We quantitatively compare the efficacy of two approved EGFR tyrosine kinase inhibitors, erlotinib and gefitinib, based on in vivo and in vitro data and show how a modeling approach can be used to scale from animal to humans. Gefitinib shows a higher tumor uptake in cancer patients, and we explored the potential impact on pharmacologic and antitumor activity in in vitro and in xenograft mice. Tumor growth inhibition was monitored, and the pharmacokinetics (PK) in plasma and tumor, as well as temporal changes of phospho-Erk (pErk) signals were examined in patient-derived tumor xenograft mice. These data were integrated in a translational PKPD model, allowing us to project an efficacious human dose, which we retrospectively compared with prescribed doses for cancer patients. In vitro experiments showed that cell-cycle arrest was similar for erlotinib and gefitinib. Similar pERK biomarker responses were obtained despite a 6.6-fold higher total tumor exposure for gefitinib. The PKPD model revealed a 3.7-fold higher in vivo potency for gefitinib, which did not translate into a lower anticipated efficacious dose in humans. The model-based dose prediction matched the recommended clinical doses well. These results suggest that despite having lower total tumor-to-plasma ratios, active drug exposure at target site is higher for erlotinib. Considering the PK properties, this translates in a 50% lower recommended daily dose of erlotinib in cancer patients. In summary, total exposure at target site is not suitable to rank compounds, and an integrated modeling and experimental approach can assess efficacy more accurately. Mol Cancer Ther; 15(12); 3110–9. ©2016 AACR.

The success rate of developing anticancer drugs is reported to be around 5% in clinical development, which is considerably lower than in other therapeutic areas (1). Model-based drug development (MBDD) can increase the likelihood of compound development success (2). It combines well-designed in vitro and in vivo experiments with extensive data analysis and mathematical pharmacokinetic/pharmacodynamic (PKPD) modeling, providing a robust rationale for comparing the chance of success of cancer drugs candidates. Thus, the integration of experimental work and mathematical modeling is a valuable tool to improve effectiveness and reduce attrition rates during drug development (2–5). The model-based approach to predict the efficacious dose in humans based on animal data is useful to design early clinical trials which are mainly conducted in patients whose disease conditions are progressive and fatal. However, the optimal dose needs to be further refined as clinical data become available (6).

EGFR is a widely explored target for cancer treatment as it plays a major role in tumor cell proliferation, regeneration, differentiation, and development (7, 8). Tyrosine kinase inhibitors (TKI) are drugs blocking the phosphorylation and activation of the EGFR tyrosine kinase by binding to the ATP pocket of the kinase (9). In this study, we compare two TKIs, erlotinib and gefitinib (10), based on their PKPD properties (11). Both compounds are orally administered to treat epidermoid cancers with high dependence on the EGFR pathway but without eliciting constitutive kinase or downstream activation (12). They were both first approved for the treatment of non–small cell lung cancer (NSCLC; refs. 13–15) and subsequently of various solid epidermoid cancers (16, 17).

Erlotinib and gefitinib exhibit a similar mechanism of action (18) with different physicochemical properties (e.g., pKa) and different drug disposition characteristics (11, 18–20). A thorough review of biology, PK, and clinical aspects of erlotinib and gefitinib was recently reported by Bronte and colleagues (10). In the mouse, rat, and human, gefitinib has a higher tumor tissue-to-plasma exposure ratio than erlotinib (21, 22). In the present study, we explore the relevance of high tumor-to-plasma ratio with regard to the pharmacologically active drug at the target site and its impact on efficacy. In vivo experiments were conducted in animal models of the human disease of interest, namely in patient-derived LXF A677 and cell line–derived VXF A431 tumor xenograft mice representing NSCLC and vulvar carcinoma, respectively. With regard to B-raf mutational status, the cell line–derived A431 is wild-type, and the patient-derived A677 tumor is reported as BRAF D594A heterozygous. The effect of both drugs on a cellular level was explored in a cell-cycle assay with A431 cells only because in vitro experiments were not feasible with the heterogeneous patient-derived A677 tumor explant. pAkt and pErk, effector proteins for proliferation and survival downstream of the EGFR pathway, are upregulated in many cancer types with sustained EGFR signaling (23–26). Drug exposure in tumors and respective changes downstream signaling represented by pERK levels were assessed. In addition, the long-term relationship between tumor size changes and exposure was characterized in a dynamic tumor growth inhibition (TGI) model.

In this study, we combined experimentation with modeling and simulation to quantitatively compare the PKPD properties of erlotinib and gefitinib and explore how these findings translate to humans. We critically discuss the relevance of tumor-to-plasma ratio and how modeling and simulation can be applied to profile and select compounds in early drug discovery and development. The proposed approach illustrates how combining in vitro and in vivo experiments with a modeling approach provides insights into the dynamics of tumor growth as a function of drug dosage and the relevance of free drug concentration at the target site. The presented model-based approach suggests a holistic way to profile drug candidates and is applicable to other anticancer drugs.

In vitro cell-cycle analysis

In vitro tests to monitor cell-cycle arrest were conducted at Oncotest. A431 cells (ATCC CRL-1555TM) were acquired from the ATCC in 2009 and have been tested and confirmed for authentication shortly after the experiments by short tandem repeat (STR) analysis at DSMZ. A431 cells in subconfluent cultures were exposed to different concentrations (0.2 or 2 μmol/L) of erlotinib or gefitinib at 37°C for 24, 48, or 72 hours. After washing, fixation with 70% ethanol/15% PBS/15% H20 and staining with 10 μg/mL RNAse A/10 μg/mL propidium iodide in PBS, cells were analyzed by flow cytometry on a FACS using Cytomics FC500 MPL (Beckman Coulter). The percentage of cells within defined cell-cycle stages were calculated as a function of fluorescence intensity (excitation 488 nm, emission 585 nm/FL-2), using Cytomics RXP software as well as MultiCycle AV (Phoenix Flow Systems).

In vivo xenograft mouse model

Female NMRI nu/nu mice were used for all in vivo studies and were supplied by Charles River. Cell line–derived A431 and patient-derived A677 tumor xenografts were grown s.c. in the mice by implantation of tumor fragments. Mice with tumor volume (TV) of 100 to 500 mm3 were randomized to different treatment groups and were sacrificed when TV reached 2,000 mm3. At time of randomization, the mice were between 7 and 11 weeks old with a body weight ranging from 18.2 to 34.3 g. A short-term PK/Target modulation (PK/TM) study to monitor TM after single and repeat dosing and a long-term PK/TGI to monitor tumor growth after repeat dosing were conducted. The treatment schedules of both studies are listed in Supplementary Table S1. In general, both drugs and vehicle were administered daily per os (p.o.) at a volume of 5 to 20 mL/kg. All procedures were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and European Union directives and guidelines and are in compliance with the Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, and the Office of Animal Welfare. The in vivo studies were performed at Oncotest.

In vivo PK/Biomarker–TM study

A short-term PK/biomarker study was conducted to characterize the phosphor- and total Erk1/2 profile after treatment with erlotinib or gefitinib after single and repeat dosing in patient tumor A677-derived xenografts. Mice were treated daily with high-dose erlotinib (100 mg/kg/d in A677 or 75 mg/kg/d in A431) or gefitinib (100 mg/kg/d for both xenografts). Plasma PK, tumor PK, and tumor pERK levels were collected at eight time points within 24 hours after singe and repeat dose. A detailed overview is given in Supplementary Table S1. Because pERK baseline levels of cell line–derived A431 were 4-fold lower as compared with A677 and too close to the quantification limit, changes in pERK upon treatment could not be quantified.

TGI study in xenograft mouse

A long-term PK/TGI study was conducted to characterize tumor shrinkage after repeated oral drug administration. The animals were randomized into seven groups (7 and 8 animals/group for A431 and A677 tumor xenografts, respectively), with one vehicle group and three groups each for erlotinib and gefitinib with low, mid, and high dose. For both compounds, doses of 6.25, 25, or 100 mg/kg/d were administered to A677 xenografts from day 3 to 16 with an interruption for high-dose erlotinib from day 10 to 13. A431 xenografts received doses of 18.75, 37.5, or 75 mg/kg/d for erlotinib or 25, 50, or 100 mg/kg/d for gefitinib, respectively, from day 3 to 10, 14 to 20, and 25 to 38. Plasma PK samples, one sample per mouse, were collected over 24 hours on days 10 and 16 for A677 and on day 3 for A431, as described in Supplementary Table S1. Tumor size was successively measured via two-dimensional caliper measurement, and TV was calculated by|\ ( {a \times {b^2}} ) \times 0.5$|⁠, where “a” is the largest tumor diameter and “b” its perpendicular.

Biomarker (pERK) measurements

Native total protein lysates were prepared from patient-derived A677 xenografted tumors upon terminal sampling. Tumors were explanted, and necrotic areas, large blood vessels, and surrounding mouse tissue were removed. Lysates from tumor samples were prepared under native conditions using lysate buffer containing protease and phosphatase inhibitors. Protein lysates were obtained from tumor tissues after homogenization and centrifugation. Supernatant was collected with a protein concentration of 5 mg/mL. Aliquots were measured in an ELISA to determine pErk expression levels. The kit ELISA #DYC 1018 (R&D Systems GmbH) was used to determine phosphor-Erk1/2 levels, and the assay was performed according to the manufacturer's instructions. Lysates from 2 mice at one time point were pooled, and duplicates were measured. Absorbance was determined at 450 nm (correction wavelength 570 nm) using a plate reader (Perkin Elmer 1420 Multilabel Counter Victor3).

PK analysis

PK measurements were conducted at Accelera using the LC-MS/MS method. Blood and tumor samples were collected from xenograft mice during in vivo antitumor studies, and plasma and tumor concentrations of erlotinib and gefitinib were measured. Tumor homogenates were prepared by ultrasonication. The lower limits of quantification (LLOQ) were 0.488 or 0.952 ng/mL in plasma and 4.88 or 9.52 ng/g in tumor tissue, respectively. For erlotinib, the precision of the assay was 2.9% to 16.2% and 1.5% to 8.4% for the intraday and interday variability, respectively. For gefitinib, the intraday variability was 4.5% to 18% and 1.7% to 6.2% for the interday variability.

PK assessment was performed via noncompartmental analysis (NCA) with Phoenix WinNonlin 6.2 (Phoenix WinNonlin Copyright 1998–2011, Tripos L.P., Phoenix Build 6.2.0.495). Concentrations below the limit of quantitation excluded. Areas under the concentration-time curves (AUC), quantifying the drug exposure in plasma and tumor, were calculated according to the linear trapezoidal rule. Tumor-to-plasma exposure ratios (AUCTumor/AUCPlasma) were computed for both tumor xenografts using short-term PK data.

PK model structure

An oral, 1-compartment, model was fitted to the PK data from the short- and the long-term experiments.

The PK model can be described with the following equations:

formula
formula
formula

where |{q_a}\ $|represents the quantity of drug in the depot compartment, |Ap$| the amount of drug in plasma, |{C_p}$| the plasma concentration, and |D$| is the dose.|\ {k_a}$|⁠, |{k_e}$|⁠, and |V$| correspond to the absorption and elimination constant rates and apparent volume of distribution, respectively. PK parameter estimates are later fixed in the PKPD models.

Target modulation PD model

A direct inhibitory effect model was used to fit the pErk profiles in A677 patient–derived tumor xenografts after single dose using the corresponding PK data. A nonlinear concentration–response relationship was assumed:

formula
formula

where |pErk$| is the phosphorylated downstream biomarker in the tumor according to |{C_p}\ $|values, and |pEr{k_0}$| is the baseline level of biomarker in the absence of drug. |{I_{max}}$| and |I{C_{50}}$| represent maximal inhibitory effect and total plasma concentration producing half of the maximum effect, respectively. |f{u_{mouse}}$| is the unbound fraction of drug in plasma and is 0.055 or 0.06 for erlotinib or gefitinib, respectively (27). |I{C_{{{50}_{free}}}}$|⁠, corresponding to the free plasma concentration producing half of the maximal effect, was derived by correcting for plasma protein binding in mice.

In order to have a metric for efficiency of the two drugs, the |\frac{{AUCE}}{{AU{C_{Tu}}}}$| ratio was computed. AUCE is the area under the effect-time profile (pErk inhibition), and AUCTu is the drug exposure in tumor tissue. AUC was calculated by NCA in Phoenix WinNonlin 6.2 as described earlier, based on the data provided in Fig. 1B, D, and F.

Figure 1.

Plasma and tumor PK profiles for erlotinib and gefitinib in cell line–derived A431 (A and C) and patient-derived A677 (B and D) tumor xenograft models. E, Tumor-to-plasma exposure ratio for both drugs in A677 and A431 xenografts. F, Overlay of pErk levels measured patient-derived A677 tumor xenografts after erlotinib and gefitinib administration (Obs. Erlo and Obs. Gefi) and pErk levels predicted by the model (Pred. Erlo and Pred. Gefi).

Figure 1.

Plasma and tumor PK profiles for erlotinib and gefitinib in cell line–derived A431 (A and C) and patient-derived A677 (B and D) tumor xenograft models. E, Tumor-to-plasma exposure ratio for both drugs in A677 and A431 xenografts. F, Overlay of pErk levels measured patient-derived A677 tumor xenografts after erlotinib and gefitinib administration (Obs. Erlo and Obs. Gefi) and pErk levels predicted by the model (Pred. Erlo and Pred. Gefi).

Close modal

TGI model

TV was described by the sum of a growth- and killing function. A nonlinear growth model was fitted to the tumor growth data of the untreated mice. This model assumes a continuous switch from exponential to linear growth behavior as the TV increases (28). We estimated the drug effect via a killing function involving a direct and linear concentration effect model (29). A schematic representation of the TGI PKPD model is given in Fig. 2, and it is described with the following equation:

formula
Figure 2.

Structure of the TGI model, including the 1-compartmental PK-, the tumor growth-, and the linear direct effect models. The PK model includes an absorbtion rate (ka), elimination rate (ke), and the volume of distribution in plasma (Vpl). TV denotes the tumor volume and k2 the potency parameter of the compound.

Figure 2.

Structure of the TGI model, including the 1-compartmental PK-, the tumor growth-, and the linear direct effect models. The PK model includes an absorbtion rate (ka), elimination rate (ke), and the volume of distribution in plasma (Vpl). TV denotes the tumor volume and k2 the potency parameter of the compound.

Close modal

where |T{V_0}$| is the initially observed TV, |{\lambda _0}$| and |{\lambda _1}$| are the exponential and linear growth parameters, respectively, and |k2$| represents the potency parameter of the compound.

A steady-state concentration for tumor stasis, Css_stasis, was derived as a secondary parameter. It represents the total plasma concentration resulting in a net tumor stasis in mice. To derive this parameter, only the exponential growth phase is considered, which represents the most aggressive scenario of tumor growth. Using the nonlinear growth function causes Css_stasis to be TV dependent (Supplementary Fig. S1). The use of Css_stasis and its correlation to clinical doses in humans was introduced by Rocchetti and colleagues (30). This parameter allows for quantitative comparison of both drugs and is used for translation to humans to derive the anticipated efficacious dose.

Predicting human efficacious dose

We assume that free Css_stasis in plasma is a surrogate for free Css_stasis at the target site and that it is the same between mouse and human (denoted as Css_stasisfree), allowing us to convert this into a human efficacious dose by taking into account species differences in PK (e.g., plasma binding, clearances, and volume of distribution; refs. 20, 27). The translation from the simulated total Css_stasis in plasma of mouse to a human daily dose was done as follows:

formula

where Css_stasisfree is assumed to be the same in mouse and human and is derived by correction for the unbound fraction (fu) accounting for the species differences in plasma protein binding:

formula

where Css-stasismouse is the simulated total plasma concentration in mouse at steady state to achieve tumor stasis, |f{u_{human}}$| the unbound fraction in plasma, and Css-stasishuman the total plasma concentration to achieve tumor stasis in human. The total plasma exposure in human required for tumor stasis, AUCss-stasishuman, could then be derived:

formula

where |\tau $| is the dosing interval (i.e., 1 day in our case). Eventually, a projected human daily dose, |D/{d_{human,}}$| was derived by accounting for clearance in humans, |C{L_{human}}$|⁠:

formula

Values for |f{u_{human}}$| and for |C{L_{human}}$| were derived from literature. |f{u_{human}}$| was 0.084 or 0.089 (27), and |C{L_{human}}$| was 127 or 857 L/d (20) for erlotinib and gefitinib, respectively.

Parameter estimation

PKPD parameters were estimated using a population approach with Monolix Version 4.2.2. (Lixoft), allowing for estimation of fixed and random effects for each model parameter and a residual error in one step (31). Residual errors for the PK were assumed to be proportional to predicted concentrations. For the PD, a combined error model was selected. Diagnostic plots were inspected to select the appropriate error model.

PK parameters were estimated separately from PD parameters to fix the individual PK input in the PKPD modeling. PD parameters were estimated by combining erlotinib and gefitinib datasets while assuming similar tumor growth for both compounds, which allowed us to better differentiate efficacy from tumor growth parameters. In case of the target modulation experiment, estimated random effects were the combination of inter- and intraindividual variability. To allow a precise and stable estimation of fixed effects, random effects were fixed to low values if necessary. Model evaluation and selection were based on model convergence, precision of the parameter estimates, fitting criteria (Akaike Information Criteria, AIC), and visual inspection of diagnostic tools (Visual Predictive checks, residuals, and observed vs. predicted plots).

Cell cycle

The effects of both compounds on cell-cycle arrest at a cellular level were explored in vitro. Cell-cycle analysis was performed by flow cytometry with the A431 cells. Flow cytometry histograms after 24-hour incubation time are depicted in Fig. 3A and B. Both compounds show a similar in vitro cell-cycle arrest between the G1 and S phase. The cell fraction increased in the G1 phase and decreased in the S- and the G2–M phases in a time- and concentration-dependent manner. The changes in the percentage of cells observed in specific cell-cycle phases were similar between erlotinib and gefitinib (Fig. 3C). These observations suggest that both drugs exhibit the same pharmacologic response in vitro.

Figure 3.

A, Flow cytometry data after 24-hour incubation time for control and erlotinib at 0.2 and 2 μmol/L. B, Flow cytometry data after 24-hour incubation time for control and gefitinib at 0.2 and 2 μmol/L. C, Proportion of cells in the different phases of cell cycle with and without tyrosine kinase inhibitor drugs. Experiment was conducted in cells from the A431 cell line.

Figure 3.

A, Flow cytometry data after 24-hour incubation time for control and erlotinib at 0.2 and 2 μmol/L. B, Flow cytometry data after 24-hour incubation time for control and gefitinib at 0.2 and 2 μmol/L. C, Proportion of cells in the different phases of cell cycle with and without tyrosine kinase inhibitor drugs. Experiment was conducted in cells from the A431 cell line.

Close modal

Tumor-to-plasma exposure ratio

After the in vitro assessment, the PK behavior of the two compounds was investigated in vivo on the basis of their total plasma and tumor exposure. Plasma and tumor PK time course profiles for both compounds and xenograft models are presented in Fig. 1A–D. NCA was done to derive the AUC as readout of exposure in plasma and tumor for the two drugs. In both cell line–derived A431 and patient-derived A677 tumor models, gefitinib shows higher tumor-to-plasma exposure ratio as presented in Fig. 1E. The ratio of tumor-to-plasma exposure (AUCTumor/AUCPlasma) was 21 times higher for gefitinib than erlotinib in the A677 xenograft model and 11-fold higher in the A431 tumor xenograft model.

Target modulation in tumor versus tumor and plasma concentrations

Next, we explored how the differences in the observed disposition properties compared with the downstream effects in the tumor. Plasma and tumor PK and biomarker time course profiles were collected in a short-term, biomarker in vivo study. The observed data were used to relate the plasma and tumor exposure to the in vivo target modulation effects of the two compounds on a molecular biomarker. Dose-normalized concentration-time profiles (Supplementary Fig. S2A–S2D) superimposed and, therefore, dose linear PK was assumed. Nonlinearity was observed only for gefitinib PK in the long-term study in patient-derived A677 xenografts at the end of treatment on day 16. The observed slightly higher Cmax and lower central clearance upon lower dose was well descriped by a descriptive model structure assuming a dose-dependent volume of distribution. Other nonlinear models such as a Michaelis–Menten model assuming a saturable CL were not applicable because the parameters could not be identified. PK profiles after single and repeat dose were similar (Supplementary Fig. S2E and S2F). Based on the principles of parsimony, a 1-compartmental linear PK model was selected, and goodness-of-fit plots (Supplementary Fig. S3) demonstrate that the model captured adequately the PK profiles.

Total Erk levels in LXF A677 were constant at 1,001|\pm $|126 pg/mL over time in control and treatment groups (Supplementary Fig. S4). Measured pErk data in A677 after multiple dosing did not pass quality control due to unexplained fluctuations in the control group. Both drugs caused a rapid decrease in pErk levels (Fig. 1F). A maximal inhibition of pErk of about 70% was observed after 30 minutes for both drugs and was associated with a tumor concentration of 4,965 μg/L for erlotinib and 4,865 μg/L of gefitinib. For gefitinib, the tumor PK increased further and remained above the tumor concentration associated with maximal pERK inhibition for 30 to 1,440 minutes. However, during this period, the pERK profile started to increase back to baseline, suggesting inactive gefitinib concentration in tumor tissue. The tumor PK concentration of erlotinib decreased after 30 minutes, and this decrease was associated with the recovery of the pERK signal. The computed AUCE/AUCTu ratio is 8.3 times higher for erlotinib, indicating an 8.3-fold higher efficiency for erlotinib than for gefitinib.

A direct inhibitory effect model was fitted to the observed data. Baseline pErk concentration and the maximal inhibitory effect of both drugs were fixed. An overlay of the model predictions and the observed pErk downregulation (Fig. 1F) suggests that the model captures most of the observed data well. Gefitinib appears to be five times more potent when the estimated IC50 values based on total plasma concentration for gefitinib (107 μg/L or 0.23 μmol/L) and erlotinib (513 μg/L or 1.31 μmol/L) are compared. Parameter estimates are presented in Supplementary Table S2.

Modeling PK/TGI

After investigating the effects of erlotinib and gefitinib at a cellular and molecular level in vitro and in vivo, we compared the effects of the two compounds in vivo on tumor tissue. The time course of tumor growth in patient-derived A677 and cell line–derived A431 xenograft mice was monitored in a long-term experiment in TKI-treated and -untreated mice. A TGI model was fitted to the data in order to relate total plasma concentration to the in vivo TGI effect and quantify the effect of the respective drug. The diagnostic plots corresponding to the TGI PK modeling are shown in Supplementary Fig. S5. Prediction distribution and observed versus predicted plots for the A677 and A431 xenograft mice are presented in Fig. 4 and Supplementary Fig. S6, respectively. Overall, the model prediction describes the observed data well particularly in the low- and mid-dose group of both erlotinib and gefitinib. A bias is observed in the high-dose group, and the effect of erlotinib is underpredicted, whereas the effect of gefitinib is overpredicted in A677 (Fig. 4). For high dose in A431, the drug effect was overpredicted for both drugs (Supplementary Fig. S6). Despite this model misspecification in the high-dose group, the model is still considered suited to address the main question on how higher tumor uptake translates to efficacy.

Figure 4.

Diagnostic plots of the TGI PKPD model fitted to the A677 TGI data. A, Prediction distribution plots: black dots represent observed TVs, lines represent median prediction of the model, and gray zones represent 95% confidence interval of the prediction. B, Observed versus predicted plots: dots represent observed TVs and lines represent identity line.

Figure 4.

Diagnostic plots of the TGI PKPD model fitted to the A677 TGI data. A, Prediction distribution plots: black dots represent observed TVs, lines represent median prediction of the model, and gray zones represent 95% confidence interval of the prediction. B, Observed versus predicted plots: dots represent observed TVs and lines represent identity line.

Close modal

The PKPD parameters were precisely estimated with relative standard error (RSE) values lower than 40%. High interindividual variability was estimated on growth and efficacy parameters (>35% variability), which is also reflected in the wide 95% confidence interval around the predicted population profile in the diagnostic plots. The high interindividual variability is precisely estimated and is indicative for the variability observed in the experimental data. This observation is most likely attributed to heterogeneous growth kinetics as a result of the xenograft tumor procedure which involved implanting of tumor fragments in the mice. The detailed parameter estimates are presented in Table 1. In both tumor xenograft models, in vivo potency parameter estimates (k2) were about 3-fold higher for gefitinib than for erlotinib (Table 1). The secondary parameter Css_stasis, which represents the concentration at steady state needed to reach tumor stasis, was predicted to be 80 to 90 μg/L for erlotinib and 27 to 31 μg/L for gefitinib (Table 2).

Table 1.

PKPD parameter estimates of TGI studies

ErlotinibGefitinib
ParametersEstimateRSE (%)VarRSE (%)EstimateRSE (%)VarRSE (%)
PK modeling (TGI)–A677 
 Ka (1/d) 55.0 55.0 
 V (L) 0.127 15 0.251 1.40 11 0.0278 
 Ke (1/d) 7.56 10 0.332 39 3.87 10 0.352 20 
 Cl = V x Ke (L/d) 0.96 5.42 
 I (V modifier) 0.00772 22 
 b 0.585 15 0.373 12 
PK modeling (TGI)–A431 
 Ka (1/d) 55.0 55.0 
 V (L) 0.120 0.190 23 0.472 14 0.466 20 
 Ke (1/d) 7.85 0.235 26 6.10 10 0.228 22 
 Cl = V x Ke (L/d) 0.942 2.88 
 A677 A431 
Parameters Estimate RSE (%) Var RSE (%) Estimate RSE (%) Var RSE (%) 
PKPD modeling (TGI) 
 K2_erlo (L/μg/d) 0.000117 17 0.654 21 0.0000401 16 0.585 21 
 K2_gefi (L/μg/d) 0.000430 12 0.361 26 0.000114 15 0.568 20 
 λ0 (1/d) 0.0971 0.456 13 0.0290 0.321 14 
 λ1 (mm3/d) 127 13 0.710 298 37 0.680 
 TV0 (mm3122 0.368 11 114 0.423 11 
 a (mm314.1 4.32 30 
 b 0.0907 0.139 
ErlotinibGefitinib
ParametersEstimateRSE (%)VarRSE (%)EstimateRSE (%)VarRSE (%)
PK modeling (TGI)–A677 
 Ka (1/d) 55.0 55.0 
 V (L) 0.127 15 0.251 1.40 11 0.0278 
 Ke (1/d) 7.56 10 0.332 39 3.87 10 0.352 20 
 Cl = V x Ke (L/d) 0.96 5.42 
 I (V modifier) 0.00772 22 
 b 0.585 15 0.373 12 
PK modeling (TGI)–A431 
 Ka (1/d) 55.0 55.0 
 V (L) 0.120 0.190 23 0.472 14 0.466 20 
 Ke (1/d) 7.85 0.235 26 6.10 10 0.228 22 
 Cl = V x Ke (L/d) 0.942 2.88 
 A677 A431 
Parameters Estimate RSE (%) Var RSE (%) Estimate RSE (%) Var RSE (%) 
PKPD modeling (TGI) 
 K2_erlo (L/μg/d) 0.000117 17 0.654 21 0.0000401 16 0.585 21 
 K2_gefi (L/μg/d) 0.000430 12 0.361 26 0.000114 15 0.568 20 
 λ0 (1/d) 0.0971 0.456 13 0.0290 0.321 14 
 λ1 (mm3/d) 127 13 0.710 298 37 0.680 
 TV0 (mm3122 0.368 11 114 0.423 11 
 a (mm314.1 4.32 30 
 b 0.0907 0.139 

NOTE: Parameter estimates of absorption rate (Ka), elimination rate (Ke), volume of distribution (V), Clearance (Cl), potency parameter (K2), exponential (λ0) and linear (λ1) growth rate of tumor, initial tumor volume (TV0), relative standard errors (RSE), and interindividual variability (Var). “I” indicates a dose-dependent change in V for gefitinib. The residual error is given (a) additive error model and (b) proportional error model.

Table 2.

Projected and prescribed human doses

Simulations/translations
A677A431
ParameterErlotinibGefitinibErlotinibGefitinib
Css_stasisfree (μg/L) 91 (0.23 μmol/L) 27 (0.06 μmol/L) 80 (0.2 μmol/L) 31 (0.07 μmol/L) 
Projected human dose (mg/d) 138 261 120 294 
Parameter Erlotinib Gefitinib 
Clinically prescribed dose 
Prescribed human dose (mg/d) 150 250 
Simulations/translations
A677A431
ParameterErlotinibGefitinibErlotinibGefitinib
Css_stasisfree (μg/L) 91 (0.23 μmol/L) 27 (0.06 μmol/L) 80 (0.2 μmol/L) 31 (0.07 μmol/L) 
Projected human dose (mg/d) 138 261 120 294 
Parameter Erlotinib Gefitinib 
Clinically prescribed dose 
Prescribed human dose (mg/d) 150 250 

NOTE: Comparison of simulated free concentration to achieve tumor stasis (Css_stasisfree) derived from xenograft mouse data and the resulting daily dose needed to reach tumor stasis in humans for gefitinib and erlotinib. The clearance and plasma protein binding values used in calculations for both species were retrieved from literature (20, 27).

Translation of nonclinical efficacious concentration into a human efficacious dose

After characterizing the PKPD relationship in xenograft mice, we translated our results from mouse to human. Css_stasisfree values, human clearance, and the predicted doses in human for both compounds and for both tumor types are presented in Table 2. The projected doses by our modeling approach were approximately 140 and 120 mg for Erlotinib and approximately 260 and 290 mg for gefitinib for A677 and A431, respectively. The derived doses from the two xenograft models are very similar, and they correspond well to the recommended standard therapeutic doses of 150 mg and 250 mg for erlotinib and gefitinib, respectively (10, 18).

The goal of this study was to differentiate erlotinib and gefitinib, both small molecular tyrosine kinase inhibitors of the human EGFR. The two compounds were selected based on the strong tumor uptake differences observed for erlotinib and gefitinib despite similar molecular specificity. The main purpose of the presented study was to better understand the impact of tumor uptake on in vivo efficacy. Gefitinib shows a higher tumor uptake in cancer patients, and we explored the potential impact on pharmacologic and antitumor activity in xenograft mice. We therefore combine experimentation with modeling to enable robust comparison of the pharmacologic responses of both drugs. In vitro cell-cycle analysis demonstrated that both drugs exhibit similar antiproliferative effects and induce a G1–S arrest. A major difference between both drugs was detected when comparing disposition in vivo. An 11- to 21-fold higher tumor-to-plasma ratio was observed with gefitinib in cell line– and patient-derived xenograft mice, which is in line with other in vivo studies and clinical studies (21, 22). Furthermore, despite the higher total exposure in tumors achieved with gefitinib, the effects on target modulation in tumors, monitored through pERK inhibition profiles, were similar for both drugs; the maximal effect on pErk inhibition was quickly achieved after administration of erlotinib and gefitinib. After the maximal effect was observed, the total concentration of gefitinib in tumor tissue further increased while pERK inhibition decreased. In contrast to erlotinib, the tumor PK profile of gefitinib did not follow the time course of pERK inhibition in tumor. These findings suggest that the high tumor-to-plasma ratio is not necessarily associated with a higher concentration of pharmacologically active drug at the target site and may be explained by a higher concentration bound to the tumor tissue. In summary, this short-term biomarker experiment enabled to quantify target engagement as an indirect approach to estimate the free drug concentration at the site of action. Because this experiment requires terminal sampling, it was not feasible to include this in the long-term studies due to ethical constraints.

The higher tumor-to-plasma ratio observed with gefitinib can also be explained by the physicochemical properties of those drugs, especially the higher lipophilicity of gefitinib (10). Furthermore, the tumor microenvironment becomes progressively hypoxic and the pH decreases when moving away from the blood supply in the tumor tissue (32, 33). Erlotinib (pKa 5.4) and gefitinib (pKa = 5.4 and 7.2) are both weak bases (19). Basic compounds ionize when pH is lower than their pKa value. Gefitinib will ionize faster in the acidic tumor microenvironment. Ionized compounds cannot cross membranes and might bind to negatively charged phospholipids of cell membranes (ionic trapping), decreasing the pharmacologically active fraction of drug in the tumor environment. This might explain why despite the higher total concentration of gefitinib than erlotinib at the target site, similar efficacy was observed for both compounds. The microenvironment at the target site is an important aspect in drug development that needs to be taken into account early on to distinguish two candidates on target accessibility as well as target concentrations.

PKPD modeling was conducted to quantitatively compare in vivo potency on biomarker and TGI response of two molecules with similar molecular specificity and distinct disposition characteristics. The direct linear effect PKPD model describes the data reasonably well despite the bias observed for the highest dose group. Various PKPD models (nonlinear effect-, Emax model, additional cytostatic drug effect,…) were tested (data not shown) and did not improve the model performance. It has been reported that resistance would occur under erlotinib or gefitinib treatment (13, 34). Integration of time-varying effect due to resistance in the model structure may improve the model performance. Genotyping of xenograft mice tumors to assess genetic alterations leading to TKI drug resistance (35) was not performed in this study.

Assuming passive exchanges from plasma to target site and an absence of active transporters, we considered free exposure in plasma to be a surrogate of free exposure at the target site when reaching steady-state equilibrium (36–38). Based on free concentration, gefitinib was around 5-fold more potent in target modulation and 3-fold more potent in TGI in mice. However, this 3-fold higher potency in TGI experiments is a best case scenario, given the fact that the TGI response of the high-dose group was slightly overpredicted with gefitinib and under-predicted with erlotinib. However, the predicted human dose is in the same range as the human recommended doses for both compounds (10, 18) due to gefitinib's higher clearance in humans. It should be noted that the recommended standard therapeutic dose of gefitinib (250 mg daily) is administered at approximately one third of its MTD (39) and is considered close to the biological dose. However, the recommended dose for Erlotinib (150 mg daily) meets the classical definition of MTD (40). Therefore, the striking agreement between projected and recommended dose may be coincidental for erlotinib. PKPD models accounting for species-specific differences are valuable tools to profile and select compounds with a high probability of success (41).

During nonclinical drug development, optimization of compound properties goes beyond the assessment of single-drug metabolism and pharmacokinetic (DMPK) properties to consider the exposure–response relationship for both safety and efficacy. Knowledge about the active drug at the target site is crucial to optimize compound selection (41, 42). Active drug at the target site is assumed to be the fraction of the compound available to bind to the target, the free fraction (43). As seen in our comparison of erlotinib and gefitinib, for optimizing or comparing compounds, the total concentration exposure at the target site can be misleading because the free fraction at the site of action varies from compound to compound according to physicochemical properties and the target microenvironment.

Free drug concentration at tissue level is commonly measured by microdialysis (44, 45). If the free fraction at the target site is not available, it might be predicted with compound properties or by generating appropriate in vivo data. Side partition coefficients are commonly used to predict the free fraction at a target. Partition coefficients (Kp,u) define the ratio between plasma and tissue concentration and are valuable indicators of a drug's tissue distribution. The use of partition coefficients and their application in drug discovery and development has been reviewed by Mariappan and colleagues (46).

In the present study, we have shown how the combination of in vitro and in vivo experiments with mathematical modeling can be used to quantitatively characterize the PKPD properties of drugs and propose a way how these findings could be translated to humans. These projections are useful to select of most favorable dose in the early clinical trials. The FDA guideline for anticancer pharmaceuticals defines “the goal of selection a starting dose is to identify a dose that is expected to have pharmacologic effects and is reasonably safe” (47). Preclinical PK/PD modeling helps to project the pharmacologic response in humans by quantifying the exposure response relationship in nonclinical species and subsequently accounting for the species differences. Moreover, profiling the therapeutic index in preclinical studies with subsequent prediction to humans is a critical step particularly owing to the narrow therapeutic index of most anticancer drugs. However, dose optimization needs to be further refined during the clinical trials. The comparative assessment done in this paper provides already valuable insights into the link between PK, tumor disposition in in vivo activity. Further studies including additional TKI inhibitors could be of value to further evaluate the impact of differences in molecular specificity on efficacy and to get mechanistic insights in what is driving the variability in patients' response. However, the main purpose of the proposed study was to better understand the impact of tumor uptake on in vivo efficacy and suggests a holistic view to profile compounds in preclinical studies. This proposed model-based approach is broadly applicable to other anticancer drugs. We demonstrated that total exposure at the target site is not sufficient to explain the pharmacologic effect. Despite having a lower tumor-to-plasma ratio and a 3- to 5-fold lower in vivo potency compared with gefitinib, the prescribed daily dose of erlotinib in cancer patients is 50% lower. Our results suggest that the total exposure at target site may not be suited to rank compounds for efficacy, and an integrated modeling and experimental approach can assess efficacy more accurately. The presented model-based approach suggests how to leverage a broad range of experimental information using mathematical modeling in order to profile drug candidates and support dose finding for entry in human trials (EIH) in late-stage patients and can potentially be applied and integrated in the drug discovery and development process of other anticancer drugs.

G. Hoffmann has ownership interest (including patents) in Roche non-voting equity securities. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Lavé, A.-C. Walz

Development of methodology: A.-C. Walz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Hoffmann

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.J. Eigenmann, N. Frances, G. Hoffmann, A.-C. Walz

Writing, review, and/or revision of the manuscript: M.J. Eigenmann, N. Frances, T. Lavé, A.-C. Walz

Study supervision: A.-C. Walz

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Kola
I
,
Landis
J
. 
Can the pharmaceutical industry reduce attrition rates?
Nat Rev Drug Discov
2004
;
3
:
711
5
.
2.
Zhou
Q
,
Gallo
JM
. 
The pharmacokinetic/pharmacodynamic pipeline: translating anticancer drug pharmacology to the clinic
.
AAPS J
2011
;
13
:
111
20
.
3.
Aarons
L
,
Karlsson
MO
,
Mentre
F
,
Rombout
F
,
Steimer
JL
,
van Peer
A
. 
Role of modelling and simulation in phase I drug development
.
Eur J Pharm Sci
2001
;
13
:
115
22
.
4.
Sheiner
LB
. 
Learning versus confirming in clinical drug development
.
Clin Pharmacol Ther
1997
;
61
:
275
91
.
5.
Danhof
M
,
de Lange
EC
,
Della Pasqua
OE
,
Ploeger
BA
,
Voskuyl
RA
. 
Mechanism-based pharmacokinetic-pharmacodynamic (PK-PD) modeling in translational drug research
.
Trends Pharm Sci
2008
;
29
:
186
91
.
6.
Mould
DR
,
Walz
AC
,
Lave
T
,
Gibbs
JP
,
Frame
B
. 
Developing exposure/response models for anticancer drug treatment: special considerations
.
CPT Pharmacometrics Syst Pharmacol
2015
;
4
:
e00016
.
7.
Yano
S
,
Kondo
K
,
Yamaguchi
M
,
Richmond
G
,
Hutchison
M
,
Wakeling
A
, et al
Distribution and function of EGFR in human tissue and the effect of EGFR tyrosine kinase inhibition
.
Anticancer Res
2003
;
23
:
3639
50
.
8.
Hynes
NE
,
Horsch
K
,
Olayioye
MA
,
Badache
A
. 
The ErbB receptor tyrosine family as signal integrators
.
Endocr Relat Cancer
2001
;
8
:
151
9
.
9.
Wakeling
AE
,
Guy
SP
,
Woodburn
JR
,
Ashton
SE
,
Curry
BJ
,
Barker
AJ
, et al
ZD1839 (Iressa): An orally active inhibitor of epidermal growth factor signaling with potential for cancer therapy
.
Cancer Res
2002
;
62
:
5749
54
.
10.
Bronte
G
,
Rolfo
C
,
Giovannetti
E
,
Cicero
G
,
Pauwels
P
,
Passiglia
F
, et al
Are erlotinib and gefitinib interchangeable, opposite or complementary for non-small cell lung cancer treatment? Biological, pharmacological and clinical aspects
.
Crit Rev Oncol Hematol
2014
;
89
:
300
13
.
11.
Gill
AL
,
Verdonk
M
,
Boyle
RG
,
Taylor
R
. 
A comparison of physicochemical property profiles of marketed oral drugs and orally bioavailable anti-cancer protein kinase inhibitors in clinical development
.
Curr Top Med Chem
2007
;
7
:
1408
22
.
12.
Guo
A
,
Villen
J
,
Kornhauser
J
,
Lee
KA
,
Stokes
MP
,
Rikova
K
, et al
Signaling networks assembled by oncogenic EGFR and c-Met
.
Proc Natl Acad Sci U S A
2008
;
105
:
692
7
.
13.
Mok
TS
,
Wu
YL
,
Thongprasert
S
,
Yang
CH
,
Chu
DT
,
Saijo
N
, et al
Gefitinib or carboplatin-paclitaxel in pulmonary adenocarcinoma
.
N Engl J Med
2009
;
361
:
947
57
.
14.
Shepherd
FA
,
Rodrigues Pereira
J
,
Ciuleanu
T
,
Tan
EH
,
Hirsh
V
,
Thongprasert
S
, et al
Erlotinib in previously treated non-small-cell lung cancer
.
N Engl J Med
2005
;
353
:
123
32
.
15.
Paez
JG
,
Janne
PA
,
Lee
JC
,
Tracy
S
,
Greulich
H
,
Gabriel
S
, et al
EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy
.
Science
2004
;
304
:
1497
500
.
16.
Albanell
J
,
Gascon
P
. 
Small molecules with EGFR-TK inhibitor activity
.
Curr Drug Targets
2005
;
6
:
259
74
.
17.
Yewale
C
,
Baradia
D
,
Vhora
I
,
Patil
S
,
Misra
A
. 
Epidermal growth factor receptor targeting in cancer: a review of trends and strategies
.
Biomaterials
2013
;
34
:
8690
707
.
18.
Rukazenkov
Y
,
Speake
G
,
Marshall
G
,
Anderton
J
,
Davies
BR
,
Wilkinson
RW
, et al
Epidermal growth factor receptor tyrosine kinase inhibitors: similar but different?
Anti-cancer Drugs
2009
;
20
:
856
66
.
19.
Scheffler
M
,
Di Gion
P
,
Doroshyenko
O
,
Wolf
J
,
Fuhr
U
. 
Clinical pharmacokinetics of tyrosine kinase inhibitors: focus on 4-anilinoquinazolines
.
Clin Pharmacokinet
2011
;
50
:
371
403
.
20.
van Erp
NP
,
Gelderblom
H
,
Guchelaar
HJ
. 
Clinical pharmacokinetics of tyrosine kinase inhibitors
.
Cancer Treat Rev
2009
;
35
:
692
706
.
21.
Armour
AA
,
Watkins
CL
. 
The challenge of targeting EGFR: Experience with gefitinib in nonsmall cell lung cancer
.
Eur Respir Rev
2010
;
19
:
186
96
.
22.
McKillop
D
,
Partridge
EA
,
Kemp
JV
,
Spence
MP
,
Kendrew
J
,
Barnett
S
, et al
Tumor penetration of gefitinib (Iressa), an epidermal growth factor receptor tyrosine kinase inhibitor
.
Mol Cancer Ther
2005
;
4
:
641
9
.
23.
Ng
SS
,
Tsao
MS
,
Nicklee
T
,
Hedley
DW
. 
Effects of the epidermal growth factor receptor inhibitor OSI-774, Tarceva, on downstream signaling pathways and apoptosis in human pancreatic adenocarcinoma
.
Mol Cancer Ther
2002
;
1
:
777
83
.
24.
Emery
IF
,
Battelli
C
,
Auclair
PL
,
Carrier
K
,
Hayes
DM
. 
Response to gefitinib and erlotinib in non-small cell lung cancer: a retrospective study
.
BMC Cancer
2009
;
9
:
333
.
25.
Buck
E
,
Eyzaguirre
A
,
Brown
E
,
Petti
F
,
McCormack
S
,
Haley
JD
, et al
Rapamycin synergizes with the epidermal growth factor receptor inhibitor erlotinib in non-small-cell lung, pancreatic, colon, and breast tumors
.
Mol Cancer Ther
2006
;
5
:
2676
84
.
26.
Zhou
JP
,
Chen
X
,
Feng
S
,
Luo
SD
,
Pan
YL
,
Zhong
L
, et al
Systems biology modeling reveals a possible mechanism of the tumor cell death upon oncogene inactivation in EGFR addicted cancers
.
PLoS One
2011
;
6
:
e28930
.
27.
Hoshino-Yoshino
A
,
Kato
M
,
Nakano
K
,
Ishigai
M
,
Kudo
T
,
Ito
K
. 
Bridging from preclinical to clinical studies for tyrosine kinase inhibitors based on pharmacokinetics/pharmacodynamics and toxicokinetics/toxicodynamics
.
Drug Metab Pharmacokinet
2011
;
26
:
612
20
.
28.
Koch
G
,
Walz
A
,
Lahu
G
,
Schropp
J
. 
Modeling of tumor growth and anticancer effects of combination therapy
.
J Pharmacokinet Pharmacodyn
2009
;
36
:
179
97
.
29.
Jusko
WJ
. 
Pharmacodynamics of chemotherapeutic effects: dose-time-response relationships for phase-nonspecific agents
.
J Pharm Sci
1971
;
60
:
892
5
.
30.
Rocchetti
M
,
Simeoni
M
,
Pesenti
E
,
De Nicolao
G
,
Poggesi
I
. 
Predicting the active doses in humans from animal studies: a novel approach in oncology
.
Eur J Cancer
2007
;
43
:
1862
8
.
31.
Sheiner
LB
,
Beal
SL
. 
Evaluation of methods for estimating population pharmacokinetics parameters. I. Michaelis-Menten model: routine clinical pharmacokinetic data
.
J Pharmacokinet Biopharm
1980
;
8
:
553
71
.
32.
Jain
RK
. 
Transport of molecules, particles, and cells in solid tumors
.
Annu Rev Biomed Eng
1999
;
1
:
241
63
.
33.
Fang
JS
,
Gillies
RD
,
Gatenby
RA
. 
Adaptation to hypoxia and acidosis in carcinogenesis and tumor progression
.
Semin Cancer Biol
2008
;
18
:
330
7
.
34.
Chmielecki
J
,
Foo
J
,
Oxnard
GR
,
Hutchinson
K
,
Ohashi
K
,
Somwar
R
, et al
Optimization of dosing for EGFR-mutant non-small cell lung cancer with evolutionary cancer modeling
.
Sci Transl Med
2011
;
3
:
90ra59
.
35.
Yu
HA
,
Arcila
ME
,
Rekhtman
N
,
Sima
CS
,
Zakowski
MF
,
Pao
W
, et al
Analysis of tumor specimens at the time of acquired resistance to EGFR-TKI therapy in 155 patients with EGFR-mutant lung cancers
.
Clin Cancer Res
2013
;
19
:
2240
7
.
36.
Smith
DA
,
Di
L
,
Kerns
EH
. 
The effect of plasma protein binding on in vivo efficacy: misconceptions in drug discovery
.
Nat Rev Drug Discov
2010
;
9
:
929
39
.
37.
Trainor
GL
. 
The importance of plasma protein binding in drug discovery
.
Expert Opin Drug Discov
2007
;
2
:
51
64
.
38.
Bohnert
T
,
Gan
LS
. 
Plasma protein binding: From discovery to development
.
J Pharm Sci
2013
;
102
:
2953
94
.
39.
Baselga
J
,
Rischin
D
,
Ranson
M
,
Calvert
H
,
Raymond
E
,
Kieback
DG
, et al
Phase I safety, pharmacokinetic, and pharmacodynamic trial of ZD1839, a selective oral epidermal growth factor receptor tyrosine kinase inhibitor, in patients with five selected solid tumor types
.
J Clin Oncol
2002
;
20
:
4292
302
.
40.
Hidalgo
M
,
Siu
LL
,
Nemunaitis
J
,
Rizzo
J
,
Hammond
LA
,
Takimoto
C
, et al
Phase I and pharmacologic study of OSI-774, an epidermal growth factor receptor tyrosine kinase inhibitor, in patients with advanced solid malignancies
.
J Clin Oncol
2001
;
19
:
3267
79
.
41.
Cook
D
,
Brown
D
,
Alexander
R
,
March
R
,
Morgan
P
,
Satterthwaite
G
, et al
Lessons learned from the fate of AstraZeneca's drug pipeline: a five-dimensional framework
.
Nat Rev Drug Discov
2014
;
13
:
419
31
.
42.
Morgan
P
,
Van Der Graaf
PH
,
Arrowsmith
J
,
Feltner
DE
,
Drummond
KS
,
Wegner
CD
, et al
Can the flow of medicines be improved? Fundamental pharmacokinetic and pharmacological principles toward improving Phase II survival
.
Drug Discov Today
2012
;
17
:
419
24
.
43.
Wright
JD
,
Boudinot
FD
,
Ujhelyi
MR
. 
Measurement and analysis of unbound drug concentrations
.
Clin Pharmacokinet
1996
;
30
:
445
62
.
44.
Bito
L
,
Davson
H
,
Levin
E
,
Murray
M
,
Snider
N
. 
The concentrations of free amino acids and other electrolytes in cerebrospinal fluid, in vivo dialysate of brain, and blood plasma of the dog
.
J Neurochem
1966
;
13
:
1057
67
.
45.
Muller
M
,
v Osten
B
,
Schmid
R
,
Piegler
E
,
Gerngross
I
,
Buchegger
H
, et al
Theophylline kinetics in peripheral tissues in vivo in humans
.
Naunyn Schmiedebergs Arch Pharmacol
1995
;
352
:
438
41
.
46.
Mariappan
TT
,
Mandlekar
S
,
Marathe
P
. 
Insight into tissue unbound concentration: utility in drug discovery and development
.
Curr Drug Metab
2013
;
14
:
324
40
.
47.
U.S. Department of Health and Human Services, FDA, Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation and Research (CBER)
. 
Guidance for Industry: S9 Nonclinical Evaluation for Anticancer Pharmaceuticals March 2010
.