Optimization of DDI study design: Comparison of minimal PBPK models on prediction of

Transcription

1 DMD This Fast article Forward. has not been Published copyedited on and formatted. April 12, The 2013 final as version doi: /dmd may differ from this version. Optimization of DDI study design: Comparison of minimal PBPK models on prediction of CYP3A inhibition by ketoconazole. Bing Han, Jialin Mao, Jenny Y. Chien and Stephen D. Hall Department of Drug Disposition, Lilly Research Labs (B.H., J.M., J.Y.C., S.D.H), Eli Lilly and Co., Indianapolis, IN Department of Drug Metabolism and Pharmacokinetics (J.M.), Genentech Inc., South San Francisco, CA d Copyright 2013 by the American Society for Pharmacology and Experimental Therapeutics.

2 Running title: Ketoconazole-midazolam DDI prediction Address correspondence to: Stephen D. Hall, Department of Drug Disposition, Lilly Research Labs, Eli Lilly and Co., Indianapolis, IN, Tel: Fax: Number of text pages: 23 Number of tables: 1 Number of figures: 9 Number of reference: 23 Number of words in the abstract: 250 Number of words in the introduction: 450 Number of words in the discussion: 1426 Abbreviation: AUC: area under the curve; CL int : intrinsic clearance; CYP3A: cytochrome P450 3A; DDIs: drug-drug interactions; PBPK: Physiologically based pharmacokinetic. d

3 Abstract Ketoconazole is a potent CYP3A inhibitor used to assess the contribution of CYP3A to drug clearance and quantify the increase in drug exposure due to a strong inhibitor. Physiologically based pharmacokinetic (PBPK) models have been employed to evaluate dosing regimens resulting in maximal CYP3A inhibition by ketoconazole but have reached different conclusions. We compare two PBPK models of the ketoconazole midazolam interaction, Model 1 (Chien et al., 2006) and Model 2 implemented in Simcyp (version 11) to predict 16 published dosing regimens. Using Model 2, 41% of the study point estimates of AUC ratio and 71% of the 90% confidence intervals were predicted within 1.5- fold of the observed but these increased to 82% and 100% respectively with Model 1. For midazolam, Model 2 predicted a maximal midazolam AUC ratio of 8 and a hepatic fraction metabolized by CYP3A of 0.97 whereas Model 1 predicted 17 and 0.90 respectively, which are more consistent with observed data. Based on Model 1, ketoconazole 400mg QD for at least 3 days and substrate administration within 2 hours is required for maximal CYP3A inhibition. Ketoconazole dosing regimens that employ 200 mg BID underestimate the systemic fraction metabolized by CYP3A (0.86 vs. 0.90) for midazolam. The systematic underprediction also applies to CYP3A substrates with high bioavailability and long half-lives. The superior predictive performance of Model 1 reflects the need for accumulation of ketoconazole at enzyme site and protracted inhibition. Model 2 is not recommended for inferring optimal study design and estimation of fraction metabolized by CYP3A. d

4 Introduction Midazolam is one of the most selective and commonly used probe substrates of CYP3A activity in vitro and in vivo. To estimate the contribution of CYP3A enzymes to systemic clearance and bioavailability, ketoconazole is often employed despite a number of frequently acknowledged drawbacks to this approach (Ohno et al., 2007; Zhao et al., 2009). The most prominent shortcoming of ketoconazole as the preferred in vivo strong inhibitor of CYP3A is the poor understanding of the determinants of inhibition in whole cells and at the enzyme site. Ketoconazole is often described as a reversible inhibitor of CYP3A but several observations suggest that rapid reversal of inhibition does not occur (Chien et al., 2006). Gibbs et al. demonstrated a persistent inhibition of midazolam in CYP3A4-expressing Caco-2 cells even when albumin was present in the basolateral media (Gibbs et al., 2000). In view of the potential for ketoconazole inhibition to be prolonged in vivo it is important that pharmacokinetic models designed to predict in vivo DDIs are validated prior to be used in study design and drug development decision making. We previously described a minimal physiologically based pharmacokinetic (PBPK) model of the ketoconazole-midazolam interaction that included a saturable efflux of ketoconazole from the hepatic enzyme compartment and consequently a duration of inhibition that was prolonged relative to plasma ketoconazole concentrations (Chien et al., 2006). This model will be referred to as Model 1 henceforth. The commonly used PBPK model implemented using the Simcyp platform employs the same minimal PBPK structure but a rapidly reversible inhibition of hepatic CYP3A that parallels ketoconazole plasma concentrations (Guest et al., 2011). This model will be referred to as Model 2 henceforth. Preliminary assessments of the ketoconazole-midazolam interaction using that model concluded that predictive performance was poor (Guest et al., 2011). Despite this conclusion, Model 2 has been used to determine whether static or dynamic plasma concentration approaches are most appropriate for predicting clinically important DDIs (Einolf, 2007; Guest et al., 2011; Peters et al., 2012), validate in vitro technologies (Youdim et al., 2008) and to estimate the fraction metabolized by CYP3A for new drugs following in vivo studies (Rakhit et al., 2008; Yang et al., 2012). Importantly, Model 2 was used by the FDA to recommend specific clinical study designs (Zhao et al., 2009). Specifically, 200 mg d

5 BID ketoconazole was recommended as superior to 400 mg QD for long half-life drugs but equivalent for short half-life drugs and this appeared in the FDA draft guidance on DDIs (FDA, 2012). In this study we determine the predictive performance of Model 1 and Model 2 of the ketoconazolemidazolam interaction for 16 dosage regimens published in literature and provide updated study design recommendations based on the more predictive Model 1. d

6 Materials and Methods Ketoconazole-Midazolam Interaction Models Model 1 We employed the ketoconazole-midazolam interaction model previously described in detail (Chien et al., 2006) and referred to herein as Model 1. In brief, the model is composed of two minimal PBPK models for midazolam and ketoconazole and pools of gut wall and hepatic CYP3A that metabolize midazolam and are inhibited by ketoconazole. The ketoconazole model is a two compartment model with first order absorption. In addition, a hepatic compartment and a hepatic enzyme compartment are included. The midazolam model is a three compartment model with first order absorption. The changes of CL int of midazolam at gut and hepatic site are functions of the concentration of ketoconazole at gut and enzyme site, respectively. The inter-subject and within-subject variability and the effects of gender, age and body weight as covariates on PK parameters were estimated using nonlinear mixed-effect modeling techniques. Model 2 The model of ketoconazole-midazolam interaction implemented using the physiologically based population ADME simulator Simcyp (version 11, Simcyp Inc, Sheffield, UK) was used to simulate the 16 study designs using all standard parameter values. Briefly, the ketoconazole model is one compartment model with first order absorption. The concentration of ketoconazole in the intestine and in the liver are used to simulate the inhibition of the metabolism of substrate in the intestine and liver over time. Model Validation and Simulation A complete search of the literature using the Metabolism and Transport Drug Interaction Database (University of Washington, Seattle, WA) indentified 11 clinical DDI studies and 16 dosing scenarios from 17 published clinical results using midazolam as CYP3A probe (IV and oral) and ketoconazole as inhibitor (400 mg QD and 200 mg QD and BID for various durations) as of 2012 (Chung et al., 2006; Eap et al., 2004; Goh et al., 2002; Krishna et al., 2009; Lam et al., 2003; Lee et al., 2002; McCrea et al., 1999; Olkkola et al., 1994; Stoch et al., 2009; Tham et al., 2006; Tsunoda et al., 1999; Yong et al., 2008). The e

7 specific dosing regimens of ketoconazole and midazolam used in the literature reports were reproduced in the simulations. For Model 1, a crossover design with sufficient separation between treatment periods to allow complete washout between midazolam alone and midazolam with ketoconazole was employed and studies designs replicated the population demographics (age, body weight and gender) and drug concentration sampling times. For Model 2, virtual subjects were randomly selected from the healthy human volunteer population of Simcyp. The oral drug doses were assumed to be administered with a fluid volume of 250mL. In addition to midazolam, we examined a hypothetical substrate (HS) with a long half-life (19 hours) and high F (0.83) for use with Model 1. This hypothetical substrate shared all PK parameters with midazolam except that hepatic and intestinal CYP3A V max values were reduced by 90%. Simulation Data Analysis The ratio of midazolam AUC in the presence of ketoconazole to midazolam AUC in the absence of ketoconazole was used as a measurement of magnitude of interaction. The midazolam plasma concentrations versus time data simulated by Trial Simulator (Pharsight Inc., Mountain View, CA) were analyzed using S-PLUS 7.0 (Insightful Corp., Seattle, WA). AUC was calculated using the linear trapezoidal rule with extrapolation to infinity. Each trial was simulated 400 times and the distribution of midazolam AUC ratios was described using a probability density function. The performance or quantitative success of the two models to predict the clinical study results were compared by examining the percentage of study predictions that fell within 1.5 fold of the observed result. e

8 Results Predictive performance of the models The 16 dosing scenarios from 17 published clinical results are listed in Table 1. These studies varied in route of midazolam administration, size and duration of ketoconazole doses and relative dose times (Fig. 1). Fig. 2 shows the probability density distribution of the mean midazolam AUC ratios for 8 representative studies simulated using Model 1 along with the observed means from the literature. These 8 representative studies were chosen to cover the complete spectrum types of study designs, i.e., single and multiple doses of ketoconazole, midazolam given intravenously and orally and different relative dosing times. Model 1 has excellent predictability, as reflected in the comparison of the mean of predicted AUC ratio. The probability density distributions of the midazolam AUC ratio obtained from the two models were plotted in Fig. 3 for representative studies. For studies with single dosing both models perform well but in studies with multiple dosing there is a clear disparity between models. To quantify the predictive performance of the two models, we compared the geometric means of 400 trial replicates and 90% confidence intervals of the 400 geometric means with the observed value and a 1.5-fold deviation. Model 1 resulted in more successful predictions than Model 2. Using Model 2, 41% of the study point estimates of AUC ratio and 71% of the 90% confidence intervals were predicted within 1.5-fold of the observed but these numbers increased to 82% and 100% respectively with Model 1. Furthermore, Model 2 demonstrates considerable bias by systematically under-predicting AUC ratios in the range of the observed values that correspond to studies with multiple doses of ketoconazole and designs with a staggered ketoconazole and midazolam intake (Fig. 4). Comparison of model predicted enzyme inhibition time profiles The predicted ketoconazole plasma concentration-time profile following 12 doses of ketoconazole (400mg QD), and the corresponding changes in CL int, hep and CL int,gut are illustrated in Fig. 5 for the two models. Both models predicted ketoconazole plasma concentration-time profiles well with minimal systemic ketoconazole accumulation of trough concentrations due to a short plasma half-life (3-5 hours). However, the predicted inhibition of hepatic CL int is greater in magnitude and of greater duration for e

9 Model 1 relative to Model 2 due to the incorporation of a saturable ketoconazole efflux from the site of enzyme inhibition in the liver in Model 1. This major difference in model structure makes Model 1 mechanistically appropriate because it consistently captured the dose dependent accumulation of ketoconazole concentration at the enzyme site on multiple dosing. This feature of Model 1 is consistent with the persistent inhibition of hepatic metabolism observed in vitro by ketoconazole (Gibbs et al., 2000). In both models, the concentration of ketoconazole in the gut is used to drive the change in CYP3A enzyme activity, and therefore the first pass through the gut. In the published studies, ketoconazole is generally administered shortly before midazolam and the concentration of ketoconazole in the gut is high enough to inhibit all CYP3A activity for both models. Simulation of inhibitory effect of timing and dose of ketoconazole Administration of ketoconazole 400mg QD and 200mg BID are commonly used with the goal of reaching maximal CYP3A inhibition in vivo but it is unclear what duration and relative timing of dosing is necessary. We simulated the effect of ketoconazole 400mg QD and 200mg BID from 1 day up to 12 days on the midazolam AUC, using both models and evaluated the effect of time of midazolam administration relative to last dose of ketoconazole (Fig. 6). For both models the inhibitory effect of ketoconazole 200mg BID is lower than that of ketoconazole 400mg QD, but the difference is even more dramatic for Model 1. However, in addressing the issue of whether SD or MD of ketoconazole should be used to achieve maximal inhibitory effect, the two models reached different conclusions. At least 3 doses of ketoconazole QD 400mg are needed for maximal effect using Model 1, while a single dose is adequate based on the simulation using Model 2. To identify the optimal clinical ketoconazole-midazolam DDI study design(s), we simulated a series of studies using 1, 2, 3 and 7 days of ketoconazole (200mg QD, 200mg BID, 400mg QD) with Model 1; midazolam dosing times were 1 hour prior, simultaneously, 1, 2, 4, 12 and 24 hours post last ketoconazole dose. Simulations were performed with inter- and within-subject variability and uncertainty in model parameters for 250 trial replicates of 10 subjects (Fig. 7). The inhibitory effect generally increased with ketoconazole dose and duration but the effect of duration of ketoconazole 200mg QD on e

10 AUC ratio change is minimal. Ketoconazole 200mg BID did not attain the maximal inhibitory effect predicted for ketoconazole 400mg QD. In addition, the timing of midazolam administration shows similar trend across all three dose regimen of ketoconazole; simultaneous dosing with midazolam and ketoconazole resulted in maximal inhibition and this was maintained for administration of midazolam up to 2 hours after ketoconazole. When non-optimal study designs are assumed to result in complete inhibition of hepatic CYP3A then an erroneous estimate of fm may be derived (Guest et al., 2011; Youdim et al., 2008). When 200mg QD, 200mg BID and 400mg QD regimens are considered with simultaneous administration of midazolam and ketoconazole, the estimated fm (assuming complete inhibition of gut wall CYP3A) is 0.76, 0.86 and 0.90 respectively compared to actual value of 0.90 used in Model 1. Zhao et al. (2009) demonstrated that Model 2 predicts a greater inhibitory effect of ketoconazole 200mg BID compared to 400mg QD for a CYP3A substrate with longer t 1/2 and higher F than midazolam. To test this hypothesis with Model 1, we constructed a hypothetical substrate sharing all PK parameters with midazolam except that hepatic and intestinal CYP3A V max values were reduced by 90%. HS has a longer t 1/2 (19 hours) and higher F (0.83) than midazolam. We simulated dosing scenarios where ketoconazole was dosed for one day, 12 days prior to HS and predosing and subsequent dosing for multiple days (12 days prior to and 15 days post HS), for either ketoconazole 200mg BID or ketoconazole 400mg QD. On the day of HS, it was administered simultaneously with the last dose of ketoconazole (Fig. 8). For all dosing schedules, ketoconazole 400mg QD had a greater inhibitory effect than ketoconazole 200mg BID, on a CYP3A substrate with long t 1/2 and high F. Multiple doses of ketoconazole pre- and post-hs dosing were necessary to reach the maximal inhibitory effect. To illustrate the difference between ketoconazole 400mg QD and 200mg BID with Model 1, the time course of change of CL int, hep and CL int,gut following ketoconazole 200mg BID were simulated and compared to those of ketoconazole 400mg QD (Fig. 9). For both dosing regimens, the CL int,gut is essentially reduced to zero for the entire ketoconazole dosing interval even for the first dose. However the nadir of CL int, hep does not correspond to complete CYP3A inhibition and requires up to 3 days of dosing dd

11 to be achieved. The difference in maximal reduction of CL int, hep (11% remaining for ketoconazole 400mg QD vs. 14% remaining for ketoconazole 200mg BID), appears to be small but results in approximately a 30% greater substrate AUC ratio due to the inverse relationship between AUC and clearance. dd

12 Discussion Modeling and simulation is routinely used by the clinical pharmacology community to describe drug pharmacokinetics and pharmacodynamics and predicting optimal dosing regimens in a stochastic framework. As the scope and complexity of drug-drug interactions (DDIs) has increased, the benefits of applying modeling and simulation to understand mechanisms and predict many possible combinations of drugs have become clear. Recent DDI guidances from the Food and Drug Administration (FDA) and European Medicines Agency have recognized this potential and recommended the use of physiologically based pharmacokinetic models for predicting DDI potential in drug development. DDI models are inherently nonlinear with respect to dose and are time dependent when mechanism based inhibition and/or induction of enzymes is involved. Consequently the rigorous validation of a DDI model requires quantitation of the extent of interaction over a range of perpetrator dosing regimens. The prototypical CYP3A inhibitor, ketoconazole, displays both dose and time dependency of inhibition in vivo. For example a 36% higher midazolam AUC was observed after 5 days of ketoconazole (400mg QD) relative to a single dose (Stoch et al., 2009). This time dependence provides an unusual challenge for DDI predictions because ketoconazole does not accumulate in plasma on multiple dosing and data obtained from human liver microsomes is generally characterized as rapidly reversible and competitive in nature. The nature of this predictive challenge is reflected in the fact that there at least 16 ketoconazole midazolam DDI dosing scenarios in the literature (Table 1). In the current study we took advantage of these published ketoconazole midazolam studies to quantify the capability of two minimal PBPK models to describe this DDI. We conclude that the commonly employed rapidly reversible, competitive inhibitor model for ketoconazole inhibition of gut wall and hepatic CYP3A performs poorly (Fig. 2) with 41% of the study point estimates within 1.5-fold of the observed result. Using the model implemented in Simcyp v11, a similar conclusion was reached by Guest et al., who noted that 54% DDI point estimates involving ketoconazole and well-characterized CYP3A substrates were within 2-fold of the observed AUC ratio in vivo (Guest et al., 2011). This conclusion is not surprising given that this model is linear with respect to duration of ketoconazole dosing and by extension this approach should not dd

13 be employed to simulate DDIs arising from multiple dose regimens. In Simcyp v12 the blood half-life of ketoconazole has been arbitrarily prolonged for the 200 mg BID and 400 mg QD regimens to enhance the AUC ratio when midazolam and ketoconazole are administered simultaneously but provides little improvement in prediction for staggered dosing or prolonged ketoconazole dosing. In contrast, our model previously described accounts for both the time and dose dependent inhibition of CYP3A and predicts 82% of the ketoconazole midazolam study point estimates within 1.5-fold of the observed result and 100% are captured by the 90% confidence intervals of the predictions (Fig. 2). The construction and validation of non-linear DDI models across a range of substrates and inhibitors is a daunting task and therefore the use of turn-key software packages that perform this function is particularly attractive to drug metabolism scientists but is potentially counterproductive. To understand the relative merits of using static inhibitor concentrations versus dynamic PBPK models, several investigators have employed the current Simcyp (version 11) ketoconazole model (Einolf, 2007; Guest et al., 2011; Peters et al., 2012). The general conclusion has been that dynamic PBPK models were not substantially better than static inhibitor concentration models. Our data would not support this conclusion because there is no single concentration of ketoconazole that can predict the observed outcomes of a time dependent DDI model. Youdim et al. concluded that incubations with cdna-expressed CYPs can accurately estimate the fraction of a drug systemically metabolized by CYP3A because this substrate property correctly predicted (within 2-fold) the inhibition by ketoconazole in vivo using the ketoconazole model implemented in Simcyp v11 (Youdim et al., 2008). In view of the tendency of the model to under predict the maximal DDI extent (Fig. 2) it is advisable to revisit this conclusion. Similarly studies that have used the ketoconazole model to estimate the true fraction metabolized by CYP3A in vivo should be re-evaluated (Rakhit et al., 2008; Yang et al., 2012). A particularly prominent application of modeling simulation of DDIs involving ketoconazole was conducted by FDA scientists with a view to recommending optimal study designs for estimating the maximum extent of CYP3A inhibition. Zhao et al. (2009) employed the ketoconazole model to conclude that a single 400mg dose of ketoconazole results in maximal inhibition of CYP3A substrates, such as dd

14 midazolam, with short terminal half-lives in plasma and low bioavailability due to gut wall metabolism. Furthermore, for a longer half-life drug multiple ketoconazole dosing of 200mg BID following victim drug administration provides maximal CYP3A inhibition and is superior to a corresponding 400mg QD regimen. These conclusions have been included in the draft DDI guidance issued by the FDA. In view of the lack of accumulation of ketoconazole at the enzyme active site in the model these conclusions are expected. Our simulations with Model 1 do not support these conclusions (Fig. 7 and 8). We clearly show that multiple dosing of ketoconazole prior to the administration of midazolam is required for maximal inhibitory effect; in the absence of this prior dosing a 36% lower AUC ratio of midazolam is predicted. When a single 400mg dose of ketoconazole is assumed to result in complete hepatic CYP3A inhibition, the estimated fm of midazolam (assuming complete inhibition of gut wall CYP3A) is 0.84 compared to actual value of Under all dosing scenarios 400mg QD ketoconazole results in a greater midazolam AUC ratio compared to 200mg BID (Fig. 7). These conclusions are independent of the half-life of the victim drug (Fig. 8). In addition to the duration ketoconazole dosing prior to and post victim drug administration it is also important to consider the timing of drug administration on the day of coadministration. In this context it is important to ensure that maximal inhibition of gut wall CYP3A exists during the time interval required for the majority of the victim drug to be absorbed. In the case of midazolam we conclude that coadministration of ketoconazole and midazolam results in a maximal AUC ratio and that this is maintained when ketoconazole is given up to 2 hours prior to midazolam and little affected up to 4 hours (Fig. 6 and 7). We recommend administration of ketoconazole 1 hour prior to the victim drug to ensure that even the most rapidly absorbed drugs will encounter complete inhibition of gut wall CYP3A. This conclusion is in agreement with the analysis conducted by Zhao et al. (Zhao et al., 2009). The effect of ketoconazole on the Cmax and AUC of a putative CYP3A substrate is determined to provide specific guidance to physicians on the expected pharmacologic response in the presence of a strong CYP3A inhibitor. An additional benefit derived from the ketoconazole study is that appropriate modeling can be employed to estimate the parameters needed to predict the DDI with other strong, dd

15 moderate and weak inhibitors of CYP3A. If the chosen ketoconazole dosing regimen causes complete inhibition of intestinal and hepatic CYP3A then observed changes in Cmax and AUC can be used to estimate the fm and Fg (after correcting for hepatic extraction ratio when appropriate) that are essential for predicting additional DDI outcomes. The simulations using our model indicate that both the 200mg BID and 400mg QD ketoconazole dosing regimens cause complete inhibition of gut wall CYP3A for at least 5 hours. However, these dosing regimens differ in the extent of hepatic inhibition such that remaining CL int, hep is 14% for 200mg BID and 11% for 400mg QD after 7 days of pretreatment (Fig. 9). If a 200mg BID ketoconazole regimen is employed and assumed to cause complete inhibition of hepatic CYP3A then the fm of midazolam would be estimated as 0.86 rather than 0.90 if 400mg QD was employed (model set at 0.90). When these values are used to predict the effect of 200mg QD the AUC ratios for midazolam are 7.48 and 8.46 for fm values of 0.86 and 0.90, respectively. Although the use of the 200mg BID regimen would clearly identify a strong interaction that may adequately alert prescribers, we would suggest that to minimize corruption of subsequent predictions the 400mg QD regimen should be employed. In conclusion we have demonstrated that our mechanistic PBPK model that allows for accumulation of ketoconazole at enzyme site is capable of reproducing the dose and time dependent inhibition of midazolam by ketoconazole. The current model implemented in Simcyp v11 should not be used for simulating the effect of ketoconazole on CYP3A activity in vivo. Contrary to current FDA guidance, a 400mg QD for at least 3 days should be employed for all substrates to estimate maximum AUC ratios and predict additional DDIs for CYP3A substrates. dd

16 Authorship Contributions Participated in research design: Han, Mao, Chien and Hall. Conducted experiments: Han, Mao, Chien and Hall. Performed data analysis: Han, Mao and Chien. Wrote or contributed to the writing of the manuscript: Han, Mao, Chien and Hall. de

17 References Chien JY, Lucksiri A, Ernest CS, 2nd, Gorski JC, Wrighton SA and Hall SD (2006) Stochastic prediction of CYP3A-mediated inhibition of midazolam clearance by ketoconazole. Drug metabolism and disposition: the biological fate of chemicals 34: Chung E, Nafziger AN, Kazierad DJ and Bertino JS, Jr. (2006) Comparison of midazolam and simvastatin as cytochrome P450 3A probes. Clinical pharmacology and therapeutics 79: Eap CB, Buclin T, Cucchia G, Zullino D, Hustert E, Bleiber G, Golay KP, Aubert AC, Baumann P, Telenti A and Kerb R (2004) Oral administration of a low dose of midazolam (75 microg) as an in vivo probe for CYP3A activity. European journal of clinical pharmacology 60: Einolf HJ (2007) Comparison of different approaches to predict metabolic drug-drug interactions. Xenobiotica; the fate of foreign compounds in biological systems 37: FDA (2012) Drug Interaction Studies Study Design, Data Analysis, Implications for Dosing, and Labeling Recommendations. M292362pdf. Gibbs MA, Baillie MT, Shen DD, Kunze KL and Thummel KE (2000) Persistent inhibition of CYP3A4 by ketoconazole in modified Caco-2 cells. Pharmaceutical research 17: Goh BC, Vokes EE, Joshi A and Ratain MJ (2002) Phase I study of the combination of losoxantrone and cyclophosphamide in patients with refractory solid tumours. British journal of cancer 86: Guest EJ, Rowland-Yeo K, Rostami-Hodjegan A, Tucker GT, Houston JB and Galetin A (2011) Assessment of algorithms for predicting drug-drug interactions via inhibition mechanisms: comparison of dynamic and static models. British journal of clinical pharmacology 71: Krishna G, Moton A, Ma L, Savant I, Martinho M, Seiberling M and McLeod J (2009) Effects of oral posaconazole on the pharmacokinetic properties of oral and intravenous midazolam: a phase I, randomized, open-label, crossover study in healthy volunteers. Clinical therapeutics 31: de

18 Lam YW, Alfaro CL, Ereshefsky L and Miller M (2003) Pharmacokinetic and pharmacodynamic interactions of oral midazolam with ketoconazole, fluoxetine, fluvoxamine, and nefazodone. Journal of clinical pharmacology 43: Lee JI, Chaves-Gnecco D, Amico JA, Kroboth PD, Wilson JW and Frye RF (2002) Application of semisimultaneous midazolam administration for hepatic and intestinal cytochrome P450 3A phenotyping. Clinical pharmacology and therapeutics 72: McCrea J, Prueksaritanont T, Gertz BJ, Carides A, Gillen L, Antonello S, Brucker MJ, Miller-Stein C, Osborne B and Waldman S (1999) Concurrent administration of the erythromycin breath test (EBT) and oral midazolam as in vivo probes for CYP3A activity. Journal of clinical pharmacology 39: Ohno Y, Hisaka A and Suzuki H (2007) General framework for the quantitative prediction of CYP3A4- mediated oral drug interactions based on the AUC increase by coadministration of standard drugs. Clinical pharmacokinetics 46: Olkkola KT, Backman JT and Neuvonen PJ (1994) Midazolam should be avoided in patients receiving the systemic antimycotics ketoconazole or itraconazole. Clinical pharmacology and therapeutics 55: Peters SA, Schroeder PE, Giri N and Dolgos H (2012) Evaluation of the use of static and dynamic models to predict drug-drug interaction and its associated variability: impact on drug discovery and early development. Drug metabolism and disposition: the biological fate of chemicals 40: Rakhit A, Pantze MP, Fettner S, Jones HM, Charoin JE, Riek M, Lum BL and Hamilton M (2008) The effects of CYP3A4 inhibition on erlotinib pharmacokinetics: computer-based simulation (SimCYP) predicts in vivo metabolic inhibition. European journal of clinical pharmacology 64: Stoch SA, Friedman E, Maes A, Yee K, Xu Y, Larson P, Fitzgerald M, Chodakewitz J and Wagner JA (2009) Effect of different durations of ketoconazole dosing on the single-dose pharmacokinetics of midazolam: shortening the paradigm. Journal of clinical pharmacology 49: de

19 Tham LS, Lee HS, Wang L, Yong WP, Fan L, Ong AB, Sukri N, Soo R, Lee SC and Goh BC (2006) Ketoconazole renders poor CYP3A phenotype status with midazolam as probe drug. Therapeutic drug monitoring 28: Tsunoda SM, Velez RL, von Moltke LL and Greenblatt DJ (1999) Differentiation of intestinal and hepatic cytochrome P450 3A activity with use of midazolam as an in vivo probe: effect of ketoconazole. Clinical pharmacology and therapeutics 66: Yang Z, Vakkalagadda B, Shen G, Ahlers CM, Has T, Christopher LJ, Kurland JF, Roongta V, Masson E and Zhang S (2012) Inhibitory Effect of Ketoconazole on the Pharmacokinetics of a Multireceptor Tyrosine Kinase Inhibitor BMS in Healthy Participants: Assessing the Mechanism of the Interaction With Physiologically-Based Pharmacokinetic Simulations. Journal of clinical pharmacology. Yong WP, Wang LZ, Tham LS, Wong CI, Lee SC, Soo R, Sukri N, Lee HS and Goh BC (2008) A phase I study of docetaxel with ketoconazole modulation in patients with advanced cancers. Cancer chemotherapy and pharmacology 62: Youdim KA, Zayed A, Dickins M, Phipps A, Griffiths M, Darekar A, Hyland R, Fahmi O, Hurst S, Plowchalk DR, Cook J, Guo F and Obach RS (2008) Application of CYP3A4 in vitro data to predict clinical drug-drug interactions; predictions of compounds as objects of interaction. British journal of clinical pharmacology 65: Zhao P, Ragueneau-Majlessi I, Zhang L, Strong JM, Reynolds KS, Levy RH, Thummel KE and Huang SM (2009) Quantitative evaluation of pharmacokinetic inhibition of CYP3A substrates by ketoconazole: a simulation study. Journal of clinical pharmacology 49: de

20 Figure legends FIG. 1. The schematic representation of ketoconazole dosing relative to midazolam dose from literature. Midazolam dose was at time 0. The numbers to the right of the study correspond to the study number in Table 1. FIG. 2. Probability density distribution of the study mean ratio of midazolam AUC (indicated by MDZ AUC (+KTZ): AUC (-KTZ)) for 8 representative studies using Model 1. The histograms represent the distribution of the geometric means of 400 trial replicates using the ketoconazole dosing regimens of the published studies. The dashed vertical lines represent the 5 th and 95 th percentile of the predictions. The solid vertical lines represent the literature reported mean of each study. The labels in the upper-right corner correspond to the study number in Table 1. FIG. 3. Comparison of probability density distribution of the study mean ratio of midazolam AUC (indicated by MDZ AUC (+KTZ): AUC (-KTZ)) for 8 representative studies using Model 1 (the curve with a solid line) and Model 2 (the curve with a dashed line). The probability density curves represent the distribution of the geometric means of 400 trial replicates using the ketoconazole dosing regimens of the published studies. The shaded regions represent the 5 th and 95 th percentile of the predictions. The solid vertical lines represent the literature reported mean of each study. The labels in the upper-right corner correspond to the study number in Table 1. FIG. 4. Plots of predicted ratio of midazolam AUC (indicated by MDZ AUC (+KTZ): AUC (-KTZ)) vs. observed values using Model 1 (A, red symbol ) and Model 2 (B, blue symbol ). Symbols are the geometric means of 400 trial replicates of each published study. The error bars denote the 5 th and 95 th percentiles of the 400 geometric means. Solid lines represent lines of unity, and the area between the dashed lines represents an area within 1.5-fold error. The labels on the right of the symbols correspond to the study number in Table 1. FIG. 5. Prediction of ketoconazole plasma concentration-time profiles and the change of CL int, hep and CL int,gut over time following ketoconazole 400mg QD for 12 days with Model 1 (A) and Model 2 (B). dd

21 FIG. 6. Effect of duration of ketoconazole and dosing time of midazolam on midazolam AUC ratio after ketoconazole QD 400mg (A) and BID 200mg (B). The top planes in both plots represent the simulation results using Model 1 while the lower planes represent Model 2. Dosing time of midazolam is expressed as the time relative to the last dose of ketoconazole. FIG. 7. Comparison of effects of the timing of the midazolam dose and duration of ketoconazole pretreatment (200mg QD, 200mg BID and 400mg QD) on midazolam AUC ratio (MDZ AUC (+KTZ): AUC (-KTZ)) using Model 1. The arrow indicates the time at which ketoconazole is given (time 0 in our trial design). 250 trial replicates with 10 subjects were simulated. The error bars denote the 5 th and 95 th percentiles of the 250 geometric means. FIG. 8. The effect of ketoconazole dose regimen and duration on the HS AUC ratio (HS AUC (+KTZ): AUC (-KTZ)) using Model 1. HS is administered with last ketoconazole simultaneously on the day of HS administration. FIG. 9. Prediction of ketoconazole plasma concentration-time profiles and the change of CL int, hep and CL int,gut over time with Model 1. A, ketoconazole 400mg QD for 12 days; B, ketoconazole 200mg BID for 12 days. dd

22 TABLE 1 Ketoconazole and midazolam dosing regimens and observed midazolam AUC ratio in ketoconazolemidazolam studies from the literature. Study No. Dose Ketoconazole Dosing Dosing Time Interval (h) Number of Doses Midazolam Dose Dosing Route Dosing Time Observed AUC ratio Subjects (N) Reference 1 200mg Day mg i. v. bolus Day 1, +12hr Tsunoda et al., mg Day mg / 1mg w/ktz i. v. 30 mins Day 1, +19hr Lee et al., mg Day mg / 1mg w/ktz i. v. bolus Day a 29 Tham et al., mg Day mg / 1mg w/ktz i. v. bolus Day b 41 Yong et al.,2008, Goh et al., mg Day mg i. v. 30 mins Day Krishna et al., mg Day 1 SD 1 2mg p.o. Day McCrea et al., mg Day 1 SD 1 2mg p.o. Day 1, +2hr McCrea et al., mg Day mg p.o. Day 1, +12hr Tsunoda et al., mg Day mg / 2mg w/ktz p.o. Day 1, +13hr Lee et al., mg Day mg p.o. Day a 21 Eap et al., mg Day mg p.o. Day 12, +1hr Lam et al., mg Day 1 SD 1 2mg p.o. Day Stoch et al., mg Day mg p.o. Day Stoch et al., mg Day mg p.o. Day 4, +1hr 16.7 c 9 Olkkola et al., mg Day mg p.o. Day Stoch et al., mg Day mg p.o. Day Krishna et al., mg Day mg/kg p.o. Day 6 or 9, -2hr Chung et al.,2006 a The ratio of mean midazolam AUC in presence and in absence of ketoconazole. b The ratio of mean midazolam AUC in presence and in absence of ketoconazole. AUCs of midazolam in presence and in absence of ketoconazole are calculated from clearance of midazolam, which are dd

23 digitalized from two studies, Yong et al., 2008 and Goh et al., 2002, respectively (as reported in Yong et al., 2008). c The geometric mean of individual AUC ratios, which are digitalized from the plot in reference. dd

24

25

26

27

28

29

30 DMD Fast Forward. Published on April 12, 2013 as DOI: /dmd

31

32