Exploring Predictors of Response to Dacomitinib in EGFR-Amplified Recurrent Glioblastoma

Abstract

PURPOSE

Despite the high frequency of EGFR genetic alterations in glioblastoma (GBM), EGFR-targeted therapies have not had success in this disease. To improve the likelihood of efficacy, we targeted adult patients with recurrent GBM enriched for EGFR gene amplification, which occurs in approximately half of GBM, with dacomitinib, a second-generation, irreversible epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor that penetrates the blood-brain barrier, in a multicenter phase II trial.

PATIENTS AND METHODS

We retrospectively explored whether previously described EGFR extracellular domain (ECD)–sensitizing mutations in the context of EGFR gene amplification could predict response to dacomitinib, and in a predefined subset of patients, we measured post-treatment intratumoral dacomitinib levels to verify tumor penetration.

RESULTS

We found that dacomitinib effectively penetrates contrast-enhancing GBM tumors. Among all 56 treated patients, 8 (14.3%) had a clinical benefit as defined by a duration of treatment of at least 6 months, of whom 5 (8.9%) remained progression free for at least 1 year. Presence of EGFRvIII or EGFR ECD missense mutation was not associated with clinical benefit. We evaluated the pretreatment transcriptome in circulating extracellular vesicles (EVs) by RNA sequencing in a subset of patients and identified a signature that distinguished patients who had durable benefit versus those with rapid progression.

CONCLUSION

While dacomitinib was not effective in most patients with EGFR-amplified GBM, a subset experienced a durable, clinically meaningful benefit. Moreover, EGFRvIII and EGFR ECD mutation status in archival tumors did not predict clinical benefit. RNA signatures in circulating EVs may warrant investigation as biomarkers of dacomitinib efficacy in GBM.

INTRODUCTION

Alterations of the epidermal growth factor receptor (EGFR) are frequent in glioblastoma (GBM; WHO grade 4), with gene amplification and/or mutation occurring in 40%-50% of patients.¹ EGFR coding sequence mutations cluster in the extracellular domain (ECD) in GBM and include in-frame deletions, such as EGFRvIII, the most common EGFR mutation in GBM, and missense mutations.^1-3 EGFRvIII and certain hotspot EGFR ECD missense mutations result in constitutive, ligand-independent kinase activity and sensitize to EGFR-selective small-molecule tyrosine kinase inhibitors in vitro.^2-6 Of note, EGFR mutations in GBM differ from lung cancer in distribution because lung cancer EGFR mutations cluster in the intracellular kinase domain⁷ and, in context, because EGFR mutations occur exclusively in the setting of EGFR gene amplification in GBM.^1-3

CONTEXT

• Key Objective

• We evaluated whether RNA in circulating extracellular vesicles (EVs) of patients with recurrent EGFR-amplified glioblastoma (GBM) correlated with clinical outcome to dacomitinib, an irreversible epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor, because predictive biomarkers to targeted agents are lacking in GBM.

• Knowledge Generated

• In this phase II trial of 56 adult patients with EGFR-amplified recurrent GBM, a subset (14.3%) had clinical benefit. Intratumoral factors, such as EGFRvIII and EGFR extracellular domain missense mutations, were not associated with clinical benefit. In a subset analysis, pretreatment RNA signature within circulating EVs was associated with durable clinical benefit compared with rapid progression.

• Relevance

• Development of a circulating biomarker could improve patient selection for targeted therapies in GBM because sampling of GBM tumors is limited by high-risk neurosurgical procedures. We identified an RNA signature within circulating EVs that was associated with benefit to dacomitinib in a prospective clinical trial, which demonstrated the potential of EVs as circulating response biomarkers.

Therefore, EGFR is considered a compelling therapeutic target in GBM. However, a number of early trials in unselected patients with GBM reported limited efficacy with EGFR-selective small-molecule tyrosine kinase inhibitors.^8,9 Reasons for this lack of efficacy remain poorly understood and include limited brain penetration of drugs,¹⁰ redundancy in intracellular signaling pathways,¹¹ intratumoral heterogeneity,^12,13 and limited potency of first-generation EGFR-targeting agents against the common variant EGFRvIII.^14,15

Dacomitinib (PF-00299804) is a second-generation, irreversible, small-molecule EGFR tyrosine kinase inhibitor that is approved by the US Food and Drug Administration for first-line treatment of patients with EGFR exon 19 deletion or L858R mutant metastatic non–small-cell lung cancer (NSCLC). Dacomitinib has pharmacologic properties that potentially enable efficacy in patients with GBM. Dacomitinib is an orally available, highly selective, and potent inhibitor of the human epidermal growth factor receptor (HER) family of tyrosine kinases (HER1 [EGFR], HER2, and HER4), with half maximal inhibitory concentration (IC₅₀) values against EGFR, HER2, and HER4 of 6.0 nM, 45.7 nM, and 74 nM, respectively. Dacomitinib is not a substrate of human P-glycoprotein or breast cancer resistance protein, readily passes the blood-brain barrier of mice, and is not metabolized by CYP3A4 and is, therefore, not affected by enzyme-inducing anticonvulsants.^16-19 In addition, dacomitinib has potent activity against EGFRvIII.¹⁵

Here, we hypothesized that dacomitinib would have clinical benefit in a GBM cohort enriched for EGFR genetic alterations. We conducted a multi-arm, multi-institutional, phase II study of dacomitinib in adult patients with recurrent GBM with EGFR amplification in their archival tumor specimens with the primary objective of estimating the progression-free survival at 6 months (PFS6). The primary efficacy arm was a phase II trial of 30 patients with EGFR-amplified GBM at first recurrence (arm B). Additional arms evaluated tumor tissue penetration of dacomitinib in a cohort of 10 patients with first-recurrent EGFR-amplified GBM eligible for surgery (arm A) and explored dacomitinib in a cohort of 16 patients with recurrent EGFR-amplified GBM after bevacizumab treatment (arm C). To explore potential molecular predictors of response, we analyzed known EGFR-activating mutations in archival tumor samples and the transcriptome of circulating extracellular vesicles (EVs), which contain tumor-derived transcripts in patients with GBM.²⁰

METHODS

For complete methods, see the Appendix. This multicenter, open-label, 3-arm nonrandomized study enrolled adult patients with recurrent EGFR-amplified GBM. Arm A comprised anti–vascular endothelial growth factor (VEGF)–naïve surgical candidates (n = 10); arm B comprised phase II trial patients with first-recurrent EGFR-amplified GBM who were anti-VEGF naïve (n = 30); and arm C comprised patients with unlimited prior therapies at first recurrence from a bevacizumab-containing regimen (n = 16). The study primary end point was PFS6 in arm B. Responses were assessed by MacDonald criteria.²¹ This trial was conducted in compliance with the Declaration of Helsinki and with the International Conference on Harmonization Good Clinical Practice Guidelines protocol and was approved by the institutional review boards at each participating institution. Dacomitinib measurements were performed by Intertek Pharmaceutical Services (San Diego, CA). EGFR gene amplification, EGFRvIII, and MGMT promoter methylation assays were performed as previously described (Appendix Table A1).^22-24 EGFR ECD mutations were detected by polymerase chain reaction (PCR)–based sequencing of coding exons where recurrent somatic or sensitizing mutations have been reported.^2,6,25 RNA sequencing (RNA-seq) was performed on RNA extracted from serum EVs using Exosome Diagnostics’s total RNA-seq platform²⁶.

RESULTS

Patient and Tumor Characteristics

Between March 2012 and September 2015, 56 patients with GBM who harbored EGFR gene amplification by fluorescence in situ hybridization were treated in the 3 arms of the study (Table 1). EGFRvIII status was determined by reverse transcription PCR in 20 (67%) of the 30 patients in arm B (the phase II trial) and 29 (52%) of the 56 patients in the overall study. The status of EGFR ECD hotspot mutations was determined by Sanger sequencing in all 30 of the patients in arm B (100%) and 48 (85%) of all 56 patients (Table 1; Appendix Table A2). Similar to previous reports,^1-3,22 EGFRvIII was detected in 45% and ECD point mutations were detected in 31% of this EGFR-amplified GBM cohort. One patient had an EGFR kinase domain mutation (18–base pair insertion in exon 19) in the archival tumor without EGFR amplification and was enrolled in arm C. EGFR copy number and mutation status were determined from the initial diagnostic tumor specimen in 48 (86%) of 56 of patients and from recurrent, immediate pretreatment tumor tissue in 8 (14%) of 56 patients. All tumors were IDH1/2 wild type, consistent with previously described EGFR amplification and IDH mutation inverse correlation.^27-29

TABLE 1.

Baseline Patient Demographics and Disease Characteristics

Efficacy

At the data cutoff, all patients were off study treatment, and 54 (96%) of 56 patients had died, including all patients in arms A and B and 14 of 16 patients in arm C. Two patients were lost to follow-up after coming off the study; all other patients were followed up for PFS and overall survival (OS).

In arm B (the phase II trial), five (17%) of 30 patients achieved PFS6, and thus, this arm did not meet the primary efficacy end point of 30% PFS6. The median PFS was 8.9 weeks, and the median OS was 43 weeks (Figs 1A and 1B). The median number of 4-week cycles received was 2.0 (range, 0.5-30 cycles). Of the five patients who achieved PFS6, the number of cycles completed were 10, 12, 14, 18, and 30. In response-evaluable patients, the best confirmed responses included one with a complete response (3%) and five with stable disease (SD; 17%).

FIG 1.

Kaplan-Meier curves for progression-free survival (PFS) and overall survival (OS). (A) PFS and (B) OS for patients in arm B, the phase II trial of bevacizumab-naïve patients at first recurrence. (C) PFS and (D) OS for patients in arm C, the bevacizumab refractory arm.

The primary objective of arm A was to evaluate post-treatment tumor tissue dacomitinib concentrations, although patients continued on dacomitinib after surgery. Two of 10 patients remained on study for > 14 and 17 cycles, while the other patients progressed before cycle 6 (median PFS, 18 weeks; median OS, 39 weeks). The best confirmed responses were SD in three patients. In arm C, the postbevacizumab cohort, no patient achieved PFS6 (median PFS, 7.8 weeks; median OS, 17 weeks; Figs 1C and 1D). Only two response-evaluable patients (17%) had confirmed SD.

EGFR Mutation and Association With Clinical Benefit

We tested the association between the presence of EGFRvIII and/or ECD mutations in the archival tumor specimen and clinical benefit as defined by achievement of PFS6 or treatment duration of at least 6 months with a confirmed partial or complete response. This cohort included eight patients (arm A [n = 2], arm B [n = 6]) who completed 6, 10, 12, 14, 14, 17, 18, and 30 cycles (median, 13 cycles; Fig 2). Seven of these patients achieved PFS6. Of note, one patient with biopsy-confirmed recurrent GBM before enrollment had a complete response (patient 16); however, a distant, nontarget recurrent lesion appeared at the end of cycle 6 and thus, did not meet the PFS6 end point (Fig 3; Appendix Fig A1). The immediate pretreatment tumor had high-level EGFR amplification (EGFR: centromere 7 ratio > 25:1) and EGFR G598V mutation.

FIG 2.

Swimmer’s plot showing number of cycles completed for all patients (N = 56). Blue bars represent patients in arm B (phase II trial of bevacizumab-naïve patients at first recurrence). Red bars represent patients in arm C, the bevacizumab refractory arm. Teal bars represent patients in arm A, the surgical cohort. (▾) Disease progression time point. (+) Patient came off study as a result of toxicity and not disease progression. (*) Patient withdrew consent without disease progression. ID, identifier.

FIG 3.

Magnetic resonance imaging (MRI) scans that depict complete target lesion response and distant failure. Shown are (A) contrast-enhanced MRI scans of patient 16 at baseline and (B) at the end of cycle 6. Yellow arrows indicate the target lesion. The red arrow shows the lesion that recurred during dacomitinib treatment.

Presence of EGFRvIII was not associated with clinical benefit; 3 of 13 patients with EGFRvIII mutation and three of 16 patients without EGFRvIII mutation had clinical responses (P = 1.0 by two-tailed Fisher’s exact test). We also did not observe an association between EGFRvIII and clinical response when limiting the analysis to only arm B or excluding arm C (Appendix Table A3), a cohort of poor-prognosis patients as evidenced by the very short median OS and a relatively high rate of toxicity-induced treatment discontinuation.

We tested the association between clinical benefit and two ECD mutant cohorts on the basis of the type of ECD mutation identified, one including only ECD mutations confirmed to be sensitizing in vitro^2,6,25 and one including all previously reported recurrent, somatic ECD mutations. There was no association between either ECD mutant cohort and clinical benefit when analyzing all patients or patients in arm B only or excluding arm C (Appendix Table A3). In addition, there was no association between clinical response and the presence of any EGFR mutation (EGFRvIII and/or ECD hotspot mutation; P = .2391; Appendix Table A3). Although not statistically significant, we noted that most patients who had clinical benefit had an EGFR mutation (6 of 8 patients in arms A + B who had clinical benefit). The lone patient (patient 29) with an EGFR kinase domain sensitizing mutation treated in arm C progressed before cycle 2.

In the subset of eight patients with recurrent GBM from whom we had immediate pretreatment tumor specimens and EGFR status, three patients had clinical benefit, and all three had either EGFRvIII or ECD mutation. One patient had EGFRvIII and an L62R (patient 3) mutation and the other two had G598V mutations (patients 11 and 16). However, three nonresponders also had EGFR mutations (EGFRvIII mutations [n = 2], and D46Y mutation [n = 1]; P = .46 by two-tailed Fisher’s exact test). Of note, MGMT promoter methylation in the archival tumor specimen was not associated with clinical response, although the prognostic and predictive value of prechemoradiation therapy MGMT promoter methylation in the recurrent disease setting is uncertain.

Association of Serum EV RNA Signature and Treatment Benefit

We analyzed transcripts contained within serum EVs at baseline (pretreatment) from 14 patients by RNA-seq (see Methods). We compared the mRNA abundance profiles of seven patients who achieved PFS6 (responders) with those of seven patients who experienced disease progression before cycle 3 (rapid progressors). In excess of 10,000 unique mRNAs were reliably detected across all samples, of which 32 genes were significantly differentially expressed between the responders and the rapid progressors (P [adjusted] < .05; Appendix Table A4). Using the normalized abundances of these 32 genes, the responders and rapid progressors separated into two distinct groups by unsupervised clustering (Fig 4).

FIG 4.

Serum extracellular vesicle–derived genes differentially expressed between clinical responders and rapid progressors. Shown are 32 genes that were significantly differentially expressed by RNA sequencing in 7 patients who remained stable > 6 months (responders) compared with 7 who experienced disease progression < 3 months (rapid progressors) (P [adjusted] < .05). These differentially expressed genes were subjected to unsupervised clustering. TPM, transcripts per million.

Tumor Tissue Dacomitinib Concentrations

Dacomitinib plasma pharmacokinetic exposures (C_trough) were consistent with those previously reported after 45-mg daily dosing through multiple cycles^16-19 (Table 2). In the surgical cohort (arm A), post-treatment contrast-enhancing tumor samples taken 7-9 days after daily oral administration of 45 mg dacomitinib demonstrated high levels of dacomitinib in tumor tissue, which reached estimated concentrations (314 to 5,724 nM) above the IC₅₀ values reported for EGFR in vitro kinase inhibition (6.0 nM), cellular EGFR inhibition (5.8 nM), growth inhibition of cell lines carrying sensitizing EGFR kinase domain mutations (2-7 nM), and growth inhibition for Ba/F3 cells expressing EGFRvIII (1.2 nM).¹⁵ In an exploratory analysis, there was no correlation between tumor tissue concentration and PFS (Spearman r = –0.3222; P = .36) or OS (Spearman r = 0.06667; P = .87; Table 2; Appendix Fig A2); however, these analyses are limited by small sample size.

TABLE 2.

Concentration of Dacomitinib in Brain Tumor and Plasma Samples

Safety and Tolerability

Across the 3 arms, the majority of treatment-related adverse events (AEs) were manageable with standard supportive care, and the overall toxicity profile was consistent with previous reports in solid cancers.^16-19 Common grade ≥ 2 AEs considered at least possibly related to dacomitinib included rash (41%), diarrhea (32%), fatigue (16%), and mucositis (9%; Appendix Table A5). Grade ≥ 3 related AEs included diarrhea (11%), fatigue (7%), and rash (5%; Appendix Table A6). Six patients (11%) discontinued dacomitinib because of AEs, five of whom were in arm C, and 57% required treatment interruption for AE management (Appendix Table A7).

DISCUSSION

In our surgical cohort, we confirmed that dacomitinib reaches concentrations in contrast-enhancing tumor tissue well above the IC₅₀ values for cells with sensitizing EGFR mutations. Drug penetration into nonenhancing tumor was not assessed; therefore, it remains unclear whether drug delivery into nonenhancing, leading-edge tumor contributes to the outcome of study treatment. Although the phase II cohort did not meet the PFS6 primary end point, a subset of patients experienced durable tumor control for at least 6 months. Overall, these efficacy results are comparable to a recently reported phase II trial of dacomitinib in first recurrent EGFR-amplified GBM.³⁰ In the study by Sepúlveda-Sánchez et al,³⁰ five (10.6%) of 47 patients achieved PFS6 compared with five (17%) of 30 patients in our primary efficacy population (arm B). Across all arms in our study, five (8.9%) of 56 patients remained progression free for 12 months compared with four (8.5%) of 47 patients in the other study.³⁰ In the recurrent GBM clinical trials, median OS has ranged between 5.0 and 9.8 months^31-33; therefore, prolonged PFS in a subset of patients is clinically meaningful.

In our correlative analysis, EGFR amplification did not predict clinical benefit for the majority of patients because all but 1 patient was EGFR amplified. Although EGFRvIII and EGFR ECD mutations have been reported to be sensitizing to small-molecule EGFR inhibitors preclinically, the presence of these mutations in archival tumor tissue was not associated with clinical benefit from dacomitinib in our study. Although correlations between EGFR mutations and clinical benefit were not statistically significant, we noted a trend for EGFR mutations to accumulate among those patients who did derive a clinical benefit. Of note, presence of EGFR mutation did not predict a lack of benefit. These observations may warrant evaluation of EGFR mutations in other EGFR inhibitor trials.

Several biologic features in GBM may have contributed to treatment failure and/or lack of correlation between EGFR genetic alterations and clinical outcomes in our study, including recent revelations with regard to EGFR mutation subclonality; intratumoral genetic heterogeneity, often with co-amplification of multiple receptor tyrosine kinases (RTKs)^{1,11-13,34-39}; and the ability of GBM tumors to dynamically respond to and resist EGFR inhibition by maintaining downstream signaling through other co-activated RTKs or by suppressing dependence on EGFR signaling.^11,40 Intratumoral heterogeneity may have been the reason that the lone complete responder (patient 16) eventually progressed because tumor grew in a distant area while the target lesion resolved.

Since the initiation of the trial, several comprehensive paired tumor sequencing studies have demonstrated clonal evolution of GBM tumors during tumor progression.^38,39,41-43 Recent larger series of paired GBM tumors (initial diagnostic tumor and recurrence after chemoradiation therapy) found that although EGFR amplification status is generally stable (approximately 80% retained), EGFR ECD point mutations were nearly all subclonal, and EGFRvIII and EGFR ECD point mutations were frequently lost or gained in recurrent tumor specimens.^38,39 These studies suggest that sampling of the immediate, pretreatment tumor sample may be necessary for genetically selected therapies. In our study, the majority of patients (86%) were enrolled on the basis of the EGFR status of their initial diagnosis tumor specimen, and EGFR amplification and/or ECD mutation may have been lost at the time of enrollment. Of the eight patients in whom we determined immediate pretreatment EGFR status, all were EGFR amplified, and only three patients had a clinical response, two of whom had G598V mutation. On the basis of this small sample size, it is unlikely that EGFR amplification predicts for clinical response, and the observation of two clinical responders with G598V mutations is anecdotal.

Comprehensive genome and transcriptome studies have confirmed that EGFR is frequently altered in GBM yet also have identified numerous novel EGFR variants in GBM, including extracellular exon deletions, intracellular C-terminal deletions, and fusions with adjacent genes.^1,44 The sensitivity of these novel variants to EGFR tyrosine kinase inhibitors is not known, and the presence of these variants, which occur nearly exclusively in EGFR-amplified tumors, may have affected the outcomes of our study.

In prior trials of EGFR tyrosine kinase inhibitors in GBM, clear correlations have not been identified between drug response and EGFR expression, activation, EGFR gene amplification, or EGFRvIII status or with activated Akt and PTEN status.⁹ Studies that identified preliminary predictive biomarkers, such as concurrent presence of EGFRvIII and intact PTEN, were not verified in subsequent studies.^45,46 More recent discoveries that GBM tumors and their microenvironment frequently evolve at recurrence^39,47 may explain, in part, the contradictory prior reports because earlier recurrent GBM trials analyzed EGFR pathway status in the initial GBM diagnostic tumors. In our study, initial GBM diagnostic tumor was analyzed for 86% of patients. Therefore, we did not determine PTEN or Akt status and focused on EGFR genetic alterations with the limited archival tumor available. Nevertheless, it remains possible that the status of these proteins may have affected the responses in our patients.

In addition, we explored the potential of RNA in tumor-derived EVs as biomarkers for response to dacomitinib. Tumor-derived genetic material in serum EVs has previously been identified in patients with glioma, and theoretically, serum EVs may provide greater representation of the entire tumor transcriptome compared with single-site tumor biopsy.^20,48,49 Furthermore, the transcriptome of serum EVs may reflect the response of other cells in the body to the presence of the tumor.⁵⁰ In a subset analysis, a transcriptome signature in baseline serum EVs was associated with patients who derived benefit versus patients who were intrinsically resistant to dacomitinib. Inspection of specific differentially expressed genes provide potential mechanisms of dacomitinib activity to explore. For example, LAMTOR2 (late endosomal/lysosomal adaptor, mitogen-activated protein kinase [MAPK], and mammalian target of rapamycin [mTOR] activator 2), which is an activator of MAPK and mTOR signaling, and CSF1, which encodes macrophage colony-stimulating factor (M-CSF), were elevated in rapid progressors. M-CSF is implicated in the recruitment and polarization of microglia and is overexpressed in GBM, and high microglia content has been associated with poor prognosis.^51,52 A combination strategy with CSF1 inhibition could therefore be explored in GBM.

In summary, we observed that dacomitinib penetrates into contrast-enhancing GBM tumor tissue and accumulates in high concentrations. However, only a small subset of patients with EGFR-amplified GBM derived a clinically meaningful benefit from dacomitinib. We did not find EGFR amplification, EGFRvIII or ECD mutation status to be associated with clinical response. Ongoing analysis of circulating EV-derived RNA obtained longitudinally in this study may reveal novel markers of clinical response. Identification of a predictive biomarker could warrant future investigations of novel dacomitinib combinations in genetically enriched GBM populations.

Appendix

Methods

Trial design.

In this multicenter, open-label study, adult patients with recurrent glioblastoma (GBM) with EGFR gene amplification in their tumor and available archival tumor material were enrolled in 1 of 3 arms. Arm A examined the tumor tissue penetration of dacomitinib in 10 patients who were candidates for surgical resection, had experienced their first recurrence, and had not had prior anti–vascular endothelial growth factor (VEGF) therapy. Arm B was a two-stage design phase II trial that included 30 patients at first recurrence who had not had prior anti-VEGF therapy. Arm C was a single-stage, 16-patient exploratory arm that included patients who had an unlimited number of prior therapies but had experienced a first recurrence after a bevacizumab-containing regimen. Arm C was triggered when a predetermined efficacy threshold was achieved in stage I of arm B (progression-free survival at 6 months [PFS6] achieved in 2 of 10 patients). Patients were enrolled at Massachusetts General Hospital (MGH), Dana-Farber Cancer Institute (DFCI)/Brigham and Women’s Hospital, Beth Israel Deaconess Medical Center, Cleveland Clinic, and Henry Ford Hospital.

The study primary end point was PFS6 for patients in arm B. Secondary end points were safety and tolerability in all arms, PFS, overall survival (OS), and radiographic response rate. Exploratory studies included evaluation of dacomitinib concentrations in post-treatment tumor tissue, dacomitinib plasma trough concentrations after repeated dosing, EGFR mutations in archival tumor tissue, and analysis of serum extracellular vesicle (EV) RNA. Tumor assessments were performed by contrast-enhanced brain magnetic resonance imaging (MRI) at baseline and at the end of every even-numbered cycle using MacDonald criteria.²¹

This trial was conducted in compliance with the Declaration of Helsinki and with the International Conference on Harmonization Good Clinical Practice Guidelines protocol and was approved by the institutional review boards and/or independent ethics committees at each of the participating investigational centers. All patients provided written, informed consent before study participation.

Patient population.

Main inclusion criteria for all patients were age ≥ 18 years, histologically confirmed diagnosis of primary or secondary GBM (WHO grade 4), presence of EGFR gene amplification by fluorescence in situ hybridization (FISH) in the most recent available tumor specimen, unequivocal evidence of progressive disease by MacDonald criteria,²¹ Karnofsky performance status ≥ 70%, interval of at least 12 weeks from radiotherapy and 2 weeks from any neurosurgical resection, sufficient time from recovery from prior therapy, availability of archival tumor sample, and adequate hematologic (hematocrit ≥ 29%, leukocytes > 3,000/μL, absolute neutrophil count ≥ 1,500 cells/μL, platelets ≥ 100,000 cells/μL), liver (total bilirubin ≤ 2 times the institutional upper limit of normal [ULN], AST ≤ 2.5 times the ULN), and kidney (creatinine below or within the ULN or creatinine clearance > 60 mL/min/1.73 m² for patients with creatinine levels above institutional normal) function. Exclusion criteria for all patients included prior receipt of therapy with a targeted agent known or proposed to inhibit any component of the epidermal growth factor receptor (EGFR), insulin-like growth factor 1 receptor, mammalian target of rapamycin, or c-Met pathways, presence of extracranial metastatic disease, prior Gliadel wafer therapy, stereotactic radiotherapy, convection enhanced delivery or brachytherapy, clinically significant GI abnormality, interstitial lung disease, significant cardiovascular disease, and concomitant receipt of any medication or food known to induce or inhibit CYP2D6.

Treatment.

Patients received 45 mg dacomitinib once daily with or without food on a continuous basis during a 28-day cycle. Dose interruptions for toxicity of < 2 weeks were allowed without discontinuation from the study; two dose attenuation levels of 30 mg and then 15 mg were allowed. Treatment was discontinued for disease progression, intolerance (grade 3 or 4 toxicity or intolerable grade 2 toxicity that did not resolve to grade 1 or baseline after 2-week interruption), global deterioration, protocol noncompliance, or patient withdrawal.

Evaluation of safety and tolerability.

Safety and tolerability were assessed by standard methods from initiation of study treatment until ≥ 28 days after the last dose of study drug. Adverse events (AEs) were graded by National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE; version 4.0). AEs were graded by worst CTCAE grade and summarized by cycle and by relatedness to dacomitinib. CTCAE grade ≥ 2 CNS hemorrhage or CTCAE grade ≥ 4 nonhematologic toxicity were defined as being unacceptable. For arm A, enrollment was held after three patients were sequentially enrolled to evaluate safety and tolerability during the first postsurgery cycle of dacomitinib. Criteria to discontinue arm A included any CTCAE grade ≥ 3 wound infection, wound dehiscence, or wound complication or grade ≥ 2 head soft tissue necrosis attributable to dacomitinib among the first 3 patients in arm A during this period.

Pharmacokinetic analyses and biomarker studies.

Plasma pharmacokinetic analysis.

Measurement of plasma and tumor tissue dacomitinib was performed by Intertek Pharmaceutical Services (San Diego, CA) using liquid chromatography (LC)-mass spectrometry and LC-tandem mass spectrometry (MS/MS). For all patients, plasma samples for pharmacokinetic (PK) analysis were obtained at baseline before initiation of dacomitinib dosing and predose at day 1 of cycles 1-4. For plasma concentration analysis, dacomitinib was extracted from 200 μL of human plasma (K₂EDTA) by solid-phase extraction (SPE). After evaporation to dryness and reconstitution, the extracts were analyzed by LC-MS/MS.

Tumor tissue PK analysis.

To evaluate post-treatment tumor material for dacomitinib concentrations, 10 patients in arm A received 45 mg dacomitinib daily for 7-9 days before tumor resection, with the last dose given 6-12 hours before craniotomy. No dose reductions were allowed before surgery. On the day of surgery, a plasma sample was obtained for dacomitinib concentration measurement. Tumor material of up to 1 cm² was obtained during surgery and flash frozen and stored at –80°C until analysis. After at least 10 days for recovery from surgery, patients in arm A resumed continuous daily dacomitinib. For tumor tissue dacomitinib concentration analysis, tissue specimens were homogenized using 3 parts 2 M MgCl₂ to 1 part tumor specimen, and homogenization included both shear homogenization and ultrasonication. Dacomitinib was subsequently extracted from the homogenate using acetonitrile protein precipitation and then SPE. After evaporation to dryness and reconstitution, the extracts were analyzed by LC-MS/MS. All tumor specimens were analyzed in one batch.

EGFR amplification.

Confirmation of EGFR gene amplification for study eligibility was performed in Clinical Laboratory Improvement Amendments–certified laboratories at MGH, Cleveland Clinic, or Henry Ford Hospital. EGFR gene amplification was evaluated by FISH using formalin-fixed paraffin-embedded (FFPE) tissue from the most recent available tumor specimen. Amplification was defined as EGFR: centromere 7 copy number control ratio of at least 2:1 as previously defined.²³

EGFR extracellular domain mutation and EGFRvIII analysis.

DNA from FFPE tissue was isolated for genomic polymerase chain reaction (PCR)–based sequencing of coding exons 2, 3, 6, 7, 8, 15, 18, and 21 of the EGFR gene (exons where either missense mutations that sensitize to EGFR inhibitors or recurrent somatic mutations have been reported).^2,6,25 Only previously reported recurrent somatic mutations were considered positive for extracellular domain (ECD) mutation. PCR products were amplified from genomic DNA templates with Platinum Taq polymerase (Thermo Fisher Scientific, Waltham, MA) per manufacturer’s protocol using intron-based primers that span the expressed coding sequences (Appendix Table A1) and then Sanger sequenced (Beckman Coulter Genomics, Chaska, MN). Presence of EGFRvIII from FFPE tissue was determined by real-time reverse transcription (RT) PCR using previously described methods.²² Specificity of RT-PCR product was assessed by direct sequencing using the same primers used for PCR amplification. For some patients, there was either insufficient tumor material, poor-quality nucleic acid after extraction, or technical failure during the assay, and EGFRvIII or ECD mutation status was not determined.

MGMT promoter methylation.

For patients treated at MGH and DFCI, MGMT promoter methylation status was obtained from diagnostic tumor specimens by methylation-specific PCR of bisulfite-treated DNA as previously described.²⁴

Serum EV analysis.

From all patients, 20 mL of whole blood was drawn at baseline (predose) into 10-mL BD Vacutainer SST tubes (BD, Franklin Lakes, NJ). In brief, serum was separated within 2 hours of blood draw, filtered through a 0.8-μm filter, and then frozen immediately and stored at –80°C until EVs were isolated. EVs were isolated from serum and RNA extracted using Exolution²⁶ and subjected to RNA sequencing (RNA-seq) using Exosome Diagnostics’s total RNA-seq platform. Briefly, this procedure subjects extracted total RNA to DNase treatment to remove genomic DNA followed by the addition of an exogenous ERCC Spike-In Mix (Thermo Fisher Scientific). Total RNA was converted to cDNA using random primers, and adapters were annealed using a PCR-based approach. Libraries were cleaned using AMPure XP beads (Beckman Coulter); ribosomal sequences were depleted; and finally, the library was subjected to a final round of amplification. Libraries were again subjected to AMPure XP bead cleanup and quantified using an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA) and Qubit software (Qubit, New York, NY). Paired-end 151 nt sequencing was performed on a NextSeq system (Illumina, San Diego, CA), which yielded an average of 50 million reads per sample.

RNA-seq analysis.

Paired-end RNA-seq reads were aligned to the human genome (hg38) and transcriptome (Gencode version 25) as well as to the ERCC Spike-In sequences and UniVec contaminant sequences, using Spliced Transcript Alignment to a Reference (STAR) version 2.5.2a.^26a Alignments to the transcriptome (in transcriptome coordinates output by STAR –quantMode TranscriptomeSAM) were used as input to Salmon version 0.8.1 (http://gensoft.pasteur.fr/docs/salmon/0.8.2/salmon.html) to obtain transcriptome-level read count and transcripts per million (TPM) expression quantification. These transcript-level counts and TPM quantifications were then aggregated to the gene level using the tximport package in R/Bioconductor (https://www.bioconductor.org/). Differential expression analysis was performed on raw read counts per gene using DESeq2 by fitting a model with treatment response as the only dependent variable and the betaPrior parameter set to false. A gene was reported as statistically significant at Benjamini-Hochberg–corrected P < .05.

Statistical design and analyses.

In a pooled, historical database of patients with recurrent GBM, the PFS6 rate was determined to be 9%³¹; therefore, for the primary end point of PFS6 in arm B, a Simon two-stage design was used to test the null hypothesis (H₀) with 5% significance level that the PFS6 rate is 9% versus the alternative hypothesis (H_a) that the PFS6 rate is 30%. The optimal Simon two-stage design required 10 patients to enroll in stage I. If ≥ two patients were alive and progression free at 6 months, an additional 20 participants would be enrolled into stage II for a total of 30 participants in arm B. In arm B, if ≥ 25 total patients experience progression or death within 6 months, the H_a is rejected, and further investigation may be unwarranted. The expected sample size was 14.5, and the probability of early termination was 0.775, with a type I error rate of 0.034 and 81.5% power. Baseline characteristics, PFS, PFS6, and OS, in arm B were evaluated in the intention-to-treat population; safety was evaluated in the as-treated population; and response was assessed in response-evaluable patients.

In addition, if ≥ two patients in arm B were alive and progression free at 6 months, enrollment in arm C was triggered. In arm C, a Fleming-A’Hern one-stage design tested the H₀ that the PFS6 rate is 5% versus the H_a that the PFS6 is 20%, with exact type I error rate of 0.19 and 86% power. If ≥ 2 patients lived at least 6 months without progression at the end of arm C, the H₀ is rejected, with a target error rate of 0.19 and an actual error rate of 0.189. If 1 or 0 patients live at least 6 months without progression, the hypothesis that P ≥ .20 is rejected, with a target error rate of 0.16 and an actual error rate of 0.141.

For safety, unacceptable toxicity rates ≥ 40% were considered as undesirable. The statistical hypothesis testing for safety in arms B and C differentiated between a 15% and 40% rate of unacceptable toxicity. All patients who received any study medication were included in the safety analyses.

FIG A1.

Axial postcontrast T1 magnetic resonance imaging (MRI) scans of patient 16, who achieved a complete response of the target lesion. Each row contains sequential images (caudal to cranial from left to right) of one MRI time point. The top row shows the baseline MRI time point where the red arrows indicate the target lesion. Note that anterior to the target lesion is a right-side frontal resection cavity that remains after removal of a lesion that appeared simultaneously with the target lesion. The pathology of the resected lesion was recurrent glioblastoma. The bottom row shows MRI scans after 3 cycles of dacomitinib, which show complete response of the target lesion. The yellow arrows depict the location of the prior target lesion. Note that there is collapse of the right-side frontal resection cavity with ex vacuo dilatation of the right-side frontal ventricle horn at the end of the cycle-3 time point.

FIG A2.

Tumor tissue concentrations and clinical outcomes in patients in arm A. Tumor tissue concentrations are plotted for each patient in arm A (cohort of patients pretreated with dacomitinib 45 mg/d for 7-9 days before planned resection of first recurrent tumor) against progression-free survival (PFS) and overall survival (OS).

TABLE A1.

Primers Used for EGFR and IDH1/2, Expressed Coding Sequence Sanger Sequencing, and EGFRvIII Polymerase Chain Reaction

TABLE A2.

Patients in Whom EGFRvIII or EGFR Missense/Kinase Domain Mutations Were Detected

TABLE A3.

Association of EGFR Mutation and Clinical Response to Dacomitinib

TABLE A4.

mRNA Transcripts Within Serum Extracellular Vesicles at Baseline That Were Significantly Differentially Expressed Among Selected Patients Who Clinically Benefited and Rapidly Progressed on Dacomitinib

TABLE A5.

Adverse Events of Grade ≥ 2 Considered Possibly, Probably, or Definitely Related to Dacomitinib

TABLE A6.

Adverse Events of Grade ≥ 3 Considered Possibly, Probably, or Definitely Related to Dacomitinib

TABLE A7.

Dacomitinib Dose Modifications

AUTHOR CONTRIBUTIONS

Conception and design: Andrew S. Chi, Daniel P. Cahill, Patrick Y. Wen, Xandra O. Breakefield, A. John Iafrate, Tracy T. Batchelor

Financial support: Xandra O. Breakefield, A. John Iafrate

Administrative support: A. John Iafrate

Provision of study material or patients: Patrick Y. Wen, David M. Peereboom, Eric T. Wong, Jorg Dietrich, Scott R. Plotkin, Eudocia Q. Lee, Lakshmi Nayak, Hiroaki Wakimoto, Leonora Balaj

Collection and assembly of data: Andrew S. Chi, Daniel P. Cahill, David A. Reardon, Tom Mikkelsen, David M. Peereboom, Eric T. Wong, Elizabeth R. Gerstner, Jorg Dietrich, Scott R. Plotkin, Andrew D. Norden, Eudocia Q. Lee, Hiroaki Wakimoto, Nina Lelic, Mara V. Koerner, Lindsay K. Klofas, Mia S. Bertalan, Isabel C. Arrillaga-Romany, William T. Curry, Darrell R. Borger, Sudipto K. Chakrabortty, Michael D. Valentino, Johan Skog, A. John Iafrate, Tracy T. Batchelor

Data analysis and interpretation: Andrew S. Chi, Daniel P. Cahill, David A. Reardon, Patrick Y. Wen, Elizabeth R. Gerstner, Jorg Dietrich, Lakshmi Nayak, Shota Tanaka, Lindsay K. Klofas, Rebecca A. Betensky, William T. Curry, Leonora Balaj, Robert R. Kitchen, Sudipto K. Chakrabortty, Johan Skog, A. John Iafrate, Tracy T. Batchelor

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

Andrew S. Chi

Employment: Neon Therapeutics, Mirati Therapeutics

Consulting or Advisory Role: Cota Healthcare

Patents, Royalties, Other Intellectual Property: Fused Bicyclic Heterocycles as Therapeutic Agents (inventor) provisional patent application No. 62/742,041 filed on October 5, 2018

Travel, Accommodations, Expenses: AbbVie

Daniel P. Cahill

Honoraria: Merck

Consulting or Advisory Role: Eli Lilly

Travel, Accommodations, Expenses: Merck

David A. Reardon

Honoraria: Merck, Novartis, Novocure, Regeneron Pharmaceuticals, Bristol-Myers Squibb, Oncorus, Agenus, EMD Serono, Merck KGaA, Taiho Pharmaceutical, Advantagene, Bayer AG, Delmar Pharmaceuticals, Imvax

Consulting or Advisory Role: Merck, Novartis, Novocure, Regeneron Pharmaceuticals, Bristol-Myers Squibb, Oncorus, Agenus, EMD Serono, Merck KGaA, Taiho Pharmaceutical, Delmar Pharmaceuticals, Advantagene, Bayer AG, Imvax

Research Funding: Celldex (Inst), Incyte (Inst), Agenus (Inst), EMD Serono (Inst), Acerta Pharma (Inst), Omniox (Inst)

Patrick Y. Wen

Consulting or Advisory Role: Agios, AstraZeneca, Vascular Biogenics, Immunomic Therapeutics, Kayatec, Puma Biotechnology, Taiho Pharmaceutical, Deciphera, VBI Vaccines, Tocagen, Bayer AG, Blue Earth Diagnostics, Karyopharm, Deciphera, Voyager, Taiho Pharmaceutical, QED, Imvax, Elevate Bio, Integral Health

Speakers’ Bureau: Merck, Prime Oncology

Research Funding: Agios (Inst), AbbVie (Inst), AstraZeneca (Inst), Merck (Inst), Novartis (Inst), Oncoceutics (Inst), Eli Lilly (Inst), AstraZeneca (Inst), BeiGene (Inst), Kazia (Inst), MediciNova (Inst), Vacular Biogenics (Inst), VBI Vaccines (Inst), Puma Biotechnology (Inst), Celgene (Inst), Bayer AG (Inst)

Tom Mikkelsen

Honoraria: Roche, Genentech

Travel, Accommodations, Expenses: Roche, Genentech

David M. Peereboom

Consulting or Advisory Role: Orbus Therapeutics

Research Funding: Stemline Therapeutics (Inst), Pfizer (Inst), Novartis (Inst), Neonc Technologies (Inst), Orbus Therapeutics (Inst), Bristol-Myers Squibb (Inst), Genentech (Inst), Roche (Inst), Mylan (Inst)

Travel, Accommodations, Expenses: Stemline Therapeutics

Eric T. Wong

Honoraria: Novocure, UpToDate, Advanced Medical, ZaiLab

Consulting or Advisory Role: Turning Point Therapeutics

Research Funding: Novocure, Orbus, Vascular Biogenics, Five Prime Therapeutics, Plexxikon

Elizabeth R. Gerstner

Consulting or Advisory Role: Blue Earth Diagnostics, MyoKardia (I), Array BioPharma (I)

Jorg Dietrich

Honoraria: Unum Therapeutics, Blue Earth Diagnostics

Consulting or Advisory Role: Blue Earth Diagnostics

Patents, Royalties, Other Intellectual Property: UpToDate

Scott R. Plotkin

Stock and Other Ownership Interests: NFlection Therapeutics, NF2 Therapeutics

Consulting or Advisory Role: AstraZeneca, NFlection Therapeutics

Research Funding: Takeda Pharmaceuticals (Inst)

Travel, Accommodations, Expenses: NFlection Therapeutics

Uncompensated Relationships: Pfizer (Inst)

Andrew D. Norden

Employment: Cota Healthcare, Oncology Analytics

Leadership: Cota Healthcare, Oncology Analytics

Stock and Other Ownership Interests: Cota Healthcare, Oncology Analytics

Eudocia Q. Lee

Honoraria: Medlink

Consulting or Advisory Role: Eli Lilly, Prime Oncology

Patents, Royalties, Other Intellectual Property: Royalties from Wolters Kluwer for Up to Date

Lakshmi Nayak

Travel, Accommodations, Expenses: Bristol-Myers Squibb

Shota Tanaka

Research Funding: Ono Pharmaceutical, Sumitomo Dainippon, Eisai

Patents, Royalties, Other Intellectual Property: P2019-095102

Mara V. Koerner

Employment: Cygnal Therapeutics, Ultivue

Stock and Other Ownership Interests: Cygnal Therapeutics, Ultivue

Research Funding: Cygnal Therapeutics, Ultivue

Isabel C. Arrillaga-Romany

Honoraria: Merck

Consulting or Advisory Role: Insys Therapeutics, Karus Therapeutics, Agios, Boehringer Ingelheim, FORMA Therapeutics

Research Funding: Astex Pharmaceuticals

Rebecca A. Betensky

Consulting or Advisory Role: Biogen, Reata, Alexion Pharmaceuticals, Quark, Intracellular Therapies

Expert Testimony: TEVA Pharmaceuticals Industries, Amarin

Darrel R. Borger

Employment: Takeda Pharmaceuticals

Stock and Other Ownership Interests: Takeda Pharmaceuticals

Leonora Balaj

Employment: Exosome Diagnostics (I)

Leadership: Exosome Diagnostics (I)

Stock and Other Ownership Interests: Exosome Diagnostics (I)

Patents, Royalties, Other Intellectual Property: Massachusetts General Hospital (MGH) has intellectual property on exosome cancer diagnostics and inventor and receives royalties (I); co-inventor and receive royalties from MGH; many patents through Exosome Diagnostics (I)

Robert R. Kitchen

Employment: Exosome Diagnostics

Stock and Other Ownership Interests: Exosome Diagnostics

Patents, Royalties, Other Intellectual Property: Several patents pending for inventions while employed at Exosome Diagnostics

Sudipto K. Chakrabortty

Employment: Exosome Diagnostics

Stock and Other Ownership Interests: Exosome Diagnostics

Research Funding: Exosome Diagnostics

Patents, Royalties, Other Intellectual Property: Patents pending for work at Exosome Diagnostics

Travel, Accommodations, Expenses: Exosome Diagnostics

Michael D. Valentino

Employment: Exosome Diagnostics

Stock and Other Ownership Interests: Exosome Diagnostics

Research Funding: Exosome Diagnostics

Patents, Royalties, Other Intellectual Property: RNA-seq patents as co-discoverer while working at Exosome Diagnostics

Travel, Accommodations, Expenses: Exosome Diagnostics

A. John Iafrate

Stock and Other Ownership Interests: Archer Biosciences

Consulting or Advisory Role: Debiopharm Group, Chugai Pharma, Roche, Repare Therapeutics,

Research Funding: Sanofi

Patents, Royalties, Other Intellectual Property: ArcherDx exclusive license to AMP technology

Tracy T. Batchelor

Honoraria: UpToDate

Consulting or Advisory Role: GenomiCare, Amgen

Travel, Accommodations, Expenses: GenomiCare

No other potential conflicts of interest were reported.