© The Author 2006. Published by Oxford University Press.
ARTICLE |
CYP3A Phenotyping Approach to Predict Systemic Exposure to EGFR Tyrosine Kinase Inhibitors
Affiliations of authors: The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Baltimore, MD (JL, JB, AS, MZ, MH, SDB); Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden (MOK)
Correspondence to: Sharyn D. Baker, PharmD, PhD, Pharmaceutical Sciences Department, St Jude Children's Research Hospital, 332 North Lauderdale St., DTRC Rm. D1034, Mail Stop 314, Memphis, TN 38105 (e-mail: sharyn.baker{at}stjude.org).
 |
ABSTRACT |
|---|
 TopNotes Abstract IntroductionPatients and methodsResultsDiscussionReferences |
|---|
Background: Gefitinib is an orally active inhibitor of epidermal growth factor receptor (EGFR) tyrosine kinase (TK) with activity in non–small-cell lung cancer. The response to gefitinib is variable, possibly because of interindividual variation in the activity of cytochrome P450 3A (CYP3A), the principal enzyme that metabolizes gefitinib. We prospectively assessed the influence of CYP3A activity on gefitinib disposition and toxicity. Methods: Twenty-seven patients with advanced cancer were treated with daily oral gefitinib at 250 mg (n = 13) or 500 mg (n = 14) for 28 days. Concentration–time profiles of midazolam and geftinib were constructed based on measurement of their concentration in serial blood samples using high-performance liquid chromatography and mass spectroscopy. CYP3A activity was determined at baseline by assessment of midazolam apparent oral clearance. Pharmacokinetic studies were performed for a period of 28 days, and population modeling was performed using NONMEM software. A structural pharmacokinetic model was developed to describe the concentration–time profiles of unbound and total gefitinib plasma concentrations, and patient-specific covariates were added to the model to account for unexplained interindividual variability in pharmacokinetic parameters. Statistical tests were two-sided. Results: Gefitinib pharmacokinetics exhibited wide interindividual variability (interindividual variability on total and unbound gefitinib apparent oral clearance was 79% and 74%, respectively). Midazolam clearance (mean = 40 L/h, range = 10–111) was highly correlated with that of total and unbound gefitinib (R2 = .60 and R2 = .68, respectively) and with steady-state plasma trough concentrations of gefitinib (R2 = .58 and R2 = .60, respectively), and it accounted for approximately 40% of interindividual variability in gefitinib clearance in the pharmacokinetic model. Both total and unbound gefitinib steady-state plasma trough concentrations were associated with the development of diarrhea (P<.05), but not skin rash. At a dose of 250 mg gefitinib, 11 of 13 patients achieved steady-state plasma trough concentrations above the IC50 for inhibition of mutant EGFR in vitro (0.015 µM), but only one achieved a steady-state plasma trough concentration above the IC50 for inhibition of wild-type EGFR (0.1 µM). Conclusions: As an in vivo phenotypic probe of CYP3A, midazolam oral clearance may have utility for prediction of gefitinib exposure and dose selection. A pharmacokinetic model incorporating this indicator of CYP3A activity has potential for optimization of treatment with gefitinib and other TK inhibitors that are metabolized in a similar manner.
|
INTRODUCTION |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Gefitinib (marketed as Iressa) is an orally active inhibitor of epidermal growth factor receptor (EGFR) tyrosine kinase (TK) (1). This drug was conditionally approved by the Food and Drug Administration in May 2003 based on results from several phase II studies that demonstrated overall tumor response rates of 10%–18% in patients with refractory locally advanced or metastatic non–small-cell lung cancer (2,3). A large randomized phase III study found increased survival among patients treated with gefitinib compared with those treated with placebo, although this difference failed to reach statistical significance (4). Furthermore, patients treated with erlotinib, another orally active EGFR TK inhibitor, showed a statistically significant improvement in survival of 2 months relative to placebo (5).
The apparent difference in efficacy between the two trials may be attributable, in part, to the fact that erlotinib was administered at its maximum tolerated dose (5), whereas gefitinib was administered at about one-third of its maximum tolerated dose (4). In addition, genetic differences in tumors of patients receiving these drugs may also underlie the difference in outcomes—the presence of somatic mutations in the ATP-binding site of the EGFR gene and an increased copy number of EGFR have been associated with better tumor response and longer survival in patients treated with gefitinib (6,7).
Studies to evaluate the pretreatment characteristics of tumors from patients receiving gefitinib or erlotinib in placebo-controlled phase III trials are in progress and may explain differences in treatment outcome (8). However, the presence of mutations in the EGFR gene or increased gene copy number does not account for all responders and those receiving clinical benefit from treatment, suggesting that other factors may be involved in patients' response to the drug (6). These factors may include differences in the extent of systemic exposure and delivery of the drug to the tumor. In vitro studies have shown that the gefitinib concentrations required for 50% inhibition (IC50) of wild-type and mutant EGFR are 0.1 and 0.015 µM, respectively, whereas the respective concentrations for 100% inhibition are 2.0 and 0.2 µM (9). Achieving these gefitinib concentrations in vivo may be necessary for antitumor activity.
The plasma pharmacokinetics of gefitinib have been described previously in cancer patients enrolled in phase I studies (10–12). Gefitinib exhibits wide interindividual pharmacokinetic variability that is likely attributable to variation in the expression and/or activity, in the intestines and liver, of cytochrome P450 3A (CYP3A), which plays a predominant role in the first-pass metabolism and systemic clearance of the drug (13–15). Although the association between gefitinib exposure and clinical efficacy and toxicity has not been evaluated comprehensively, substantial pharmacokinetic variability is likely to impact treatment outcomes. The use of a noninvasive phenotypic probe to measure CYP3A could be useful for therapeutic optimization of the dosage of gefitinib and other TK inhibitors with similar absorption and disposition characteristics.
The objective of this study was to prospectively evaluate the relationship between CYP3A activity, as indicated by midazolam oral clearance, and gefitinib absorption and disposition in cancer patients. A population pharmacokinetic model was developed to assess the influence of CYP3A activity and other patient covariates on gefitinib disposition as well as to determine the association between gefitinib exposure and toxicity.
|
PATIENTS AND METHODS |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Study Population and Treatment
Patients with cancer were treated with gefitinib as part of a biologic effects study to evaluate the relationships among gefitinib dose, exposure, and tissue pharmacodynamic activity (16). A secondary objective of the study, and the focus of the work reported here, was to develop a population model for gefitinib and to assess the relationship between CYP3A activity, gefitinib exposure, and toxicity. Patients with a histologically confirmed malignant solid tumor that was susceptible to gefitinib therapy (as determined by the treating physician) and lesions amenable to serial tumor biopsies were enrolled in this study. Additional eligibility criteria included age above 18 years, an Eastern Cooperative Oncology Group performance status of 2 or less, adequate marrow function (leukocyte count 
3 x 109/L, absolute neutrophil count 1.5 x 109/L, platelet count 100 x 109/L), adequate liver and kidney functions (total bilirubin 
2 mg/dL, aspartate aminotransferase/alanine aminotransferase 2.5 times the institution upper limit of normal, serum creatinine 1.5 mg/dL), no anticancer treatment during the 30 days previous to enrollment, and no concomitant use of drugs that would induce CYP3A, including phenytoin, carbamazepine, rifampicin, phenobarbital, and St John's wort.
Gefitinib was provided by AstraZeneca (Wilmington, DE) as 250 mg brown, film-coated tablets. Cohorts of 12 patients were treated once daily with 250 or 500 mg gefitinib, and additional patients were treated until 12 in each cohort were assessable for serial tumor and skin biopsies (before treatment and after 28 days of daily treatment). After cycle one, patients underwent additional cycles of treatment until treatment was discontinued upon disease progression or unacceptable or unmanageable drug-related adverse events. All patients were treated at The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins (Baltimore, MD). The Institutional Review Board of The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins approved the protocol, and written informed consent was obtained from each patient.
Gefitinib Pharmacokinetics
Serial blood samples were obtained from each patient at baseline (before treatment with gefitinib) and at 1, 2, 3, 4, 5, 6, and 8 hours following administration of the first dose of gefitinib on day 1. Additional blood samples were obtained before administration of gefitinib on days 2, 3, 8, 15, 22, and 28. Blood samples were collected in heparinized tubes, placed on ice, and processed within 1 hour of collection. Plasma was isolated by centrifugation at 1000g at 4 °C for 10 minutes, divided into two aliquots, and stored at –20 °C until analysis. Total gefitinib plasma concentrations were determined in one aliquot of plasma using a validated high-performance liquid chromatography (HPLC) and tandem mass spectrometric method (LC/MS/MS), as described previously (17). Because unbound drug concentrations are believed to be more relevant to pharmacologic and toxicologic responses than total drug concentrations, unbound gefitinib pharmacokinetics were also evaluated. Unbound gefitinib plasma concentrations were determined in the second aliquot of plasma on a 96-well equilibrium dialyzer with a 5-kDa cutoff regenerated cellulose membrane (Harvard Apparatus, Holliston, MA) using a validated equilibrium dialysis method, as described previously (18).
Oral Midazolam Test
Within 72 hours before beginning gefitinib treatment, patients were given a single 3-mg dose of oral midazolam (Versed Syrup, Roche, Basel, Switzerland). Serial blood samples were collected before treatment and at 5, 20, and 45 minutes and 1, 1.5, 2, 3, 4, 5, and 7 hours after administration of midazolam. Blood samples were collected in tubes containing EDTA, placed on ice, and processed within 1 hour of collection. Plasma was isolated by centrifugation at 1000g at 4 °C for 10 minutes and stored at –70 °C or below until analysis.
Midazolam was quantif ied using a validated HPLC and tandem mass spectrometric method. Samples were prepared for liquid chromatography by a solvent extraction procedure for which recovery exceeded 90%. Briefly, 200 µL of plasma was extracted by adding 5 mL of ethyl ether containing temazepam (100 µg/mL) as the internal standard, followed by vortex mixing and centrifugation at 2000g for 10 minutes at ambient temperature. The top organic layer was withdrawn and evaporated to dryness at 30 °C under a gentle stream of nitrogen. The residue was reconstituted in 100 µL of acetonitrile/water (50 : 50, vol/vol), and an aliquot of 10 µL was applied to a Waters Model 2690 separations system (Milford, MA). The compound of interest was separated on a Waters X-Terra MS C18 analytical column (50 x 2.1 mm, i.d. 3.5 µm) and eluted with acetonitrile/0.1% formic acid (70 : 30, vol/vol) at an isocratic flow rate of 0.15 mL/min for 5 minutes. The effluent was monitored with a Micromass Quattro LC triple-quadrupole mass-spectrometric detector (Beverly, MA) using the electrospray positive ionization mode. The linear calibration curve of midazolam was generated over the range of 0.2–100 ng/mL with a coefficient of determination of greater than 0.99, and the within- and between-day precision and accuracy for a period of 4 days of validation were within 15%.
Pharmacokinetic Analysis
Noncompartmental analysis. Individual total and unbound gefitinib concentrations and midazolam plasma concentrations were analyzed by standard noncompartmental methods using the software program WinNonlin (version 5.0) (Pharsight Corporation, Cary, NC). Gefitinib pretreatment trough concentration (Cmin) was considered evaluable if the sample was obtained between 22 and 26 hours after the previous dose and within 2 hours before the next dose. Gefitinib Cmin at steady state (Css,min) was determined as the average of the pretreatment concentrations on days 8, 14, 22, 28, and (when obtained) 29. Apparent oral clearance for midazolam was calculated as dose divided by area under the concentration–time curve (AUC).
Population pharmacokinetic analysis. The population pharmacokinetic model for gefitinib was developed in two stages: structural (covariate-free) model development followed by covariate model development. All analyses were performed with a first-order conditional estimation method with interaction using NONMEM program software (version V) (University of California, San Francisco, CA). Xpose 3.1/ S-PLUS 6.0 (19) software was used for graphic diagnostics and covariate screening.
The structure model was built to fit total (Cp) and unbound (Cu) gefitinib plasma concentrations from all patients simultaneously. Because both in vitro and in vivo data have shown that the binding of gefitinib to plasma proteins is linear over the therapeutic concentration range (1–5000 ng/mL) (18), the unbound fraction (Fu) was used to link unbound and total concentrations in the model. When the model was intended to predict total gefitinib pharmacokinetic parameters, the expression Cu = Cp x Fu was used to link the total to unbound concentrations in the model; when the model was intended to predict unbound gefitinib pharmacokinetics, the expression Cp = Cu/Fu was used to link the unbound to total concentrations in the model. One- and two-compartment models with first-order absorption (with or without lag time) and first-order elimination and a two-compartment model with sequential zero-order input (mimicking dissolution) and first-order absorption and first-order elimination were tested to fit the multiple-dose plasma concentration versus time profiles. Model selection for nonhierarchic models was guided by the decrease in Akaike Information Criterion (AIC) and by graphic goodness-of-fit analyses with Xpose 3.0. AIC is calculated as AIC = (–2LL) + 2 x p, where –2LL is the NONMEM objective function value (OFV, –2log likelihood) and p is the number of pharmacokinetic parameters in the model. Interindividual variability of pharmacokinetic parameters was included as an exponential function. Residual error was modeled with a combination method including an additive and a proportional part, each of which could be excluded if it was estimated to be negligible.
A screen for potentially statistically significant covariates was performed using S-PLUS 6.2/Xpose 3.1 software (Uppsala University, Uppsala, Sweden) with a generalized additive model (GAM), a stepwise multiple regression method that allows covariates to be entered into the model in either linear or nonlinear fashion according to a natural cubic spline function with one internal break point (19). The following covariates were screened on all pharmacokinetic parameters in the structural model: sex, age, body size (i.e., weight, height, and body surface area), hematocrit, serum albumin, alpha-1-acid glycoprotein concentration, liver function (as indicated by levels of alanine aminotransferase, aspartate transaminase, and total bilirubin), kidney function (as indicated by serum creatinine), and midazolam apparent oral clearance. Potentially statistically significant covariates selected from GAM analysis were introduced into the covariate model as linear, exponential, or power functions according to the following discrimination criteria: 1) a decrease in the objective function value of greater than 3.875 (P<.05) during the forward full covariate model building; 2) an increase in the objective function value of greater than 10.828 (P<.001) during the stepwise backward model reduction (20); 3) minimization of the relative standard error of the parameter estimates; 4) minimization of the model-estimated parameter interindividual variability and an improvement in the precision of the parameter estimate; and 5) random scatter of points around a horizontal line of identity at 0 in plots of weighted residue versus predicted concentrations.
Assessment of Toxicity
Diarrhea and skin rash, the two most common drug-related toxicities caused by gefitinib treatment, were graded and scored on a point scale of 0–4 in accordance with the National Cancer Institute Common Toxicity Criteria version 3.0 (Cancer Therapy Evaluation Program, Division of Cancer Treatment & Diagnosis, National Cancer Institute, National Institutes of Health, Bethesda, MD). The grade of diarrhea or skin rash assigned to a patient was the worst grade of diarrhea or skin rash experienced during the first 28-day cycle of treatment.
Exposure–Toxicity Analysis
Relationships between pharmacokinetic parameters and the incidence of the principal toxicities (diarrhea or rash) were examined. Pharmacokinetic parameters assessed were 1) parameters of total and unbound gefitinib exposure calculated from noncompartmental analysis (AUC and Cmax on day 1; observed Css,min), 2) gefitinib apparent oral clearance (estimated from the final model), and 3) midazolam apparent oral clearance. Because the worst grade of diarrhea or rash was scored as a categorical variable (i.e., 0, 1, 2, 3, or 4), the probability of each individual score of diarrhea or rash might be linked to drug exposure by the use of a logistic regression model, as described previously (21). However, based on data from a small population of 27 patients, a statistically significant association between drug exposure and individual scores for diarrhea or rash was not observed. Hence, in the final model, diarrhea and rash were treated as dichotomous categorical variables (0 and 1), with 0 representing the absence of toxicity during the first cycle of treatment and 1 representing the occurrence of toxicity (including grade 1, 2, and 3) during the first cycle of treatment. A logistic regression model was applied to predict the probability of toxicity (P) with a single predictor (x), i.e., observed gefitinib Css,min.
The probability of diarrhea was expressed as: 
 |
Log odds was expressed as a linear function of a single predictor [g(x)] as:
 |
Thus, the probability of diarrhea (P) was determined as:
 |
where 
represents the logit baseline of probability and 
represents the drug effect introduced in the model either as a linear function (i.e., slope x Css) or as a maximum effect model [i.e., 
].
The exposure–toxicity model was analyzed using the NONMEM program. An increase of the objective function value greater than 3.84 (P<.05) with the exclusion of the drug effect from the model (i.e., slope or Emax set as 0) was adopted to indicate the statistically significant association between the drug exposure and incidence of diarrhea or rash.
Statistical Analysis
Interindividual variation in gefitinib exposure parameters (AUC, Cmax, Css,min) and midazolam clearance (CL/F_MDZ) that was derived from noncompartmental analysis was described as fold difference, determined from the ratio of the maximum and minimum value for each parameter. Bivariate correlations were examined with the Pearson test, and P<.05 was regarded as statistically significant Pearson correlation. Statistical analyses were performed with SPSS version 10.0 (SPSS Inc, Chicago, IL). All P values were based on two-sided statistical tests.
|
RESULTS |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Plasma Exposure to Gefitinib
Pharmacokinetic studies were performed in 27 patients, of whom 13 were treated with 250 mg of gefitinib once daily and 14 with 500 mg once daily. Patient characteristics are summarized in Table 1. Twelve patients were female, and the median age was 61 years. Individual observed total gefitinib plasma concentration versus time profiles for day 1 (after administration of a single dose) in patients receiving 250 or 500 mg daily oral gefitinib were plotted (Fig. 1, A and B). Pretreatment total and unbound concentrations as measured at day 2, 3, 8, 15, 22, and 28 during treatment of patients receiving these doses of daily oral gefitinib were also plotted (Fig. 1, C–F). The mean (and associated 95% confidence interval) values for pharmacokinetic parameters are listed in Table 2.
|
|
|
At the dose of 250 mg, the maximal total gefitinib concentration observed on day 1 (C1,max) and the area under the plasma concentration–time curve during the first dosing interval (24 hours) (AUC0–
) varied 10-fold (geometric mean = 265 ng/mL, range = 79–757 ng/mL, and geometric mean = 3653 h·ng/mL, range = 1376–13 167 h·ng/mL, respectively [Fig. 1]). During the 28-day course of treatment, the average steady-state trough plasma concentration (Css,min) varied 18-fold among individuals treated with the 250-mg dose (geometric mean = 406 ng/mL, range = 104–1846 ng/mL). At the dose of 500 mg, total gefitinib C1,max (geometric mean = 368 ng/mL, range = 60–1128 ng/mL) and AUC0– (geometric mean = 5112 h·ng/mL, range = 851–13 191 h·ng/mL) varied up to 19-fold on day 1, and Css,min (geometric mean = 502 ng/mL, range = 145–1590 ng/mL) varied 11-fold. A similar extent of variability at both doses was observed for exposure to unbound gefitinib; unbound and total concentrations were highly correlated (R2 = .91, P<.001). Wide overlap was observed in individual gefitinib pharmacokinetic parameters between the 250- and 500-mg dose levels, likely due to the substantial interindividual pharmacokinetic variability (10- to 19-fold). Following chronic daily gefitinib treatment, steady-state concentrations of total and unbound gefitinib were reached between days 8 and 14 (Fig. 1). AUC0– on day 1 and Css,min were highly correlated for both total (R2 = .61, P<.001) and unbound (R2 = .58, P<.001) gefitinib. Population Pharmacokinetic Model
A two-compartment model with sequential zero-order input (mimicking dissolution) and first-order absorption and elimination resulted in a decrease in the AIC of 32.5 (P<.001) compared with the traditional first-order absorption model and was chosen as the structural model.
Mean (range) midazolam apparent oral clearance was 46 L/h (10–111 L/h) and 34 L/h (11–62 L/h) for the cohorts receiving 250 and 500 mg gefitinib, respectively. Among all patients, midazolam clearance varied approximately 11-fold (mean = 40 L/h, range = 10–111 L/h), similar to previously reported values in cancer patients (22). Midazolam clearance was the only potentially statistically significant covariate that was retained in the final model. The inclusion of midazolam clearance as a linear, power, or exponential function on total gefitinib apparent oral clearance resulted in a decrease of the objective function value by 21 (from –1354.8 to –1377.3; Table 3) (P<.001) and a decrease of the unexplained interindividual variability of gefitinib clearance from 79% (covariate-free model) to 50% (final model). Midazolam clearance was incorporated in the final model as an exponential function, which provided the most precise estimate for total gefitinib clearance (relative standard error of estimation = 24%). Similar to its relationship to total gefitinib, midazolam clearance was a statistically significant covariate (decrease of the objective function value by 21; P<.001) on unbound gefitnib clearance, resulting in a 39% decrease in interindividual variability for this parameter from 74% (covariate-free model) to 45% (final model).
|
The population pharmacokinetic parameter estimates for total and unbound gefitinib are presented in Table 3. The population values for apparent oral clearance for total and unbound gefitinib were estimated as 21.4 and 612 L/h, respectively. Midazolam oral clearance accounted for 37% and 39% of unexplained interindividual variability for total and unbound gefitinib oral clearance, respectively. Midazolam clearance was correlated with model-predicted total gefitinib clearance (R2 = .60, P<.001) and unbound gefitinib clearance (R2 = .68, P<.001) (Fig. 2, A and B). Midazolam clearance was also strongly associated with observed total (R2 = .58, P<.001) and unbound (R2 = .60, P<.001) gefitinib Css,min (Fig. 2, C and D).
|
Exposure–Toxicity Relationships
Diarrhea and skin rash were dose dependent, with more severe toxicity occurring at the dose of 500 mg. Grade 1, 2, and 3 diarrhea occurred in 46%, 8%, and 0%, respectively, of patients treated with 250 mg gefitinib and 50%, 14%, and 7%, respectively, of patients receiving 500 mg gefitinib. Grade 1, 2, and 3 skin rash occurred in 46%, 0%, and 0%, respectively, of patients treated with 250 mg gefitinib and 21%, 50%, and 7%, respectively of patients receiving 500 mg gefitinib.
Logistic regression analysis indicated that observed total and unbound gefitinib Css,min were the only statistically significant predictors for the incidence of diarrhea (Fig. 3, A and B). Inclusion of Css,min as a linear function on the probability of diarrhea resulted in a decrease in the objective function value of 4.56 (P<.05) compared with the model that excluded drug effect. The values of the slope were estimated as 0.003 µM–1 and 0.078 µM–1 for the total and unbound gefitinib Css,min, respectively, and the respective relative standard error of estimation was 49% and 46%. No statistically significant associations were found between gefitinib pharmacokinetic parameters or midazolam oral clearance and the incidence of rash.
|
|
DISCUSSION |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
In this study, interindividual variability in gefitinib plasma exposure was extensive (for example, the average steady-state trough plasma concentration during the first 28 days of dosing varied up to 18-fold). Based on the pharmacologic principle that adequate systemic exposure is a prerequisite for sufficient drug delivery to tissue, this degree of pharmacokinetic variability has the potential to result in two kinds of negative treatment outcomes: in some patients, the drug exposure achieved may be too low for adequate inhibition of EGFR, whereas in others it may be higher than desired, increasing the likelihood that some patients will experience toxicity. Identification of factors that predict gefitinib exposure may allow optimization of gefitinib treatment in a particular individual. This study prospectively assessed the value of oral midazolam as an in vivo phenotypic probe for CYP3A using a population pharmacokinetic approach. Midazolam clearance accounted for a substantial (approximately 40%) portion of the unexplained intraindividual variation in gefitinib clearance. Thus, the oral midazolam test may be useful to identify patients predisposed to higher gefitinib exposure as well as those that are potentially underdosed due to low drug absorption and/or high systemic clearance. In fact, in our study the two patients with the lowest unbound gefitinib plasma trough concentrations had the highest midazolam oral clearance values (approximately 100 L/h [Fig. 2, D]) and could have been identified prospectively using the oral midazolam test. Because the association of gefitinib exposure with the risk of developing the principal toxicity of diarrhea was statistically significant, the test may also be valuable in identifying those patients whose higher gefitinib exposure would predispose them to toxicity.
In the randomized phase II IDEAL 1 and 2 trials, where gefitinib doses of 250 and 500 mg were evaluated, the higher gefitinib dose resulted in a greater incidence of drug-related toxicity but not efficacy (2,3). However, it is unknown if genetic and biologic characteristics of patient tumors and drug exposure were similar between the two dosing arms of these trials, which makes any observations of dose–efficacy relationships problematic. In this study, substantial interindividual pharmacokinetic variability was observed, resulting in wide overlap in gefitinib plasma concentrations between the 250- and 500-mg dose levels. The strong and statistically significant correlation between the oral midazolam test and gefitinib concentrations suggests that a priori assessment of CYP3A activity with this test might allow for individualized adjustment of gefitinib doses to achieve adequate systemic exposure for inhibition of EGFR and downstream signaling in target tissue.
The strong correlation observed between the apparent oral clearances of midazolam and gefitinib (R2 = .6, P<.001) (Fig. 2) is not surprising, given the predominant role of CYP3A (including CYP3A4 and 3A5 isozymes) in the metabolism of both drugs (13,14,23,24) and the fact that both are poor substrates of ABCB1 (or P-glycoprotein) (25,26). Further contributing to the usefulness of midazolam as a predictor of drug exposure is the fact that, after oral administration, it undergoes substantial first-pass intestinal metabolism in the intestines and liver (27,28). Thus, midazolam oral clearance is a useful indicator of the combined hepatic and intestinal CYP3A activity that may play a principal role in the first-pass metabolism and systemic clearance of gefitinib.
Erlotinib, another orally active EGFR-TK inhibitor, also undergoes extensive CYP3A-mediated metabolism, and, like gefitinib, it exhibits wide interindividual pharmacokinetic variability (29). A population pharmacokinetic analysis of erlotinib involving 708 cancer patients demonstrated an interindividual variability of 52% in apparent clearance, of which 24% could be explained by a combined effect of total bilirubin, alpha-1-acid glycoprotein, smoking status, sex, albumin, and creatinine clearance (29). Due to the involvement of CYP3A in erlotinib metabolism, it is likely that the oral midazolam test will account for a larger percentage of unexplained interindividual variability in erlotinib clearance than that accounted for by the covariates tested in the population pharmacokinetic analysis, and the usefulness of the test in predicting erlotinib exposure warrants investigation. Likewise, CYP3A plays a predominant role in the metabolism of three other kinase inhibitors—imatinib (30), sorafenib (31), and sunitinib (32)—and the oral midazolam test may also be useful for prediction of exposure to these drugs as well.
Assessment of exposure–toxicity relationships using logistic regression analysis demonstrated that both total and unbound gefitinib Css,min predicted the probability of the development of diarrhea (Fig. 3). This observation agrees with previous reports that demonstrated that gefitinib-related gastrointestinal toxicity resulting in diarrhea and skin toxicity occurs at a higher frequency at higher doses (33,34), where drug exposure tended to be higher (10–12,35); however, in this study, where pharmacokinetic data were obtained in a larger cohort of patients treated with gefitinib 250 or 500 mg, a dose-dependent increase in exposure (e.g., Css,min) was not observed (Table 2), likely due to substantial interindividual pharmacokinetic variability. Because both total and unbound Css,min equally predicted diarrhea, total and unbound concentrations are highly correlated (R2 = .91) (18), and interindividual pharmacokinetic variability is similar for both unbound and total gefitinib, monitoring of unbound gefitinib plasma concentrations may not be necessary for future exposure–response and pharmacodynamic analyses.
Consistent with previous studies, no statistically significant relationship was observed between the incidence of rash and total or unbound gefitinib exposure in this study. One possible explanation is the fact that gefitinib-related skin toxicity is more strongly associated with inhibition of EGFR in normal epidermal keratinocytes and other cells residing in the skin than with systemic exposure, and factors related to inhibition of EGFR in these cells (e.g., EGFR expression, EGFR polymorphisms) and downstream signaling may better predict this toxicity (36,37).
Because gefitinib has a favorable safety profile and is generally well tolerated (33), exposure-related adverse events may not be a principal concern as they are for cytotoxic chemotherapeutics. On the contrary, underdosing may be a more clinically relevant concern with targeted therapy. At equilibrium, when steady state is achieved, unbound plasma concentrations are thought to reflect free, pharmacologically active drug concentrations in interstitial fluid and target cells. In this study, 11 of 13 patients treated with 250 mg gefitinib once daily achieved unbound steady-state plasma trough concentrations that were above the IC50 of gefitinib for mutant EGFR (0.015 µM), but only one patient had unbound trough concentrations above the IC50 for inhibition of wild-type EGFR (0.1 µM) (Fig. 1, E). It is reasonable to assume that maximum tumor growth inhibition would be achieved only if EGFR-TK and its downstream signaling pathway were blocked effectively, and therefore, EGFR-TK inhibitors should be given at a dose that achieves appropriate inhibitory concentrations in tumors.
The potential value of the population pharmacokinetic model incorporating midolozam clearance in optimizing gefitinib dose is illustrated in Table 4, which lists estimated values for unbound average steady-state concentrations in patients with varying midazolam oral clearance and receiving different gefitinib doses. As shown, patients with high midazolam clearance (e.g., 100 L/h) may benefit from a higher gefitinib dose of 750 mg to achieve unbound concentrations above the IC50 for mutant EGFR, whereas patients with low midazolam clearance (e.g., 10 L/h) or an average midazolam clearance (40 L/h) would achieve unbound steady-state concentrations of approximately 0.1 µM, the IC50 for wild-type EGFR, at gefitinib doses of 250 and 500 mg, respectively. Because the frequency of somatic mutations in EGFR is approximately 14% in Caucasian populations and approximately 27% in Asians (7), administration of gefitinib at a dose of 250 mg probably does not achieve adequate intratumoral exposure for inhibition of EGFR in the majority of patients who do not have EGFR mutations. However, based on exposure–toxicity assessments (Fig. 3), increasing the gefitinib dose to achieve exposure for inhibition of wild-type EGFR is also likely to increase the probability of experiencing diarrhea. Individualized gefitinib therapy with dose selection based on CYP3A activity could optimize systemic and tumor exposure to maximize EGFR inhibition.
|
The pharmacokinetic model described has several limitations. The first is the relatively small sample size used in its development. Although the relative standard error for the model-estimated population clearance for gefitinib was acceptable (<25%), the error for some of the other pharmacokinetic parameters approached 50% (Table 3). This, combined with the oral midazolam test, which accounted for 40% (rather than a higher percentage) of interindividual gefitinib pharmacokinetic variability, may limit the ability of the model to predict a priori gefitinib exposure and thus the extent to which the gefitinib dose can be individualized successfully in the majority of patients. Furthermore, the oral midazolam test used an intensive pharmacokinetic sampling scheme that required the patient to be in the clinic for up to 7 hours. For an oral midazolam test to be feasible and have wide clinical applicability, a more limited blood sampling scheme will be required. Finally, the clinical applicability of an individualized dosing strategy based on the oral midazolam test would require prospective evaluation, preferably in a suitably sized clinical trial that included patients who might be more likely to respond to EGFR-TK inhibitor therapy, such as those with activating EGFR mutations, high EGFR gene copy number, and/or inhibition of relevant downstream markers.
In conclusion, a population pharmacokinetic model that adequately described the multiple-dose plasma concentration versus time profiles for total and unbound gefitinib was developed. The model indicated that midazolam oral clearance accounted for approximately 40% of the interindividual variability in gefitinib oral clearance, supporting further investigation of the oral midazolam test as a potential in vivo CYP3A-phenotyping probe to predict gefitinib exposure and, importantly, to identify those patients who may be underdosed due to low oral bioavailability or high systemic clearance. The knowledge gained from the population model we have presented may also be useful for the optimization of dosing of other kinase inhibitors, for which CYP3A plays a principal role in drug absorption and systemic clearance.
|
NOTES |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
Supported, in part, by the Commonwealth Foundation for Cancer Research and AstraZeneca Pharmaceuticals. AstraZeneca reviewed the final version of the manuscript.
Presented at the 41st Annual Meeting of the American Society of Clinical Oncology, Orlando, FL, May 13–17, 2005.
|
REFERENCES |
|---|
|
TopNotesAbstractIntroductionPatients and methodsResultsDiscussionReferences |
|---|
(1) Baselga J, Averbuch SD. ZD1839 (‘Iressa’) as an anticancer agent. Drugs 2000;60 Suppl 1:33–40; discussion 41–2.
(2) Fukuoka M, Yano S, Giaccone G, Tamura T, Nakagawa K, Douillard JY, et al. Multi-institutional randomized phase II trial of gefitinib for previously treated patients with advanced non-small-cell lung cancer. J Clin Oncol 2003;21:2237–46.
(3) Kris MG, Natale RB, Herbst RS, Lynch TJ Jr, Prager D, Belani CP, et al. Efficacy of gefitinib, an inhibitor of the epidermal growth factor receptor tyrosine kinase, in symptomatic patients with non-small cell lung cancer: a randomized trial. JAMA 2003;290:2149–58.
(4) Thatcher N, Chang A, Parikh P, Rodrigues Pereira J, Ciuleanu T, von Pawel J,et al. Gefitinib plus best supportive care in previously treated patients with refractory advanced non-small-cell lung cancer: results from a randomised, placebo-controlled, multicentre study (Iressa Survival Evaluation in Lung Cancer). Lancet 2005;366:1527–37.[CrossRef][Web of Science][Medline]
(5) 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.
(6) Johnson BE, Janne PA. Selecting patients for epidermal growth factor receptor inhibitor treatment: a FISH story or a tale of mutations? J Clin Oncol 2005;23:6813–6.
(7) Janne PA, Engelman JA, Johnson BE. Epidermal growth factor receptor mutations in non-small-cell lung cancer: implications for treatment and tumor biology. J Clin Oncol 2005;23:3227–34.
(8) Tsao MS, Sakurada A, Cutz JC, Zhu CQ, Kamel-Reid S, Squire J, et al. Erlotinib in lung cancer—molecular and clinical predictors of outcome. N Engl J Med 2005;353:133–44.
(9) Lynch TJ, Bell DW, Sordella R, Gurubhagavatula S, Okimoto RA, BranniganBW, et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med 2004;350:2129–39.
(10) Ranson M, Hammond LA, Ferry D, Kris M, Tullo A, Murray PI, et al. ZD1839, a selective oral epidermal growth factor receptor-tyrosine kinase inhibitor, is well tolerated and active in patients with solid, malignant tumors: results of a phase I trial. J Clin Oncol 2002;20:2240–50.
(11) 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.
(12) Nakagawa K, Tamura T, Negoro S, Kudoh S, Yamamoto N, Takeda K, et al. Phase I pharmacokinetic trial of the selective oral epidermal growth factor receptor tyrosine kinase inhibitor gefitinib (‘Iressa’, ZD1839) in Japanese patients with solid malignant tumors. Ann Oncol 2003;14:922–30.
(13) McKillop D, Hutchison M, Partridge EA, Bushby N, Cooper CM, Clarkson-Jones JA, et al. Metabolic disposition of gefitinib, an epidermal growth factor receptor tyrosine kinase inhibitor, in rat, dog and man. Xenobiotica 2004;34:917–34.[CrossRef][Web of Science][Medline]
(14) McKillop D, McCormick AD, Millar A, Miles GS, Phillips PJ, Hutchison M. Cytochrome P450-dependent metabolism of gefitinib. Xenobiotica 2005;35:39–50.[CrossRef][Web of Science][Medline]
(15) Swaisland HC, Ranson M, Smith RP, Leadbetter J, Laight A, McKillop D, et al. Pharmacokinetic drug interactions of gefitinib with rifampicin, itraconazole and metoprolol. Clin Pharmacokinet 2005;44:1067–81.[CrossRef][Web of Science][Medline]
(16) Kulesza P, Brahmer JR, Jimeno A, Baker SD, Spitz A, Li J, et al. Evaluation of gefitinib biological effects in patients with solid tumors amenable to sequential biopsies—final results. Proc Am Soc Clin Oncol 2006;24:Abstract 3090.
(17) Zhao M, Hartke C, Jimeno A, Li J, He P, Zabelina Y, et al. Specific method for determination of gefitinib in human plasma, mouse plasma and tissues using high performance liquid chromatography coupled to tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2005;819:73–80.[Web of Science][Medline]
(18) Li J, Brahmer J, Messersmith W, Hidalgo M, Baker SD. Binding of gefitinib, an inhibitor of epidermal growth factor receptor-tyrosine kinase, to plasma proteins and blood cells: in vitro and in cancer patients. Invest New Drugs 2006;24:291–7.[CrossRef][Web of Science][Medline]
(19) Jonsson EN, Karlsson MO. Xpose—an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM. Comput Methods Programs Biomed 1999;58:51–64.[CrossRef][Web of Science][Medline]
(20) Wahlby U, Jonsson EN, Karlsson MO. Comparison of stepwise covariate model building strategies in population pharmacokinetic-pharmacodynamic analysis. AAPS PharmSci 2002;4:E27.[CrossRef][Medline]
(21) Xie R, Mathijssen RH, Sparreboom A, Verweij J, Karlsson MO. Clinical pharmacokinetics of irinotecan and its metabolites in relation with diarrhea. Clin Pharmacol Ther 2002;72:265–75.[CrossRef][Web of Science][Medline]
(22) Wong M, Balleine RL, Collins M, Liddle C, Clarke CL, Gurney H. CYP3A5 genotype and midazolam clearance in Australian patients receiving chemotherapy. Clin Pharmacol Ther 2004;75:529–38.[CrossRef][Web of Science][Medline]
(23) Kronbach T, Mathys D, Umeno M, Gonzalez FJ, Meyer UA. Oxidation of midazolam and triazolam by human liver cytochrome P450IIIA4. Mol Pharmacol 1989;36:89–96.[Abstract]
(24) Wandel C, Bocker R, Bohrer H, Browne A, Rugheimer E, Martin E. Midazolam is metabolized by at least three different cytochrome P450 enzymes. Br J Anaesth 1994;73:658–61.
(25) Kim RB, Wandel C, Leake B, Cvetkovic M, Fromm MF, Dempsey PJ, et al. Interrelationship between substrates and inhibitors of human CYP3A and P-glycoprotein. Pharm Res 1999;16:408–14.[CrossRef][Web of Science][Medline]
(26) Ozvegy-Laczka C, Hegedus T, Varady G, Ujhelly O, Schuetz JD, Varadi A, et al. High-affinity interaction of tyrosine kinase inhibitors with the ABCG2 multidrug transporter. Mol Pharmacol 2004;65:1485–95.
(27) Thummel KE, O'Shea D, Paine MF, Shen DD, Kunze KL, Perkins JD, et al. Oral first-pass elimination of midazolam involves both gastrointestinal and hepatic CYP3A-mediated metabolism. Clin Pharmacol Ther 1996;59:491–502.[CrossRef][Web of Science][Medline]
(28) Paine MF, Shen DD, Kunze KL, Perkins JD, Marsh CL, McVicar JP, et al. First-pass metabolism of midazolam by the human intestine. Clin Pharmacol Ther 1996;60:14–24.[CrossRef][Web of Science][Medline]
(29) Lu J-F, Eppler S, Hamilton M, Rakhit A, Gaudreault J, Bruno R, et al. A population pharmacokinetic (PK) model for erlotinib (E), a small molecule Inhibitor of the epidermal growth factor receptor (EGFR). American Society of Clinical Oncology Annual Meeting; June 2–6, 2005; Orlando. Orlando (FL): American Society of Clinical Oncology; 2005.
(30) Gschwind HP, Pfaar U, Waldmeier F, Zollinger M, Sayer C, Zbinden P, et al. Metabolism and disposition of imatinib mesylate in healthy volunteers. Drug Metab Dispos 2005;33:1503–12.
(31) Lathia C, Lettieri J, Cihon F, Gallentine M, Radtke M, Sundaresan P. Lack of effect of ketoconazole-mediated CYP3A inhibition on sorafenib clinical pharmacokinetics. Cancer Chemother Pharmacol 2006;57:685–92.[CrossRef][Web of Science][Medline]
(32) Faivre S, Delbaldo C, Vera K, Robert C, Lozahic S, Lassau N, et al. Safety, pharmacokinetic, and antitumor activity of SU11248, a novel oral multitarget tyrosine kinase inhibitor, in patients with cancer. J Clin Oncol 2006;24:25–35.
(33) Cohen MH, Williams GA, Sridhara R, Chen G, McGuinn WD Jr, Morse D, et al. United States Food and Drug Administration Drug Approval summary: gefitinib (ZD1839; Iressa) tablets. Clin Cancer Res 2004;10:1212–8.
(34) Rothenberg ML, Lafleur B, Levy DE, Washington MK, Morgan-Meadows SL, Ramanathan RK, et al. Randomized phase II trial of the clinical and biological effects of two dose levels of gefitinib in patients with recurrent colorectal denocarcinoma. J Clin Oncol 2005;23:9265–74.
(35) Herbst RS, Maddox AM, Rothenberg ML, Small EJ, Rubin EH, Baselga J,et al. Selective oral epidermal growth factor receptor tyrosine kinase inhibitor ZD1839 is generally well-tolerated and has activity in non-small-cell lung cancer and other solid tumors: results of a phase I trial. J Clin Oncol 2002;20:3815–25.
(36) Perez-Soler R, Saltz L. Cutaneous adverse effects with HER1/EGFR-targeted agents: is there a silver lining? J Clin Oncol 2005;23:5235–46.
(37) Amador ML, Oppenheimer D, Perea S, Maitra A, Cusati G, Iacobuzio-Donahue C, et al. An epidermal growth factor receptor intron 1 polymorphism mediates response to epidermal growth factor receptor inhibitors. Cancer Res 2004;64:9139–43.
Manuscript received March 2, 2006; revised September 7, 2006; accepted October 20, 2006.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. D Baker Pharmacokinetic Considerations for Molecularly Targeted Therapy Am. Assoc. Cancer Res. Educ. Book, April 12, 2008; 2008(1): 705 - 709. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. M. Rudin, W. Liu, A. Desai, T. Karrison, X. Jiang, L. Janisch, S. Das, J. Ramirez, B. Poonkuzhali, E. Schuetz, et al. Pharmacogenomic and Pharmacokinetic Determinants of Erlotinib Toxicity J. Clin. Oncol., March 1, 2008; 26(7): 1119 - 1127. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. V. Karamouzis, J. R. Grandis, and A. Argiris Therapies Directed Against Epidermal Growth Factor Receptor in Aerodigestive Carcinomas JAMA, July 4, 2007; 298(1): 70 - 82. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Li, M. Zhao, P. He, M. Hidalgo, and S. D. Baker Differential Metabolism of Gefitinib and Erlotinib by Human Cytochrome P450 Enzymes Clin. Cancer Res., June 15, 2007; 13(12): 3731 - 3737. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||