Prediction of drug‐drug interactions using physiologically‐based pharmacokinetic models of CYP450 modulators included in Simcyp software

In recent years, physiologically based PharmacoKinetic (PBPK) modeling has received growing interest as a useful tool for the assessment of drug pharmacokinetics. It has been demonstrated to be informative and helpful to quantify the modification in drug exposure due to specific physio‐pathological conditions, age, genetic polymorphisms, ethnicity and particularly drug–drug interactions (DDIs). In this paper, the prediction success of DDIs involving various cytochrome P450 isoenzyme (CYP) modulators namely ketoconazole (a competitive inhibitor of CYP3A), itraconazole (a competitive inhibitor of CYP3A), clarithromycin (a mechanism‐based inhibitor of CYP3A), quinidine (a competitive inhibitor of CYP2D6), paroxetine (a mechanism‐based inhibitor of CYP2D6), ciprofloxacin (a competitive inhibitor of CYP1A2), fluconazole (a competitive inhibitor of CYP2C9/2C19) and rifampicin (an inducer of CYP3A) were assessed using Simcyp^® software. The aim of this report was to establish confidence in each CYP‐specific modulator file so they can be used in the future for the prediction of DDIs involving new victim compounds. Our evaluation of these PBPK models suggested that they can be successfully used to evaluate DDIs in untested scenarios. The only noticeable exception concerned a quinidine inhibitor model that requires further improvement. Additionally, other important aspects such as model validation criteria were discussed.     

儿童生理药代动力学模型及其在儿科药物研究中的应用

生理药代动力学(physiologically based pharmacokinetic, PBPK)模型是预测药物在特殊人群中的药代动力学、药效学和安全性的重要工具。尤其对于儿童这类不易开展临床试验的人群, PBPK模型的应用更是能有效促进儿科药物的开发以及儿童的临床用药。目前, PBPK模型在儿科药物开发中的主要应用有以下几种:临床试验设计、药物相互作用(drug-drug interaction, DDI)的风险评估和儿童给药剂量的确立等。本综述简介了儿童生理药动学模型在儿科药物研究中的优越性,总结了PBPK模型如何实现从成人到儿童的外推,儿童生理药动学模型的理论基础,建模过程及所要注意的重要生理参数,列举了目前PBPK模型在儿科药物研究中的一些应用实例。最后简述了儿童PBPK模型当前的局限性和未来发展方向。     

A physiologically based pharmacokinetic modeling approach to predict drug–drug interactions between domperidone and inhibitors of CYP3A4

Domperidone is a dopamine receptor antagonist and a substrate of CYP3A4, hence there is a potential for CYP3A inhibition‐based drug–drug interactions (DDI). A physiologically based pharmacokinetic model was developed to describe DDIs between domperidone and three different inhibitors of CYP3A4. Simcyp V13.1 was used to simulate human domperidone pharmacokinetics and DDIs. Inputs included domperidone chemical and physical properties (LogP, pKa, etc.), in vitro human liver microsomal data and pharmacokinetic parameters from single‐dose intravenous clinical studies in healthy participants. The simulated mean maximum domperidone plasma concentration and AUC after single‐ and multiple‐oral doses under diverse conditions were within 1.1–1.4 fold of the observed values. The simulated intestinal availability, hepatic availability and the fraction absorbed were 0.45 ± 0.14, 0.31 ± 0.10 and 0.89 ± 0.11, respectively, and comparable to observed in vivo values. The simulated ratios of AUC and C_max in the presence of ketoconazole, erythromycin or itraconazole to baseline were consistent with the observed ratios. Simulated ketoconazole, erythromycin, itraconazole and C_max,ss and AUC_ss were within 1.5‐fold of the observed values. Copyright © 2016 John Wiley & Sons, Ltd.     

Dealing with the complex drug–drug interactions: Towards mechanistic models

Unmanageable severe adverse events caused by drug‐drug interactions (DDIs), leading to market withdrawals or restrictions in the clinical usage, are increasingly avoided with the improvement in our ability to predict such DDIs quantitatively early in drug development. However, significant challenges arise in the evaluation and/or prediction of complex DDIs caused by inhibitor drugs and/or metabolites that affect not one but multiple pathways of drug clearance. This review summarizes the discussion topics at the 2013 AAPS symposium on “Dealing with the complex drug‐drug interactions: towards mechanistic models”. Physiologically based pharmacokinetic (PBPK) models, in combination with the established in vitro‐to‐in vivo extrapolations of intestinal and hepatic disposition, have been successfully applied to predict clinical pharmacokinetics and DDIs, especially for drugs with CYP‐mediated metabolism, and to explain transporter‐mediated and complex DDIs. Although continuous developments are being made towards improved mechanistic prediction of the transporter‐enzyme interplay in the hepatic and intestinal disposition and characterizing the metabolites contribution to DDIs, the prediction of DDIs involving them remains difficult. Regulatory guidelines also recommended use of PBPK modeling for the quantitative prediction and evaluation of DDIs involving multiple perpetrators and metabolites. Such mechanistic modeling approaches culminate to the consensus that modeling is helpful in predicting DDIs or quantitatively rationalizing the clinical findings in complex situations. Furthermore, they provide basis for the prediction and/or understanding the pharmacokinetics in populations like patients with renal impairment, pediatrics, or various ethnic groups where the conduct of clinical studies might not be feasible in early drug development stages and yet some guidance on management of dosage is necessary. Copyright © 2014 John Wiley & Sons, Ltd.     

Verification of a physiologically based pharmacokinetic model of ritonavir to estimate drug–drug interaction potential of CYP3A4 substrates

Ritonavir is one of several ketoconazole alternatives used to evaluate strong CYP3A4 inhibition potential in clinical drug–drug interaction (DDI) studies. In this study, four physiologically based pharmacokinetic (PBPK) models of ritonavir as an in vivo time‐dependent inhibitor of CYP3A4 were created and verified for oral doses of 20, 50, 100 and 200 mg using the fraction absorbed (F_a) and oral clearance (CL_oral) values reported in the literature, because transporter and CYP enzyme reaction phenotyping data were not available. The models were used subsequently to predict and compare the magnitude of the AUC increase in nine reference DDI studies evaluating the effect of ritonavir at steady‐state on midazolam (CYP3A4 substrate) exposure. Midazolam AUC and C_max ratios were predicted within 2‐fold of the respective observations in seven studies. Simulations of the hepatic and gut CYP3A4 abundance after multiple oral dosing of ritonavir indicated that a 3‐day treatment with ritonavir 100 mg twice daily is sufficient to reach maximal CYP3A4 inhibition and subsequent systemic exposure increase of a CYP3A4 substrate, resulting in the reliable estimation of f_m,CYP3A4. The ritonavir model was submitted as part of the new drug application for Kisqali® (ribociclib) and accepted by health authorities.     

Physiologically based pharmacokinetic modelling to predict exposure differences in healthy volunteers and subjects with renal impairment: Ceftazidime case study

Ceftazidime is a widely used β‐lactam antibiotic and almost entirely excreted via glomerular filtration in kidney. The objective of this analysis was to assess the ability of physiologically based pharmacokinetic (PBPK) model to predict ceftazidime exposure in healthy volunteers and subjects with renal impairment. A full PBPK model of ceftazidime was developed using physiochemical properties and clinical data. The total clearance of 115 mL/min and renal clearance of 100 mL/min were obtained from ceftazidime package insert. Healthy and chronic kidney disease (CKD) populations were applied for sampling of virtual subjects. The established PBPK model predicted mean plasma AUC_inf were 138.5 ± 19.6, 230.7 ± 22.2, 369.3 ± 53.1 and 561.8 ± 92.4 h   µg/mL in healthy, mild, moderate and severe renal impairment subjects, respectively, after administration of 1 g ceftazidime intravenous bolus dose. The predicted values were in close agreement with the weighted mean of the five reported clinical studies. The exposure was slightly under predicted in subjects with severely impaired renal function, but still within 1.5‐fold range. The concentration‐time profiles of ceftazidime were also well captured in healthy volunteers and subjects with renal impairment. The developed PBPK model along with systems pharmacokinetics (PK) (renal impaired populations) well predicted the ceftazidime exposure. PBPK models verified with clinical study in healthy volunteers could be potentially applied to predict PK and recommend dose adjustment for CKD patients.     

The application of physiologically based pharmacokinetic modelling to assess the impact of antiretroviral‐mediated drug–drug interactions on piperaquine antimalarial therapy during pregnancy

Antimalarial therapy during pregnancy poses important safety concerns due to potential teratogenicity and maternal physiological and biochemical changes during gestation. Piperaquine (PQ) has gained interest for use in pregnancy in response to increasing resistance towards sulfadoxine–pyrimethamine in sub‐Saharan Africa. Coinfection with HIV is common in many developing countries, however, little is known about the impact of antiretroviral (ARV) mediated drug–drug interaction (DDI) on piperaquine pharmacokinetics during pregnancy. This study applied mechanistic pharmacokinetic modelling to predict pharmacokinetics in non‐pregnant and pregnant patients, which was validated in distinct customised population groups from Thailand, Sudan and Papua New Guinea. In each population group, no significant differences in day 7 concentrations were observed during different gestational weeks (GW) (weeks 10–40), supporting the notion that piperaquine is safe throughout pregnancy with consistent pharmacokinetics, although possible teratogenicity may limit this. Antiretroviral‐mediated DDIs (efavirenz and ritonavir) had moderate effects on piperaquine during different gestational weeks with a predicted AUC_ratio in the range 0.56–0.8 and 1.64–1.79 for efavirenz and ritonavir, respectively, over GW 10–40, with a reduction in circulating human serum albumin significantly reducing the number of subjects attaining the day 7 (post‐dose) therapeutic efficacy concentrations under both efavirenz and ritonavir DDIs. This present model successfully mechanistically predicted the pharmacokinetics of piperaquine in pregnancy to be unchanged with respect to non‐pregnant women, in the light of factors such as malaria/HIV co‐infection. However, antiretroviral‐mediated DDIs could significantly alter piperaquine pharmacokinetics. Further model refinement will include collation of relevant physiological and biochemical alterations common to HIV/malaria patients.     

A physiologically based pharmacokinetic model to predict the pharmacokinetics of highly protein‐bound drugs and the impact of errors in plasma protein binding

Predicting the pharmacokinetics of highly protein‐bound drugs is difficult. Also, since historical plasma protein binding data were often collected using unbuffered plasma, the resulting inaccurate binding data could contribute to incorrect predictions. This study uses a generic physiologically based pharmacokinetic (PBPK) model to predict human plasma concentration–time profiles for 22 highly protein‐bound drugs. Tissue distribution was estimated from in vitro drug lipophilicity data, plasma protein binding and the blood: plasma ratio. Clearance was predicted with a well‐stirred liver model. Underestimated hepatic clearance for acidic and neutral compounds was corrected by an empirical scaling factor. Predicted values (pharmacokinetic parameters, plasma concentration–time profile) were compared with observed data to evaluate the model accuracy. Of the 22 drugs, less than a 2‐fold error was obtained for the terminal elimination half‐life (t_1/2, 100% of drugs), peak plasma concentration (C_max, 100%), area under the plasma concentration–time curve (AUC_0‐t, 95.4%), clearance (CL_h, 95.4%), mean residence time (MRT, 95.4%) and steady state volume (V_ss, 90.9%). The impact of f_up errors on CL_h and V_ss prediction was evaluated. Errors in f_up resulted in proportional errors in clearance prediction for low‐clearance compounds, and in V_ss prediction for high‐volume neutral drugs. For high‐volume basic drugs, errors in f_up did not propagate to errors in V_ss prediction. This is due to the cancellation of errors in the calculations for tissue partitioning of basic drugs. Overall, plasma profiles were well simulated with the present PBPK model. Copyright © 2016 John Wiley & Sons, Ltd.     

Physiologically based pharmacokinetic modeling revealed minimal codeine intestinal metabolism in first‐pass removal in rats

The physiologically based model with segregated flow to the intestine (SFM‐PBPK; partial, lower flow to enterocyte region vs. greater flow to serosal region) was found to describe the first‐pass glucuronidation of morphine (M) to morphine‐3β‐glucuronide (MG) in rats after intraduodenal (i.d.) and intravenous (i.v.) administration better than the traditional model (TM), for which a single intestinal flow perfused the whole of the intestinal tissue. The segregated flow model (SFM) described a disproportionately greater extent of intestinal morphine glucuronidation for i.d. vs. i.v. administration. The present study applied the same PBPK modeling approaches to examine the contributions of the intestine and liver on the first‐pass metabolism of the precursor, codeine (C, 3‐methylmorphine) in the rat. Unexpectedly, the profiles of codeine, morphine and morphine‐3β‐glucuronide in whole blood, bile and urine, assayed by LCMS, were equally well described by both the TM‐PBPK and SFM‐PBPK. The fitted parameters for the models were similar, and the net formation intrinsic clearance of morphine (from codeine) for the liver was much higher, being 9‐ to 13‐fold that of the intestine. Simulations, based on the absence of intestinal formation of morphine, correlated well with observations. The lack of discrimination of SFM and TM with the codeine data did not invalidate the SFM‐PBPK model but rather suggests that the liver is the only major organ for codeine metabolism. Because of little or no contribution by the intestine to the metabolism of codeine, both the TM‐ and SFM‐PBPK models are equally consistent with the data. Copyright © 2016 John Wiley & Sons, Ltd.     

Direct and metabolism‐dependent cytochrome P450 inhibition assays for evaluating drug–drug interactions

We developed methods for evaluating the ntial inhibition of human cytochrome P450 (CYP) enzymes, including CYP1A2, CYP2A6, CYP2B6, CYP2 C9, CYP2 C19, CYP2D6, CYP2E1 and CYP3A4, using pooled human liver microsomes (HLMs). The CYP inhibition assay used substrate cocktail sets [set A: phenacetin for CYP1A2, coumarin for CYP2A6, (S)‐(+)‐mephenytoin for CYP2C19, dextromethorphan for CYP2D6 and midazolam for CYP3A4; set B: bupropion for CYP2B6, tolbutamide for CYP2C9, chlorzoxazone for CYP2E1, and testosterone for CYP3A4] with quantitation by liquid chromatography–tandem mass spectrometry. A direct inhibition assay was performed with the substrate cocktails without β‐nicotinamide adenine dinucleotide phosphate (NADPH) pre‐incubation, and a metabolism‐dependent inhibition (MDI) assay was performed after 30 min of pre‐incubation with NADPH in HLMs. MDI was identified based on the half‐maximal inhibitory concentration (IC₅₀) shifts. The IC₅₀ values of the direct inhibitors determined using the probe substrate cocktails were in good agreement with previously reported values. Eight metabolism‐dependent inhibitors including furafylline, 8‐methoxypsoralen, tienilic acid, ticlopidine, fluoxetine, paroxetine, disulfiram and verapamil against CYP1A2, CYP2A6, CYP2B6, CYP2C9, CYP2C19, CYP2D6, CYP2E1 and CYP3A4, respectively, resulted in significant IC₅₀ shifts (≥2.5‐fold) after pre‐incubation. Thus, these CYP inhibition assays are considered to be useful tools for evaluating both direct inhibition and MDI at an early stage of the drug discovery and development process. Copyright © 2011 John Wiley & Sons, Ltd.     

Assessment of cytochrome P450‐mediated drug–drug interaction potential of orteronel and exposure changes in patients with renal impairment using physiologically based pharmacokinetic modeling and simulation

Orteronel is a nonsteroidal, selective inhibitor of 17,20‐lyase that was recently in phase 3 clinical development as a treatment for castration‐resistant prostate cancer. In humans, the primary clearance route for orteronel is renal excretion. Human liver microsomal studies indicated that orteronel weakly inhibits CYP1A2, 2C8, 2C9 and 2C19, with IC₅₀ values of 17.8, 27.7, 30.8 and 38.8 µm , respectively, whereas orteronel does not inhibit CYP2B6, 2D6 or 3A4/5 (IC₅₀ > 100 µm ). Orteronel also does not exhibit time‐dependent inhibition of CYP1A2, 2B6, 2C8, 2C9, 2C19, 2D6 or 3A4/5. The results of a static model indicated an [I]/K_i ratio >0.1 for CYP1A2, 2C8, 2C9 and 2C19. Therefore, a physiologically based pharmacokinetic (PBPK) model was developed to assess the potential for drug–drug interactions (DDIs) between orteronel and theophylline, repaglinide, (S)‐warfarin and omeprazole, which are sensitive substrates of CYP1A2, 2C8, 2C9 and 2C19, respectively. Simulation of the area under the plasma concentration–time curve (AUC) of these four CYP substrates in the presence and absence of orteronel revealed geometric mean AUC ratios <1.25. Therefore, in accordance with the 2012 US FDA Draft Guidance on DDIs, orteronel can be labeled a ‘non‐inhibitor’ and further clinical DDI evaluation is not required. In PBPK models of moderate and severe renal impairment, the AUC of orteronel was predicted to increase by 52% and 83%, respectively. These results are in agreement with those of a clinical trial in which AUC increases of 38% and 87% were observed in patients with moderate and severe renal impairment, respectively. Copyright © 2014 John Wiley & Sons, Ltd.     

Physiologically Based Pharmacokinetic Modeling to Supplement Nilotinib Pharmacokinetics and Confirm Dose Selection in Pediatric Patients

In adult patients, nilotinib is indicated for chronic myeloid leukemia at an approved oral dose of 300 or 400 mg BID. Physiologically based pharmacokinetic (PBPK) model was developed to describe and supplement limited PK data in the pediatric population ranging from 2 to less than 6 years of age and ultimately inform dosing regimen. An adult Simcyp PBPK model was established and verified with clinical pharmacokinetic data after a single or multiple oral doses of 400 mg nilotinib (230 mg/m²). The model was then applied to a pediatric PBPK model, taking account of ontogeny profiles of metabolizing enzymes and pediatric physiological parameters. The model was further verified using observed pediatric PK data in 12- to <18-year-old and from 6- to <12-year-old patients. The PBPK models were able to recover, describe, and supplement the limited nilotinib concentration-time data profile in 2- to <6-year-old patients after a single dose and C_min,ss after BID dosing. The exposure (C_max,ss, C_min,ss, and AUC_tau,ss) was predicted to be similar across age groups. PBPK model simulations confirmed that body surface area–normalized dosing regimen of 230 mg/m² is considered appropriate for pediatric patients >2 to <18 years of age.     

Physiologically based pharmacokinetic model for ethyl tertiary‐butyl ether and tertiary‐butyl alcohol in rats: Contribution of binding to α2u–globulin in male rats and high‐exposure nonlinear kinetics to toxicity and cancer outcomes

In cancer bioassays, inhalation, but not drinking water exposure to ethyl tertiary‐butyl ether (ETBE), caused liver tumors in male rats, while tertiary‐butyl alcohol (TBA), an ETBE metabolite, caused kidney tumors in male rats following exposure via drinking water. To understand the contribution of ETBE and TBA kinetics under varying exposure scenarios to these tumor responses, a physiologically based pharmacokinetic model was developed based on a previously published model for methyl tertiary‐butyl ether, a structurally similar chemical, and verified against the literature and study report data. The model included ETBE and TBA binding to the male rat‐specific protein α2u–globulin, which plays a role in the ETBE and TBA kidney response observed in male rats. Metabolism of ETBE and TBA was described as a single, saturable pathway in the liver. The model predicted similar kidney AUC_0–∞ for TBA for various exposure scenarios from ETBE and TBA cancer bioassays, supporting a male‐rat‐specific mode of action for TBA‐induced kidney tumors. The model also predicted nonlinear kinetics at ETBE inhalation exposure concentrations above ~2000 ppm, based on blood AUC_0–∞ for ETBE and TBA. The shift from linear to nonlinear kinetics at exposure concentrations below the concentration associated with liver tumors in rats (5000 ppm) suggests the mode of action for liver tumors operates under nonlinear kinetics following chronic exposure and is not relevant for assessing human risk. Copyright © 2016 The Authors Journal of Applied Toxicology Published by John Wiley & Sons Ltd     

Estimation of placental and lactational transfer and tissue distribution of atrazine and its main metabolites in rodent dams,fetuses, and neonates with physiologically based pharmacokinetic modeling

Atrazine (ATR) is a widely used chlorotriazine herbicide, a ubiquitous environmental contaminant, and a potential developmental toxicant. To quantitatively evaluate placental/lactational transfer and fetal/neonatal tissue dosimetry of ATR and its major metabolites, physiologically based pharmacokinetic models were developed for rat dams, fetuses and neonates. These models were calibrated using pharmacokinetic data from rat dams repeatedly exposed (oral gavage; 5 mg/kg) to ATR followed by model evaluation against other available rat data. Model simulations corresponded well to the majority of available experimental data and suggest that: (1) the fetus is exposed to both ATR and its major metabolite didealkylatrazine (DACT) at levels similar to maternal plasma levels, (2) the neonate is exposed mostly to DACT at levels two-thirds lower than maternal plasma or fetal levels, while lactational exposure to ATR is minimal, and (3) gestational carryover of DACT greatly affects its neonatal dosimetry up until mid-lactation. To test the model's cross-species extrapolation capability, a pharmacokinetic study was conducted with pregnant C57BL/6 mice exposed (oral gavage; 5 mg/kg) to ATR from gestational day 12 to 18. By using mouse-specific parameters, the model predictions fitted well with the measured data, including placental ATR/DACT levels. However, fetal concentrations of DACT were overestimated by the model (10-fold). This overestimation suggests that only around 10% of the DACT that reaches the fetus is tissue-bound. These rodent models could be used in fetal/neonatal tissue dosimetry predictions to help design/interpret early life toxicity/pharmacokinetic studies with ATR and as a foundation for scaling to humans.     

Predicting points of departure for risk assessment based on in vitro cytotoxicity data and physiologically based kinetic (PBK) modeling: The case of kidney toxicity induced by aristolochic acid I

Aristolochic acids are naturally occurring nephrotoxins. This study aims to investigate whether physiologically based kinetic (PBK) model-based reverse dosimetry could convert in vitro concentration-response curves of aristolochic acid I (AAI) to in vivo dose response-curves for nephrotoxicity in rat, mouse and human. To achieve this extrapolation, PBK models were developed for AAI in these different species. Subsequently, concentration-response curves obtained from in vitro cytotoxicity models were translated to in vivo dose–response curves using PBK model-based reverse dosimetry. From the predicted in vivo dose–response curves, points of departure (PODs) for risk assessment could be derived. The PBK models elucidated species differences in the kinetics of AAI with the overall catalytic efficiency for metabolic conversion of AAI to aristolochic acid Ia (AAIa) being 2-fold higher for rat and 64-fold higher for mouse than human. Results show that the predicted PODs generally fall within the range of PODs derived from the available in vivo studies. This study provides proof of principle for a new method to predict a POD for in vivo nephrotoxicity by integrating in vitro toxicity testing with in silico PBK model-based reverse dosimetry.