Advertisement

A Simple Decision Tree Suited for Identification of Early Oral Drug Candidates With Likely Pharmacokinetic Nonlinearity by Intestinal CYP3A Saturation

Published:October 30, 2020DOI:https://doi.org/10.1016/j.xphs.2020.10.050

      Abstract

      To identify oral drugs that likely display nonlinear pharmacokinetics due to saturable metabolism by intestinal CYP3A, our previous report using CYP3A substrate drugs proposed an approach using thresholds for the linear index number (LIN3A = dose/Km; Km, Michaelis-Menten constant for CYP3A) and the intestinal availability (FaFg). Here, we aimed to extend the validity of the previous approach using both CYP3A substrate and non-substrate drugs and to devise a decision tree suited for early drug candidates using in vitro metabolic intrinsic clearance (CLint, vitro) instead of FaFg. Out of 152 oral drugs (including 136 drugs approved in Japan, US or both), type I nonlinearity (in which systemic drug exposure increases in a more than dose-proportional manner) was noted with 82 drugs (54%), among which 58 drugs were identified as CYP3A substrates based on public information. Based on practical feasibility, 41 drugs were selected from CYP3A substrates and subjected to in-house metabolic assessment. The results were used to determine the thresholds for CLint, vitro (0.45 μL/min/pmol CYP3A4) and LIN3A (1.0 L). For four drugs incorrectly predicted, potential mechanisms were looked up. Overall, our proposed decision tree may aid in the identification of early drug candidates with intestinal CYP3A-derived type I nonlinearity.

      Keywords

      Introduction

      Nonlinear pharmacokinetics (PK) refer to deviations from dose-proportional (linear) profiles and are often observed with oral drugs.
      • Lappin G.
      • Noveck R.
      • Burt T.
      Microdosing and drug development: past, present and future.
      ,
      • Bosgra S.
      • Vlaming M.L.
      • Vaes W.H.
      To apply microdosing or not? Recommendations to single out compounds with non-linear pharmacokinetics.
      Depending on the upward or downward direction of such deviations, nonlinearity can be categorized into two different types: type I for an increasing trend and type II for a decreasing trend when the dose-normalized systemic exposure (e.g., the maximal blood/plasma concentration, Cmax or the area under the concentration-time curve, AUC) of drugs is compared over the increasing dose range. Drug candidates with type I nonlinearity are often considered undesirable for further development and marketing due to the concerns for unpredictable changes in drug exposure and subsequent safety issues as varying and multiple doses are administered. Thus, it is important to identify drug candidates with type I nonlinear PK during the early stages of drug development.
      For oral drugs displaying nonlinear PK, type I nonlinearity is more common than type II. The main contributing factors of type I nonlinearity include the saturation of (1) metabolic enzymes and/or uptake transporters in the liver, and (2) metabolic enzymes and/or efflux transporters in the intestine. Considering the importance of cytochrome P450 (CYP) 3A in drug metabolism and its abundant expression in the intestine,
      • Canaparo R.
      • Finnström N.
      • Serpe L.
      • Nordmark A.
      • Muntoni E.
      • Eandi M.
      • et al.
      Expression of CYP3A isoforms and P-glycoprotein in human stomach, jejunum and ileum.
      ,
      • Berggren S.
      • Gall C.
      • Wollnitz N.
      • Ekelund M.
      • Karlbom U.
      • Hoogstraate J.
      • et al.
      Gene and protein expression of P-glycoprotein, MRP1, MRP2, and CYP3A4 in the small and large human intestine.
      drug candidates are routinely tested for their likelihood of undergoing saturable metabolism by CYP3A in the intestine. CYP3A also plays an important role in drug metabolism in the liver. But, from the viewpoint of nonlinear PK of oral drugs, the intestine has been considered more important than the liver, considering the drug concentration in the intestinal epithelial cells is generally higher than that in the liver parenchymal cells.
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      To identify oral drugs that likely undergo saturable first-pass metabolism by intestinal CYP3A, Tachibana et al. proposed a decision tree based on the linear index number (LIN3A = dose/Km where Km is the Michaelis-Menten constant for CYP3A) and the intestinal availability (FaFg).
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      In doing so, Tachibana et al. utilized a panel of CYP3A substrates and identified the threshold values for LIN3A (≧2.8 L) and FaFg (<0.8). Given that this particular study included only CYP3A substrates and employed the FaFg threshold value (often not available for early drug candidates), further investigations are necessary to extend the validity of the threshold-based approach using a diverse panel of drugs (including both CYP3A substrates and non-substrates), and to devise a decision tree suited for early stages of drug development.
      The current study evaluated 152 oral drugs (including both CYP3A substrates and non-substrates; 136 small-molecule drugs that had been approved in Japan, US or both and 16 drugs previously compiled
      • Imawaka H.
      • Ito K.
      • Kitamura Y.
      • Sugyama K.
      • Sugiyama Y.
      Prediction of human bioavailability from human oral administration data and animal pharmacokinetic data without data from intravenous administration of drugs in humans.
      ). Based on commercial availability and practical feasibility, we selected 41 drugs and assessed their metabolic rates in-house. We then developed a decision tree to predict nonlinear PK based on the thresholds for the two parameters, LIN3A (dose/Km) and intrinsic metabolic clearance obtained in vitro (CLint, vitro) obtained using human CYP3A4 expression system instead of FaFg. For four drugs that were incorrectly assigned based on our current thresholds, possible mechanisms were looked up and discussed. Overall, our refined decision tree may serve as a simple method to identify early drug candidates with CYP3A-derived type I nonlinearity.

      Materials and Methods

       Compilation of the Systemic Exposure Information and Classification Per Type and Extent of Nonlinear PK

      The current study evaluated 152 oral small-molecule drugs including both CYP3A substrates and non-substrates [16 drugs were previously compiled
      • Imawaka H.
      • Ito K.
      • Kitamura Y.
      • Sugyama K.
      • Sugiyama Y.
      Prediction of human bioavailability from human oral administration data and animal pharmacokinetic data without data from intravenous administration of drugs in humans.
      and 136 drugs were selected from those approved from 2003 to 2013 in Japan, US or both (as of 2013, 116 out of 136 drugs had been approved in both Japan and US; as of 2020, 129 out of 136 drugs approved in both Japan and US)]. The dose-normalized systemic exposure (i.e., Cmax/dose and AUC/dose) was compared over the dose range (mainly from Phase 1 clinical trials that evaluated more than 3 doses) by surveying common technical documents, interview forms, and Pharmaceuticals and Medical Devices Agency (PMDA) review reports. For drugs not approved in Japan, Food and Drug Administration (FDA) review reports were used. For some drugs whose PK data was not available from the above sources, other scientific publications were used.
      • Cyong J.C.
      • Kodama K.
      • Yafune A.
      • Takebe M.
      • Takayanagi H.
      Phase I study of azithromycin single dose and multiple-dose for 3 days-.
      • Nakashima M.
      • Kanamaru M.
      • Takiguchi Y.
      • Mizuno A.
      • Watanabe I.
      Phase I study of betaxolol hydrochloride (MCI-144).
      • Nilsen O.G.
      • Dale O.
      • Husebø B.
      Pharmacokinetics of trazodone during multiple dosing to psychiatric patients.
      • Sahajwalla C.G.
      • Ayres J.W.
      Multiple-dose acetaminophen pharmacokinetics.
      • Bornemann L.D.
      • Min B.H.
      • Crews T.
      • Rees M.M.
      • Blumenthal H.P.
      • Colburn W.A.
      • et al.
      Dose dependent pharmacokinetics of midazolam.
      • Lacey L.F.
      • Hussey E.K.
      • Fowler P.A.
      Single dose pharmacokinetics of sumatriptan in healthy volunteers.
      Our analysis excluded prodrugs and drugs with sustained release formulations.
      In assessing the trend of the dose-normalized systemic exposure, we calculated fold differences of Cmax/dose and AUC/dose values to the smallest respective values over the dose range tested in Phase 1 clinical trials. When the results from multiple Phase 1 studies were available, preference was given to those covering low doses and/or a wide dose range. Based on the trend observed with Cmax/dose or AUC/dose, drugs were classified into 5 different groups. If at least one of the two parameters (i.e., Cmax/dose, AUC/dose) displayed deviations from unity with fold differences exceeding 25%, the drug was deemed to have nonlinear PK. Detailed criteria for group classification were as follows:
      Group 1 (linear PK, no evidence of nonlinearity): Neither of the two parameters showed fold differences exceeding 25% over the dose range. In the case of ruxolitinib, the results from Japanese and Non-Japanese populations varied, but no clear trend for nonlinearity was observed (thus classified as Group 1).
      Group 2 (a modest extent of type I nonlinearity): At least one of the two parameters showed an increasing trend with fold differences of 25–65% over the dose range.
      Group 3 (a pronounced extent of type I nonlinearity): At least one of the two parameters showed an increasing trend with fold differences exceeding 65% over the dose range as well as statistical significance (p < 0.05 compared to the parameters for the lowest dose, Dunnett's test).
      Group 4 (a pronounced extent of type I nonlinearity, but lacking a uniform pattern or statistical significance including the cases with no variability data): At least one of the two parameters showed a trend of an increase with fold differences exceeding 65%, but the cases lacked a uniform pattern (typically showing an initial increase, followed by a decreasing trend) or the fold differences did not reach statistical significance (p > 0.05 compared to the parameters for the lowest dose, Dunnett's test; some cases with no information on data variability).
      Group 5 (type II nonlinearity): At least one of the two parameters showed a decreasing trend with fold differences exceeding 25% over the dose range. This may be attributable mostly to the limited dissolution in the intestine or the interaction with intestinal components (e.g., food, bile acids, lipids).

       In Vitro CYP3A4 Metabolic Studies and Calculation of Approximate Km Values

      From Groups 2–4 (type I nonlinearity), 58 drugs were identified as CYP3A substrates based on their publicly available data. For the verification purpose, 41 drugs were selected from Groups 1–4 (based on commercial availability and practical feasibility to establish the analytical methods) and their metabolic rates were assessed in vitro using the human CYP3A4 expression system (Supersome®, BD Gentest Co., Woburn, MA).
      Metabolic rates were evaluated by measuring the depletion of the parent drug (the substrate depletion method, commonly used for early drug candidates and adaptable for high-throughput screening). Briefly, each drug (0.1 μM) was added to the incubation mixture containing recombinant human CYP3A4 (30 pmol/mL). After preincubation for 5 min, an NADPH regenerating system was added to initiate the metabolic reaction. An aliquot of the incubation mixture was taken at 1, 5, and 30 min, and mixed with acetonitrile containing an internal standard (chlorpropamide). After centrifugation at 15,000 rpm for 10 min, the resulting supernatant was diluted with 50% acetonitrile containing 0.1% formic acid, and the remaining drug levels were quantified by liquid chromatography-tandem mass spectrometry (LC-MS/MS). For compounds displaying rapid depletion, the content of CYP3A4 in the incubation mixture was reduced to ensure that at least 20% of the initial drug concentration remains after 30 min incubation.
      For 32 drugs with metabolic rates faster than erlotinib (0.11 μL/min/pmol CYP3A4), their Km values were estimated by performing in vitro metabolic reaction using another drug concentration corresponding to “clinical dose/2.8 L” (the concentration likely to saturate intestinal CYP3A-mediated metabolism according to Tachibana et al.
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      ; denoted as “CH” in subsequent sections to distinguish from the initial test concentration of 0.1 μM denoted as “CL”). The CLint, vitro values were calculated for both concentrations (CH and CL) and used to obtain approximate Km values using Equations ((1), (2), (3), (4)).
      ν=Vmax×CKm+C
      (1)


      CLint,vitro("CL")=νc=VmaxKm+CL
      (2)


      CLint,vitro("CH")=νc=VmaxKm+CH
      (3)


      When CL is much lower than Km, the approximate Km values can be obtained using Equation (4);
      CLint,vitro("CL")CLint,vitro("CH")=VmaxKm+CLVmaxKm+CHKm+CHKm=1+CHKm
      (4)


      When the results at CH were deemed to reach nearly complete saturation, the metabolic reaction was performed with another concentration below CH and the results were included in order to obtain more reliable Km values.

       Determination of Drug Concentrations

      The drug levels were quantified by LC-MS/MS (LCMS8040 or LCMS8050, Shimadzu, Tokyo, Japan) using a C18 column (Kinetex 1.6 mm × 150 mm, Shimadzu, Tokyo, Japan) and a gradient mobile phase that consisted of acetonitrile and 0.1% formic acid at the flow rate of 0.2 mL/min and column temperature of 40 °C (detailed MS/MS analytical conditions in Supplemental Table 3).

      Results

      When the dose-normalized Cmax and AUC were assessed for 152 oral drugs, a vast majority (n = 130, 85.5%) of drugs were deemed to have nonlinear PK: type I (n = 82, 53.9%) and type II (n = 48, 31.6%). When the drugs with type I nonlinearity were further classified based on the extent and pattern of nonlinearity, 69 drugs (Groups 2 and 3) showed a monophasic increase in dose-normalized exposure (Table 1; classification flow shown in Scheme 1; profiles of dose-normalized Cmax and AUC of individual drugs provided in Supplemental Fig. 1). Thirteen drugs (Group 4) were identified to have a pronounced extent of type I nonlinearity, but lacking a uniform pattern or statistical significance.
      Table 1Classification of 152 Oral Drugs Based on Their Linear/Non-Linear PK Profiles Based on the Dose-Normalized Cmax and AUC Values (Including 136 Drugs Approved in Japan, US or Both From 2003 to 2013 and 16 Oral Drugs Compiled by Imawaka et al.
      • Imawaka H.
      • Ito K.
      • Kitamura Y.
      • Sugyama K.
      • Sugiyama Y.
      Prediction of human bioavailability from human oral administration data and animal pharmacokinetic data without data from intravenous administration of drugs in humans.
      ).
      Group CalcificationTotal
      12345
      Linear PKNonlinear PK (type I)Nonlinear PK (Type II)
      Modest ExtentPronounced ExtentPronounced Extent, But Lacking a Uniform Pattern or Statistical Significance
      Number of drugs (% from the total 152 drugs surveyed)22 (14.5%)36 (23.7%)33 (21.7%)13 (8.6%)48 (31.6%)152 (100%)
      82 (53.9%)
      CYP3A substrates (% in the respective group as per publicly available data)14 (63.6%)24 (66.7%)25 (75.8%)9 (69.2%)27 (56.3%)99 (65.1%)
      58 (70.7%)
      Selected for in-house metabolic assessment713174ND41
      Correctly assigned for their linear or non-linear PK using the threshold values from the current study (% of positive predictive values)6 (85.7%)10 (76.9%)14 (82.4%)3 (75.0%)ND33 (80.5%)
      ND = Not determined. Each drug was classified based on Cmax/dose and AUC/dose values over the dose range tested in the Phase 1 clinical trials. Fold differences of Cmax/dose or AUC/dose relative to the smallest respective values were calculated. Detailed classification criteria are provided in the text.
      Figure thumbnail sc1
      Scheme 1Classification scheme of 152 oral small-molecule drugs for their linear/nonlinear pharmacokinetic profiles based on dose-normalized Cmax and AUC. Classification scheme of 152 oral drugs based on their linear/nonlinear pharmacokinetic profiles based on the normalized Cmax and AUC values (including 136 drugs approved in Japan, United States or both from 2003 to 2013 and 16 oral drugs compiled by Imawaka et al.
      • Imawaka H.
      • Ito K.
      • Kitamura Y.
      • Sugyama K.
      • Sugiyama Y.
      Prediction of human bioavailability from human oral administration data and animal pharmacokinetic data without data from intravenous administration of drugs in humans.
      ). Pie chart shows the % from the total 152 drugs surveyed.
      Toward our goal of establishing a simple, quantitative approach that can predict the occurrence of type I nonlinearity among early drug candidates, we selected 41 drugs [identified as CYP3A substrates; 7 drugs displaying linear PK (Group 1), and 34 drugs displaying nonlinear PK (Groups 2–4)] based on commercial availability and practical feasibility to establish the analytical methods (Table 1, Scheme 1). By performing in-house measurements of the metabolic rates for 41 drugs selected, we obtained their LIN3A and CLint, vitro values. As shown in Fig. 1, visual inspection led us to identify the threshold values for LIN3A and CLint, vitro: 1.0 L and 0.45 μL/min/pmol CYP3A4, respectively. Out of 34 drugs which displayed type I nonlinearity based on publicly available data and were assessed for their metabolic rates (Groups 2–4), 27 drugs (79.4%) met the criteria of both LIN3A and CLint, vitro being above the threshold values (located in the upper right quadrant in Fig. 1). For 7 drugs classified as Group 1 (linear PK), 6 drugs were found to be below the threshold values for LIN3A4 and/or CLint, vitro (located in the lower left and right quadrants in Fig. 1; the two drugs of naratriptan and lorcaserin appear as one symbol due to their nearly overlapping values). There were also 4 drugs (mirabegron, maraviroc, voriconazole and finasteride) that displayed nonlinear PK at a pronounced extent (Groups 3–4), but were incorrectly assigned to have linear PK according to our thresholds. The dose-dependent profiles of the systemic exposure for these four drugs are shown in Fig. 2. For those drugs, possible mechanisms were looked up and discussed below.
      Figure thumbnail gr1
      Fig. 1Plot showing the LIN3A and CLint, vitro values obtained for 41 drugs (Groups 1–4) by performing in-house measurements of metabolic rates. For 41 drugs selected from Groups 1–4, their CLint, vitro values (metabolic activity at 0.1 μM) were determined using the human CYP3A4 expression system. For 32 drugs whose CLint, vitro values faster than 0.11 μL/min/pmol CYP3A4, approximate Km values were determined, subsequently calculating LIN3A. In the cases of voriconazole and maraviroc, the LIN3A values were calculated using the Km values from the published reports (11 μM for voriconazole,
      • Murayama N.
      • Imai N.
      • Nakane T.
      • Shimizu M.
      • Yamazaki H.
      Roles of CYP3A4 and CYP2C19 in methyl hydroxylated and N-oxidized metabolite formation from voriconazole, a new anti-fungal agent, in human liver microsomes.
      ,
      • Damle B.
      • Varma M.V.
      • Wood N.
      Pharmacokinetics of voriconazole administered concomitantly with fluconazole and population-based simulation for sequential use.
      10 μM for maraviroc
      • Tseng E.
      • Fate G.D.
      • Walker G.S.
      • Goosen T.C.
      • Obach R.S.
      Biosynthesis and identification of metabolites of maraviroc and their use in experiments to delineate the relative contributions of cytochrome P450 3A4 versus 3A5.
      ). For a subset of drugs whose CLint, vitro values were slower than 0.11 μL/min/pmol CYP3A4, their LIN3A values were plotted as zero. Dotted lines represent the thresholds for CLint, vitro (0.45 μL/min/pmol CYP3A4) and LIN3A (1.0 L) identified in the current study. When a drug meets the criteria of exceeding the threshold values for both CLint, vitro and LIN3A (being located in the upper right quadrant; grey-colored shaded region), the drug is predicted to have nonlinear pharmacokinetics from saturation of the intestinal CYP3A. The two drugs of naratriptan and lorcaserin appear as a single, closed circle due to their nearly overlapping values. Four drugs that were not correctly assigned based on the current threshold-based approach are marked as follows. F, finasteride; MA, maraviroc; MI, mirabegron; V, voriconazole.
      Figure thumbnail gr2
      Fig. 2Plots showing dose-normalized drug exposure parameters (Cmax/dose and AUC/dose) of mirabegron, maraviroc, finasteride and voriconazole (oral and intravenous administration) over the dose ranges Results are shown as mean and standard deviation except for maraviroc which lacked the information from the source data. The Cmax/dose and AUC/dose values are shown as closed circle and triangle, respectively. PO = oral, IV = intravenous, ∗ = p < 0.05 compared to the parameters for the lowest dose, Dunnett's test.

      Discussion

      For oral drugs, type I nonlinearity in their PK is often caused by saturation of intestinal metabolism (mainly mediated by CYP3A). To assess the likelihood of such events, the previous report by Tachibana et al.
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      proposed a decision tree using the LIN3A and FaFg values. To extend the validity of the previous approach, our current study surveyed an expanded panel of 152 oral drugs (including both CYP3A substrates and non-substrates) and identified the threshold values for the LIN3A and CLint, vitro values (instead of FaFg).
      The threshold values for LIN3A and CLint, vitro identified in the current study appear in line with those reported previously. The report by Kato examined the relationship between FaFg values and hepatic intrinsic clearance (CLint,h) values.
      • Kato M.
      Intestinal first-pass metabolism of CYP3A4 substrates.
      In that report, Fa values of CYP3A4 substrate were close to unity, thereby their FaFg values mainly representing Fg values. The plot showed that the FaFg values were mostly above 0.8 when the CLint,h values were less than 100 mL/min/kg. As CLint,h exceeded an apparent cut-off value of 100 mL/min/kg, the FaFg values markedly decreased from 0.8 (likely leading to type I nonlinearity).
      • Kato M.
      Intestinal first-pass metabolism of CYP3A4 substrates.
      When appropriate physiological scaling factors were applied (131 pmol CYP3A4/mg microsomal protein
      • Rowland Yeo K.
      • Rostami-Hodjegan A.
      • Tucker G.T.
      Abundance of cytochromes P450 in human liver: a meta-analysis. 225P GKT, University of London Winter meeting December 2003.
      ; 52.5 mg microsomal protein/g liver
      • Iwatsubo T.
      • Hirota N.
      • Ooie T.
      • Suzuki H.
      • Shimada N.
      • Chiba K.
      • et al.
      Prediction of in vivo drug metabolism in the human liver from in vitro metabolism data.
      ; 25.7 g liver/kg
      • Davies B.
      • Morris T.
      Physiological parameters in laboratory animals and humans.
      ), our threshold value of CLint, vitro (0.45 μL/min/pmol CYP3A4) corresponds to CLint,h of 79.5 mL/min/kg, comparable to the apparent cut-off value of 100 mL/min/kg with the corresponding FaFg value of 0.8 as reported previously.
      • Kato M.
      Intestinal first-pass metabolism of CYP3A4 substrates.
      The threshold value for LIN3A (1.0 L) obtained in the current study were within 3-fold to LIN3A (2.8 L) obtained by Tachibana et al.
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      Although further validation may be necessary using a larger sample size, we cautiously interpret that our threshold values of LIN3A (1.0 L) and CLint, vitro (0.45 μL/min/pmol CYP3A4) may offer an improved prediction accuracy, given that the current threshold values were determined from 41 CYP3A substrate drugs, which were selected from an initial panel of 152 drugs including both CYP3A substrates and non-substrates (different from Tachibana et al.
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      using 11 CYP3A substrates only).
      Scheme 2 depicts a simple decision tree to predict type I nonlinearity based on the results of the current study. The first step is to determine CLint, vitro at a concentration of 0.1 μM (“CL”) by performing in vitro metabolic study using the human CYP3A4 expression system by the substrate depletion method. If the measured metabolic rates are slower than 0.45 μL/min/pmol CYP3A4, a decision can be made that drug candidates are likely to display linear PK with the estimated FaFg value close to unity. For drug candidates which display metabolic rates faster than 0.45 μL/min/pmol CYP3A4, the second step is to determine Km and LIN3A values by measuring CLint, vitro at another higher concentration of clinical dose/2.8 L (“CH”). If the calculated LIN3A values exceed 1.0 L, a decision can be made that drug candidates are likely to display type I nonlinear PK. Considering that early drug candidates are routinely examined whether or not they are CYP3A substrates using human CYP3A expression system at low concentrations, the proposed decision tree is expected to aid in the assessment of type I nonlinearity of drug candidates.
      Figure thumbnail sc2
      Scheme 2Decision tree to identify early drug candidates with the likelihood to display nonlinear pharmacokinetics by intestinal CYP3A saturation based on the results from in vitro metabolic studies. The first step is to perform in vitro metabolic study using the human CYP3A4 expression system by the substrate depletion method and determine CLint, vitro (“CL”) at a concentration of 0.1 μM. For drug candidates which display metabolic rates faster than 0.45 μL/min/pmol CYP3A4, CLint, vitro (“CH”) is determined at another concentration of CH (“clinical dose/2.8 L”). More detailed description is provided in the main text.
      When the criteria using the threshold values from our current study were applied, we noted that 4 drugs did not satisfy the criteria even though they were classified to Groups 3 or 4 (type I nonlinearity). We looked up potential mechanisms of their nonlinear PK as summarized below and in Table 2.
      Table 2Pharmacokinetic Properties of 4 Drugs Which Belong to Groups 3 or 4, But Were Not Correctly Assigned Based on the Current Threshold-Based Approach.
      ParametersCalculation FormulaMirabegronMaravirocVoriconazoleFinasteride
      Molecular weight397514349373
      Maximum clinical dose (mg)50300400 (Japan), 200 (US)1 (AGA), 5 (BPH, US)
      Dose range in clinical study (mg)
      The numbers in parentheses are dose levels at or beyond which nonlinearity was observed (1.25-folds increase). Dose range in Phase1 study for mirabegron was 10–340 mg, but 10 mg was excluded because AUCinf was not calculated. Dose range in Phase1 study for maraviroc was 1–1200 mg. 1 mg was excluded because n = 2. Three mg was excluded because plasma profile was not provided.
      30–340 (100)10–1200 (30)100–400 (200)0.2–100 (1)
      Cmax, plasma (μM)
      Maximum concentrations at dose where nonlinearity was observed.
      0.2800.03002.610.0265
      Blood/Plasma (Rb)1.420.5911
      fb0.1920.4240.4180.160
      Cmax, blood, u (μM)
      Maximum concentrations at dose where nonlinearity was observed.
      Cmax, blood × fb0.07630.007441.090.00424
      AUCinf, plasma (ng·h/mL)230 (15 mg IV)57.6
      AUC0-t for maraviroc and AUC0–24 for finasteride.
      (3 mg IV)
      2388 (1.5 mg/kg IV)PP49.29
      AUC0-t for maraviroc and AUC0–24 for finasteride.
      ,
      • Suzuki R.
      • Satoh H.
      • Ohtani H.
      • Hori S.
      • Sawada Y.
      Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
      (1 mg PO)
      CLtot (L/h)Dose/AUCinf×Rb45.988.344.020.3
      • Suzuki R.
      • Satoh H.
      • Ohtani H.
      • Hori S.
      • Sawada Y.
      Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
      fe (%)22.223.22
      • Roffey S.J.
      • Cole S.
      • Comby P.
      • Gibson D.
      • Jezequel S.G.
      • Nedderman A.N.R.
      • et al.
      The disposition of voriconazole in mouse, rat, rabbit, guinea pig, dog, and human.
      0.04
      fe (%) was after oral administration.
      CLh (L/h)CLtot×(100fe)/10035.767.843.120.3
      fe (%) after IV dosing was estimated very low because fe (%) after PO dosing was 0.04% and F is 0.8. Thus, CLh was assumed to be CLtot.
      Fh
      The numbers on the left and right represent the calculation results using the Qh values of 1.24 and 1.53 L/h/kg, respectively (assuming the body weight of 70 kg).
      1CLhQh0.59, 0.670.22, 0.370.50, 0.600.77, 0.81
      F (PK data used for calculation)AUC(PO)×Dose(IV)AUC(IV)×Dose(PO)0.224 (IV:15 mg, PO: 30 mg)0.0896 (IV:3 mg, PO: 10 mg)0.799 (IV:1.5 mg/kg, PO:100 mg)0.8
      • Suzuki R.
      • Satoh H.
      • Ohtani H.
      • Hori S.
      • Sawada Y.
      Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
      FaRg
      The numbers on the left and right represent the calculation results using the Qh values of 1.24 and 1.53 L/h/kg, respectively (assuming the body weight of 70 kg).
      F/Fh0.38, 0.340.41, 0.241.58, 1.341.04, 0.99
      Cmax, in, u (μM)
      Maximum concentrations at dose where nonlinearity was observed.
      ,
      Calculation was done using the Qh value of 1.24 L/h/kg (assuming the body weight of 70 kg). The ka values for all drugs were assumed to be 0.1 min-1. FaFg = 1 was used for voriconazole and finasteride though calculated FaFg is over 1. Rb = 1 was used for voriconazole and finasteride because no available data.
      Cmax,blood,u+ka×FaFg×DoseQh6.681.6640.60.189
      Kd for target binding (nM)Not applicableNot applicableNot applicable0.631
      • Suzuki R.
      • Satoh H.
      • Ohtani H.
      • Hori S.
      • Sawada Y.
      Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
      CLint, vitro
      In-house data.
      (μL/min/pmol CYP3A4)
      0.1260.08130.1160.0581
      Km of CYP3A4 (μM)1.3
      In-house data.
      , 39.9
      10.0–10.9
      • Tseng E.
      • Fate G.D.
      • Walker G.S.
      • Goosen T.C.
      • Obach R.S.
      Biosynthesis and identification of metabolites of maraviroc and their use in experiments to delineate the relative contributions of cytochrome P450 3A4 versus 3A5.
      , 13
      • Hyland R.
      • Dickins M.
      • Collins C.
      • Jones H.
      • Jones B.
      Maraviroc: in vitro assessment of drug-drug interaction potential.
      , 11.1
      • Lu Y.
      • Hendrix C.W.
      • Bumpus N.N.
      Cytochrome P450 3A5 plays a prominent role in the oxidative metabolism of the anti-human immunodeficiency virus drug maraviroc.
      15
      • Damle B.
      • Varma M.V.
      • Wood N.
      Pharmacokinetics of voriconazole administered concomitantly with fluconazole and population-based simulation for sequential use.
      , 11
      • Murayama N.
      • Imai N.
      • Nakane T.
      • Shimizu M.
      • Yamazaki H.
      Roles of CYP3A4 and CYP2C19 in methyl hydroxylated and N-oxidized metabolite formation from voriconazole, a new anti-fungal agent, in human liver microsomes.
      ,
      • Damle B.
      • Varma M.V.
      • Wood N.
      Pharmacokinetics of voriconazole administered concomitantly with fluconazole and population-based simulation for sequential use.
      , 16
      • Murayama N.
      • Imai N.
      • Nakane T.
      • Shimizu M.
      • Yamazaki H.
      Roles of CYP3A4 and CYP2C19 in methyl hydroxylated and N-oxidized metabolite formation from voriconazole, a new anti-fungal agent, in human liver microsomes.
      , 235
      • Hyland R.
      • Jones B.C.
      • Smith D.A.
      Identification of the cytochrome P450 enzymes involved in the N-oxidation of voriconazole.
      , 834.7
      • Qi F.
      • Zhu L.
      • Li Na
      • Ge T.
      • Xu G.
      • Liao S.
      Influence of different proton pump inhibitors on the pharmacokinetics of voriconazole.
      No Data
      Km of CYP2C9 (μM)No DataNo Data20
      • Hyland R.
      • Jones B.C.
      • Smith D.A.
      Identification of the cytochrome P450 enzymes involved in the N-oxidation of voriconazole.
      ,
      • Qi F.
      • Zhu L.
      • Li Na
      • Ge T.
      • Xu G.
      • Liao S.
      Influence of different proton pump inhibitors on the pharmacokinetics of voriconazole.
      No Data
      Km of CYP2C19 (μM)No DataNo Data3.5
      • Damle B.
      • Varma M.V.
      • Wood N.
      Pharmacokinetics of voriconazole administered concomitantly with fluconazole and population-based simulation for sequential use.
      ,
      • Hyland R.
      • Jones B.C.
      • Smith D.A.
      Identification of the cytochrome P450 enzymes involved in the N-oxidation of voriconazole.
      , 9.3
      • Qi F.
      • Zhu L.
      • Li Na
      • Ge T.
      • Xu G.
      • Liao S.
      Influence of different proton pump inhibitors on the pharmacokinetics of voriconazole.
      , 14
      • Murayama N.
      • Imai N.
      • Nakane T.
      • Shimizu M.
      • Yamazaki H.
      Roles of CYP3A4 and CYP2C19 in methyl hydroxylated and N-oxidized metabolite formation from voriconazole, a new anti-fungal agent, in human liver microsomes.
      No Data
      Km of CYP2D6 (μM)0.23No DataNo DataNo Data
      Km of P-gp (μM)294
      • Takusagawa S.
      • Ushigome F.
      • Nemoto H.
      • Takahashi Y.
      • Li Q.
      • Kerbusch V.
      • et al.
      Intestinal absorption mechanism of mirabegron, a potent and selective β₃-adrenoceptor agonist: involvement of human efflux and/or influx transport systems.
      37
      • Walker D.K.
      • Abel S.
      • Comby P.
      • Muirhead G.J.
      • Nedderman A.N.R.
      • Smith D.A.
      Species differences in the disposition of the CCR5 antagonist, UK-427,857, a new potential treatment for HIV.
      No DataNo Data
      Km of OATP1B1 (μM)No Data5
      • Kimoto E.
      • Vourvahis M.
      • Scialis R.J.
      • Eng H.
      • Rodorigues A.D.
      • Varma M.V.S.
      Mechanistic evaluation of the complex drug-drug interactions of maraviroc: contribution of cytochrome P450 3A, P-glycoprotein and organic anion transporting polypeptide 1B1.
      No DataNo Data
      LIN3A4 (L)Dose
      Maximum dose in Phase I study (340 mg for mirabegron, 1200 mg for maraviroc, 400 mg for voriconazole) were used for calculation.
      /Km
      Calculation used the Km values of 1.3 μM, 10 μM, and 11 μM for mirabegron, maraviroc, and voriconazole, respectively.
      659233104Not calculated
      LINp-gp (L)Dose
      Maximum dose in Phase I study (340 mg for mirabegron, 1200 mg for maraviroc, 400 mg for voriconazole) were used for calculation.
      /Km
      2.9163.1Not calculatedNot calculated
      Possible mechanisms for nonlinear PKSaturation of CYP3A/2D6 mediated metabolism during first-pass effectSaturation of P-gp/OATP1B1Saturation of CYP2C19 mediated metabolismSaturation of the high-affinity binding to steroid 5α-reductase 2
      Basis for reasoning
      • Cmax, in, u > Km of CYP3A4/2D6
      • Cmax, u < Km of CYP3A4/2D6
      • LINp-gp > 0.77L
      • Cmax, in, u > Km of OATP1B1
      • Cmax, u < Km of OATP1B1
      • Cmax, in, u > Km of CYP2C19
      • Cmax,u~Km of CYP2C19
      • Nonlinearity was also observed after intravenous administration.
      • Mechanism of nonlinearity was investigated by Suzuki et al
        • Suzuki R.
        • Satoh H.
        • Ohtani H.
        • Hori S.
        • Sawada Y.
        Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
      AGA = androgenetic alopecia; BPH = benign prostatic hyperplasia; Cmax = blood concentration; Cmax, u = unbound blood concentration of Cmax; Cmax, in, u = unbound blood concentration of Cmax at inlet liver; F = bioavailability; IV = intravenous; PO = oral.
      Unless references are provided, the data are from the common technical documents provided by PMDA (Pharmaceuticals and Medical Devices Agency) at the following URL addressees.
      a The numbers in parentheses are dose levels at or beyond which nonlinearity was observed (1.25-folds increase). Dose range in Phase1 study for mirabegron was 10–340 mg, but 10 mg was excluded because AUCinf was not calculated. Dose range in Phase1 study for maraviroc was 1–1200 mg. 1 mg was excluded because n = 2. Three mg was excluded because plasma profile was not provided.
      b Maximum concentrations at dose where nonlinearity was observed.
      c The numbers on the left and right represent the calculation results using the Qh values of 1.24 and 1.53 L/h/kg, respectively (assuming the body weight of 70 kg).
      d Calculation was done using the Qh value of 1.24 L/h/kg (assuming the body weight of 70 kg). The ka values for all drugs were assumed to be 0.1 min-1. FaFg = 1 was used for voriconazole and finasteride though calculated FaFg is over 1. Rb = 1 was used for voriconazole and finasteride because no available data.
      e In-house data.
      f AUC0-t for maraviroc and AUC0–24 for finasteride.
      g fe (%) was after oral administration.
      h fe (%) after IV dosing was estimated very low because fe (%) after PO dosing was 0.04% and F is 0.8. Thus, CLh was assumed to be CLtot.
      i Maximum dose in Phase I study (340 mg for mirabegron, 1200 mg for maraviroc, 400 mg for voriconazole) were used for calculation.
      j Calculation used the Km values of 1.3 μM, 10 μM, and 11 μM for mirabegron, maraviroc, and voriconazole, respectively.

      Mirabegron (LIN3A≧1.0 L, but CLint, vitro less than 0.45 μL/min/pmol CYP3A4)

      The FaFg and Fh values of mirabegron were calculated to be 0.34–0.38 and 0.59–0.67, respectively (Table 2). The calculated FaFg values were relatively low, suggesting that the observed nonlinearity may arise from the saturation of transporter-mediated efflux in the intestine. The Km value of mirabegron for- P-glycoprotein (P-gp) is reported to be 294 μM
      • Takusagawa S.
      • Ushigome F.
      • Nemoto H.
      • Takahashi Y.
      • Li Q.
      • Kerbusch V.
      • et al.
      Intestinal absorption mechanism of mirabegron, a potent and selective β₃-adrenoceptor agonist: involvement of human efflux and/or influx transport systems.
      and LINp-gp is 2.91 L. LINp-gp of mirabegron is larger than the criteria of nonlinear PK by P-gp which Tachibana et al. suggested (0.77 L).
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      In addition, the Km values of mirabegron were assessed to be 1.3 μM for CYP3A4 (our in-house data) and 0.23 μM for CYP2D6, lower than Cmax,in,u (6.7 μM) (Table 2). Taken together, saturable first-pass metabolism by CYP3A4/CYP2D6 in the liver and/or P-gp-mediated efflux in the intestine may account for the observed nonlinear PK.

      Maraviroc (LIN3A≧1.0 L, but CLint, vitro Less Than 0.45 μL/min/pmol CYP3A4)

      The FaFg and Fh values of maraviroc were calculated to be 0.24–0.41 and 0.22–0.37, respectively (Table 2). As per the results from Phase 1 study, the Cmax/dose and AUC/dose of orally administered maraviroc displayed the respective increases by 8- and 5-fold over the oral dose range of 10–1200 mg (Fig. 2). Nonlinearity was observed even at low oral doses (10 mg), but not after intravenous administration (3–30 mg).
      Common technical document of maraviroc for new drug application in Japan.
      Thus, the nonlinearity after oral administration of maraviroc is likely mediated by saturable components in the intestine and/or the liver during the first-pass effect. Maraviroc is a substrate of P-gp with the Km of 37 μM.
      • Walker D.K.
      • Abel S.
      • Comby P.
      • Muirhead G.J.
      • Nedderman A.N.R.
      • Smith D.A.
      Species differences in the disposition of the CCR5 antagonist, UK-427,857, a new potential treatment for HIV.
      When the LIN value was calculated for P-gp in a similar manner to LIN3A, the resulting value was 63 L, much greater than the threshold value (0.77 L) that Tachibana et al. proposed for substrates of P-gp.
      • Tachibana T.
      • Kato M.
      • Sugiyama Y.
      Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
      In addition to saturable interactions with P-gp, maraviroc may also undergo saturable metabolism and/or uptake in the liver. Maraviroc is reported to be a substrate of organic anion transporting polypeptide (OATP)1B1 with its Km value of 5 μM.
      • Kimoto E.
      • Vourvahis M.
      • Scialis R.J.
      • Eng H.
      • Rodorigues A.D.
      • Varma M.V.S.
      Mechanistic evaluation of the complex drug-drug interactions of maraviroc: contribution of cytochrome P450 3A, P-glycoprotein and organic anion transporting polypeptide 1B1.
      After administration of the maraviroc in a Phase 1 study, the Cmax, bood, u were approximately 1.4 μM even at 1200 mg. In addition, the maximal concentration of unbound maraviroc at the hepatic inlet (Cmax,in,u) at 100 and 1200 mg were calculated to be 5.6 and 67.4 μM, respectively.
      Common technical document of maraviroc for new drug application in Japan.
      These results suggest that saturable hepatic uptake of maraviroc during the first-pass effect may also contribute to its nonlinear PK. Taken together, nonlinear PK of maraviroc may involve the saturation of both intestinal P-gp and hepatic OATP1B1.

      Voriconazole (LIN3A≧1.0 L, but CLint, vitro Less Than 0.45 μL/min/pmol CYP3A4)

      The FaFg and Fh values of voriconazole were calculated to be close to 1 and 0.50–0.60, respectively (Table 2). As per the results from a Phase 1 study over the dose range of 100–400 mg, the Cmax/dose and AUC/dose of orally administered voriconazole increased by 1.9- and 4.3-fold, respectively (Fig. 2). Interestingly, nonlinearity was also observed after intravenous administration of voriconazole: the AUC/dose increased by 1.9-fold over the dose range of 1.5–6 mg/kg.
      Common technical document of voriconazole for new drug application in Japan.
      These results suggest that nonlinearity after oral administration may involve saturable elimination in the liver, considering a minor contribution (~1.5%) of renal excretion. Voriconazole is metabolized mainly by CYP2C19 (with some contributions by CYP2C9 and CYP3A4). The Km values of voriconazole for CYP2C19 are reported to be 3.5–14 μM,
      • Murayama N.
      • Imai N.
      • Nakane T.
      • Shimizu M.
      • Yamazaki H.
      Roles of CYP3A4 and CYP2C19 in methyl hydroxylated and N-oxidized metabolite formation from voriconazole, a new anti-fungal agent, in human liver microsomes.
      • Damle B.
      • Varma M.V.
      • Wood N.
      Pharmacokinetics of voriconazole administered concomitantly with fluconazole and population-based simulation for sequential use.
      • Hyland R.
      • Jones B.C.
      • Smith D.A.
      Identification of the cytochrome P450 enzymes involved in the N-oxidation of voriconazole.
      • Qi F.
      • Zhu L.
      • Li Na
      • Ge T.
      • Xu G.
      • Liao S.
      Influence of different proton pump inhibitors on the pharmacokinetics of voriconazole.
      and Cmax, blood, u, and Cmax, in, u after oral administration of 200 mg was about 1.09 μM and 40.6 μM, respectively. Taken together, the nonlinearity observed after oral administration of voriconazole is likely from its saturable metabolism by CYP2C19 in the liver (both during first-pass and in the systemic circulation).

      Finasteride (LIN3A Less Than 1.0 L, CLint, vitro Less Than 0.45 μL/min/pmol CYP3A4)

      The FaFg and Fh values of finasteride were calculated to be close to 1 and 0.77–0.81, respectively (Table 2). Therefore, the first-pass elimination in the intestine and liver is unlikely to be responsible for non-linearity. Saturable, high-affinity binding of finasteride to the molecular target (steroid 5α-reductase, type II) has been already identified as the underlying mechanism for nonlinear PK.
      • Suzuki R.
      • Satoh H.
      • Ohtani H.
      • Hori S.
      • Sawada Y.
      Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
      In fact, the dissociation constant (Kd) for the binding of finasteride to the molecular target is known as 0.6 nM, which is much lower than Cmax, blood, u of 4.2 nM, supporting this idea. This type of nonlinear PK is referred to as target-mediated drug disposition (TMDD) and has been frequently associated with biologics in recent years. However, the early TMDD examples were in fact with small-molecule drugs which feature high-affinity, low-capacity interactions with their targets (e.g., warfarin, bosentan).
      • Levy G.
      Pharmacologic target-mediated drug disposition.
      ,
      • Mager D.E.
      • Jusko W.J.
      General pharmacokinetic model for drugs exhibiting target-mediated drug disposition.
      If small-molecule drugs of certain features (i.e. high potency, high target-specificity, limited non-target-mediated tissue distribution) display nonlinear PK, the potential for the TMDD may be considered.
      In conclusion, the current study utilized an expanded panel of CYP3A substrates and non-substrates and developed a simple decision tree which allows for the prediction of PK nonlinearity based on the two parameters, LIN3A and CLint, vitro. Our simple and refined decision tree may serve as a useful tool for predicting CYP3A-mediated PK nonlinearity of orally administered drugs during the early stages of drug development.

      Appendix A. Supplementary Data

      References

        • Lappin G.
        • Noveck R.
        • Burt T.
        Microdosing and drug development: past, present and future.
        Expert Opin Drug Metab Toxicol. 2013; 9: 817-834
        • Bosgra S.
        • Vlaming M.L.
        • Vaes W.H.
        To apply microdosing or not? Recommendations to single out compounds with non-linear pharmacokinetics.
        Clin Pharmacokinet. 2016; 55: 1-15
        • Canaparo R.
        • Finnström N.
        • Serpe L.
        • Nordmark A.
        • Muntoni E.
        • Eandi M.
        • et al.
        Expression of CYP3A isoforms and P-glycoprotein in human stomach, jejunum and ileum.
        Clin Exp Pharmacol Physiol. 2007; 34: 1138-1144
        • Berggren S.
        • Gall C.
        • Wollnitz N.
        • Ekelund M.
        • Karlbom U.
        • Hoogstraate J.
        • et al.
        Gene and protein expression of P-glycoprotein, MRP1, MRP2, and CYP3A4 in the small and large human intestine.
        Mol Pharm. 2007; 4: 252-257
        • Tachibana T.
        • Kato M.
        • Sugiyama Y.
        Prediction of nonlinear intestinal absorption of CYP3A4 and P-glycoprotein substrates from their in vitro Km values.
        Pharm Res. 2012; 29: 651-668
        • Imawaka H.
        • Ito K.
        • Kitamura Y.
        • Sugyama K.
        • Sugiyama Y.
        Prediction of human bioavailability from human oral administration data and animal pharmacokinetic data without data from intravenous administration of drugs in humans.
        Pharm Res. 2009; 26: 1881-1889
        • Cyong J.C.
        • Kodama K.
        • Yafune A.
        • Takebe M.
        • Takayanagi H.
        Phase I study of azithromycin single dose and multiple-dose for 3 days-.
        Jpn J Chemother. 1995; 43: 139-163
        • Nakashima M.
        • Kanamaru M.
        • Takiguchi Y.
        • Mizuno A.
        • Watanabe I.
        Phase I study of betaxolol hydrochloride (MCI-144).
        J Clin Ther Med. 1989; 5: 1349-1382
        • Nilsen O.G.
        • Dale O.
        • Husebø B.
        Pharmacokinetics of trazodone during multiple dosing to psychiatric patients.
        Pharmacol Toxicol. 1993; 72: 286-289
        • Sahajwalla C.G.
        • Ayres J.W.
        Multiple-dose acetaminophen pharmacokinetics.
        J Pharm Sci. 1991; 80: 855-860
        • Bornemann L.D.
        • Min B.H.
        • Crews T.
        • Rees M.M.
        • Blumenthal H.P.
        • Colburn W.A.
        • et al.
        Dose dependent pharmacokinetics of midazolam.
        Eur J Clin Pharmacol. 1985; 29: 91-95
        • Lacey L.F.
        • Hussey E.K.
        • Fowler P.A.
        Single dose pharmacokinetics of sumatriptan in healthy volunteers.
        Eur J Clin Pharmacol. 1995; 47: 543-548
        • Kato M.
        Intestinal first-pass metabolism of CYP3A4 substrates.
        Drug Metab Pharmacokinet. 2008; 23: 87-94
        • Rowland Yeo K.
        • Rostami-Hodjegan A.
        • Tucker G.T.
        Abundance of cytochromes P450 in human liver: a meta-analysis. 225P GKT, University of London Winter meeting December 2003.
        (Available at:)
        • Iwatsubo T.
        • Hirota N.
        • Ooie T.
        • Suzuki H.
        • Shimada N.
        • Chiba K.
        • et al.
        Prediction of in vivo drug metabolism in the human liver from in vitro metabolism data.
        Pharmacol Ther. 1997; 73: 147-171
        • Davies B.
        • Morris T.
        Physiological parameters in laboratory animals and humans.
        Pharm Res. 1993; 10: 1093-1095
        • Takusagawa S.
        • Ushigome F.
        • Nemoto H.
        • Takahashi Y.
        • Li Q.
        • Kerbusch V.
        • et al.
        Intestinal absorption mechanism of mirabegron, a potent and selective β₃-adrenoceptor agonist: involvement of human efflux and/or influx transport systems.
        Mol Pharm. 2013; 10: 1783-1794
      1. Common technical document of maraviroc for new drug application in Japan.
        (Available at:)
        • Walker D.K.
        • Abel S.
        • Comby P.
        • Muirhead G.J.
        • Nedderman A.N.R.
        • Smith D.A.
        Species differences in the disposition of the CCR5 antagonist, UK-427,857, a new potential treatment for HIV.
        Drug Metab Dispos. 2005; 33: 587-595
        • Kimoto E.
        • Vourvahis M.
        • Scialis R.J.
        • Eng H.
        • Rodorigues A.D.
        • Varma M.V.S.
        Mechanistic evaluation of the complex drug-drug interactions of maraviroc: contribution of cytochrome P450 3A, P-glycoprotein and organic anion transporting polypeptide 1B1.
        Drug Metab Dispos. 2019; 47: 493-503
      2. Common technical document of voriconazole for new drug application in Japan.
        (Available at:) (In Japanese)
        • Murayama N.
        • Imai N.
        • Nakane T.
        • Shimizu M.
        • Yamazaki H.
        Roles of CYP3A4 and CYP2C19 in methyl hydroxylated and N-oxidized metabolite formation from voriconazole, a new anti-fungal agent, in human liver microsomes.
        Biochem Pharmacol. 2007; 73: 2020-2026
        • Damle B.
        • Varma M.V.
        • Wood N.
        Pharmacokinetics of voriconazole administered concomitantly with fluconazole and population-based simulation for sequential use.
        Antimicrob Agents Chemother. 2011; 55: 5172-5177
        • Hyland R.
        • Jones B.C.
        • Smith D.A.
        Identification of the cytochrome P450 enzymes involved in the N-oxidation of voriconazole.
        Drug Metab Dispos. 2003; 31: 540-547
        • Qi F.
        • Zhu L.
        • Li Na
        • Ge T.
        • Xu G.
        • Liao S.
        Influence of different proton pump inhibitors on the pharmacokinetics of voriconazole.
        Int J Antimicrob Agents. 2017; 49: 403-409
        • Suzuki R.
        • Satoh H.
        • Ohtani H.
        • Hori S.
        • Sawada Y.
        Saturable binding of finasteride to steroid 5α-reductase as determinant of nonlinear pharmacokinetics.
        Drug Metab Pharmacokinet. 2010; 25: 208-213
        • Levy G.
        Pharmacologic target-mediated drug disposition.
        Clin Pharmacol Ther. 1994; 56: 248-252
        • Mager D.E.
        • Jusko W.J.
        General pharmacokinetic model for drugs exhibiting target-mediated drug disposition.
        J Pharmacokinet Pharmacodyn. 2001; 28: 507-532
        • Tseng E.
        • Fate G.D.
        • Walker G.S.
        • Goosen T.C.
        • Obach R.S.
        Biosynthesis and identification of metabolites of maraviroc and their use in experiments to delineate the relative contributions of cytochrome P450 3A4 versus 3A5.
        Drug Metab Dispos. 2018; 46: 493-502
        • Hyland R.
        • Dickins M.
        • Collins C.
        • Jones H.
        • Jones B.
        Maraviroc: in vitro assessment of drug-drug interaction potential.
        Br J Clin Pharmacol. 2008; 66: 498-507
        • Lu Y.
        • Hendrix C.W.
        • Bumpus N.N.
        Cytochrome P450 3A5 plays a prominent role in the oxidative metabolism of the anti-human immunodeficiency virus drug maraviroc.
        Drug Metab Dispos. 2012; 40: 2221-2230
        • Roffey S.J.
        • Cole S.
        • Comby P.
        • Gibson D.
        • Jezequel S.G.
        • Nedderman A.N.R.
        • et al.
        The disposition of voriconazole in mouse, rat, rabbit, guinea pig, dog, and human.
        Drug Metab Dispos. 2003; 31: 731-741