Prediction of Active Drug Plasma Concentrations Achieved in Cancer Patients by Pharmacodynamic Biomarkers Identified from the Geo Human Colon Carcinoma Xenograft Model

Materials and Methods

Chemical reagents and antibodies. Complete protease inhibitor tablets were from Roche Diagnostics (Indianapolis, IN). Rabbit anti-mouse IgG and MicroBCA reagents were from Pierce (Rockford, IL). Anti-phosphotyrosine antibody (PY20) and anti-MAPK antibody (sc-94) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse monoclonal anti-phospho-EGFR[pY1068] antibody (catalogue #2236) and rabbit polyclonal anti-phospho-MAPK (catalogue #9101) were purchased from Cell Signaling (Beverly, MA). Mouse monoclonal anti-EGFR antibody was purchased from Zymed Laboratories, Inc. (South San Francisco, CA). Anti-rabbit IgG antibody conjugated with horseradish peroxidase was purchased from BD Bioscience (Lexington, KY). Unless otherwise specified, all other chemicals and reagents were from Sigma (St. Louis, MO). Sterile buffers and solutions were obtained from Life Technologies (Carlsbad, CA). Sterile tissue culture ware was obtained from Fisher Scientific Co. (Hanover Park, IL).

Animals. Female, nude mice, 5 to 6 weeks of age, were obtained from Harlan Sprague-Dawley Co. (Indianapolis, IN), and maintained in an ammonia-free environment in a defined and pathogen-free colony. Animals were quarantined for ∼3 weeks prior to their use for tumor propagation and drug efficacy testing. They were given food and water ad libitum. Tumors were propagated in nude mice as s.c. transplants using tumor fragments obtained from donor mice. Tumor passage occurred approximately every 2 to 4 weeks. Tumors were then allowed to grow to the predetermined size range (usually between 100 and 200 mg, tumors outside the range were excluded) and animals were evenly distributed to various treatment and control groups. All studies were done in accordance with Bristol-Myers Squibb and the American Association for Accreditation of Laboratory Animal Care protocols.

Drug formulation and administration. Cetuximab (lot #00C01178) was supplied by Imclone Systems, Inc. (New York, NY) at a concentration of 2 mg/mL in a buffer consisting of 10 mmol/L sodium phosphate and 145 mmol/L sodium chloride (pH 7). For the pharmacodynamic study requiring lower concentrations of cetuximab, the stock solution was diluted with sterile PBS (pH 7.4). Cetuximab was administered i.p. at a constant volume of 0.5 mL.

Western analysis of the phospho-EGFR[pY1068] and total EGFR in tumor xenograft samples. Mice bearing Geo xenografts (n = 3) were given a single i.p. dose of cetuximab. Tumors were surgically removed and split in half at the indicated time points. One-half of the tumor was immediately snap-frozen in liquid nitrogen and stored at −80°C until analysis. Frozen tumors were then lysed in ice-cold TTG lysis buffer [1% Triton X-100, 5% glycerol, 20 mmol/L Tris (pH 7.7), 1 mmol/L EDTA, 0.15 mol/L NaCl, 1 mmol/L sodium orthovanadate, 40 μmol/L ammonium molybdate, and 2% complete protease inhibitor]. The total protein concentration of tumor lysate was determined using the MicroBCA method (Pierce). Prior to SDS-PAGE, 10 μL of lysate with 10 μg of total proteins were denatured by boiling with 10 μL of Laemmli SDS sample buffer (Bio-Rad Laboratories, Hercules, CA), further resolved on 7.5% SDS-PAGE gels and transferred to Immobilon-P polyvinyl membrane (Millipore Corp, Bedford, MA) at a constant voltage at 4°C overnight. To detect phospho-EGFR, the membrane was blocked with 5% blotto (Bio-Rad Laboratories) for 30 minutes at 37°C in TBST buffer [10 mmol/L Tris (pH 7.7), 120 mmol/L NaCl, 1 mmol/L EDTA, 0.2% Tween 20], and was subsequently probed with a rabbit polyclonal anti-phospho-EGFR[pY1068] (1:1,000 dilution) for 30 minutes at 37°C, then with an anti-rabbit IgG antibody conjugated with horseradish peroxidase. The phosphoprotein was visualized by chemiluminescence reagent ECL Plus (Amersham Pharmacia Biotech, Piscataway, NJ). The molecular size of EGFR was estimated by comparison with prestained protein precision markers. To detect the total EGFR expression, the same blot was stripped with a stripping buffer [7 mol/L guanidine-HCl, 50 mmol/L glycine (pH 10.3), 0.05 mmol/L EDTA, 0.1 mol/L KCl, 20 mmol/L 2-mercaptoethanol] at room temperature for 10 minutes and probed with rabbit polyclonal anti-EGFR antibody (1:1,000 dilution) under the same conditions described above.

Quantitation of the total EGFR and phospho-EGFR[pY1068] levels in tumor xenograft samples. Tumor lysates were adjusted to a concentration of 3 mg/mL protein. An equal amount of total protein (30 μg) from each tumor sample was used to analyze the levels of the total EGFR and phospho-EGFR by the ELISA assay according to the protocol recommended by the manufacturer (Biosource International Inc., Camarillo, CA). The level of phospho-EGFR was normalized by the level of total EGFR in each tumor sample. The percentage of inhibition of phospho-EGFR was calculated using the normalized values of the phospho-EGFR.

Immunohistochemical analysis. Upon surgical removal and splitting, the tumor was immediately fixed in 10% neutral formalin and embedded in paraffin. Five micrometer sections were prepared and mounted on positively charged glass slides (VWR, Plainfield, NJ). After deparaffinization in xylene and graded alcohol, the tissue sections were rehydrated and processed for antigen retrieval. For staining with the anti-phospho-MAPK antibody, the sections were submerged in 1× High pH Target Retrieval Solution (DakoCytomation, Carpinteria, CA) and were treated in a pressure cooker for 15 minutes. Sections to be stained with antibodies to MAPK and Ki67 were treated similarly, except that antigen retrieval was done in 1 mmol/L EDTA for MAPK or 10 mmol/L citrate buffer (pH 6) for Ki67. Endogenous peroxidase activity was quenched by 3% hydrogen peroxide in PBS for 10 minutes. The slides were then incubated overnight at 4°C with the following primary antibodies: anti-MAPK at 1:1,500 dilution, anti-phospho-MAPK at 1:250 dilution, or anti-Ki67 at 1:50 dilution. This was followed by incubation with biotinylated goat anti-mouse or anti-rabbit secondary antibodies (Vector Labs, Burlingame, CA) at 1:200 dilution for 30 minutes at room temperature. The slides were further incubated with Vectastain ABC Elite Kit and then with the peroxidase substrate 3,3′-diaminobenzidine (Vector Labs).

Pharmacokinetic and pharmacodynamic modeling. The pharmacokinetic and pharmacodynamic data were analyzed sequentially using the SAAM II software (Seattle, WA). The plasma concentrations of cetuximab following i.v. and i.p. administration were simultaneously fitted using a one-compartment model with first-order absorption kinetics, which can be described by the following integrated equation:where k_el is the elimination rate constant, V_d is the volume of distribution, D stands for dose, k_a is the absorption rate constant, and F stands for bioavailability. The estimated pharmacokinetic variables were then linked with the pharmacodynamic model as well as to simulate the pharmacokinetic profile for the multiple dosing schedule.

The percentage of inhibition of phospho-EGFR was used as the pharmacodynamic response of cetuximab in pharmacodynamic modeling with an indirect inhibitory E_max model. In the absence of cetuximab, the rate of changes in the phospho-EGFR level can be described by the following equation:where R₀ is the baseline value of phospho-EGFR, k_in is the zero-order rate constant for the formation of phospho-EGFR, k_out is the first-order rate constant of dephosphorylation of phospho-EGFR. In the absence of cetuximab, the phospho-EGFR level was assumed to be at steady sate, at which the rate of phospho-EGFR formation (k_in) was equal to the rate of dephosphorylation of phospho-EGFR (k_out × R₀).

In the presence of cetuximab, it was assumed that the reduction of phospho-EGFR was primarily due to the inhibition by drug instead by endogenous phosphatases. The percentage of phospho-EGFR relative to the baseline value (i.e., R/R₀) in the presence of cetuximab was described by the following equation:where E_max was the maximum inhibitory response and was assumed to be 100% in the model, and EC₅₀ is the drug concentration corresponding to 50% of E_max. Substituting Eq. D into Eq. E, it yielded:

Eq. F was then used to simultaneously fit the pharmacodynamic data. Goodness of fit was assessed by the minimization of the objective function, Akaike and Schwarz-Bayesian information criteria, visual inspection of the fitting and residual plots, and the precision of the variables estimated.

Results

Pharmacodynamics of cetuximab in nude mice bearing Geo tumor xenografts. We have conducted efficacy studies to define the optimal biological dose and the minimum effective dose of cetuximab in nude mice bearing Geo human colon xenografts (23). It was observed that cetuximab was equally active at dose levels ranging from 1 to 0.25 mg, whereas inhibiting the tumor growth dose-dependently at levels of 0.25, 0.1, and 0.04 mg. Therefore, cetuximab dose levels of 0.25 mg or higher were suggested to be optimal for the antitumor activity. The dose of 0.04 mg was active but much less efficacious than 0.25 mg, and therefore was considered suboptimal or close to the minimum effective dose. Our observation was consistent with the observations by other investigators using the same tumor model (16, 24). In the present study, pharmacodynamic biomarkers of cetuximab were determined using the same tumor model at dose levels of 0.25 and 0.04 mg.

Tumoral phospho-EGFR was analyzed by Western blot analysis with a phospho-specific antibody anti-EGFR[pY1068]. It was shown that tumoral phospho-EGFR was inhibited at both doses, but the extent and duration of tumoral phospho-EGFR inhibition appeared different (Fig. 1). The inhibition of tumoral phospho-EGFR was quantitated by the ELISA assay and plotted versus time (Fig. 2). At 0.25 mg, tumoral phospho-EGFR was maximally inhibited by 91% at 24 hours post-dose, whereas the level of inhibition declined progressively to 72% by 72 hours post-dose. Because the in vivo efficacy study of cetuximab was conducted with a multiple dosing schedule (q3dx5; ref. 23), the inhibition of tumoral phospho-EGFR would be conceivably ≥72% throughout the treatment period at dose levels of ≥0.25 mg. At 0.04 mg, the maximum inhibition of tumoral phospho-EGFR was 53% at 24 hours, whereas the level of the inhibition was reversed to 37% by 72 hours. In general, the time course of tumoral phospho-EGFR inhibition and recovery seemed to correlate with plasma and tumoral levels of cetuximab at 0.25 and 0.04 mg (Fig. 2). The expression level of the total EGFR was not significantly modified at either dose, indicating that the inhibition of tumoral phospho-EGFR was primarily due to the reduction of ligand-induced EGFR activation in tumor.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Inhibition of the phosphorylation of EGFR in the Geo tumor xenograft following a single i.p. injection of cetuximab at 0.04 and 0.25 mg. Tumors were surgically removed, immediately snap-frozen in liquid nitrogen and stored at −80°C until Western analysis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

A and B, pharmacodynamics of cetuximab in the Geo tumor-bearing mice following a single i.p. administration at 0.25 and 0.04 mg. Nude mice bearing human Geo tumor following a single i.p. administration at dose levels of 0.25 and 0.04 mg. Tumors were surgically removed at 0, 1, 3, 6, 24, 48, and 72 hours and immediately frozen in liquid nitrogen and stored at −80°C until Western analysis. The inhibition of the activated EGFR was quantitated by an ELISA assay. Points, means; bars, ±SD. Pharmacokinetic data were quoted from our previous publication (23).

MAPK is one of the most well-characterized downstream components of the EGFR signaling pathway, and therefore, could serve as an important surrogate biomarker of the activated EGFR kinase (25). Hence, we measured the phosphorylation status of MAPK in tumor by immunohistochemistry (Fig. 3A). It was observed that tumoral phospho-MAPK was inhibited at both 0.25 and 0.04 mg, paralleling the inhibition of tumoral phospho-EGFR by cetuximab. The time course of tumoral phospho-MAPK seemed, in general, consistent with the pharmacokinetics of cetuximab at 0.25 and 0.04 mg. The inhibition of tumoral phospho-MAPK was believed to be mediated by the inhibition of the EGFR pathway because cetuximab is a potent and specific inhibitor of EGFR kinase (12, 14). To determine the tumor growth inhibitory effect of cetuximab, tumor cell proliferation was assessed with a commonly used marker Ki-67, a proliferation-associated antigen present only in proliferating cells (26, 27). The inhibition of Ki67 expression was also observed at both 0.25 and 0.04 mg (Fig. 3B). It seemed that Ki67 expression was similarly inhibited at 24 and 48 hours for both doses, whereas, at 72 hours, the inhibition of Ki67 expression became less for 0.04 mg than for 0.25 mg, suggesting a greater tumor growth inhibition was likely achieved at 0.25 than at 0.04 mg. Essentially, the tumor growth-inhibitory effect of cetuximab, assessed by Ki67 expression, is consistent with the antitumor activity of cetuximab determined in vivo (23). Collectively, our study suggested that cetuximab is able to modulate EGFR kinase activity in tumor and to inhibit tumor proliferation consequently; whereas the tumoral levels of phospho-EGFR/phospho-MAPK could serve as pharmacodynamic biomarkers assessing target inhibition by cetuximab.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Inhibition of the phosphorylation of MAPK (A) and Ki-67 expression (B) in the Geo tumor xenograft following a single i.p. injection of cetuximab at 0.25 or 0.04 mg. Tumors were surgically removed, immediately fixed in 10% neutral formalin, embedded in paraffin and analyzed immunohistochemically.

Estimation of active drug concentration by pharmacokinetic/pharmacodynamic modeling. Because cetuximab does not bind to mouse EGFR (9), its distribution is likely confined in the plasma compartment. Therefore, a one-compartment model with the first-order absorption kinetics was applied in pharmacokinetic modeling. The pharmacokinetic model was simultaneously fitted to the plasma concentrations of mice given a single dose of cetuximab i.v. and i.p. at 1.0, 0.25, and 0.04 mg (Fig. 4). The constants of k_a and k_d were estimated to be 0.44 ± 0.09 and 0.017 ± 0.002 hour⁻¹, respectively, whereas the V_d was 0.094 ± 0.003 L/kg. The percentage inhibition of tumoral phospho-EGFR by cetuximab was used as the pharmacodynamic response in pharmacodynamic modeling. It was assumed that the level of tumoral phospho-EGFR was at steady state in the absence of cetuximab, whereas the reduction of tumoral phospho-EGFR when cetuximab was administered was primarily due to the inhibition of phospho-EGFR by drug. When the inhibition of tumoral phospho-EGFR was plotted versus the plasma concentration of cetuximab at both 0.25 and 0.04 mg, hysteresis was observed, suggesting that there was a delay between the plasma concentration and the inhibition of tumoral phospho-EGFR by cetuximab (data not shown). Therefore, an indirect inhibitory E_max model was applied in pharmacodynamic modeling. The pharmacodynamic model linked with the defined pharmacokinetic model was simultaneously fitted to the pharmacodynamic data to estimate the active plasma concentration of cetuximab (Fig. 5). The value of EC₅₀, the plasma concentration required to inhibit 50% of tumoral phospho-EGFR, was 7.5 ± 0.8 μg/mL.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Pharmacokinetic modeling of cetuximab in mice following a single i.v. (A) or i.p. (B) injection at 1, 0.25, and 0.04 mg. The pharmacokinetic data were simultaneously fitted using a one-compartment model. The estimated disposition variables were then linked with the pharmacodynamic model.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Pharmacokinetic and pharmacodynamic modeling of cetuximab in Geo tumor-bearing mice following a single i.p. injection at 0.25 and 0.04 mg. The pharmacokinetic data were simultaneously fitted using a one-compartment model with first-order absorption kinetics. The estimated pharmacokinetic variables were then used to link with the pharmacodynamic model. The pharmacodynamic data were simultaneously fitted using an indirect pharmacodynamic model with an assumption that the phosphorylated EGFR level was assumed to be at steady state in the absence of cetuximab. Also, the reduction of the phosphorylated EGFR measured in the experiment was primarily due to the inhibition by drug.

When the EC₅₀ was compared with the simulated plasma profiles of cetuximab with the schedule of q3dx5 (Fig. 6), it was found that the trough plasma levels of cetuximab, when given at the suboptimal dose of 0.04 mg, were around the EC₅₀, implying ∼50% of phospho-EGFR at this dose level. Because 0.04 mg is close to the minimum effective dose in our studies, it can be conceived that 50% of target inhibition is probably minimally required for cetuximab in order to be active against Geo tumor. On the other hand, the peak plasma levels of cetuximab at 0.04 mg were well below the EC₉₀ (67.5 μg/mL), implying significantly >90% of phospho-EGFR inhibition at 0.04 mg. Therefore, the optimal antitumor activity may not be achieved at 0.04 mg, which was the observation in our in vivo efficacy study of cetuximab (23). For 0.25 mg, the trough plasma levels of cetuximab were around or above the EC₉₀, implying that nearly complete, if not 100% inhibition of phospho-EGFR was achieved and a consequent optimal activity could be expected at dose levels of ≥0.25 mg, which was also consistent with the outcome of our efficacy study (23).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Simulation of plasma levels of cetuximab in mice with a dosing schedule of q3dx5 at 0.25 and 0.04 mg. The pharmacokinetic data were simultaneously fitted using a one-compartment model with first-order absorption kinetics. The plasma profile was simulated based on an assumption of linear pharmacokinetics for cetuximab with multiple administrations at dose range of 0.25 to 0.04 mg. (•) 0.25 mg, (○) 0.04 mg.

Discussion

For many molecularly targeted agents, the clinical dose continues to be defined by traditional means. For small molecule inhibitors of EGFR kinase such as gefitinib or erlotinib, investigators were able to define a maximum tolerated dose associated with dose-limiting toxicity. The further studies were conducted at a lower but tolerable dose, which still has evidence of clinical response (28, 29). However, there remain agents which are associated with minimal toxicity in the clinical setting, such as cetuximab. In these cases, the decision regarding optimal dose needs to be based on other criteria, for example, biomarkers, especially the mechanism-based biomarkers. In general, cetuximab seems to be well-tolerated in clinical studies, whereas a clear maximum tolerated dose has never been reached in contrast to other small molecule inhibitors. The current dosing regimen was recommended based on the clinical observation of a complete saturation of systemic clearance at efficacious doses (20, 30). However, there has not been any rigorous experimental evidence to equate the saturation of systemic clearance to complete inhibitory status of EGFR in tumors under the current clinical dosing regimen. Moreover, the functionally activated EGFR, the predominant driving force for cell proliferation in cetuximab-responsive tumors, could be a portion of total EGFR. This would further imply that the dose required to saturate systemic clearance or total EGFR in tissues could be different from the one required to inhibit the activity of the functional EGFR in tumor.

In the present study, on the basis of the thorough understanding of EGFR signaling and the clinical experience of cetuximab, we proposed phospho-EGFR/phospho-MAPK as the mechanism-based pharmacodynamic biomarker to assess target exposure. At dose levels of 0.25 and 0.04 mg, a dose-dependent inhibition of phospho-EGFR was observed for cetuximab. The time course of phospho-EGFR inhibition and recovery correlated with the plasma levels of cetuximab at both doses. This essentially supported the hypothesis that the level of tumoral phospho-EGFR could be used as an in vivo biomarker assessing target exposure of cetuximab. MAPK is one of the most well-characterized cell proliferative components of the EGFR signaling pathway. The inhibition of phospho-EGFR by cetuximab has been clearly linked with the phosphorylation status of MAPK in cells and with subsequent pharmacologic end points such as tumor cell proliferation or apoptosis (31–34). Particularly, in patients with head and neck tumors, a strong correlation has been shown, upon treatment with cetuximab, between the inhibition of phospho-MAPK and the inhibition of EGFR kinase activity in tumor and skin tissues (33). Therefore, we further investigated the level of tumoral phospho-MAPK in the present study. It was found that tumoral phospho-MAPK was inhibited at both 0.25 and 0.04 mg, which correlated with the inhibition of tumoral phospho-EGFR. In general, the time course of phospho-MAPK inhibition and recovery seemed to correlate with the plasma levels of cetuximab, similar to that of phospho-EGFR. Because the inhibition of phospho-MAPK has been reported to correlate with antiproliferation in tumor assessed by Ki67 expression (33), we further determined the tumoral Ki67 expression. It seemed that the tumoral Ki67 expression was inhibited dose-dependently, which paralleled the inhibition of tumoral phospho-EGFR/phospho-MAPK. Essentially, our study suggested that cetuximab is able to modulate the activity of the EGFR kinase in tumor and consequently to inhibit tumor proliferation. In general, phospho-MAPK has been more extensively studied as a biomarker to assess EGFR kinase signaling and inhibition. Take into consideration that MAPK may be activated by several pathways (35–37), phospho-MAPK itself, as a biomarker, may not be sufficient to predict EGFR targeting. Our study suggests that the tumoral levels of phospho-EGFR/phospho-MAPK should be evaluated together to assess the EGFR inhibition by cetuximab.

Our pharmacokinetic/pharmacodynamic model predicted the EC₅₀, which is probably required to achieve a minimal antitumor activity of cetuximab, whereas the EC₉₀ is more likely to be required for the optimal response. We would recommend that the EC₉₀ (67.5 μg/mL) could serve as a conservative measure of the active plasma concentration of cetuximab. A complete inhibition of the active EGFR could be achieved at plasma concentrations of cetuximab >75 μg/mL, i.e., EC₁₀₀, which can be suggested as active plasma concentrations. When cancer patients were treated with cetuximab at a multiple weekly dose of 400 mg/m², the C_ss was estimated to be 56 μg/mL (20). When the treatment schedule was a 400 mg/m² loading dose followed by a 250 mg/m² weekly maintenance dose, the mean peak and trough plasma levels of cetuximab in patients were estimated to be 192 and 75 μg/mL, respectively, whereas the C_ss was ∼100 μg/mL (38). Therefore, the C_ss of cetuximab in cancer patients is likely within the range of 56 to 100 μg/mL under the current clinical dosing regimen. These reported plasma levels of cetuximab have been confirmed in phase II clinical studies being conducted by Bristol-Myers Squibb and its alliances (data not shown). It is interesting that our predicted active plasma concentration of 75 μg/mL is fairly comparable to the C_ss achieved in cancer patients. In contrast, when predicting the clinical active dose from the preclinical data, the traditional allometric dose scaling method seemed problematic. The preclinical optimal dose of 0.25 mg, if scaled up based on the formula recommended by Freireich et al. (39) with an assumption of a 70 kg average body weight of patients, is equivalent to a clinical dose of 37.5 mg/m², which is significantly below the current clinical dose. This discrepancy could be due to the different routes of drug clearance, i.e., cetuximab only binds to human but not to mouse EGFR, therefore, extra amounts of cetuximab are required for patients to saturate EGFR in a number of tissues, e.g., the skin and liver. Because the inhibition of EGFR directly correlates to the plasma concentration of drug, the different affinities of cetuximab to human and mouse EGFR or the consequent different routes of drug clearance should not affect EGFR inhibition once appropriate concentrations are achieved regardless of species. Therefore, the active plasma concentration, predicted by biomarker, phospho-EGFR inhibition in human tumors grown s.c. in mouse, is more readily translated into clinical studies due to limited impact from affinity difference of the antibody.

It is usually believed that biomarker should be identified and validated in the preclinical stage. The optimal exposure or active drug concentration could therefore be predicted before the drug candidate enters into clinical studies. One may argue the necessity of the present study because cetuximab has been approved by the Food and Drug Administration. However, there have not been enough convincing data to prove that the active drug concentration determined in animal models could provide a useful guide for clinical trials. The present study showed evidence that the active drug concentration, predicted by preclinical pharmacokinetics/biomarkers, correlated with the active concentrations achieved in cancer patients. Our study also suggests that EGFR is likely to be significantly, if not completely, blocked by cetuximab in cancer patients under the current clinical dosing regimen. Hence, the clinical dose of cetuximab, even though defined not based on biomarkers, probably approximates to the optimal biological dose. The current study was conducted by no means to tackle the clinical regimen of cetuximab, but to strongly advocate that more preclinical pharmacokinetics/biomarker studies should be conducted to define the optimal drug exposure to facilitate clinical drug development. The model established in the present study has been applied in phase I studies in chronic myelogenous leukemia patients treated with a novel SRC/BCR-ABL kinase inhibitor, BMS-354825. In this case, the predicted active concentration from preclinical models was in line with those achieved by patients who also had clinical response to the treatment (40).

Having mentioned some success with our pharmacokinetics/biomarker model, we would also suggest that this model should be used with certain caution before it is extensively evaluated. Phospho-EGFR/phospho-MAPK, as pharmacodynamic biomarkers, are probably adequate to predict the target exposure. In the present study, they probably coincidentally correlated with cetuximab efficacy in the Geo model, which is a sensitive model for cetuximab treatment. In resistant tumor models, one could conceive that EGFR might still be inhibited by cetuximab, but the lack of efficacy could be attributable to either the lower degree of modulation of signaling pathway or activation of survival pathways. Therefore, the current model may not be suitable to predict the overall treatment response of cetuximab. To facilitate clinical drug development, pharmacogenomic and proteomic approaches should also be applied to identify gene or protein biomarkers, which could predict the responsiveness of patients to drug treatment. It has been shown that certain proteins are potentially involved in resistance mechanisms of cetuximab therapy, whereas the different expression levels of proteins may serve as biomarkers predicting resistance (41). Our colleagues have also identified a set of gene and protein biomarkers predicting response to cetuximab by profiling cetuximab-sensitive and -resistant tumors using pharmacogenomics and proteomics approaches (42, 43). To develop novel molecularly targeted agents, the optimal clinical exposure and regimen should be determined by an array of biomarkers, including target exposure biomarkers and patient selection biomarkers, in combination with clinical responses and toxicity profiles.