Curr Fungal Infect Rep (2014) 8:67–71 DOI 10.1007/s12281-014-0174-1

PEDIATRIC FUNGAL INFECTIONS (T LEHRNBECHER, SECTION EDITOR)

Physiology-Based Pharmacokinetic Modeling—Promise for Pediatric Drug Development? Georg Hempel

Published online: 21 January 2014 # Springer Science+Business Media New York 2014

Abstract Physiologically based modeling has gained much interest from pharmaceutical industry. Prediction of pharmacokinetic properties based on physicochemical properties and in vitro data appears possible. Two applications are of high interest: the prediction of the pharmacokinetics before conducting first in man studies based on preclinical data, and the prediction of pharmacokinetics in children based on the pharmacokinetics in adults. To date, only a few investigations describing the prediction of the pharmacokinetics in children have been published. Some of these investigations showed data on the precision of these predictions by comparing the model with experimental pharmacokinetic data form clinical investigations in children. However, the method holds the promise of speeding up drug development in children, by rationalizing study planning and thus avoiding unnecessary clinical studies. This would ultimately save costs and more importantly, reduce the risks for children in clinical studies. Unfortunately, most of the work done in this field is not published, as investigations are conducted by the pharmaceutical industry during drug development. Therefore, it is difficult to assess the success rate of this approach. However, as practical experience is gained and knowledge on drug-metabolizing enzymes and drug transporters increases, the value of this approach will probably increase in the next few years. Keywords Physiologically-based pharmacokinetic modeling . Pediatrics . Children . Drug development . Pharmaceuticals . Pharmacokinetic . Pharmacodynamic

G. Hempel (*) Institut für Pharmazeutische und Medizinische Chemie, Klinische Pharmazie, Westfälische Wilhelms-Universität Münster, Corrensstraße 48, 48149 Münster, Germany e-mail: [email protected]

Introduction For many years, drug development in children has gained very limited interest in the pharmaceutical industry. This was mainly due to the low number of pediatric patients, and therefore, to limited sales expectancy for drugs in children. In 1997, the Food and Drug Administration (FDA), and later the European Medicines Agency (EMA), started initiatives to support drug development in children by regulations offering 6 months longer market exclusivity for drugs if a pediatric investigation plan was developed [1]. In the following years, there has been more interest, especially in the pharmaceutical industry, in methods to predict the pharmacokinetics and pharmacodynamics in children from experience in adults. Physiologically based pharmacokinetic modeling (PbPk) is a method initially developed more than 80 years ago [2]. For many years, the method was mainly applied in toxicology to estimate the metabolic fate of pollutants in the body of different species [3]. In the field of toxicology, only limited data on plasma or tissue concentrations of the substances of interest are available, and extrapolation from estimated intake of substances to tissue concentrations is desirable. The interest for PbPk methods in drug development increased with the finding that modeling and simulation can be applied to increase the limited success in drug development [4]. Failure in late stages of drug development or early withdrawal of newer drugs from the market can be severe for drug companies, given the fact that the average costs for developing a new drug have risen up to 3.8 Billion € (5 Billion US $) [5]. In most industries, simulations are a key tool in the development of new products (airplanes, cars etc.). In drug development, modeling and simulation developed very late in comparison to other industry areas [6]. This is due to limited activities in pharmaceutical research in the area of modeling and simulation (M & S). In pharmacokinetics, scientists in the pharmaceutical industry and the regulatory authorities mostly relied on traditional

68

noncompartmental analyses, because these methods are believed to be less prone to misinterpretations. The disadvantage of noncompartmental analyses is that they only describe the data. Any extrapolations or predictions from these analyses are not suitable, for example predictions of plasma concentrations when increasing or decreasing the dose or changing the schedule of administration. Therefore, this method lacks “predictive performance” [7]. Model-based approaches in pharmacokinetics and also in pharmacodynamics have been around for many years. The pediatrician Dost published his textbook on pharmacokinetics as early as 1953 [8]. Usually, simple structural models, mostly one to three compartmental models, are applied. Compartments are characterized by their volumes and elimination processes, typically following first-order or MichaelisMenten kinetics. The compartments do not necessarily represent physiologically defined spaces in the body. In the simplest scenario, the compartment represents the whole body, for example, for a hydrophilic drug distributing in total body water. If glomerular filtration rate is the main elimination process, it can usually be described by a first-order process. For example, the decline in the plasma concentration can often be described as: . dc dt ¼ k c… with c describing the plasma concentration, t the time after administration, and k as the elimination constant. Lipophilic drugs often bind to proteins, membranes and/or tissues, and are eliminated more slowly from their binding partners. Elimination often occurs by drug metabolism through P450 enzymes and active transport into the bile by transporters of the ABC family, such as P-glycoprotein (PGP, now termed ABCB1). For such drugs, multi-compartment models with both linear and nonlinear elimination processes are required to adequately describe the pharmacokinetics. A good example is voriconazole, a drug metabolized by P450 enzymes. The description of its pharmacokinetic behavior requires a twocompartment model with first-order absorption, and both linear and saturable (nonlinear) elimination processes are required to adequately describe the kinetics of the drug [9]. Population pharmacokinetics came up in the 1980s and showed its great potential in analyzing pharmacokinetic data collected via clinical routine [10]. The application in drug development of pharmacokinetic and pharmacodynamic data from clinical studies became very popular, and is now recommended in the respective guidelines of the FDA and EMA [11]. In addition to the features of compartmental, modeldependent analysis, population pharmacokinetics also describes the variability in the pharmacokinetic parameters both inter-individually and intra-individually. Having both the pharmacokinetic parameters and their variability, simulations of different dosing

Curr Fungal Infect Rep (2014) 8:67–71

schedules can be conducted that cannot be done using classical noncompartmental analyses. By assessing variability, probabilities of achieving certain therapeutic goals can be calculated [12]. Thus, in population pharmacokinetics, for every drug, the simplest model adequately describing the fate of the drug should be used in order to get the best precision in the estimates of the pharmacokinetic parameters. In contrast, PbPk holds the promise to develop a generic model for all drugs or xenobiotics representing the body as a whole (whole-body PbPk). Only some adjustments to the generic model need to be applied, depending on the physicochemical properties of the drug and the contribution of certain drug-metabolizing enzymes or transporters. In the following section, we will describe this in more detail.

Description of the Method In a whole-body PbPk model, all important organs are represented by a separate compartment connected by a certain blood flow. Each organ is subdivided into subcompartments, i.e. intravasal, interstitial and intercellular space. Mass transport between the sub-compartments is mainly determined by the lipophilicity of the substance of interest. In addition to passive diffusion, active transport processes can be implemented, and, if available, data from in vitro experiments, such as Vmax and Km values, can be used. Absorption from the gastrointestinal tract is estimated based on acid-base equilibrium quantified as pKa values. In addition, the water solubility, as well as the size of the molecule calculated from its molecular weight and the number of halogen atoms, is taken into account for predicting its membrane permeability and, finally, absorption and distribution processes. Elimination is determined either by first-order processes—for example, glomerular filtration—or Michaelis-Menten kinetics for drug-metabolizing enzymes. As clearance of the drug cannot be estimated from its structure, information on intrinsic or overall clearance processes must be available, at least from in vitro or animal experiments Figs. 1 and 2. The predictions rely on a database where average organ weights, blood flow and water and lipid contents of the body are stored. Based on gender, body weight and age, values for these parameters are estimated and used for the predictions of the individual pharmacokinetics. This a priori information enables a more mechanistic insight into the fate of a substance in the body than classical compartmental analysis. The latter approach usually does not allow the gaining of knowledge regarding the mechanisms involved in distribution and elimination. Identifying deviations from the expected pharmacokinetics can be used to find other elimination or distribution processes (i.e. active transport, drug-metabolizing enzymes). Variability in a population of patients is estimated based on the heterogeneous distribution of the organ weight and blood flow

Curr Fungal Infect Rep (2014) 8:67–71

69

adults should be validated using individual pharmacokinetic data from a data set not used for the development of the model. Subsequently, this model can then be adapted for children by scaling down anatomical and physiological processes, as well as organ functions relevant for the Pk of the drug.

Applications A recent MEDLINE search on application of PbPk for children found only very few applications. However, it must be noted that most of the applications of PbPk in pediatric drug development are done during early drug development and are published, if at all, later. Notably, there are more reviews and comments on this topic than there are original articles, indicating the high interest in this area. In addition, it is likely that only applications with at least reasonable results regarding the agreement with experimental data are presented in the literature.

Examples for PbPk Models Various Drugs

Fig. 1 Schematic representation of a whole-body physiologically based model [23]

parameters within a population stored in a database. However, it must be pointed out that variability is mostly underestimated by PbPk models, because for most drugs, not all sources of variability are known. In addition, PbPk models cannot account for intra-individual variability in pharmacokinetics. Two scenarios are of special interest during drug development by the pharmaceutical industry: & &

prediction of the pharmacokinetics in man based on the physicochemical properties of a drug and data from in vitro and animal data prediction of the pharmacokinetics in children based on data from investigations in adults

The latter application is also of high interest in clinical pediatrics, given frequent off-label use. The method offers the chance to rationalize dose calculations in children beyond simple allometric scaling approaches. Current guidelines of the FDA and the EMA recommend PbPk methods for the development of pediatric investigation plans (PIPs). Edginton et al. suggested a workflow for developing a PbPk model for children [13, 14]. Ideally, sufficient data are available from adults to develop and evaluate a PbPk model for this patient group. Whenever possible, the PbPk model for

Edginton et al. presented the development of a generic PbPk model collecting data bodyweight, height, organ weights, blood flows, interstitial space and vascular space for different age groups from the literature [14]. In addition, data on the age-dependency of cardiac output, portal vein flow, extracellular water, total body water, lipid and protein content were collected and implemented into a database. Using PkSim®, predictions from the model drugs acetaminophen, alfentanil, morphine, theophylline and levofloxacin were analyzed, and showed sufficient predictions of the pharmacokinetics of the drugs. Similarly, Johnson presented simulations for eleven drugs (midazolam, caffeine, carbamazepine, cisapride, theophylline, diclofenac, omeprazole, S-warfarin, phenytoin, gentamicin and vancomycin) using SymCyp® over the age range of birth to 18 years [15]. Functions describing the ontogeny of drug-metabolizing enzymes and renal functions are presented. Overall, the authors could predict the pharmacokinetic parameters and their variability sufficiently. Clearance scaling from adults to children using PkSim® was presented in another work applying the method to six compounds (including caffeine, gentamicin, fentanyl and midazolam) with good results, including variability of the clearance [16]. Oseltamivir Parrott et al. presented a PbPk model for oseltamivir [17]. They used animal data, including data from newborn monkeys, to set up the model. Afterwards, the model was applied

70

Curr Fungal Infect Rep (2014) 8:67–71

Fig. 2 Workflow for the development of a PbPk model in children (modified from Edginton et al. [13])

Anatomy, Physiology (blood flow etc.) ADME data (in-vitro, animals)

Physicochemical

PbPk Model (Adults) Model Refinement

Model Evaluation

Pharmacokinetic data

PbPk Population Model (Adults)

Scaling : Protein binding, clearance physiology

PbPk Model (Children)

Model Refinement

to data from humans, including a comparison of the simulations with data from studies in children after intravenous and oral administration of oseltamivir. The model also included the active metabolite oseltamivir-carboxylate and provided reasonable predictions for children and neonates. Overall, the simulations matched the experimental data sufficiently. However, as with the other models presented here, the models cannot be regarded as “validated” in a sense that the predictions are absolutely reliable. It is questionable if pharmacokinetic models can be validated at all [18]. Model evaluation with an independent data set not used for model development is at least the best.

Model evaluation

Pharmacokinetic data

Acetaminophen Jiang et al. [20] used SymCyp® to develop a model for acetaminophen including the main metabolic pathways of glucuronidation, sulfation and oxidative toxification to the N-acetyl-p-benzoquinone imine. Model development was conducted using in vitro data and pharmacokinetic data from adults. Subsequently, the model was modified to account for maturational changes in children, and the predictions were compared with plasma levels and urinary excretion from the literature. The model sufficiently reflected the observed data in neonates, infants, children and adolescents following i.v. and oral administration.

Etoposide Lorazepam Kersting et al. developed a PbPk model for adults using PkSim® and receiving high and low dose etoposide, evaluated this model, and applied it for the prediction of the pharmacokinetics in children [19]. Predictions were sufficiently precise and drug interactions with cyclosporine A on P450 enzymes and PGP transporters could be taken into account. In addition, the reduced renal function due to coadministration of carboplatin was also reflected in the model, and the effect on the plasma concentrations of etoposide can be predicted.

Maharaj et al. [21] from the group of Edginton presented a case study with lorazepam showing estimates of clearance and volume of distribution in the range of up to twofold deviation from the experimental data in children. Input parameters such as lipophilicity and intrinsic clearances were iteratively optimized to reflect the experimental data in adults. Subsequently, the predictions were evaluated using data from 63 children. It is noted that individualized predictions generated through

Curr Fungal Infect Rep (2014) 8:67–71

PbPk modeling requires further improvement, while the predictions of the average pharmacokinetic parameters are sufficient. Rivaroxaban Willmann et al. presented a PbPk model for rivaroxaban, an oral anticoagulant drug [22]. For the clearance of rivaroxaban, metabolism by cytochrome P450 3A4/5 and CYP2J2, as well as CYP-independent hydrolysis of rivaroxaban, was taken into account. Renal clearance included the glomerular filtration rate (GFR) and tubular secretion. Data from mass balance studies were also used. In adults, the model shows good agreement with experimental data and their variability. The predictions of the model for children await confirmation by a current phase I study in children, and serve as a hypothesis for these pharmacokinetic investigations in children.

Conclusion PbPk has the potential to rationalize drug development in children. As the published applications of PbPk in children are limited, it is difficult to assess how successful the approach is in predicting the pharmacokinetics from in vitro data or in predicting pharmacokinetics in children based on adult data. However, given the high interest in systems biology and the potential financial benefits of such in silico methods, it is of high probability that the method will further improve. It is desirable that pharmacokinetics can be predicted not only for the average child, but also for patients in special situations such as overweight, cachectic patients, the critically ill, or patients after bone marrow transplantation. In addition, several examples from investigations in adults show that the impact of drug interactions as well as genetic polymorphism in drugmetabolizing enzymes can be assessed using PbPk. In the future, PbPk might even be used for dose individualization. Compliance with Ethics Guidelines Conflict of Interest G Hempel receives financial support from Bayer Technology Services. Human and Animal Rights and Informed Consent This article does not contain any studies with human or animal subjects performed by any of the authors.

References 1. Schachter AD, Ramoni MF. Paediatric drug development. Nat Rev Drug Discov. 2007;6:429–30. 2. Haggard H. The Absorption, Distribution, and Elimination of Ethyl Ether. J Biol Chem. 1924;59:753–70.

71 3. El-Masri HA. Experimental and mathematical modeling methods for the investigation of toxicological interactions. Toxicol Appl Pharmacol. 2007;223:148–54. 4. Vozeh S, Steimer JL, Rowland M, Morselli P, Mentre F, Balant LP. Aarons L The use of population pharmacokinetics in drug development. Clin Pharmacokinet. 1996;30:81–93. 5. Desmond-Hellmann S. The Cost Of Creating A New Drug Now $5 Billion, Pushing Big Pharma To Change. Forbes 11th of August, 2013 http://www.forbes.com/sites/matthewherper/2013/08/11/howthe-staggering-cost-of-inventing-new-drugs-is-shaping-the-futureof-medicine/ Accessed 16th of Sept 2013 6. Bonate PL. Clinical trial simulation in drug development. Pharm Res. 2000;17:252–6. 7. Gibaldi M, Perrier D. Pharmacokinetics. 2nd ed. New York: Marcel Dekker; 1982. 8. Der DF. Blutspiegel. Germany: Thieme Verlag Stuttgart; 1953. 9. Theuretzbacher U, Ihle F, Derendorf H. Pharmacokinetic/ Pharmacodynamic Profile of Voriconazole. Clin Pharmacokinet. 2006;45:649–63. 10. Sheiner LB, Beal S. Evaluation of methods for estimating population pharmacokineticp. I. Michaelis-menten Model: Routine clinical pharmacokinetic data. J Pharmacokinet Biopharm. 1980;8:553–71. 11. European Medicines Agency. Guidelines of reporting the results of population pharmacokinetic analyses. July 2007 http://www.ema. europa.eu/ema/index.jsp?curl=pages/regulation/general/general_ content_000370.jsp&mid=WC0b01ac0580032ec5 assessed 16th of Sept 2013 12. Holford N, Ma SC, Ploeger BA. Clinical trial simulation: a review. Clin Pharmacol Ther. 2010;88:166–82. 13. Edginton AN. Knowledge-driven approaches for the guidance of first-in-children dosing. Paediatr Anaesth. 2011;21:206–13. 14. Edginton AN, Schmitt W, Willmann S. Development and evaluation of a generic physiologically based pharmacokinetic model for children. Clin Pharmacokinet. 2006;45:1–22. 15. Johnson TN, Rostami-Hodjegan A, Tucker GT. Prediction of the clearance of eleven drugs and associated variability in neonates. Infants and Child Clin Pharmacokinet. 2006;45:931–56. 16. Edginton AN, Schmitt W, Voith B, Willmann S. A Mechanistic Approach for the Scaling of Clearance in Children. Clin Pharmacokinet. 2006;45(7):683–704. 17. Parrott N, Davies B, Hoffmann G, Koerner A, Lave T, Prinssen E, et al. Development of a physiologically based model for oseltamivir and simulation of pharmacokinetics in neonates and infants. Clin Pharmacokinet. 2011;50:613–23. 18. Sherwin CM, Kiang TK, Spigarelli MG, Ensom MH. Fundamentals of population pharmacokinetic modelling: validation methods. Clin Pharmacokinet. 2012;51:573–90. 19. Kersting G, Willmann S, Würthwein G, Lippert J, Boos J, Hempel G. Physiologically based pharmacokinetic modelling of high- and lowdose etoposide: from adults to children. Cancer Chemother Pharmacol. 2012;69:397–405. 20. Jiang XL, Zhao P, Barrett JS, Lesko LJ, Schmidt S. Application of physiologically based pharmacokinetic modeling to predict acetaminophen metabolism and pharmacokinetics in children. CPT Pharmacometrics Syst Pharmacol. 2013;2:e80. doi:10.1038/psp. 2013.55. 21. Maharaj AR, Barrett JS, Edginton AN. A workflow example of PBPK modeling to support pediatric research and development: case study with lorazepam. AAPS J. 2013;15:455–64. 22. Willmann S, Becker C, Burghaus R, Coboeken K, Edginton A, Lippert J et al. Development of a Paediatric Population-Based Model of the Pharmacokinetics of Rivaroxaban. Clin Pharmacokinet. 2013 Aug 3. [Epub ahead of print]. 23. Dickschen K. Overcoming CYP2D6-mediated tamoxifen resistance in postmenopausal breast cancer: a PbPk approach. Ph.D thesis University of Münster, Germany 2013.