data file.

Fig S2

Click here for additional data file.

Fig S3

Click here for additional data file.

Fig S4

Click here for additional data file.

Fig S5

Click here for additional data file.

Fig S6

Click here for additional data file.

Table S1

Click here for additional data file.

Table S2

Click here for additional data file.

Table S3

Click here for additional data file.

Supplementary Material

Click here for additional data file.