5-[2-Ethoxy-5-(4-ethyl-piperazine-1-sulfonyl)-pyridin-3-yl]-3-ethyl-2-(2-methoxy-ethyl)-2,6-dihydro-pyrazolo[4,3-d]pyrimidin-7-one (UK-369,003) is a phosphodiesterase-5 inhibitor in clinical development at Pfizer. UK-369,003 is predominantly metabolized by cytochrome P450 3A4 and is also a substrate for the efflux transporter P-glycoprotein. The pharmacokinetics of UK-369,003 has been profiled after oral administration of 1 to 800 mg of an immediate release formulation to healthy volunteers. Nonlinearity was observed in the systemic exposure at doses of 100 mg and greater. In addition, the pharmacokinetics of UK-369,003 has also been investigated after oral administration of the more therapeutically attractive modified release formulation. Systemic exposure was prolonged with the modified release formulation, but bioavailability was reduced in comparison with that of the immediate release formulation. Physiologically based pharmacokinetic modeling strategies are commonly used in drug discovery and development. This work describes application of the physiologically based pharmacokinetic software GastroPlus to understand the pharmacokinetics of UK-369,003. The impact of gut wall and hepatically mediated CYP3A4 metabolism, in addition to the actions of P-glycoprotein, in causing the nonlinear pharmacokinetics of the immediate release formulation and the reduced bioavailability of the modified release form, was investigated. The model accurately described the systemic exposure of UK-369,003 after intravenous and both immediate and modified release oral administration and suggested that CYP3A4 is responsible for the majority of the nonlinearity in systemic exposure observed after administration of the immediate release form. Conversely, the reduced bioavailability of the modified release formulation is believed to be caused by incomplete release from the device, incomplete absorption of released drug, and, to a lesser extent, CYP3A4 metabolism.
5-[2-Ethoxy-5-(4-ethyl-piperazine-1-sulfonyl)-pyridin-3-yl]-3-ethyl-2-(2-methoxy-ethyl)-2,6-dihydro-pyrazolo[4,3-d]pyrimidin-7-one (UK-369,003) is a phosphodiesterase-5 enzyme inhibitor in clinical development at Pfizer. UK-369,003 is a moderately lipophilic (log D7.4 = 2.4) and weakly basic (pKa = 6.6) compound, the structure of which is presented in Fig. 1.
UK-369,003 is moderately soluble and demonstrates high flux across Caco-2 cell monolayers. Consistent with these properties, UK-369,003 exhibits high absorption in preclinical pharmacokinetic studies. In vitro incubations performed in baculovirus cells engineered to individually express a range of cytochrome 450 enzymes showed CYP3A4 to be responsible for the majority of the metabolism of UK-369,003. Three major metabolic pathways were identified, namely, N-deethylation and N,N-deethylation of the N-ethylpiperazine in addition to O-demethylation of the methoxyethyl side chain. Ketoconazole (a CYP3A4 inhibitor) produced a significant, concentration-dependant reduction in the rates of formation of each of these metabolites in vitro, further substantiating the role of CYP3A4 in the metabolism of UK-369,003. In addition to those experiments performed to understand its metabolic profile, studies performed in multidrug resistance protein-1-transfected Madin-Darby canine kidney cell lines indicate that UK-369,003 is a substrate for the efflux transporter P-glycoprotein (P-gp) (Km = 24 μM).
The clinical pharmacokinetics of UK-369,003 has been profiled after intravenous and oral administration to healthy volunteers. All clinical studies discussed here were performed in the fasted state. Plasma concentration-time profiles observed after intravenous, immediate release (IR) and modified release (MR) oral administration are summarized in Fig. 2. After intravenous administration UK-369,003 demonstrated moderate clearance (CL = 60 l/h) and steady-state volume of distribution (Vdss = 2.4 l/kg). UK-369,003 was found to have a renal clearance of 4.9 l/h. After administration of 100 mg of UK-369,003 in an IR formulation, a bioavailability of 34% was observed. If the clearance of UK-369,003 (60 l/h) is considered in comparison with hepatic blood flow [approximately 87 l/h (Davies and Morris, 1993)], this result is consistent with complete absorption from the gastrointestinal tract. A supraproportional increase in systemic exposure was observed at doses greater than 100 mg of the IR formulation (Fig. 2D). The PK of UK-369,003 has also been investigated after oral administration of a therapeutically more attractive MR formulation. As anticipated, the MR formulation prolongs the exposure and extends the terminal half-life of UK-369,003 (Fig. 2E). In comparison with the IR form, the MR form shows reduced bioavailability (18 versus 34%) after administration of 50-, 100-, and 200-mg doses.
The use of physiologically based pharmacokinetic (PBPK) modeling strategies in drug discovery and development is becoming increasingly more commonplace (Parrott and Lavé, 2002; Jones et al., 2006; De Buck et al., 2007; Germani et al., 2007; Jones et al., 2009; Parrott et al., 2009; Rowland et al., 2011; Zhao et al., 2011). This work describes the PBPK modeling undertaken to examine the pharmacokinetics of UK-369,003. The role of both gut wall and hepatically mediated CYP3A4 metabolism, along with gut-mediated P-gp efflux, was investigated using the physiologically based GastroPlus simulation software. The aim of this work was to investigate the causes of the nonlinearity in oral exposure observed after IR administration, in addition to investigating the cause of the reduced bioavailability of the MR formulation. Improved understanding of the kinetics of UK-369,003 will allow more informed decision making in the clinical development of this candidate. For example, the model may be used to ensure appropriate delivery via the MR device in addition to educating the drug-drug interaction (DDI) studies required to underwrite progression of this candidate.
Materials and Methods
GastroPlus (version 6; Simulations Plus, Inc., Lancaster, CA) is a commercially available software that is routinely used to model drug absorption. The model underlying the prediction of absorption in GastroPlus is known as the advanced compartmental absorption and transit (ACAT) model (Agoram et al., 2001). The physiologically based ACAT model consists of nine compartments corresponding to different segments of the digestive tract and is based on the original compartmental absorption and transit model described by Yu and Amidon (Yu et al., 1996). Each compartment is further subdivided to describe drug that is unreleased (when modified release formulations are simulated), in addition to that which is undissolved, dissolved, and absorbed. Absorbed drug is defined as drug that has entered into the enterocyte. Movement of drug between each subcompartment is described by a series of differential equations. The rate of drug flow through sequential intestinal compartments is determined by the transit time of each compartment. All simulations were performed in GastroPlus. Input parameters used for this modeling are summarized in Table 1.
Only drug in solution is available for absorption. The solubility profile of UK-369,003 throughout the gastrointestinal tract was simulated by incorporation of the in vitro solubility in a biorelevant fasted state simulated intestinal fluid medium, in combination with the ionization state of the basic moiety as modulated by the regional pH in the gastrointestinal tract. The effective passive permeability of UK-369,003 was predicted from flux in Caco-2 cells in the presence of P-gp inhibitors. Passive absorption is assumed to be driven by the concentration gradient of the drug across the luminal membrane in an unsaturable manner. The model uses an understanding of the abundance of both intestinal and hepatic CYP3A4 and gut P-gp to allow simulation of nonlinear metabolism and P-gp efflux. It is assumed that carrier-mediated active transport is described by Michaelis-Menten kinetics and is therefore saturable. Likewise, CYP3A4 metabolism was modeled according to Michaelis-Menten kinetics. A well stirred venous equilibrium model in which the unbound exposure in the liver is assumed to be equivalent to the unbound systemic exposure at steady state was used to describe hepatocyte drug concentrations, whereas gut-mediated metabolism and active transport were driven by predicted enterocyte concentrations.
Disposition after Intravenous Administration.
The PK of UK-369,003 has been investigated after intravenous administration of 10, 30, and 50 mg to 5, 17, and 5 male healthy volunteers, respectively. The disposition of UK-369,003 at these doses was modeled using a two-compartmental fit in the GastroPlus PKPlus module. Simultaneous fitting of all dose groups was performed, assuming linearity in PK across the dose range profiled. A Hooke and Jeeves pattern search optimization with a proportional residual error structure was used.
Pharmacokinetics after IR Oral Administration.
The PK of UK-369,003 has been investigated after administration of 1, 3, 10, 30, 100, 200, 400, and 800 mg to eight healthy male volunteers, respectively. With use of the GastroPlus optimization tool, the contribution of gut and hepatic CYP3A4 metabolism and the impact of P-gp efflux on the PK of UK-369,003 were investigated. Intravenous and IR oral profiles were modeled simultaneously, ensuring compatibility of the model with each route of administration. The ACAT model uses the input parameters listed in Table 1 to estimate the solubility, dissolution, and passive absorption of UK-369,033 in each region of the gastrointestinal tract. These parameters are fixed in the model. A Hooke and Jeeves pattern search optimization and proportional residual error structure was again used. Three alternative models were investigated, namely the CYP3A4, P-gp, and combined models.
The CYP3A4 model assumes that nonlinearity in IR plasma kinetics is mediated only by nonlinear CYP3A4 metabolism. The impact of gut and liver metabolism was investigated by estimation of CYP3A4 enzyme kinetics (Vmax and Km). In this model, the distribution of UK-369,003 was assumed to be the same as that estimated after intravenous administration.
The P-gp model assumes that all nonlinearity in IR plasma kinetics is mediated by P-gp only. The impact of nonlinear gut-mediated P-gp efflux was investigated by estimation of efflux kinetics (Vmax and Km). In this model, it was assumed that metabolic extraction of UK-369,003 remains unsaturated over the dose range investigated. As a result, both the clearance and distribution of UK-369,003 were assumed to be equivalent to those estimated after intravenous administration.
Finally, the combined model assumes that the nonlinearity in IR plasma kinetics is due to a combination of both nonlinear CYP3A4 metabolism and P-gp efflux. The impact of gut and liver-mediated CYP 3A4 metabolism, in addition to gut-mediated P-gp efflux, was investigated by estimation of both metabolic and efflux kinetics (Vmax and Km). As with the CYP3A4 model, the distribution of UK-369,003 was assumed to be the same as that estimated after intravenous administration.
Pharmacokinetics after Modified Release Oral Administration.
The PK of the MR formulation of UK-369,003 has been profiled after oral administration of 50, 100, and 200 mg. Subsequent investigation of the MR profiles was undertaken, assuming that the ADME profile of UK-369,003 upon release from the MR device was as described by the combined model.
Previous clinical experience with the MR device used in these studies indicates that its passage through the gut may be considerably slower (mean = 34 h; A. Cleton, unpublished data) than the fasted gut transit time implemented within the standard ACAT model (21 h). As a result, the default ACAT model was altered to ensure that the total gut transit time was reflective of the observed transit time. The relative transit through each compartment (excluding the stomach) was conserved; i.e., the transit time of each compartment was increased proportionally. The release of UK-369,003 from the device was described by the Weibull function shown in eq. 1: Lag, shape, and parameters were estimated in parallel using the MR clinical data in the model. A Hooke and Jeeves pattern search optimization with proportional residual error structure was again used.
Implicit within each of these models are the following assumptions:
The distribution and clearance kinetics of UK-369,003 are equivalent after both oral and intravenous administration.
Where appropriate (CYP3A4 model and combined model), the metabolism of UK-369,003 is assumed to be occurring only in the liver and gut wall and is mediated solely by CYP3A4. Differences in gut and liver Vmax are accounted for by tissue weight and CYP3A4 abundance only. CYP3A4 unbound Km liver = unbound Km gut.
When P-gp is assumed to be the sole cause of the nonlinearity (P-gp model), elimination occurs only in the liver and is linear across the dose range investigated.
The abundance and distribution of CYP3A4 contained within GastroPlus are physiologically relevant (Paine et al., 1997).
Where appropriate (P-gp model and combined model), the role of P-gp is limited to the gut wall, and it operates solely as an efflux transporter.
The relative activity of P-gp through the gastrointestinal tract, which is implemented in GastroPlus, is physiologically accurate (Makhey et al., 1998).
The in vitro solubility profile is predictive of the in vivo situation.
Human passive permeability can be accurately predicted from in vitro Caco-2 data in the presence of P-gp inhibitors.
With the exception of the actions of P-gp in the gut wall, the absorption and distribution of UK-369,003 are assumed to be passive.
UK-369,003 is subject to a renal clearance (CLr) of 4.9 l/kg.
Figure 3 shows the predicted PK profiles of UK-369,003 estimated by simultaneous fitting of 10-, 30-, and 50-mg doses. Each is superimposed on the observed clinical data (±S.D.). Corresponding goodness-of-fit plots are presented in Fig. 4. When one considers both Figs. 3 and 4, it is clear that the disposition of UK-369,003 has been well described by a two-compartment kinetic model. As a result of the relatively narrow dose range investigated, adequate description of the data is achieved, assuming linear kinetics across all doses. Table 2 summarizes the PK parameters estimated from this model, confirming the moderate clearance and distribution of UK-369,003.
Immediate Release Oral Pharmacokinetics.
Figures 5, 6, and 7 show the predicted PK profiles of UK-369,003 estimated by simultaneous investigation of intravenous and IR oral kinetics by the CYP3A4, P-pg, and combined models, respectively. Each is superimposed on the observed mean clinical data (±S.D.). Corresponding goodness-of-fit plots are presented in Fig. 8.
When considered together, Figs. 5 and 8, A and B, demonstrate that the nonlinearity in systemic exposure observed after administration of IR UK-369,003 can be well described by assuming that CYP3A4 is the sole perpetrator of the observed nonlinearity. In contrast, Figs. 6 and 8D highlight the fact that P-gp is not likely to be the sole cause of the nonlinearity. This observation is exemplified by an overestimation of systemic exposure observed at a low dose (1 mg) (Fig. 6A) and underestimated exposure noted with the 800-mg dose (Fig. 6H). The increased precedence of underestimating systemic exposure using the P-gp model is highlighted by comparison of the goodness-of-fit plots presented in Fig. 8, A and B. Visual inspection of the model-predicted exposure profiles achieved using the combined model (Fig. 7) suggests little improvement over the CYP3A4 model (Fig. 5). Likewise, the goodness-of-fit plots presented in Fig. 8, A and E, suggest little difference between the models. Table 3 presents a summary of the kinetic parameters, in addition to the objective goodness-of-fit measures (R2, Akaike, and objective function) returned by each of the models. Comparison of the objective measures associated with the CYP3A4 and P-pg models reconfirms the significant improvement in the model fit offered by the CYP3A4 model (increased R2 and decreased Akaike and objective function), despite an equal number of estimated parameters. Subsequent comparison of CYP3A4 and combined models again suggests that there is no significant difference in the description of the observed data by either model. For example, the R2 value returned by each is identical, whereas the addition of two further parameters in the combined model (i.e., P-gp Vmax and Km) results in a drop in objective function of only 0.72, indicating redundancy of the additional parameters in the combined model.
Modified Release Oral Pharmacokinetics.
Figure 9 presents MR profiles generated by incorporating the predicted release profile and increased gut transit time into the combined model. Although the CYP3A4 model provides an adequate description of the IR kinetics, the MR kinetics was investigated using the combined model (see Discussion for a full explanation). The observed clinical data (±S.D.) are presented for comparison. Corresponding goodness-of-fit plots are presented in Fig. 10, confirming the accurate description of the observed systemic exposures. In addition, the predicted release profile from the MR device is presented in Fig. 11.
Modeling and simulation approaches are increasingly being used to guide the clinical development of drug candidates. Precedence for using such techniques, particularly GastroPlus, to justify clinical trial design has already been established. For example, simulations performed with multiple weak acidic and basic Biopharmaceutics Classification System class II compounds have successfully justified biowavers for those candidates, negating the regulatory requirement to perform these studies (Tubic-Grozdanis et al., 2008). In addition, similar simulations for etoricoxib have been used to support an assertion of equivalence between different IR solid oral formulations (Okumu et al., 2009).
The clinical PK of UK-369,003 has been described after both intravenous and oral administration of IR and MR formulations using a PBPK model. In vitro studies suggest that UK-369,003 is metabolized by CYP3A4 and is also a substrate for the P-gp efflux transporter. Therefore, this work was performed to investigate whether the nonlinearity in plasma exposure observed after oral administration of the IR formulation was most likely due to saturation of gut and/or hepatically mediated CYP3A4 or gut-mediated P-gp efflux. Furthermore, the cause of the reduced bioavailability observed with the MR form, in comparison with the IR form, has been investigated. In a similar manner, GastroPlus has previously been used to deconvolute the active transport mediated nonlinearity in the PK of UK-343,664, valacyclovir, gabapentin, and talinolol (Tubic et al., 2006; Bolger et al., 2009; Abuasal et al., 2010). Modulation of the PK and pharmacodynamic endpoints of adinazolam and metoprolol by MR devices has also been successfully described previously (Lukacova et al., 2009).
As discussed, comparison of Figs. 5 to 8 and the objective function measures (Table 3) indicates that the majority of the nonlinearity observed after administration of the IR formulation can be attributed to the saturation of CYP3A4 metabolism. In addition to understanding the relative importance of CYP3A4 and P-gp in causing the nonlinearity in plasma exposure, it is possible to examine the model further to determine the absolute contributions of both gut and hepatic CYP3A4 to the metabolism of UK-369,003. Figure 12 summarizes the fraction of each dose that remains unabsorbed or is removed by either gut wall or hepatic extraction. In fact, UK-369,003 is completely absorbed at all doses, except after administration of 800 mg for which 10% of the dose remains unabsorbed. The high fraction absorbed is consistent with the high permeability and solubility of UK-369,003 (Table 1) and complete absorption observed in preclinical PK studies. The simulations suggest that the relative contributions of gut and liver metabolism to the extraction of UK-369,003 change with dose (Fig. 12). At the lowest doses investigated, the gut wall plays a major role in the exclusion of UK-369,003 from the systemic circulation; approximately 50% of a 1-mg dose is removed by the gut wall. However, gut wall extraction is rapidly saturated upon increasing dose; only 10% of the administered 30-mg dose is extracted by the gut.
The largest proportion of the dose (∼60%) is predicted to be absorbed from the jejunum after administration of the IR formulation. Figure 13 contrasts the enterocyte concentrations (Cent) simulated in the jejunum with the model-predicted CYP3A4 Km. The rapid saturation that is predicted in gut wall metabolism results from the predicted enterocyte exposure profiles of UK-369,003 being significantly in excess of the Km in all but the lowest dose investigated. In contrast, saturation of hepatic metabolism occurs to a lesser extent with increasing dose; a significant reduction in the fraction of the dose removed by the liver occurs only at higher doses (Fig. 12). This observation is again reflected by comparing the free hepatic concentrations and predicted Km (Fig. 13), which demonstrate that simulated free exposure in the liver only reaches equivalence with the predicted Km at doses of 100 mg and greater. Overall, the model suggests that a rapid saturation of gastrointestinal metabolism occurs with increasing dose (1–30 mg), during which time the fraction removed by hepatic metabolism increases to compensate. Thus, no significant nonlinearity in plasma kinetics is observed until the saturation of hepatic extraction becomes prevalent (dose ≥100 mg).
The bioavailability of the MR form (18%) is reduced in comparison with that of the IR form. Previous clinical experience demonstrates that the mean transit time of the MR device used in these studies is approximately 34 h. Visual inspection of the concentration-time profiles observed after administration of the MR form (Fig. 2) indicates that the extended half-life of the MR form is maintained at times in excess of the standard gut transit time of the ACAT model (21 h), further substantiating the hypothesis that absorption is continuing in excess of this time. Therefore, the ACAT model was modified when the MR PK profiles were investigated to maintain its physiological relevance.
Although the CYP3A4 model was recognized as offering an adequate description of IR oral pharmacokinetics, the MR profiles were modeled using the combined model. The decision to use the combined model was made to reflect the different absorption profiles of both formulations. Because UK-369,003 is a highly permeable candidate, the absolute mass of drug remaining to be absorbed in the lower intestine is comparatively low when considered in comparison with the MR form, whereas delayed release from the device increases the drug present in the colon (Fig. 14). As the activity of P-gp incorporated into the GastroPlus PBPK model increases in the lower intestine (Makhey et al., 1998), the impact of P-gp on the absorption of the MR form may be greater than that on the IR form. Furthermore, because the combined model-derived estimate of P-gp Km (39 μM) is broadly similar to that measured in vitro (24 μM), the likelihood of significant model misspecification caused by the inclusion of P-gp in the interrogation of MR forms is assumed to be minimal.
By integrating the optimized release profile of the MR device, along with the modified transit time, into the combined model, the model accurately describes the PK of the MR formulation (Fig. 11). The range of MR doses investigated clinically (50–200 mg) is much narrower than the IR dose range profiled, and, as a result, no nonlinearity in plasma exposure is observed. Investigation of the simulations performed allows an understanding of the multifactorial nature of the reduced bioavailability of the MR form. In contrast to the complete absorption of the IR formulation, it is predicted that 40% of UK-369,003 remains unabsorbed at the doses investigated. When considered along with the predicted release profile of the MR device presented in Fig. 11 (∼75% released from the device by 34 h), unabsorbed UK-369,003 appears to be composed of approximately equivalent fractions of unreleased and unabsorbed drug. The incomplete absorption of released drug that is predicted is due to differences in regional permeability consistent with the physiology of the gastrointestinal tract, P-gp efflux in the lower intestine (as a result of delayed absorption, resulting in increased exposure to P-gp in the lower intestine), and the physiochemical properties of UK-369,003 (Ungell et al., 1998; Agoram et al., 2001). The simulations suggest that gut wall metabolism has a limited impact on the systemic exposure of the MR form (5% extraction at the doses investigated). Although seemingly counterintuitive (given the lower enterocyte concentrations arising from administration of MR in comparison with IR forms), this result is probably driven by the regional distribution of CYP3A4 implemented in GastroPlus (Paine et al., 1997), in which the abundance of CYP3A4 is lowest in the lower intestine and colon. As a result of the sustained release from the MR device, the absolute mass of drug being exposed to drug-metabolizing enzyme in the upper intestine is reduced, and the contribution of metabolism to the first pass is lowered accordingly. In summary, the simulations presented suggest that the decreased bioavailability of the MR form is multifactorial in nature, namely incomplete release from the MR device, incomplete absorption, P-gp-mediated efflux, and metabolic extraction (first-pass metabolism) across both the gut wall and liver.
The capacity of the proposed model to quantitatively reflect the PK of intravenous, IR, and MR forms greatly increases confidence in the physiological relevance of the simulations performed. To this end, the improved understanding of the kinetics afforded by the model allows it to be a useful tool for guiding clinical study design. For example, amalgamation of the DDI studies performed using the IR formulation along with increased understanding of the absorption and disposition of UK-369,003 afforded by this model, could enable timely estimation of the likely magnitude of DDIs that will be observed using the therapeutically preferred MR formulation.
In conclusion, the work presented here demonstrates the increased insight into the clinical kinetics of UK-369,003 that was gained by use of a PBPK model. These simulations suggest that the majority of the nonlinearity in systemic exposure observed after administration of UK-369,003 in an IR form is mediated by saturation of both gut wall and hepatic CYP3A4, whereas the impact of P-gp is limited. The relative contribution of gut wall and hepatically mediated metabolism varies, depending on the dose administered, the contribution of gut wall metabolism reducing rapidly with increasing dose. Furthermore, reduced bioavailability of MR UK-369,003 is probably caused by incomplete release from the device and reduced absorption from the colon, in addition to a first-pass extraction mediated by hepatic and, to a lesser extent, gut wall metabolism.
Participated in research design: Watson, Davis, and Jones.
Performed data analysis: Watson and Davis.
Wrote or contributed to the writing of the manuscript: Watson and Jones.
We thank Kevin Beaumont, Susan Cole, Anne Heatherington, Adriaan Cleton, and Maurice Dickins for valuable scientific discussion of this work.
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
- immediate release
- modified release
- physiologically based pharmacokinetic
- drug-drug interaction
- advanced compartmental absorption and transit.
- Received January 16, 2011.
- Accepted March 30, 2011.
- Copyright © 2011 by The American Society for Pharmacology and Experimental Therapeutics