Purpose: The aim of this pharmacokinetic–pharmacodynamic (PK–PD) analysis was to evaluate the pharmacologic characteristics of erlotinib and its main metabolite (OSI-420) in pediatric patients compared with those in adult patients.

Experimental Design: Plasma concentrations of erlotinib and OSI-420 of 46 children with malignant brain tumors included in a phase I study and 42 adults with head and neck carcinoma were analyzed by a population-pharmacokinetic method (NONMEM). The effect of several covariates and single nucleotide polymorphisms (SNP) in ABCB1, ABCG2, and CYP3A5 on pharmacokinetic parameters was evaluated. PK/PD relationships between plasma drug exposure Area Under the Curve (AUC) at day 1 and skin toxicity were studied in children and compared with the relationship observed in adults.

Results: A significant difference in erlotinib clearance (P = 0.0001), when expressed in L·h−1·kg−1, was observed between children and adults with mean values of 0.146 and 0.095, respectively (mean difference = 0.051 L·h−1·kg−1, SD = 0.0594). However, a common covariate model was obtained describing erlotinib clearance according to body weight, alanine aminotransferase, ABCB1, and CYP3A5 polymorphisms (2677G > T/A and 6986G > A) for both children and adult patients. The PK–PD relationship was very consistent between the children and adult groups with risk of skin toxicity rising with increasing erlotinib AUC.

Conclusions: The nonlinear population approach applied to pharmacokinetic data combined with a pharmacokinetic–pharmacodynamic analysis revealed that the higher recommended dose in children (125 mg/m2/day) compared with adults (90 mg/m2/day) is mainly due to pharmacokinetic rather than pharmacodynamic particularities. Clin Cancer Res; 17(14); 4862–71. ©2011 AACR.

This article is featured in Highlights of This Issue, p. 4617

Translational Relevance

Erlotinib is an oral epithelial growth factor receptor (EGFR) tyrosine kinase inhibitor approved for treatment of lung and pancreas cancers with potential applications in pediatrics particularly for the treatment of malignant brain tumors overexpressing EGFR. The recommended erlotinib dose differs between pediatric and adult patients (125 mg/m2/day versus 90 mg/m2/day, respectively).

The reason for this difference must be examined to compare the erlotinib therapeutic index in children and in adults, and to identify additional causes of interindividual pharmacokinetic or pharmacodynamic variability.

By using a population pharmacokinetic approach to simultaneously analyze data from two clinical trials (i.e., neoadjuvant pilot study in adults with head and neck carcinoma and phase 1 study in pediatrics), we show that the difference is related to pharmacokinetics, and not to the pharmacokinetic–pharmacodynamic relationship.

These results are useful not only for the clinical use of erlotinib, but also in terms of methodology to encourage future analysis of combined data from children and adults.

Erlotinib (Tarceva) is an orally active, potent, and selective inhibitor of the epithelial growth factor receptor (EGFR) tyrosine kinase. From a pharmacokinetic point of view, erlotinib is metabolized by CYP3A4, CYP3A5, CYP1A1, and CYP1A2 isoforms of the P450 enzyme (1). The main active metabolite OSI-420 is produced by O-demethylation of the side chains and represents approximately 5% of the circulating erlotinib (2). It has been suggested that erlotinib is a substrate of both ABCB1 (P-glycoprotein) and ABCG2 (BCRP) transporters (3, 4). Erlotinib is registered for treatment of locally advanced or metastatic nonsmall cell lung cancer with stable disease after standard platinum-based first-line chemotherapy or after failure of at least one prior chemotherapy regimen, according to a daily dose of 150 mg per day (5). Erlotinib is also indicated at a daily dose of 100 mg in combination with gemcitabine for the treatment of patients with metastatic pancreatic cancer (6). Several promising preclinical studies have led to the exploration of its use in pediatrics, particularly in children with malignant brain tumors overexpressing EGFR (7, 8). Broniscer and colleagues reported on erlotinib pharmacokinetics in plasma of an 8-year old child with glioblastoma treated with oral erlotinib and found it to be similar to that of published adult studies (9). A Phase I study was set up with 46 children with refractory solid tumors treated with erlotinib alone or erlotinib and temozolomide (10). The authors determined that the maximum tolerated dose (MTD) was 85 mg/m2/day, however significant interpatient variability in erlotinib disposition was also found at all dose levels. Another Phase I study in 23 patients with high-grade glioma aged 3 to 22 including erlotinib pharmacokinetics administered concurrently with radiotherapy found a much higher MTD of 120/mg/m2/day (11). The authors found that the PK variables of erlotinib and OSI-420 were similar to those described in adults. A third multicenter phase I study was carried out to establish the recommended dose of erlotinib, when given alone or combined with radiotherapy to children with malignant brain tumors (12). The recommended dose (125 mg/m2/day) was found to be the same for both treatment strata and is higher than that in adults (corresponding to around 90 mg/m2/day). The dosing of tyrosine kinase inhibitors has become an issue. Relationships between imatinib dose and efficacy have been observed in some subgroups of patients with gastrointestinal stromal tumors (GIST) defined according to c-Kit genotype (13). However Delbaldo and colleagues found a wide dispersion of imatinib plasma concentrations in 34 patients with advanced GIST receiving the same dose, with no relationship between imatinib exposure and response rate or between total or free unbound imatinib and clinical activity, but a positive correlation between exposure and hematologic toxicity (14). Other studies described close relationships between plasma imatinib concentrations and both efficacy and toxicity in patients with either chronic myeloid leukemia (15, 16) or GIST (17, 18). Similar relationships have been observed for sunitinib (19).

The aim of this pharmacokinetic–pharmacodynamic (PK–PD) analysis was to evaluate the pharmacologic characteristics of erlotinib in paediatric patients compared with those in adult patients. The adult pharmacokinetic data was provided by a neoadjuvant pilot study carried out in patients with head and neck carcinoma (20, 21) and the pediatric data come from a phase 1 study in children with malignant brain-tumors (12). First, both datasets were combined to evaluate PK characteristics. Secondly, relationships between individual pharmacokinetic parameters, such as erlotinib plasma exposure, and adverse events were examined to compare child and adult patients' pharmacodynamic sensitivity to the drug.

Patients

Adult patients (n = 42) with nonmetastatic head and neck tumors awaiting surgical resection were enrolled in a clinical trial. The study design and eligibility criteria have been described previously in detail (20). Patients were treated with erlotinib at 150 mg/day for a variable period corresponding to the time between pan-endoscopy and surgical resection. In the event of grade 2 or more diarrhea or skin rash that were symptomatically unacceptable to the patient, treatment was withheld until resolution to grade 1, and then erlotinib was restarted at a dose of 100 mg/day. If toxicity reoccurred, erlotinib was stopped.

Fifty young patients (age range 2 to 19 years) with malignant brain tumors were included within a multicenter, dose escalation phase I study (NCT00418327; ITCC-003/MO18461) of the Innovative Therapies with Children with Cancer (ITCC) European Consortium. Children with histologically/cytologically confirmed malignant brain tumors refractory to, or relapsing after, first line therapy, for which no effective treatment exists, were enrolled into Group 1 (erlotinib alone). Children with newly diagnosed, histologically confirmed brainstem glioma were enrolled into Group 2 (erlotinib and radiotherapy). Eligibility criteria are described in the article reporting the clinical results (12). Erlotinib tablets were administered orally once daily in 3-week cycles at four dose levels; 1 = 75 mg/m2, 2 = 100 mg/m2, 3 = 125 mg/m2, 4 = 150 mg/m2. According to the recommendations for the conduct of phase I trials in children with cancer (22), the starting dose was approximately 80% of the adult recommended dose (RD) (150 mg/day). In Group 2, patients received local irradiation to the brainstem region of 54 Gy over 6 weeks (1.8 Gy/fraction/day). The first erlotinib dose was administered within 4 hours of the first irradiation. Conventional 3+3 dose escalation methodology was used for Group 1 (23), with an extension of 10 patients at the RD. In Group 2, dose escalation decisions were made using a continual reassessment method with likelihood-based inference (24).

Both populations of patients (adults and children) were fasted when taking the drug as recommended in the protocols. For both children and adults, complete medical histories, a physical examination, and the following laboratory tests were done at baseline and at each scheduled visit: complete differential blood count, albumin (ALB), α1 acid glycoprotein (AGP), alanine aminotransaminase (ALAT). The concomitant administration of enzymatic inducers (such as carbamazepine, phenytoin, and glucocorticoids) was also reported in both populations. Both studies were approved by independent institutional review board/ethical committees and conducted in compliance with the Declaration of Helsinki and the Guidelines for Good Clinical Practice. Written informed consent was obtained from all subjects or parents/legal representatives in the case of children under 18.

Blood sampling and mass spectrometry analysis

Serial blood samples were taken at various intervals to determine the pharmacokinetic profile of erlotinib and its principal metabolite (OSI-420). For adults, on day 1 of the treatment, three blood samples were collected from each patient: before erlotinib administration, 2 hours and 4 to 8 hours after the dose. Patients were clinically evaluated at a median time of 14 days after starting the treatment and one blood sample 24 hours after the last administration was obtained. On the day preceding, and the day of the surgery, two additional samples were collected (6–8 hours and approximately 24 hours, respectively, after the last administration of erlotinib).

For children and young adults, plasma samples were collected before drug intake at day 1, 30 minutes, 1, 2, 4, 6, 8, and 24 hours after the erlotinib dose at day 8 and day 22 for group 1 and at day 8 and day 43 (end of irradiation) for group 2.

For both children and adults, blood samples were centrifuged (1500 g, 15 min, 4°C), the plasma frozen at −20°C until analysis and the determination of plasma erlotinib and OSI-420 was done by Advion BioServices, Inc (Ithaca). Quantitative analyses were done using a validated coupled liquid chromatography-mass spectrometry technique (25). The calibration range was 1–3000 ng/ml for erlotinib and 1–1000 ng/ml for OSI-420.

Genetic analysis

A blood sample was collected on day 1 for pharmacogenetic investigation. Genomic DNA was extracted from peripheral blood leukocytes according to standard procedures with QIAamp DNA Blood Mini Kit (Qiagen) according to the manufacturer's instructions. Genotyping was done for 5 single-nucleotide polymorphisms (SNP). ABCB1 variants (2677G>T, 2677G>A and 3435C>T) were determined by PCR-RFLP as previously described (26, 27). ABCG2 421C>A variant was determined by polymerase chain reaction (PCR) and direct sequencing as published before (21). The presence of CYP3A5*1/CYP3A5*3 alleles was determined on genomic DNA with allele specific real time PCR as previously described (28). Whenever genetic data was missing, genotypes were randomly assigned according to the frequency of the genotypes in the remaining study population. Box plots of clearance values were examined to verify that this attribution did not change the population clearance distribution.

Population pharmacokinetic analysis

Plasma erlotinib and OSI-420 concentrations were analyzed according to a nonlinear mixed effects approach using the NONMEM program (version VI, level 1.0, Icon Development Solutions running on an Intel Core 2 Quad) according to a one-compartment pharmacokinetic model for erlotinib and to a two-compartment pharmacokinetic model when both erlotinib and OSI-420 were analyzed and using the first-order conditional estimation (FOCE) method (21). A proportional error model was used for both interpatient and residual variability.

Determination of individual pharmacokinetic parameters and covariate analyses

The influence of the following variables on pharmacokinetic parameters was examined: body weight, sex (0 for male, 1 for female), age, population (coded POP = 0 for adults and POP = 1 for children), albuminemia, plasma AGP level, ALAT concentration, effect of simultaneous administration of enzymatic inducers (carbamazepine, phenytoine, dexamethasone, prednisone, prednisolone), genotype for ABCB1 (2677G > T/A and 3435C > T), ABCG2 (421C > A) and CYP3A5 (6986G > A). Moreover, interoccasion variability (IOV) was tested on all the parameters because data from 2 cycles were analyzed simultaneously. The first occasion was defined as the samples of cycle 1 (i.e., day 1 and day 8 for children) and the second occasion as those of cycle 2 (i.e., day 22 for group 1 patients) or cycle 3 (i.e. day 43 for group 2 patients). There was no IOV for adults because they only received one cycle of erlotinib.

First, the influence of each covariate on total plasma clearance of erlotinib (CL) was tested individually. For example, in the case of body weight, the following equation was used: CL = θ1 (WT/mean WT)θ2, where θ1 is the typical value of CL (TVCL) estimated by the population mean and θ2 is the estimated influential factor for body weight. The χ2 test was used to compare the objective function values (OFV = −2loglikelihood) of nested models (likelihood ratio test). A decrease of at least 3.84 (P < 0.05) was required for a covariate to be considered significantly correlated with the pharmacokinetic variable. Second, an intermediate model including all significant covariates was obtained. A stepwise backward elimination procedure was carried out. Covariates remained in the final population pharmacokinetic model when the removal of the covariate resulted in an increase of the OFV of at least 10.83 (P < 0.001). The influence of covariates on other pharmacokinetic parameters (volume of distribution V and absorption constant Ka) was examined using the same methodology. The likelihood ratio test was also used to select the structural pharmacokinetic model. Once the population pharmacokinetics model for erlotinib had been defined (including the covariates), all the parameter values were fixed while developing the OSI-420 pharmacokinetic model using the same methodology as described for erlotinib. A bioavailability (F) of 1 was assumed in the absence of IV drug administration data and CL and V corresponded to apparent parameters (i.e., CL/F and V/F).

Performance of final pharmacokinetic model

A visual predictive check (VPC) was carried out by simulating from the final model 1000 concentrations post-dose at day 1, day 8, day 22, and 43. The 50th percentile concentration (as an estimator of the population-predicted concentration) and the 5th and 95th percentile concentrations were processed using R (RfN, version 2007a) and then plotted with Stata version 10.0, Stata Co.

Pharmacokinetic/Pharmacodynamic relationship

Systemic exposure to erlotinib was calculated using individual post hoc clearance: AUC = dose/CL for erlotinib. Relationship between AUC at day 1 (i.e., dose/CL at day 1) and skin toxicity at the midterm treatment visit was evaluated in children and was compared with the relationship previously observed in adults (21). Toxicity (i.e., skin rash, folliculitis, acne, and pruritis) was evaluated at the midterm treatment visit and graded using the National Cancer Institute Common Toxicity Criteria, version 2.0 or 3.0. The relationship between initial daily AUC and skin toxicity grade was examined using a four-level ordered logistic regression for adults (with skin toxicity graded 0–3) and using a three-level ordered logistic regression for children [with skin toxicity graded 0, 1 and 2/3 (grades 2 and 3 were grouped as only 3 children developed a grade 3 skin toxicity)]. Based on this model, the cumulative-predicted probabilities of skin toxicity according to AUC were obtained and represented as curves. The analyses were conducted using Stata version 10.0, Stata Co.

Determination of cutoff point of erlotinib AUC

Receiver operating characteristic (ROC) curves were constructed to examine the value of an erlotinib AUC threshold predicting a minimum skin toxicity of grade 2. The area under ROC curves of erlotinib AUC both in children and in adults for the prediction of a minimum skin toxicity of grade 2 were calculated. The diagnostic accuracy values of sensitivity (SE) and specificity (SP) were determined. The optimal cutoff value was obtained. The analyses were conducted using Stata version 10.0, Stata Co.

Patients

Forty-two adults were assessable for pharmacokinetic investigations and plasma samples of 46 children (26 in group 1 and 20 in group 2) were analyzed. A total of 1659 plasma samples were taken into account for the population pharmacokinetic analysis. Adults were treated for a median time of 20 days (range 5–32 days) and children for 50 days (range 8–106 days). Complete pharmacogenetic data were available for 79 (42 adults and 37 children) of the 88 patients. Minor allele frequencies are given in Table 1 (footnote). The genotypes did not depart significantly from Hardy–Weinberg equilibrium (exact tests nonsignificant in all four cases P > 0.10). The patients' characteristics at baseline are summarized in Table 1.

Determination of individual pharmacokinetic parameters and covariate analyses

Erlotinib pharmacokinetics were adequately described by a one-compartment model with first-order absorption and first-order elimination. The corresponding pharmacokinetic variables were CL/F (apparent total plasma clearance of erlotinib), V/F (apparent volume of distribution of erlotinib), and Ka (absorption rate constant). The pharmacokinetic model included interoccasion variability of CL/F and V/F. During the individual testing of the 11 covariates, 7 covariates [body weight, population variable coded POP indicating whether patient is a child or an adult ALAT concentration, ABCB1 2677G > T/A and 3435C > T, ABCG2 421C > A, and CYP3A5*1 allele] were significantly (P < 0.05) associated with clearance. However, sex, albuminemia, plasma AGP level, and effect of simultaneous administration of enzymatic inducers were not found to be significantly associated with clearance. A stepwise backward elimination applied to the intermediate model identified the following covariates as remaining significant (P < 0.001): body weight, ALAT concentration, CYP3A5*1, and ABCB1 (2677G > T/A). An increase in body weight was associated with an increase in CL (positive θ coefficient) whereas an increase in ALAT concentration led to a decrease in CL (negative θ coefficient). For example, CL/F is estimated to increase, on average, by approximately 22% between a typical individual weighing 52 kg and one weighing 72 kg, all other covariates being equal. Concerning the effect of ALAT, CL/F is estimated to decrease, on average, by approximately 19% between a typical individual with ALAT concentration of 24.4 IU/L and one with ALAT concentration of 50 IU/L, all other covariates being equal. The presence of a variant allele of ABCB1 2677G > T/A was associated with a 19% decrease in CL whereas the presence of a CYP3A5*1 allele was associated with a 42% increase in CL. For V/F, body weight was the only significant (P < 0.001) covariate associated with a decrease in interindividual variability (IIV) from 56.6% (no covariate) to 13%. The final covariate models are detailed in Table 2. Of note, substantial interoccasion variability was observed for CL/F and V/F (22% and 46%, respectively). The residual variability was 31.9% for erlotinib concentrations.

Analysis of both plasma erlotinib and OSI-420 concentrations required an additional compartment with first order elimination from the metabolite compartment and the corresponding pharmacokinetics parameters: CLm/fm (apparent total clearance of OSI-420) and Vm/fm (apparent volume of distribution of OSI-420), where fm is the fraction of erlotinib converted into OSI-420. Four covariates (body weight, ALB, ALAT, and ABCB1 (2677G > T/A)) were significantly (P < 0.05) correlated with CLm/fm. Three covariates [body weight, ALAT, and ABCB1 (2677G>T/A)] remained significant (P < 0.001) after the stepwise backward elimination procedure from the intermediate model (Table 2). A sensitivity analysis excluding the 9 patients for which the pharmacogenetic data was incomplete was conducted to test the robustness of the model and all covariates found in the final model remained significant (P < 0.001) (data not shown). The proportion of circulating active metabolite OSI-420 compared with erlotinib, calculated using AUCmetabolite/(AUCmetabolite + AUCerlotinib), was 8.1% (SD = 3.2%).

About the discrepancy between the recommended doses in adults and children (90 and 125 mg/m2/day, respectively), the impact of the covariate POP on the pharmacokinetic model was carefully evaluated. The typical value of erlotinib clearance was 27% lower (P < 0.001) in children when POP was the only considered covariate, and 24% higher (P < 0.025) if POP was considered together with body weight (Table 2). Comparison of the individual POSTHOC values of clearance confirmed these differences: the mean value (± SD) was significantly lower (41%) in children (CL/F = 4.15 ± 2.22 L.h−1) in comparison with that in adults (CL/F = 6.98 ± 3.48 L.h−1) (P < 0.0001) while the mean (± SD) weighted clearance value was 54% higher in children (0.146 ± 0.070 L·h−1·kg−1 for children versus 0.095 ± 0.045 L·h−1·kg−1 for adults, P = 0.0001, Student's t-test). However, the addition of the POP covariate to the final covariate model (i.e., the model which includes body weight, ALAT, CYP3A5, and ABCB1 (2677G> T/A)) did not significantly decrease the OFV, and the exponent corresponding to POP was only 1.11 (Table 2). Moreover, when the child population was split between children (age < 12 years; n = 35) and adolescents (age ≥ 12 years; n = 11), the mean weighted clearance of adolescents (i.e., 0.099 ± 0.056 L.h−1·kg−1) was significantly (P = 0.009) lower than that of children (0.161 ± 0.068 L.h−1.kg−1) but not significantly different from the adults' weighted clearance (0.095 ± 0.045 L·h−1·kg−1). All these results indicate that although erlotinib clearance differs between children and adults, the final covariate model explains most of this difference.

Performance of final pharmacokinetic model

This final model adequately described erlotinib and OSI-420 pharmacokinetic profiles as shown by the goodness-of-fit plots (Fig. 1) and by the VPC (Fig. 2).

Pharmacokinetic/Pharmacodynamic relationships

The relationship between drug exposure and the severity of skin toxicity was examined. The box plot in Fig. 3 shows an increase in AUC with increasing grade of toxicity in both children and adults. This positive association between exposure and toxicity was significant in the four-level ordered logistic regression analysis for adults (P = 0.014; ref. 13) and a trend is observed in the three-level ordered logistic regression analysis for children (P = 0.06). Figure 4 shows the cumulative probabilities of having skin toxicity according to AUC using the probabilities predicted by the ordered logistic regression models [adult data analyzed with the model previously published (13) and children data analyzed with the present model]. There is within each population, adults or children, a strong PK/PD relationship, which seems very similar. It is noteworthy that skin toxicity was not significantly correlated with trough erlotinib plasma levels observed at steady-state (data no shown).

Determination of a cutoff of erlotinib AUC

Using the ROC curve, the AUC threshold predicting a minimum skin toxicity of grade 2 was 27 ng·mL·h−1 with the area under the curve at 70% (CI 95% = [59%–81%], P < 0.001) in our population. The diagnostic accuracy values of sensitivity (SE) and specificity (SP) were 64.9% and 64.7%, respectively.

Pharmacokinetic–pharmacodynamic (PK–PD) studies combining children and adults are scarce (29, 30). This PK–PD analysis carried out on data from two independent clinical trials contributes important information toward explaining the observed difference in erlotinib tolerance between pediatric and adult patients (i.e., daily recommended dose of 125 mg/m2 vs. 150 mg corresponding to approximately 90 mg/m2, respectively).

Although the skin symptoms were rather different between children and adults, the relationship between daily erlotinib AUC and the probability of skin toxicity was very consistent between the two subgroups. Previous phase I studies in children concluded that erlotinib and OSI-420 pharmacokinetics values were similar to those in adults (11, 31). However, in our study significant differences in pharmacokinetics were observed between children and adults. By considering the covariate POP (i.e., POP = 1 for children or POP = 0 for adults) alone or in combination with other covariates, the population approach applied to this dataset revealed and quantified this difference. A higher mean value of erlotinib clearance (i.e., + 54%) was observed in children compared with adults when expressed in L·h−1·kg−1. Comparison of the individual POSTHOC values confirmed this higher elimination capacity in children: CL = 0.146 ± 0.070 L·h−1·kg−1 for children versus CL = 0.095 ± 0.045 L·h−1·kg−1for adults. This pharmacokinetic particularity explains why the recommended dose in children is higher than in adults. The lack of data prevented us from including smoking status as a covariate in our study. However, a large proportion of the adults smoked (55%) whereas one can assume most of the children did not, which makes the higher clearance in children particularly remarkable and unexpected, as one can assume that the adults' clearance is increased due to the large proportion of smokers in that population. Specific consideration of the adolescents showed that their mean erlotinib clearance was lower than that in children younger than 12 years old, and was similar to that in adults. However, it is interesting to note that by adding the covariate POP to the final covariate model (i.e., which included body weight, ALAT, CYP3A5, and ABCB1 (2677G>T/A)), the typical value of CL was only 11% higher in pediatrics (versus 24% when only the POP covariate is considered). This result showed that the final covariate model explains most of the difference between the two subgroups.

The erlotinib and OSI-420 clearance were negatively correlated with ALAT. An increase of this covariate might be the reflection of hepatic dysfunction, which leads to a decrease in erlotinib and OSI-420 clearance. The impact of hepatic dysfunction on erlotinib clearance has been confirmed in a clinical study using a more conventional approach (32).

Our results also highlight the relevance of taking into account pharmacogenetic information for erlotinib pharmacokinetics analysis. By analyzing children and adults together, we found that carriers of at least one allele *1 of CYP3A5 had a 1.4 times higher clearance than patients harboring the CYP3A5 *3/*3 genotype (which represents about 80% in Caucasians (33)). This observation is in agreement with the functional consequence of the 6986G > A SNP (allele CYP3A5*3), which encodes an aberrantly spliced mRNA with a premature stop codon, leading to the absence of the protein (34). A similar trend was observed for ABCB1 2677 G>T/A SNP. Although a recent work suggested that OSI-420, but not erlotinib, is a substrate for ABCB1 (35), the presence of one variant allele (T or A) was associated with an average 20% decrease of the clearance of both erlotinib and OSI-420, suggesting the role of ABCB1 in erlotinib elimination as described in another study (36).

These results confirm the relationship between plasma concentrations of tyrosine kinase inhibitors and their dose limiting toxicity. Besides, it has been suggested that patients experiencing skin rash during erlotinib treatment have an improved survival and rash has been proposed as a surrogate marker for favorable outcome (37). The correlation observed between exposure and rash in both adults and children and the previously observed relationship between rash and efficacy strongly supports the existence of an exposure/response correlation in addition to efficacy observed in EGFR amplified and mutant tumours (38, 39). There is now a need to evaluate the benefit of therapeutic drug monitoring in patients treated with erlotinib. The methodology we used here, consisting in AUC calculation from pharmacokinetic data processed with population pharmacokinetics followed by a ROC analysis to determine the cutoff AUC, is a reasonable approach. Due to the heterogeneity between patients in terms of disease, it was not possible to examine the relationship between AUC and clinical response. Because previous clinical studies conducted with erlotinib in lung and pancreas cancer showed that a grade 2 rash was associated with a better outcome (37), we chose to determine the cutoff AUC associated with a grade 2 rash which is a surrogate marker for efficacy. In this study, we found a cutoff AUC of 27 ng/ml/h but further studies with larger datasets and in other groups of patients treated with erlotinib (e.g., lung-cancer patients) are necessary to determine optimal cutoff values and subsequently suggest their use in prospective clinical studies.

In conclusion, this analysis of combined data from two clinical trials showed that the higher recommended dose in children (125 mg/m2/day) than in adults (90 mg/m2/day) is mainly due to pharmacokinetic rather than pharmacodynamic particularities.

No potential conflicts of interest were disclosed.

We thank all the patients and parents who participated in the trials, and the clinical teams of the centers (Institut Gustave Roussy, Villejuif, Centre Leon-Berard, Lyon, Institut Curie, Paris, Centre Oscar-Lambret, Lille, Royal Marsden Hospital, Sutton, CHU LaTimone, Marseille, Birmingham Children's Hospital, Birmingham, Catholic Gemelli University, Rome).

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.
Li
J
,
Zhao
M
,
He
P
,
Hidalgo
M
,
Baker
SD
. 
Differential metabolism of gefitinib and erlotinib by human cytochrome P450 enzymes
.
Clin Cancer Res
2007
;
13
:
3731
7
.
2.
Ling
J
,
Johnson
KA
,
Miao
Z
,
Rakhit
A
,
Pantze
MP
,
Hamilton
M
, et al
Metabolism and excretion of erlotinib, a small molecule inhibitor of epidermal growth factor receptor tyrosine kinase, in healthy male volunteers
.
Drug Metabol Dispos: Biol Fate Chem
2006
;
34
:
420
6
.
3.
Li
J
,
Cusatis
G
,
Brahmer
J
,
Sparreboom
A
,
Robey
RW
,
Bates
SE
, et al
Association of variant ABCG2 and the pharmacokinetics of epidermal growth factor receptor tyrosine kinase inhibitors in cancer patients
.
Cancer Biol Ther
2007
;
6
:
432
8
.
4.
Shi
Z
,
Peng
XX
,
Kim
IW
,
Shukla
S
,
Si
QS
,
Robey
RW
, et al
Erlotinib (Tarceva, OSI-774) antagonizes ATP-binding cassette subfamily B member 1 and ATP-binding cassette subfamily G member 2-mediated drug resistance
.
Cancer Res
2007
;
67
:
11012
20
.
5.
Johnson
JR
,
Cohen
M
,
Sridhara
R
,
Chen
YF
,
Williams
GM
,
Duan
J
, et al
Approval summary for erlotinib for treatment of patients with locally advanced or metastatic non-small cell lung cancer after failure of at least one prior chemotherapy regimen
.
Clin Cancer Res
2005
;
11
:
6414
21
.
6.
Moore
MJ
,
Goldstein
D
,
Hamm
J
,
Figer
A
,
Hecht
JR
,
Gallinger
S
, et al
Erlotinib plus gemcitabine compared with gemcitabine alone in patients with advanced pancreatic cancer: a phase III trial of the National Cancer Institute of Canada Clinical Trials Group
.
J Clin Oncol
2007
;
25
:
1960
6
.
7.
Geoerger
B
,
Gaspar
N
,
Opolon
P
,
Morizet
J
,
Devanz
P
,
Lecluse
Y
, et al
EGFR tyrosine kinase inhibition radiosensitizes and induces apoptosis in malignant glioma and childhood ependymoma xenografts
.
Int J Cancer
2008
;
123
:
209
16
.
8.
Sarkaria
JN
,
Yang
L
,
Grogan
PT
,
Kitange
GJ
,
Carlson
BL
,
Schroeder
MA
, et al
Identification of molecular characteristics correlated with glioblastoma sensitivity to EGFR kinase inhibition through use of an intracranial xenograft test panel
.
Mol Cancer Ther
2007
;
6
:
1167
74
.
9.
Broniscer
A
,
Panetta
JC
,
O'Shaughnessy
M
,
Fraga
C
,
Bai
F
,
Krasin
MJ
, et al
Plasma and cerebrospinal fluid pharmacokinetics of erlotinib and its active metabolite OSI-420
.
Clin Cancer Res
2007
;
13
:
1511
5
.
10.
Jakacki
RI
,
Hamilton
M
,
Gilbertson
RJ
,
Blaney
SM
,
Tersak
J
,
Krailo
MD
, et al
Pediatric phase I and pharmacokinetic study of erlotinib followed by the combination of erlotinib and temozolomide: a Children's Oncology Group Phase I Consortium Study
.
J Clin Oncol
2008
;
26
:
4921
7
.
11.
Broniscer
A
,
Baker
SJ
,
Stewart
CF
,
Merchant
TE
,
Laningham
FH
,
Schaiquevich
P
, et al
Phase I and pharmacokinetic studies of erlotinib administered concurrently with radiotherapy for children, adolescents, and young adults with high-grade glioma
.
Clin Cancer Res
2009
;
15
:
701
7
.
12.
Geoerger
B
,
Hargrave
D
,
Thomas
F
,
Ndiaye
A
,
Frappaz
D
,
Andreiuolo
F
, et al
Innovative Therapies for Children with Cancer pediatric phase I study of erlotinib in brainstem glioma and relapsing/refractory brain tumors
.
Neuro-oncology
2010
.
13.
Debiec-Rychter
M
,
Sciot
R
,
Le Cesne
A
,
Schlemmer
M
,
Hohenberger
P
,
van Oosterom
AT
, et al
KIT mutations and dose selection for imatinib in patients with advanced gastrointestinal stromal tumours
.
Eur J Cancer
2006
;
42
:
1093
103
.
14.
Delbaldo
C
,
Chatelut
E
,
Re
M
,
Deroussent
A
,
Seronie-Vivien
S
,
Jambu
A
, et al
Pharmacokinetic-pharmacodynamic relationships of imatinib and its main metabolite in patients with advanced gastrointestinal stromal tumors
.
Clin Cancer Res
2006
;
12
:
6073
8
.
15.
Larson
RA
,
Druker
BJ
,
Guilhot
F
,
O'Brien
SG
,
Riviere
GJ
,
Krahnke
T
, et al
Imatinib pharmacokinetics and its correlation with response and safety in chronic-phase chronic myeloid leukemia: a subanalysis of the IRIS study
.
Blood
2008
;
111
:
4022
8
.
16.
Picard
S
,
Titier
K
,
Etienne
G
,
Teilhet
E
,
Ducint
D
,
Bernard
MA
, et al
Trough imatinib plasma levels are associated with both cytogenetic and molecular responses to standard-dose imatinib in chronic myeloid leukemia
.
Blood
2007
;
109
:
3496
9
.
17.
Demetri
GD
,
Wang
Y
,
Wehrle
E
,
Racine
A
,
Nikolova
Z
,
Blanke
CD
, et al
Imatinib plasma levels are correlated with clinical benefit in patients with unresectable/metastatic gastrointestinal stromal tumors
.
J Clin Oncol
2009
;
27
:
3141
7
.
18.
Widmer
N
,
Decosterd
LA
,
Leyvraz
S
,
Duchosal
MA
,
Rosselet
A
,
Debiec-Rychter
M
, et al
Relationship of imatinib-free plasma levels and target genotype with efficacy and tolerability
.
Br J Cancer
2008
;
98
:
1633
40
.
19.
Houk
BE
,
Bello
CL
,
Poland
B
,
Rosen
LS
,
Demetri
GD
,
Motzer
RJ
. 
Relationship between exposure to sunitinib and efficacy and tolerability endpoints in patients with cancer: results of a pharmacokinetic/pharmacodynamic meta-analysis
.
Cancer Chemother Pharmacol
2010
;
66
:
357
71
.
20.
Thomas
F
,
Rochaix
P
,
Benlyazid
A
,
Sarini
J
,
Rives
M
,
Lefebvre
JL
, et al
Pilot study of neoadjuvant treatment with erlotinib in nonmetastatic head and neck squamous cell carcinoma
.
Clin Cancer Res
2007
;
13
:
7086
92
.
21.
Thomas
F
,
Rochaix
P
,
White-Koning
M
,
Hennebelle
I
,
Sarini
J
,
Benlyazid
A
, et al
Population pharmacokinetics of erlotinib and its pharmacokinetic/pharmacodynamic relationships in head and neck squamous cell carcinoma
.
Eur J Cancer
2009
;
45
:
2316
23
.
22.
Smith
M
,
Bernstein
M
,
Bleyer
WA
,
Borsi
JD
,
Ho
P
,
Lewis
IJ
, et al
Conduct of phase I trials in children with cancer
.
J Clin Oncol
1998
;
16
:
966
78
.
23.
Ivanova
A
. 
Dose finding in oncology–non-parametric methods
.
New York, NY
: 
2006
.
24.
O'Quigley
J
,
Shen
LZ
. 
Continual reassessment method: a likelihood approach
.
Biometrics
1996
;
52
:
673
84
.
25.
Zhao
M
,
He
P
,
Rudek
MA
,
Hidalgo
M
,
Baker
SD
. 
Specific method for determination of OSI-774 and its metabolite OSI-420 in human plasma by using liquid chromatography-tandem mass spectrometry
.
J Chromatogr
2003
;
793
:
413
20
.
26.
Cascorbi
I
,
Gerloff
T
,
Johne
A
,
Meisel
C
,
Hoffmeyer
S
,
Schwab
M
, et al
Frequency of single nucleotide polymorphisms in the P-glycoprotein drug transporter MDR1 gene in white subjects
.
Clin Pharmacol Ther
2001
;
69
:
169
74
.
27.
Asano
T
,
Takahashi
KA
,
Fujioka
M
,
Inoue
S
,
Okamoto
M
,
Sugioka
N
, et al
ABCB1 C3435T and G2677T/A polymorphism decreased the risk for steroid-induced osteonecrosis of the femoral head after kidney transplantation
.
Pharmacogenetics
2003
;
13
:
675
82
.
28.
Yates
CR
,
Zhang
W
,
Song
P
,
Li
S
,
Gaber
AO
,
Kotb
M
, et al
The effect of CYP3A5 and MDR1 polymorphic expression on cyclosporine oral disposition in renal transplant patients
.
J Clin Pharmacol
2003
;
43
:
555
64
.
29.
Petain
A
,
Kattygnarath
D
,
Azard
J
,
Chatelut
E
,
Delbaldo
C
,
Geoerger
B
, et al
Population pharmacokinetics and pharmacogenetics of imatinib in children and adults
.
Clin Cancer Res
2008
;
14
:
7102
9
.
30.
Tod
M
,
Jullien
V
,
Pons
G
. 
Facilitation of drug evaluation in children by population methods and modelling
.
Clin Pharmacokinetics
2008
;
47
:
231
43
.
31.
Christiansen
SR
,
Broniscer
A
,
Panetta
JC
,
Stewart
CF
. 
Pharmacokinetics of erlotinib for the treatment of high-grade glioma in a pediatric patient with cystic fibrosis: case report and review of the literature
.
Pharmacotherapy
2009
;
29
:
858
66
.
32.
Miller
AA
,
Murry
DJ
,
Owzar
K
,
Hollis
DR
,
Lewis
LD
,
Kindler
HL
, et al
Phase I and pharmacokinetic study of erlotinib for solid tumors in patients with hepatic or renal dysfunction: CALGB 60101
.
J Clin Oncol
2007
;
25
:
3055
60
.
33.
VanSchaik
RHN
. 
CYP3A5 variant allele frequencies in Dutch Caucasians
.
Clin Chem
2002
;
48
:
1668
71
.
34.
Kuehl
P
,
Zhang
J
,
Lin
Y
,
Lamba
J
,
Assem
M
,
Schuetz
J
, et al
Sequence diversity in CYP3A promoters and characterization of the genetic basis of polymorphic CYP3A5 expression
.
Nat Genet
2001
;
27
:
383
91
.
35.
Elmeliegy
MA
,
Carcaboso
AM
,
Tagen
M
,
Bai
F
,
Stewart
CF
. 
Role of ATP-binding cassette and solute carrier transporters in erlotinib CNS penetration and intracellular accumulation
.
Clin Cancer Res
2011
;
17
:
89
99
.
36.
Marchetti
S
,
de Vries
NA
,
Buckle
T
,
Bolijn
MJ
,
van Eijndhoven
MA
,
Beijnen
JH
, et al
Effect of the ATP-binding cassette drug transporters ABCB1, ABCG2, and ABCC2 on erlotinib hydrochloride (Tarceva) disposition in in vitro and in vivo pharmacokinetic studies employing Bcrp1-/-/Mdr1a/1b-/- (triple-knockout) and wild-type mice
.
Mol Cancer Ther
2008
;
7
:
2280
7
.
37.
Wacker
B
,
Nagrani
T
,
Weinberg
J
,
Witt
K
,
Clark
G
,
Cagnoni
PJ
. 
Correlation between development of rash and efficacy in patients treated with the epidermal growth factor receptor tyrosine kinase inhibitor erlotinib in two large phase III studies
.
Clin Cancer Res
2007
;
13
:
3913
21
.
38.
Mellinghoff
IK
,
Wang
MY
,
Vivanco
I
,
Haas-Kogan
DA
,
Zhu
S
,
Dia
EQ
, et al
Molecular determinants of the response of glioblastomas to EGFR kinase inhibitors
.
New Engl J Med
2005
;
353
:
2012
24
.
39.
Lynch
TJ
,
Bell
DW
,
Sordella
R
,
Gurubhagavatula
S
,
Okimoto
RA
,
Brannigan
BW
, 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
.