|Home | About | Journals | Submit | Contact Us | Français|
This study estimated the maximum tolerated dose (MTD) of imatinib with irradiation in children with newly diagnosed brainstem gliomas, and those with recurrent malignant intracranial gliomas, stratified according to use of enzyme-inducing anticonvulsant drugs (EIACDs). In the brainstem glioma stratum, imatinib was initially administered twice daily during irradiation, but because of possible association with intratumoral hemorrhage (ITH) was subsequently started two weeks after irradiation. The protocol was also amended to exclude children with prior hemorrhage. Twenty-four evaluable patients received therapy before the amendment, and three of six with a brainstem tumor experienced dose-limiting toxicity (DLT): one had asymptomatic ITH, one had grade 4 neutropenia and, one had renal insufficiency. None of 18 patients with recurrent glioma experienced DLT. After protocol amendment, 3 of 16 patients with brainstem glioma and 2 of 11 patients with recurrent glioma who were not receiving EIACDs experienced ITH DLTs, with three patients being symptomatic. In addition to the six patients with hemorrhages during the DLT monitoring period, 10 experienced ITH (eight patients were symptomatic) thereafter. The recommended phase II dose for brainstem gliomas was 265 mg/m2. Three of 27 patients with brainstem gliomas with imaging before and after irradiation, prior to receiving imatinib, had new hemorrhage, excluding their receiving imatinib. The MTD for recurrent high-grade gliomas without EIACDs was 465 mg/m2, but the MTD was not established with EIACDs, with no DLTs at 800 mg/m2. In summary, recommended phase II imatinib doses were determined for children with newly diagnosed brainstem glioma and recurrent high-grade glioma who were not receiving EIACDs. Imatinib may increase the risk of ITH, although the incidence of spontaneous hemorrhages in brainstem glioma is sufficiently high that this should be considered in studies of agents in which hemorrhage is a concern.
Children with intrinsic brainstem malignant gliomas have a dismal prognosis, with one-year progression-free survival rates of less than 25% with current therapies (Jennings et al., 1996; Pollack, 1994). These lesions typically arise in the pons and have a characteristic appearance on MRI that generally obviates the need for biopsy to establish the diagnosis in children with an appropriate clinical history (Albright et al., 1993). Other than irradiation, which often provides transient symptomatic improvement, no therapy has favorably affected outcome (Jennings et al., 2002; Packer et al., 1993). Similarly, children with recurrent malignant gliomas have a poor prognosis, regardless of tumor location. Excluding the small subgroup of children with lesions amenable to extensive resection, the five-year survival rate is less than 5% (Finlay et al., 1996; Lefkowitz et al., 1988). These lesions have shown a low frequency of objective responses to conventional chemotherapeutic agents (Newlands et al., 1996).
In view of these discouraging results, there is a strong need to identify new therapeutic approaches that target features of these tumors that account for their dysregulated growth. Studies have demonstrated that malignant gliomas are driven to proliferate by aberrant activation of growth factor receptor–mediated signal transduction pathways (Aaronson, 1991; Pollack et al., 1998). One receptor pathway that appears to be particularly important in this regard involves platelet-derived growth factor (PDGF),3 which encompasses a family of ligands (AA, AB, BB, CC, and DD) that bind to a pair of receptors (α and β) (Lokker et al., 2002). Concurrent expression of one or more of these ligands and their receptors has been observed in a high percentage of malignant gliomas (Lokker et al., 2002; Mauro et al., 1991; Maxwell et al., 1990; Nistér et al., 1991 Pollack and Kawecki, 1994), allowing autocrine and paracrine stimulation (Harsh et al., 1990; Hermanson et al., 1992; Lokker et al., 2002; Vassbotn et al., 1994). Comparative analyses demonstrate that PDGF is among the most potent mitogens for malignant glioma cell lines in vitro (Pollack et al., 1991; Westphal et al., 1988). Conversely, antibody-mediated, antisense-mediated, and dominant negative-mediated inhibition of PDGF receptor (PDGFR) activation, as well as pharmacological blockade of PDGFR, has been observed to inhibit glioma growth both in vitro and in vivo (Kilic et al., 2000; Kovalenko et al., 1994; Nitta and Sato, 1994; Pollack et al., 1991; Uhrbom et al., 2000). In addition, because of the role of PDGF in supporting the angiogenic properties of malignant gliomas (Carmeliet and Jain, 2000; Jensen, 1998; Pietras et al., 2001; Wang et al., 1999), inhibition of PDGFR may provide a means for simultaneously blocking tumor growth and tumor-induced angiogenesis.
In that context, recent studies of imatinib mesylate (also known as imatinib, STI571, CGP57148B, and Gleevec) have demonstrated that this agent disrupts PDGF/PDGFR autocrine and paracrine loops and interferes with the growth of glioma cell lines in vitro and in vivo (Kilic et al., 2000; Pietras et al., 2001; Uhrbom et al., 2000). This agent initially became a focus of clinical interest because of its inhibition of bcr-abl, a fusion protein produced by Philadelphia chromosome–positive leukemias (Druker et al., 1996, 2001a, 2001b), and c-kit, which is often activated by mutation in gastrointestinal stromal tumors (Joensuu et al., 2001). Clinical trials have validated the usefulness of imatinib in treating these tumors, as well as myeloproliferative disorders driven by constitutive activation of PDGFR (Apperly et al., 2002). In a recently reported Children’s Oncology Group study for children with Philadelphia chromosome–positive leukemia, biological efficacy in terms of cytogenetic responses was observed at doses as low as 260 mg/m2, whereas doses as high as 570 mg/m2 were well tolerated, and a maximum tolerated dose (MTD) was not established (Champagne et al., 2004).
The objectives of the phase I study reported here were to define the safety of imatinib administered in conjunction with involved field irradiation in children with newly diagnosed intrinsic brainstem gliomas (stratum 1) and those with recurrent intracranial malignant gliomas (stratum 2), and to characterize the pharmacokinetics of imatinib in the aforementioned groups. Given that radiotherapy is the single therapeutic modality that has proven (albeit modest and generally transient) efficacy in brainstem gliomas, and that progression-free survival of these tumors is measured in months (Albright et al., 1993; Jennings et al., 2002; Packer et al., 1993), there was a strong rationale to administer irradiation immediately after diagnosis in conjunction with imatinib. This also took into account the potential for radiosensitization that has been observed with this agent (Russell et al., 2003). Because of the known cytochrome P-450 metabolism of imatinib and the potential for altered pharmacokinetics in children receiving other P-450–metabolized medications (Champagne et al., 2004), there were separate substrata for recurrent high-grade glioma patients, determined by whether or not enzyme-inducing anticonvulsant drugs (EIACDs) were being administered.
Patients 3 to 21 years of age with newly diagnosed malignant brainstem gliomas and recurrent malignant gliomas were eligible for this study (Pediatric Brain Tumor Consortium study PBTC006). Stratum 1 consisted of patients with newly diagnosed nonmetastatic diffuse intrinsic brainstem gliomas. A histopathological diagnosis was not required in the setting of a characteristic imaging appearance and clinical history (Albright et al., 1993; Jennings et al., 2002). Eligibility criteria were later amended to exclude patients with imaging evidence of intratumoral hemorrhage (ITH) at diagnosis. Stratum 2 consisted of patients with recurrent anaplastic astrocytoma, glioblastoma multiforme, or other high-grade glioma, as well as recurrent diffuse brainstem glioma. A histopathological diagnosis from either the initial presentation or the time of progression was required for all but malignant brainstem gliomas in this stratum. The eligibility criteria were later amended to exclude patients with an ITH not related to a previous surgical procedure.
Other eligibility criteria included an adequate performance status (Karnofsky or Lansky score ≥ 50); absolute neutrophil count > 1000/μl, platelets > 100,000/μl (transfusion independent), and hemoglobin > 8 g/dl (with or without transfusion); adequate renal function as assessed by creatinine ≤ 1.5 times the institutional normal for age or glomerular filtration rate > 70 ml/ min/1.73 m2; and adequate hepatic function as assessed by bilirubin ≤ 1.5 times the institutional normal for age, serum glutamate-pyruvate transaminase (SGPT) (alanine aminotransferase) < 3 times the institutional normal for age, and albumin ≥ 2 g/dl.
For patients in stratum 1, no prior therapy was allowed except routine corticosteroids. Concomitant anticonvulsant use was not permitted on this stratum. Furthermore, after the amendment, patients with evidence of hemorrhage after irradiation were off study and did not receive imatinib. For patients in stratum 2, at least three months must have elapsed prior to study entry for those who had previously received craniospinal irradiation; longer than eight weeks for local radiation to the primary tumor; and longer than two weeks for focal irradiation to symptomatic metastatic sites. Patients who had previously received chemotherapy must have fully recovered from the acute toxic effects of therapy and, for those who had undergone stem cell transplantation, at least three months must have elapsed following treatment. In addition, neurological deficits, if present, had to be stable for a minimum of one week prior to study entry. Patients were also ineligible if they were pregnant or breastfeeding, had an uncontrolled infection, or were receiving another investigational or anticancer agent. Patients with deep venous or arterial thrombosis within six weeks of registration were also ineligible, as were those receiving warfarin.
The institutional review boards of each PBTC institution approved the protocol before initial patient enrollment, and continuing approval was maintained throughout the study. Patients or their legal guardians gave written informed consent, and assent was obtained as appropriate at the time of enrollment.
Imatinib was provided in 50- and 100-mg capsules by the Pharmaceutical Management Branch of the National Cancer Institute (Bethesda, Md.). Because of the potential for gastric irritation, it was recommended that the drug be taken with a meal and specifically with a large glass of water (240 ml) to minimize the chance of adherence to the esophageal mucosa. For children who could not swallow capsules, up to four imatinib 50-mg capsule contents were mixed with approximately 100 ml of water or apple juice (orange juice and cola drinks were not used because of stability/incompatibility issues). If a patient vomited after taking the drug, the dose was replaced if the pills could be seen or, for younger children who took the drug dissolved in liquids, the dose was replaced if the vomiting occurred directly after swallowing, and the emesis contained evidence of the yellow drug.
Imatinib was administered twice a day with no interruptions in the absence of dose-limiting toxicity (DLT). Each 28-day period was defined as a course. Imatinib therapy was continued for up to 13 courses (52 weeks) in the absence of progression or serious toxicity. Although previous studies of imatinib in adults have generally used a fixed oral daily dose, without adjustment for body surface area, the significant size differences between adults and children of various ages provided an impetus for incorporating dosing parameters based on body surface area for MTD determinations in this cohort.
In stratum 1, children with newly diagnosed brainstem gliomas initially received imatinib beginning with the start of irradiation, at a dose of 200 mg/m2, with possible escalation between patients to doses of 265, 350, 465, 620, or 800 mg/m2 and possible de-escalation to doses of 150 and 100 mg/m2. Irradiation was administered with conventional fractionation, at 180-cGy/day fractions, five days per week, to a total dose of 5580 cGy administered to the tumor plus a 1-cm margin in the axial plane and a 2-cm margin in the sagittal plane, the latter to incorporate white matter pathways extending rostrally and caudally from the tumor. For the purpose of establishing the MTD, the DLT evaluation period was the first two 28-day courses (eight weeks) of imatinib treatment. Interpatient imatinib escalation and de-escalation were controlled by a modified continual reassessment method (CRM) (Goodman et al., 1995), as described in the Trial Design Section below. Because of concerns regarding the incidence of ITH during irradiation, the protocol was subsequently amended to enroll patients without evidence of hemorrhage prior to irradiation, but to begin imatinib treatment two weeks (± 1 week) after completion of radiotherapy, starting at a dose of 350 mg/m2/day, provided that there was no evidence of ITH on a postir-radiation MRI scan. The protocol was then amended to perform dose assignment upon completion of irradiation without evidence of hemorrhage.
In stratum 2, children with recurrent intracranial malignant gliomas received imatinib at a starting dose of 350 mg/m2 with possible escalation to doses of 465, 620, and 800 mg/m2 or de-escalation as determined by the CRM, using the same range of dose levels included in stratum 1. Initially, the DLT monitoring period was one 28-day cycle (four weeks), but because of concerns regarding hemorrhagic toxicity, the protocol was amended to increase the DLT monitoring interval to eight weeks. We planned to stop dose escalations at 800 mg/m2, even if no formal MTD was identified, because of a lack of experience with comparable dose levels in adults. Patients were stratified based on concurrent use of EIACDs into substrata 2A (not receiving) and 2B (receiving) EIACDs.
Because of concerns that drugs that significantly alter gastric pH (e.g., proton-pump inhibitors and H2 blockers) might interfere with drug absorption, every effort was made to administer these drugs at least 4 h before imatinib. In those patients who required corticosteroids to treat cerebral edema and mass effect, efforts were made to administer the lowest dose needed to achieve symptomatic control. With the exception of anticonvulsants in stratum 2B, medications known to influence cytochrome P-450 metabolism and in particular cytochrome P-3A4 and cytochrome P-2D6 activity were avoided, if possible.
A modified CRM as described by Goodman et al. (1995) was used to assign dose levels in all three strata and to estimate the MTDs, defined as the dosages at which 20% of patients are expected to experience a DLT. Rounding the CRM-estimated MTDs to the nearest protocol-prescribed dose level provides the dose-finding MTDs. A minimum of 18 evaluable patients would be studied in each stratum, and the trial would continue until at least six evaluable patients had been treated and observed for toxicity at the dose-finding MTD, which is the prespecified dose level closest to the CRM-estimated MTD. For each stratum, the prior probabilities of DLTs for these eight prespecified dose levels were assumed to be 0.005, 0.010, 0.05, 0.10, 0.15, 0.20, 0.25, and 0.35, respectively. Once the MTD was established, additional accrual to that dose level was permitted until a total of 12 patients (six younger than 12 years of age and six 12 years of age or older) had been enrolled to gain additional experience with the toxicity of imatinib.
At the dose level currently being studied in each stratum, two patients were assigned and a third could be assigned without additional toxicity information. The CRM model was continually updated, and dose-escalation decisions were made as toxicity information became known for each patient. Subsequent dose levels were determined to be the prespecified level closest to the CRM-estimated MTD without skipping a level that had been assigned to fewer than two evaluable patients. Following the amendment to begin imatinib administration after completion of irradiation, four enrollment slots were opened for stratum 1 on the first day of each month, anticipating that some patients would not receive imatinib after irradiation because of evidence of hemorrhage. This created one anomaly in dose assignment (see Table 1), because a dose of imatinib had to be assigned once eligibility was reconfirmed after completion of irradiation.
Patients who came off therapy for reasons other than toxicity before the DLT observation period was completed were replaced for purposes of estimating the MTD. Similarly, patients who missed more than seven days of imatinib (i.e., >14 doses) for reasons other than toxicity in any course during the DLT observation period were replaced.
Kaplan-Meier estimates of distributions of event-free survival (EFS) and survival are provided with Peto estimates of the standard errors. EFS is measured from the date of initiation of protocol treatment (radiation therapy or imatinib treatment) to the date of disease progression, diagnosis of second malignancy, or death from any cause. The cumulative incidence functions of ITH, as measured from date on therapy to the earliest date of a competing event (disease progression or death), diagnosis of ITH, or last contact, were estimated by methods discussed in Kalbfleish and Prentice (1980). A generalized linear model was used to compare pharmacokinetic (PK) parameters between the strata of patients receiving and not receiving EIACDs.
Toxicities were graded according to the NCI Common Toxicity Criteria (version 2.0 [CTC 2.0]) scale. DLTs were defined as any of the following events occurring during the DLT observation period with likely attribution to imatinib: grade 3 or 4 thrombocytopenia, grade 4 neutropenia, grade 3 or 4 nonhematologic toxicity with the exception of grade 3 nausea and vomiting, and grade 3 transaminase (serum glutamic-oxaloacetic transaminase [SGOT] or SGPT) elevation that returned to grade 1 or less within seven days of stopping the drug, and any grade 2 nonhematologic toxicity that was considered sufficiently intolerable or significant medically to warrant treatment interruption and/or dose reduction. In addition, the detection of hemorrhage on the MRI scan obtained after course 2 of therapy, which was performed within one week of completing this course, was considered a DLT, even though the scan was done several days after the DLT monitoring interval.
Imatinib was withheld for a minimum of seven days if a patient experienced a dose-limiting hematologic toxicity, grade 2 nonhematologic toxicity that lasted for more than three days, and any grade 3 nonhematologic toxicity (with the exception of grade 2 or 3 nausea or vomiting that was controlled with antiemetics or grade 2 or 3 elevations of SGOT or SGPT). Patients were off therapy if the toxicity did not resolve or decrease to grade 1 within 14 days. If the toxicity did decrease to grade 1 or less, the study drug could be resumed at one dose level lower. Doses reduced for imatinib-related toxicity were not re-escalated, even if there was minimal or no toxicity at the reduced dose. Patients who again experienced DLT after a single dose reduction were considered off therapy. For grade 3 elevations of serum transaminases (SGOT/SGPT), the study drug was withheld, and resumed at the same dose if the toxicity resolved or decreased to grade 1 within seven days of stopping the study drug. If the toxicity decreased to grade 1 or less between 7 to 14 days, the study drug could be resumed at one dose level lower. Patients were off therapy if the toxicity did not resolve to grade 1 or less within 14 days or recurred at the next lower dose level. For grade 4 nonhematologic toxicity, the study drug was withheld immediately, and the patient was off therapy. Patients who underwent dose level reductions but remained on therapy were counted as DLTs for estimating the MTD. Patients who again experienced DLT at the next lower dose level were taken off treatment, and this second DLT was not used in MTD estimations.
Standard imaging response criteria were used, with the caveat that both complete and partial responses required a stable or decreasing dose of corticosteroids, accompanied by a stable or improving results of neurological examination, maintained for at least six weeks. In addition, because imatinib was considered to be a cytostatic agent, for which a lag time may have been present between initiation of therapy and maximum antitumor effect, patients were allowed to continue on therapy until the maximum cross-sectional area had increased by 50% from baseline, provided the patient had no clinical symptoms of tumor progression and the treating physician and patient/family elected to continue therapy. However, to enable comparisons with previous studies, progression in such patients was defined based on the time at which a 25% increase in cross-sectional area was observed. Patients were off therapy for disease progression if the tumor increased at least 50% in area from baseline or patients exhibited clinical symptoms from tumor enlargement, even if the size increase was less than 50%.
Blood samples for PK assessments were collected during the first course of treatment, immediately before study drug administration on days 1 and 8, and at 0.5, 1, 1.5, 2, 4, 10, and 12 h after the morning dose. The 12-h sample was obtained prior to administration of the subsequent dose of imatinib. Blood samples were also obtained immediately before drug administration on days 14 ± 2, 21 ± 2, and 28 ± 2. On day 8 and later, patients did not take their imatinib dose that day until they came to the clinic for their office visit to facilitate predose and postdose blood samples. Samples were collected in heparinized Vacutainer tubes and centrifuged immediately at 1200 g for 5 min. Plasma was separated and stored at −20°C until analysis.
Plasma concentrations of imatinib and its principal and active metabolite CGP74588 (le Coutre et al., 2004) were quantified by a validated assay using liquid chromatography–mass spectrometry (Bakhtiar et al., 2002). Concentration-time curves were evaluated by noncom-partmental analysis (WinNonlin Pro 3.1; Pharsight, Mountain View, Calif.). Whenever feasible, the following PK parameters were calculated from the plasma concentration-time profiles of imatinib (Peng et al., 2004): (1) tmax, sampling time when maximum measured plasma concentration occurred; (2) Cmax, maximum measured plasma concentration; (3) t1/2, apparent terminal disposition half-life; t1/2 = ln(2)/λz; and (4) AUC(0–τ), area under the concentration-time curve during a dose interval, calculated using linear-log trapezoidal rule.
Between May 30, 2001, and May 13, 2004, a total of 84 eligible patients were enrolled at PBTC institutions: 35 on stratum 1, 33 on stratum 2A, and 16 on stratum 2B (Table 2). Median age at study entry was 7.2 years for stratum 1 and 8.9 and 15.4 years on strata 2A and 2B, respectively. The ratios of male to female patients on the three strata were 18:17, 13:20, and 7:9, respectively. The diagnosis was based on imaging criteria alone in 30 patients in stratum 1; in the remaining five patients, the diagnosis was based on clinical, imaging, and histological determination of malignant brainstem glioma. In stratum 2A, 14 patients had recurrent brainstem glioma, based on imaging characteristics, and 19 had nonbrain-stem recurrent astrocytomas, including seven classified as anaplastic astrocytoma, seven as glioblastoma multiforme or gliosarcoma, and five as other malignant gliomas. In stratum 2B, all patients had recurrent nonbrain-stem malignant gliomas, including 11 with anaplastic astrocytoma, two with glioblastoma multiforme, and three with other malignant gliomas.
Table 1 summarizes the CRM dose escalation and de-escalation decisions, patient evaluability, and DLT observations. Among the 29 eligible patients who were enrolled on stratum 1 following the amendment to begin imatinib therapy after irradiation, nine were inevaluable because they did not receive the study drug: Three were observed to have had asymptomatic ITH after radiotherapy that was not present before irradiation, two had progressive disease during irradiation, one had an infection that precluded beginning imatinib within the scheduled time interval, and three withdrew from the protocol. Of the 20 patients who received imatinib, 3 were not evaluable due to incomplete courses of treatment. All six patients enrolled on stratum 1 prior to the amendment received imatinib and were evaluable. Six of the 33 eligible patients enrolled in stratum 2A were not evaluable for dose finding because of incomplete course of treatment (five) and withdrawal prior to receiving imatinib (one); in addition, seven patients were enrolled at the MTD after dose finding (not included in Table 1) to better define toxicities, and one of these patients was not evaluable because of noncompliance. Four of the eligible patients enrolled in stratum 2B were not evaluable for dose finding because of incomplete course of treatment (three) and withdrawal prior to receiving imatinib (one).
Prior to amending the protocol to exclude patients with previous evidence of spontaneous ITH, six evaluable patients were enrolled on stratum 1, nine on stratum 2A, and six on stratum 2B. None of three patients on stratum 1 treated with a dose of 200 mg/m2 had a symptomatic hemorrhage, but one experienced a DLT associated with renal insufficiency; one of three patients treated at a dose of 150 mg/m2 had an asymptomatic hemorrhage detected on the scheduled MRI eight weeks after beginning therapy, and a second patient experienced a DLT with an absolute neutrophil count of less than 500/μl. On stratum 2A, nine patients, three each at doses of 350, 465, and 620 mg/m2, were treated, with no patient experiencing a DLT. However after the dose-finding observation period, one child treated at the 465-mg/m2 dose level had a symptomatic hemorrhage, as did two children treated with 620 mg/m2. Two of these children had evidence of previous spontaneous hemorrhages. Similarly on stratum 2B, no DLTs were observed among six evaluable patients; three children each were treated at 350- and 465-mg/m2 dose levels. One child at the 350-mg/m2 dose level had a documented hemorrhage after the dose-finding period and was taken off of therapy after having had multiple previous hemorrhages.
These observations provided an impetus to amend the protocol in October 2001 to exclude patients with evidence of previous hemorrhage, to deliver irradiation and imatinib sequentially in stratum 1, and to increase the DLT monitoring interval from four to eight weeks for patients in strata 2A and 2B. Subsequently, 17, 11, and 6 evaluable patients were enrolled in strata 1, 2A, and 2B, respectively. A summary of the dose-limiting toxicities observed is provided in Table 1. Of note, two of the five hemorrhages classified as DLTs were asymptomatic. At the time that dose-finding was stopped for stratum 1, the CRM estimated the MTD (i.e., the dose at which 20% of patients would be expected to experience a DLT during the first eight weeks of treatment) to be 334 mg/m2, implying a dose-finding MTD estimate of 350 mg/m2. However, two of six patients experienced hemorrhages at 350 mg/m2 compared with only one DLT among 11 patients treated with 265 mg/m2. (The latter patient was noted to have had an asymptomatic bleed just outside of the DLT monitoring period. Based on the protocol definitions at that time, this would have met the criteria for a DLT, but this was recognized retrospectively and therefore did not influence dose-escalation determinations, as noted in Table 1.) Because of concerns that DLTs are usually reversible but neurological deficits from ITHs often are not, PBTC investigators concluded that the maximum safe dose to recommend for study in a phase II trial for patients with newly diagnosed brainstem gliomas was 265 mg/m2. The CRM-estimated MTD for recurrent high-grade glioma patients not receiving EIACDs (stratum 2A) was 541 mg/m2, implying a dose-finding MTD of 465 mg/m2. However, an MTD was not established in the group of recurrent high-grade glioma patients receiving EIACDs (stratum 2B), with no DLTs in four patients enrolled at the 620- and 800-mg/m2 dose levels. Given the increasing trend during the time period of the study toward using non-enzyme-inducing agents in children with brain tumors who require anticonvulsant therapy and the resultant slow accrual, further escalation was not pursued in this stratum.
Non-dose-limiting adverse events during the DLT monitoring period included nausea, vomiting, headache, and fatigue, generally grade 2 or lower, in approximately half of the patients, which were symptoms that would not be unexpected as a result of the underlying disease process. Grade 3 neutropenia and lymphopenia were observed in five patients each. One additional patient who was entered on stratum 2A after the MTD had been assigned had grade 4 neutropenia during the aforementioned interval. No instance of grade 3 or greater liver function abnormalities was observed. Similarly, none of the patients had grade 3 or greater dermatological toxicity, and only one patient, who was entered on stratum 2A after the MTD had been assigned, had a grade 2 rash.
After the DLT monitoring period, the predominant adverse event that led to stopping the study drug was hemorrhage, which was observed in 10 children. Two of these hemorrhages were asymptomatic and detected only on imaging (grade 3 on CTC 2.0), whereas eight were clinically apparent as a result of new neurological symptoms (grade 4) and confirmed on imaging. A summary of these events is provided in Table 3. The six-month estimates of cumulative incidences of hemorrhage for patients treated prior to the protocol amendment, all of whom received study drug, were 33.3% ± 21.3% for the six patients in stratum 1 and 16.7% ± 9.1% for the 18 in stratum 2. Among patients enrolled after the amendment, the six- and nine-month estimates of cumulative incidences of hemorrhage were 18.2% ± 7.5% and 21.9% ± 8.2%, respectively, for the 29 patients in stratum 1 and 13.8% ± 6.6% at both time points for the 31 in stratum 2. Among the patients treated after the amendment who received study drug, six- and nine-month estimates were 10% ± 6.9% and 15% ± 8.4%, respectively, for the 20 patients in stratum 1, and 14.3% ± 6.8% at both time points for the 29 in stratum 2. Subset analyses from strata 2A and 2B showed that both groups had comparable estimates of hemorrhage frequency at both six and nine months.
Imatinib PK data for day 1 of course 1 were available for 37 patients. The median value of the maximum plasma concentration of imatinib (Cmax) ranged from 1.0 μg/ml (2.0 μM) at a dose of 150 mg/m2 on stratum 1 to 6.2 μg/ml (12.6 μM) at a dose of 800 mg/m2 on stratum 2A. The PK parameters for imatinib and its principal metabolite, including the minimum, maximum, and median values, are summarized in Tables 4 and and5.5. As indicated, there was substantial variation in the PK parameters among patients. Plasma concentrations as a function of time for imatinib are shown in Fig. 1.
Pharmacokinetic data were also available for day 8 of course 1 in 39 patients. As anticipated, the Cmax and AUC for both imatinib and its metabolite were correspondingly higher on these steady-state measurements (Tables 4 and and5).5). The mean (±SD) accumulation ratio across different doses (day 8 AUC over day 1 AUC) was 1.57 (±0.43) for strata 1 and 2A combined, and slightly lower 1.25 (±0.44) for stratum 2B. Concentrations of imatinib as a function of time are illustrated in Fig. 2. Steady-state levels obtained before drug administration on days 14, 21, and 28 were available in 32, 33, and 31 patients, respectively, and were comparable to those obtained prior to drug administration on day 8. A mixed model (Diggle et al., 1994) adjusting for strata and dose found no significant (P = 0.70) change over time, indicating that steady-state levels were reached by day 8 of drug administration.
To determine whether there was any influence of EIACDs on the metabolism of imatinib, a generalized linear model was used to compare PK parameters between strata 2A and 2B for imatinib and CGP74588 separately for both days 1 and 8, adjusting for dose level as well as potential interactions between stratum and dose. Each of the PK parameters listed in Tables 4 and and55 were analyzed in these models. The interaction term was highly nonsignificant (P > 0.50) in all models except the analysis with respect to the half-life of imatinib on day 8. Since the interaction between dose and stratum was significant (P < 0.001), it was necessary to compare the two strata within each dose level separately. These analyses were performed with exact Wilcoxon tests, and no comparisons were significant at any of the dose levels (all P > 0.10).
Significant differences between the two strata were found among parameters related to plasma concentrations of imatinib itself, and all data pointed to greater systemic exposure in the stratum not receiving EIACDs. Maximum concentrations of imatinib differed significantly between stratum 2A and 2B patients on both days 1 and 8 (P = 0.039 and P = 0.001, respectively). Also, the AUCs differed between the two strata on days 1 and 8 (P = 0.059 and P < 0.001, respectively) and were approximately 60% lower in stratum 2B compared with 2A. In contrast, maximum concentrations and AUCs for the metabolite CGP74588 did not differ significantly between stratum 2A and 2B patients on either day 1 or day 8 (P > 0.34). Accordingly, the dose-normalized CGP74588 to imatinib exposure ratio was approximately doubled in stratum 2B compared with 2A, suggesting an induced imatinib metabolism by EIACDs (Fig. 3).
Among the eight patients who experienced DLTs, five had at least one PK parameter estimated. Because the number of patients was small, only anecdotal inferences could be made regarding the association between DLTs and PK parameters. For example, the patient treated on stratum 1 at a dose level of 350 mg/m2 who experienced a CNS hemorrhage and had PK data had the highest AUC for both STI571 and its metabolite on day 1 of PK sampling, although this patient represented one of only three with PK data for that stratum and dose. One of the two patients treated on stratum 2A at a dose level of 620 mg/m2 who had a hemorrhage had the highest AUC for STI571 on day 8 among the eight patients with PK data for that stratum, dose, and day, although this patient’s parameters for the metabolite were close to the median values for days 1 and 8. When examined independently of dose level, no obvious association was apparent between either Cmax or AUC and the presence or absence of either a DLT in general or hemorrhage in particular, although the small numbers of patients precluded any statistically based conclusions.
Because this was a phase I study, assessing response and outcome was not a major objective of the analysis, even though the data about them were collected. For stratum 1, among patients treated at or above the recommended phase II dose (n = 20), six-month and one-year EFS rates were 80.0% ± 8.7% and 20.0% ± 8.0%, and survival rates were 100% and 45.0% ± 11.1%, respectively. From an intent-to-treat perspective, taking into consideration all 35 eligible patients who were entered on stratum 1, six-month and one-year EFS rates were 69.9% ± 7.8% and 24.3% ± 7.1%, and survival rates were 94.0% ± 4.1% and 45.5% ± 8.7%, respectively (Fig. 4). Among the subset of 29 who were enrolled after the amendment to receive imatinib after irradiation, including those who did not receive the study drug because of hemorrhage during irradiation, disease progression, or other factors, six-month and one-year EFS rates were 74.4% ± 8.2% and 22.3% ± 7.4%, and survival rates were 92.7% ± 4.9% and 40.8% ± 9.5%, respectively (Fig. 5). For stratum 2A patients treated at or above the MTD (n = 29), the six-month EFS rate was 17.9% ± 6.6%, and the disease of all patients had progressed by one year. For stratum 2B, an MTD was not determined, although the corresponding EFS rates for patients who were treated at dose levels at or above the MTD of stratum 2A (n = 11) were 18.2% ± 9.5% at six months, and again, the disease of all patients had progressed by one year.
Two patients from stratum 1 had objective responses, corresponding to a greater than 50% decrease in tumor cross-sectional area compared with the results of the pretreatment scan, although it is impossible to discern whether this simply related to the effect of radiotherapy, the study drug, or a combination thereof. One patient on stratum 2A, who was treated with a dose of 350 mg/m2, also had a greater than 50% reduction in tumor area noted on the scan obtained two months after beginning therapy, although this child subsequently had disease progression at 2.5 months. This response was therefore not sustained for the six-week duration specified by the protocol for declaring a partial response.
Imatinib is one of the first among a host of molecularly targeted agents that have been evaluated in the treatment of primary CNS tumors of childhood. Based on the aforementioned role of PDGF/PDGFR interactions in glial tumorigenesis, there was strong rationale to examine the safety and tolerability of a PDGFR inhibitor such as imatinib in a pediatric cohort.
Lin and Shih (2001) proved that traditional “3 + 3” phase I cohort designs do not provide targeted toxicity doses as MTD estimates. On the other hand, modified CRM designs were used in this trial to estimate MTDs in each stratum as the doses at which 20% of patients would be expected to experience DLTs. In addition to providing well-defined and interpretable MTD estimates, the modified CRM used variable-sized cohorts of two to three patients and continually made dose decisions after toxicity results for each patient. This afforded the potential for earlier trial completion with fewer patients without exposing more patients to potentially toxic doses than would the traditional cohort design. However, effectively executing these designs and gaining the theoretical advantages require a tightly integrated, attentive group of investigators that includes a central office and on-site clinical research associates, attending physicians, study chairs, computer scientists, and statisticians, who are supported by a modern, secure computing infrastructure.
The current study estimated the MTD of imatinib in children with recurrent malignant glioma who were not receiving EIACDs and recommended a phase II dose following irradiation for patients with newly diagnosed brainstem glioma. This study also highlighted the challenges in interpreting issues of toxicity in a tumor type for which the spontaneous rate of a potential adverse event is uncertain. In particular, the rate of spontaneous hemorrhage in children with brainstem gliomas has not been well described in the literature. In previous reports based on CT and clinical data, hemorrhage was noted infrequently (Jennings et al., 2002; Kaplan et al., 1996; Packer et al., 1993). In a more recent study, based on T1- and T2-weighted MR images, the cumulative hemorrhage rate was estimated at 15.5% ± 5.5% within six months after diagnosis (Broniscer et al., 2006). Moreover, given the evolution in imaging technology, minute foci of hemorrhage can now be detected by using gradient images, which would have eluded detection on either CT or conventional T1- and T2-weighted MR images (Broniscer et al., 2006; Smith et al., 1990).
Accordingly, our study called attention to a significant rate of hemorrhages, in some cases asymptomatic, in children with brainstem and recurrent malignant gliomas treated with imatinib. Our concern that concurrent radiation or a history of previous bleeding episodes might increase the risk of hemorrhage led to amendment of the protocol to begin therapy after irradiation for stratum 1 and to exclude patients with evidence of a prior hemorrhage. After doing so, we noted a significant frequency of patients (3 of 27) who were enrolled on the study, but then taken off protocol following the detection of an asymptomatic hemorrhage on MRI after irradiation, but prior to beginning imatinib therapy, as well as additional patients who were deemed ineligible because features consistent with hemorrhage were detected on preirradiation gradient echo images. These observations indicate that a substantial subset of children with brainstem gliomas exhibit imaging evidence of hemorrhage as part of the natural history of this disease. Many such events would have escaped detection in previous studies that used less sensitive imaging modalities. The NCI CTC have recently been modified to incorporate these technical advancements. The asymptomatic punctate ITHs that were classified as grade 3 toxicities based on CTC 2.0 would be considered grade 1 in CTC 3.0, and these refined guidelines have been incorporated in more recent studies by the group. During this study, institutions within the PBTC increasingly included gradient echo MRI as a routine part of their imaging battery, although these studies were not specifically mandated in the initial years of the study. In addition, it is important to emphasize that ascertainment of hemorrhages in the context of this report was based on the institutional determination, although rapid review of these scans by the PBTC Neuroimaging Center was available on request. Because central review of the imaging data for the presence of hemorrhage was not used in the prospective determination of DLTs or adverse events in the current study, it is anticipated that additional cases with small foci of asymptomatic hemorrhage will be identified on central review. Accordingly, these data will be the subject of a separate report in conjunction with the data derived from MR spectroscopy, perfusion, diffusion, and PET scans.
In addition to the five patients with asymptomatic hemorrhages, however, 11 patients in the study group experienced symptomatic hemorrhages, among 73 patients who received imatinib. Although this incidence was viewed with some concern, the six-month estimated cumulative hemorrhage rate of 18.2% ± 7.5% for patients enrolled on stratum 1 after the amendment was essentially the same as the aforementioned hemorrhage rate in patients treated with irradiation and conventional chemotherapy. A similar hemorrhage rate, 13.8% ± 6.6%, was also observed for patients enrolled on stratum 2. It is conceivable that imatinib may have enhanced an underlying tendency to spontaneous hemorrhage among the tumor types included in this study, as has been noted previously with gastrointestinal stromal tumors (Demetri et al., 2002; El Hajj et al., 2005). The mechanism for such an effect, if any, is conjectural and may reflect the impact of PDGFR inhibition on endothelial cell function (Carmeliet and Jain, 2000; Jensen, 1998; Pietras et al., 2001; Wang et al., 1999) rather than a direct effect on the tumor cells themselves.
Of note, several preliminary studies conducted in Europe for adults with recurrent malignant gliomas, in which imatinib was administered either alone (Katz et al., 2004; Raymond et al., 2004) or with hydroxyurea (Dresemann et al., 2004), noted a low incidence of symptomatic hemorrhages. Similar to the current study, these reports noted a small, but intriguing, incidence of objective responses, which were observed in 4 of 52 patients and 4 of 15 patients, respectively, in the two studies that administered imatinib alone (Katz et al., 2004; Raymond et al., 2004), and 5 of 26 patients who received imatinib with hydroxyurea (Dresemann et al., 2004). However, given the phase I nature of the current study, no conclusions can be made regarding efficacy. In recent cooperative group studies, such as Children’s Cancer Group study CCG-9882 (Kaplan et al., 1996), which examined the use of hyperfractionated irradiation, and CCG-9941 (Jennings et al., 2002), which gave three courses of dose-intensive chemotherapy prior to hyperfractionated irradiation, the one-year EFS rates were 18.8% ± 3.5% and 17% ± 5%, respectively. One-year survival for CCG-9882 was 37% ± 5%. These results are nominally similar to one-year EFS and survival rates of 24.3% ± 7.1% and 45.5% ± 8.7% observed in the current study for all eligible patients, which in conjunction with the concerns raised about bleeding risk, dampened enthusiasm for proceeding directly to a phase II study.
Previous studies have suggested that agents which influence cytochrome P-450–dependent metabolism have an impact on imatinib pharmacokinetics (Champagne et al., 2004), an observation that was also made in a North American Brain Tumor Consortium study in adults (Wen et al., 2004). Our findings support this assertion, in that an MTD was not reached in children who received enzyme-inducing anticonvulsant agents, and significantly lower Cmax and AUC values were observed at corresponding dose levels in children that received EIACDs compared with those who did not. Doses of up to 800 mg/m2 were well tolerated in the EIACD-receiving subgroup, but given the increasing trend toward using non-enzyme-inducing agents in children with brain tumors who require anticonvulsant therapy, further escalation was not pursued. In the subsets of patients with newly diagnosed brainstem and recurrent malignant gliomas who were not receiving enzyme-inducing anticonvulsants, PK parameters were comparable to results obtained in adults (Peng et al., 2004) and children with leukemia (Champagne et al., 2004). There was no evidence that Cmax or AUC levels were associated with an increased risk of hemorrhagic toxicity, although the small numbers of patients precluded any definitive conclusions in this regard.
The imatinib study has since been followed by other PBTC studies of molecularly targeted agents, which have benefited from the observations of this protocol. In particular, because many such agents also have potential effects on endothelial cell function, standardized imaging guidelines have been adopted to ensure consistency between studies, and the more clinically relevant CTC 3.0 toxicity criteria for establishing CNS hemorrhage- related DLT have been incorporated. Finally, as studies of such agents have advanced to phase II, the aforementioned imaging and toxicity caveats have been applied in the generation of sequential monitoring parameters in which hemorrhage-specific rules supplement previously used failure-based guidelines to halt accrual once a threshold of disease progression, hemorrhage, or a combination thereof is reached.
1This work was supported in part by NIH grant U01 CA81457 for the PBTC, M01 RR00188–37, and American Lebanese Syrian Associated Charities. The authors and the PBTC acknowledge statistical support of Dr. Daniel Hunt and the clinical research assistant support of Ms. Lisa Rush.
3Abbreviations used are as follows: AUC, area under the curve; Cmax, maximum measured plasma concentration; CRM, continual reassessment method; CTC, common toxicity criteria; DLT, dose-limiting toxicity; EFS, event-free survival; EIACD, enzyme-inducing anticonvulsant drug; ITH, intratumoral hemorrhage; MTD, maximum tolerated dose; PBTC, Pediatric Brain Tumor Consortium; PDGF, platelet-derived growth factor; PDGFR, PDGF receptor; PK, pharmacokinetic; SGOT, serum glutamic-oxaloacetic transaminase; SGPT, serum glutamic-pyruvic transaminase; t1/2, terminal half-life.