Phase 2 Study of Erdafitinib in Patients with Tumors with FGFR Mutations or Fusions: Results from the NCI-MATCH ECOG-ACRIN Trial (EAY131) Sub-protocol K2

Abstract

PURPOSE:

Subprotocol K2 (EAY131-K2) of the NCI-MATCH platform trial was an open-label, single-arm, phase 2 study designed to evaluate the antitumor efficacy of the oral FGFR1–4 inhibitor, erdafitinib, in patients with tumors harboring FGFR1–4 mutations or fusions.

METHODS:

Central confirmation of tumor FGFR1–4 mutations or fusions was required for outcome analysis. Patients with urothelial carcinoma were excluded. Enrolled subjects received oral erdafitinib at a starting dose of 8 mg daily continuously until intolerable toxicity or disease progression. The primary endpoint was objective response rate (ORR) with key secondary endpoints of safety, progression-free survival (PFS), and overall survival (OS).

RESULTS:

Thirty-five patients were enrolled, and 25 patients were included in the primary efficacy analysis as prespecified in the protocol. The median age was 61 years, and 52% of subjects had received ≥3 prior lines of therapy. The confirmed ORR was 16% (4 of 25, 90% confidence interval (CI) [5.7–33.0%], P = 0.034 against the null rate of 5%). An additional 7 patients experienced stable disease as best-confirmed response. Four patients had a prolonged PFS including 2 with recurrent WHO grade IV, IDH1/2-wildtype glioblastoma. The median PFS and OS were 3.6 months and 11.0 months, respectively. Erdafitinib was manageable with no new safety signals.

CONCLUSION:

This study met its primary endpoint in patients with several pretreated solid tumor types harboring FGFR1–3 mutations or fusions. These findings support advancement of erdafitinib for patients with FGFR-altered tumors outside of currently approved indications in a potentially tumor-agnostic manner.

Introduction

The fibroblast growth factor receptor (FGFR) family is comprised of 4 transmembrane receptor tyrosine kinases (FGFR1–4) where receptor activation by more than 20 known fibroblast growth factor (FGF) ligands leads to initiation of signaling cascades critical for cellular proliferation, survival, angiogenesis, and differentiation ¹. Aberrations in FGFR1-4, inclusive of gene mutations (i.e., single-nucleotide variants or SNVs), copy number amplifications (CNAs), and gene rearrangements or fusions, have been identified in approximately 5%−10% of all human cancers ². Erdafitinib was the first FGFR inhibitor approved for patients with locally advanced or metastatic urothelial carcinoma progressing on or after ≥1 prior chemotherapy with FGFR3 gene mutations or FGFR2/FGFR3 gene fusions ³. Since then, two other FGFR inhibitors, pemigatinib and infigratinib, have been approved for patients with locally advanced or metastatic cholangiocarcinoma with FGFR2 fusions or rearrangements having been treated with ≥1 prior chemotherapies ^4,5. Pemigatinib has also been approved for FGFR1-rearranged myeloid and lymphoid neoplasms ⁶.

NCI-MATCH treatment subprotocol K2 was designed as an open-label phase 2 study of the oral FGFR1–4 inhibitor, erdafitinib, in patients with tumors harboring FGFR1-4 mutations or FGFR1-3 fusions. The overall objective of this subprotocol was to evaluate the activity of FGFR inhibition across a spectrum of treatment-refractory and biomarker-selected malignancies for which erdafitinib had not been specifically approved or in which the activity of FGFR inhibition has not been established.

Methods

Patient selection.

Eligible adult patients had any solid tumor except for transitional cell carcinoma of the bladder and/or urothelial tract, lymphoma, or myeloma progressing on standard treatment or for whom no standard treatment was available. Adequate hematologic (leukocytes ≥3,000/mcL, absolute neutrophil count ≥1,500/mcL, platelets ≥100,000/mcL), liver (total bilirubin ≤1.5 and AST/ALT <2.5 X institutional upper limit of normal or ULN), and kidney function (creatinine clearance ≥45 mL/min/1.73 m² for creatinine above ULN), performance status of ECOG ≤ 1, and submission of fresh biopsy were required. Key eligibility criteria for NCI-MATCH subprotocol K2 include presence of select FGFR1–4 mutations or FGFR1–3 fusions as determined by protocol-defined tumor profiling and must not have previously received treatment with an FGFR-targeted inhibitor. Investigators obtained informed consent from each participant or each participant’s guardian and approval by local IRB was obtained for this study. Of note, the first FGFR inhibitor (pemigatinib) was approved in 2020 for cholangiocarcinoma based on the FIGHT-202 trial after enrollment of subjects into subprotocol K2 was finished ⁴.

Tumor profiling.

Tumor profiling for central testing was accomplished using the Oncomine Cancer Panel assay which uses AmpliSeq chemistry and the PGM sequencer to detect and report 4066 predefined genomic variations, encompassing 3259 SNVs, 114 indels, 435 large indels, 75 CNVs, and 183 gene fusions across 143 unique genes as described in Lih, et al. ⁷. Regarding detection of fusions, the Oncomine assay detects targeted gene fusions by sequencing cDNA converted directly from specific targeted RNA transcripts rather than sequencing intronic DNA to identify breakpoints. Actionable mutations were assessed using a next-generation sequencing (NGS) panel of 143 genes, including SNVs, indels, amplifications, and selected fusions ⁷.

The full list of FGFR fusions and activating mutations included in this study is shown in Appendix II of Protocol #EAY131-K2 for subprotocol K2. In brief, FGFR gene variants were identified if at least one of the following were met: 1) gene variant selected has an approved drug, 2) gene variant represents an eligibility criterion for an ongoing clinical trial for that treatment or gene variant has been identified as etiology of N of 1 response, or 3) gene variant has preclinical data providing biological evidence to support use of that variant for treatment selection (in vitro or in vivo). Confirmatory central testing was attempted for all patients whose tumors were deemed eligible on the basis of results from designated outside laboratories.

Study design and assessments.

Patients were assigned using a prospectively defined NCI-designed informatics rules algorithm (MATCHBOX), as previously described ⁸. Erdafitinib was administered orally, with or without food, at the prescribed dose (flat dose) daily continuously, for 28 days of every 28-day cycle. The starting dose of erdafitinib was 8 mg oral daily throughout the first 14 days of a 28-day cycle. On Day 15 of Cycle 1, erdafitinib could be increased to 9 mg oral daily continuously if serum phosphate levels were <5.5 mg/dL, and no significant toxicities were noted. Alternatively, patients were continued at 8 mg oral daily continuous if serum phosphate levels were between 5.5 and 7.0 mg/dL and no significant toxicities, or dose was reduced per protocol for serum phosphate levels ≥7.0 or presence of significant toxicity. Responses were evaluated every 2 cycles for the first 26 cycles and every 3 cycles thereafter until disease progression (PD) by investigator confirmation per RECIST version 1.1 ⁹. Toxicity was evaluated using NCI Common Terminology Criteria for Adverse Events (CTCAE) version 4.0.

Statistical considerations.

Primary analyses will be based on the group of eligible and treated patients excluding those who were entered into a subprotocol of MATCH with results from an “outside” assay that could not be confirmed by the MATCH assays or a validated orthogonal assay. For each treatment subprotocol, 90% two-sided confidence intervals will be calculated for the following two endpoints: overall response rate (ORR, the primary endpoint); proportion of patients alive and progression free at 6 months from start of treatment on that treatment step. The original accrual goal for each subprotocol was 35 patients, aiming to obtain a minimum of 31 eligible patients. A response rate of 5/31 patients (16%) or more met the primary endpoint. With this design, the power was 91.8% to conclude an agent is promising if its true response rate is 25%, and the type 1 error (one-sided) was 1.8% if its true response rate is 5%. Secondary objectives were PFS6, PFS, OS, evaluation of predictive biomarkers (co-mutations that potentially predict response to erdafitinib), and safety. If fewer than 31 patients accrued were eligible for inclusion in the primary analysis, then the primary endpoint was assessed using a 5% one-sided exact binomial test of the null hypothesis that the ORR was ≤5%.

Results

Patient and tumor characteristics

From July 3, 2018 to July 15, 2019, a total of 35 patients were enrolled into the NCI-MATCH K2 subprotocol. Figure 1 highlights how the 25 cases for inclusion in the primary analysis were reached.

Figure 1.

Patient flow diagram.

Characteristics of the 25 subjects included in the primary analysis are described in Table 1. The majority had been previously treated with 3 or more lines of prior therapies (52%). Data on the larger cohort of 33 subjects who were eligible by outside molecular profiling, treated with erdafitinib (including 8 subjects whose tumor FGFR alterations were not able to be centrally confirmed and thus were not included in the primary analysis) are presented in Supplementary Tables 1-3. The most common histologic subtypes included intrahepatic cholangiocarcinoma (n=5, 20%), WHO grade IV glioblastoma (n=4, 16%), endometrioid endometrial adenocarcinoma (n=3, 12%), carcinomas of the salivary gland (n=3, 12%), and pancreatic adenocarcinoma (n=2, 8%, Supplemental Table 9).

Table 1.

Patient Characteristics

	Centrally confirmed (n=25)
Female	20 (80%)
Age: min, 25%, med, 75%, max	26, 54, 61, 67, 83
Race:
White	18 (72%)
Black	3 (12%)
Asian	2 (8%)
Unknown/Not_reported	2 (8%)
Ethnicity:
Hispanic	1 (4%)
Non-Hispanic	20 (80%)
Unknown/Not_reported	4 (16%)
ECOG performance status:
0	10 (40%)
1	15 (60%)
Number of Prior Lines of Therapies:
0*	2 (8%)
1	4 (16%)
2	6 (24%)
3	2 (8%)
>3	11 (44%)
Types of Prior Therapies+:	16
Chemotherapy (multi-agent)	11
Chemotherapy (single agent)	4
Chemotherapy NOS	6
Immunotherapy	5
Hormonal therapy	1
Image-directed local therapy	15
Radiation therapy	19
Surgery	2
Therapy NOS
Weight loss in previous 6 month:
< 5%	20 (80.0%)
5 to < 10%	3 (12.0%)
10 to < 20%	2 (8.0%)

^*Two subjects had no prior therapy: Malignant Brenner tumor of ovary (n=1); Low-grade rosette-forming glioneuronal tumor of midline brain stem, WHO grade I (n=1). These patients were eligible as no standard therapy available.

⁺Each subject may have had more than 1 type of therapy received along treatment course

NOS, not otherwise specified

The mutational landscape of all 25 subjects, which existed at baseline prior to study treatment, by tumor histology and best overall response to erdafitinib is shown in Figure 2. Alterations in FGFR2 (48%) were most common, followed by FGFR3 alterations (32%). The most frequent non-FGFR alterations were TERT and PIK3CA mutations.

Figure 2.

Heatmap of molecular alterations

Efficacy

As of the data cut-off of June 2, 2022, all 25 patients had discontinued study treatment. In the primary analysis cohort, the ORR was 16% (4 out of 25, 90% CI [5.7–33.0%]). Seven patients (28%) experienced stable disease (SD) and 7 had PD as best confirmed response; 7 were not evaluable. The four responders in the primary efficacy cohort to erdafitinib had tumors carrying FGFR2 fusions in 2 cases of intrahepatic cholangiocarcinoma, an FGFR2 mutation in 1 case of high-grade adenoid cystic carcinoma of submandibular gland, and an FGFR3 mutation in 1 case with malignant Brenner tumor of the ovary. A waterfall plot showing best change from baseline lesion size in response to erdafitinib is shown in Figure 3. The most common reason for treatment discontinuation of erdafitinib was PD in 14 subjects (56%) followed by AEs in 4 subjects (16%). Three discontinuations were due to patient withdrawal/refusal, 3 discontinuations per investigator discretion including clinical progression, and 1 discontinuation due to patient death. For the 33-patient cohort (eligible and treated), the ORR was 12% (Supplemental Table 4 and Supplemental Figure 1).

Figure 3.

Best Lesion Size Change from Baseline

Data on all cases of PR and PFS >168 days (six 28-day cycles) in the primary analysis cohort and 33-patient cohort along with all FGFR variants are presented in Supplemental Tables 5-6. The median duration of response to erdafitinib in the primary analysis patients with best response of PR or SD is shown in Figure 4 and can be found for the 33-patient cohort (eligible and treated) in Supplemental Figure 2.

Figure 4.

Duration of treatment

Safety

A total 34 of 35 enrolled subjects who received erdafitinib treatment were assessed for safety (Supplemental Table 10). Adverse events resulted in study drug discontinuation in 4/25 subjects (16%). The most common grade 1–2 AEs were dry mouth (52.9%), diarrhea (50.0%), fatigue (47.1%), and anemia (32.4%). The most frequent grade 3 AEs were mucositis (14.7%), fatigue (2.9%), anemia (2.9%), nausea (2.9%), paronychia (2.9%), and palmar-plantar erythrodysesthesia syndrome (2.9%). No treatment-related grade 4–5 AEs were reported. The one treatment discontinuation due to death was from grade 5 disease progression deemed unrelated to study treatment (Supplemental Table 3). Of 97 treatment cycles in the primary analysis cohort, 4 subjects never had any type of dose modification, while 5/25 subjects had erdafitinib escalated to 9 mg during cycle 1 (10 out of 34 subjects assessed for safety had dose escalation as per protocol). There was a total of 14 dose reductions across 97 treatment cycles (14%) in the primary efficacy cohort (21/103 dose reductions in the 33-patient cohort). Dose modifications for both the primary efficacy and 33-patient cohorts are shown in Supplemental Tables 7-8.

Survival

At the time of data cut-off, there were 21 PFS events and 21 OS events. Twenty-five patients of the primary analysis cohort with FGFR-altered tumors were treated with erdafitinib (Figure 5). The median PFS and OS for the 33-patient cohort (eligible and treated) is shown Supplemental Figure 3.

Figure 5.

PFS and OS for patients treated with erdafitinib

Predictive Biomarkers

As part of an exploratory analysis, the top co-occurring mutations were associated with efficacy in the primary analysis cohort treated with erdafitinib (Figure 2). The 6 most frequent co-occurring mutations were TERT (n=6), PIK3CA (n=6), TP53 (n=5), ARID1A (n=5), PTEN (n=4), and BAP1 (n=3). All 5 cases with TP53 mutations experienced PD as their best response, translating to a 6-month PFS rate of 0%. The 6-month PFS rate was also 0% in 2 evaluable subjects with co-occurring PTEN mutations where no responses to erdafitinib were seen. Of 5 evaluable subjects with PIK3CA mutations, there was 1 PR in a subject with a co-occurring BAP1 mutation as well, while the other 4 did not experience any response. On the other hand, in 3 subjects with BAP1 mutations, there were 2 PRs (including 1 with a co-occurring PIK3CA mutation) and 1 SD as best response to erdafitinib, which translated to a 6-month PFS rate of 100%. Although TERT mutations were the most common co-occurring alterations in this cohort, the majority of subjects (n=4) were unevaluable for tumor response.

Discussion

Subprotocol K2 of the NCI-MATCH trial was a phase 2, single-arm study investigating the antitumor activity of the oral FGFR1–4 inhibitor, erdafitinib, in patients with treatment-refractory solid tumors that had centrally confirmed FGFR1–4 mutations or FGFR1–3 fusions. In this phase 2 study, 25 subjects had centrally confirmed FGFR alterations, and we observed an ORR of 16.0% (P = 0.034), meeting the primary endpoint for this study. Erdafitinib was well-tolerated in this cohort with no new safety signals. Most AEs were grade 1–2 in severity; there were no grade 4–5 AEs.

Our primary efficacy results are consistent with an earlier phase I dose-escalation study of erdafitinib in 187 subjects with advanced solid tumors for which standard antineoplastic therapy was no longer effective ¹⁰. Here, erdafitinib demonstrated an ORR of 21% across all 92 subjects (19/92 subjects, all with PRs) with solid tumors having FGFR mutations or gene fusions. Notably, the 19 responders to erdafitinib in the phase I study were heavily driven by 2 FGFR-driven tumor subtypes: urothelial cancer (n=12 PRs) and cholangiocarcinoma (n=3 PRs). In essence, patients with FGFR-altered urothelial cancer contributed to 63% of responders in this phase I study. The efficacy is not surprising, given that erdafitinib became the first FGFR inhibitor approved in any cancer type with its FDA approval in previously treated urothelial cancer having an FGFR3 mutation or FGFR2/3 fusion based on a 40% ORR to erdafitinib in an open-label, phase 2 study ³. This approval has been followed by two other phase 2 studies in treatment-refractory advanced cholangiocarcinoma carrying FGFR2 fusions or rearrangements that resulted in FDA approvals of the FGFR inhibitors infigratinib, pemigatinib, and futibatinib based on ORRs of 23.1%, 35.5%, and 42%, respectively, in this population ^4,5,11.

Although there were two PRs to erdafitinib in 2 cases of intrahepatic cholangiocarcinoma with FGFR2 fusions, it should be noted that our phase 2 study excluded patients with urothelial cancer. PRs were seen in other FGFR-altered tumor types with limited data (1 in a FGFR2-mutated high-grade adenoid cystic carcinoma of the submandibular gland and 1 with a FGFR3-mutated Malignant Brenner tumor of the ovary). FGFR2 signaling has been implicated in morphogenesis of submandibular glands, and a previous phase 2 study showed limited activity in patients with metastatic adenoid cystic carcinoma treated with the FGFR inhibitor dovitinib ^12,13. Although responses (<10% overall) have been characterized in patients with advanced gynecologic cancers such as ovarian and endometrial cancers with FGFR inhibitors ^10,14,15, our study describes one of the first cases, to the best of our knowledge, of activity with an FGFR inhibitor in a Malignant Brenner tumor of the ovary, for which there is no definitively established systemic therapy. Interestingly, 55% of Malignant Brenner tumors have been shown to possess FGFR1/3 genomic alterations, and FGFR3 mutations have been associated with benign or borderline Brenner tumor precursor components that support ovarian origin ¹⁶.

Additionally, there were seven SDs as best overall response to erdafitinib in our cohort, of which 4 of these cases occurred in treatment-refractory, FGFR-altered, IDH1/2-wildtype brain tumors (Figure 2). Notably, 4 subjects experienced a prolonged PFS >168 days that included 1 WHO grade I dysembryoplastic neuroepithelial tumor of brain (FGFR1 fusion), 2 WHO grade IV, IDH1/2-wildtype glioblastoma (both with FGFR3 fusions), and 1 intrahepatic cholangiocarcinoma (FGFR2 fusion). Previous phase I-II trials in patients with recurrent high-grade glioblastomas have shown limited activity with oral FGFR inhibitors when unselected for FGFR alterations ^17–19. However, recent translational studies have reported that 3.5% of IDH1/2 wildtype, but 0% of IDH1/2-mutant high-grade gliomas harbor FGFR3 fusions ²⁰. Erdafitinib showed evidence of anticancer activity in preclinical glioma models and in patients with FGFR3-rearranged gliomas ²⁰.

In an exploratory analysis of predictive biomarkers, we identified that co-occurring mutations in BAP1 were positive predictors of benefit to erdafitinib, while those with TP53, PTEN, and PIK3CA mutations were associated with lack of benefit. Interestingly, prior data has supported that BAP1 protein loss was associated with enhanced sensitivity to FGFR inhibition in a preclinical tumor model ²¹. Mutations in the PTEN/PI3K/AKT/mTOR and MAPK pathways have been increasingly described as mechanisms of acquired resistance to FGFR inhibition through bypass signaling ²². In several tumors, FGF ligand binding has been shown to downregulate TP53, while FGFR inhibition has been shown to upregulate TP53 activity; the role of TP53-mediated resistance to FGFR inhibition remains poorly described, however ^23,24. Although limited by small sample sizes, our exploratory analyses point to potential predictive biomarkers to erdafitinib and FGFR inhibition, in general, and warrant further investigation.

Beyond evaluation of FGFR inhibition in specific tumor types that are FGFR altered, our study lends support that better selection for FGFR inhibition in FGFR-altered malignancies may be indicated as well. In our cohort, activity was seen with erdafitinib across a number of solid tumors demonstrating mutations or fusions in FGFR1, FGFR2, and FGFR3. However, as shown in Figure 2, PR and SD appear to occur more frequently in those with FGFR2–3 fusions suggesting that this subset may further identify those who could benefit from FGFR inhibition among all FGFR alterations in tumor-agnostic trials. Results of the phase 2 RAGNAR study recently reported an ORR of 30% for erdafitinib in heavily pretreated, advanced solid tumors with FGFR1–4 mutations or gene fusions ²⁵. RAGNAR was a larger study enrolling 217 subjects treated with erdafitinib and enrolled a greater proportion of subjects with FGFR2–3 fusions (61%). ORR to erdafitinib was higher in the presence of FGFR fusions in RAGNAR (whereby 93% of subjects with FGFR fusions had FGFR 2–3 fusions) compared those with FGFR mutations (33% vs. 25%)²⁵. Notably, of the 6 complete responders to erdafitinib in RAGNAR, all of them had FGFR fusions. Interestingly, ORR was 56% of the 18 subjects with pancreatic cancer and 14% in the 37 subjects with gliomas in RAGNAR. We observed no responses to erdafitinib in our pancreatic cancer and low-grade or high-grade glioma subjects. We had 8 subjects that were non-evaluable for tumor response as well which could have contributed to differences in ORR observed between RAGNAR and our study where the primary efficacy cohort consisted of 25 subjects. Additionally, it should be noted that 5/25 subjects were escalated to the 9 mg dose level and given that a prior phase I has suggested a higher ORR to erdafitinib at 9 mg vs. 8 mg, it is unclear if this played a role in our study¹⁰. Despite these differences, it is reassuring that despite being a late-line therapy in many subjects with refractory solid tumors, erdafitinib resulted in a median OS of 11.0 months in subprotocol K2, which is comparable to the 10.7 months in RAGNAR²⁵. This growing evidence, along with results from this trial, offers provocative support that targeting FGFR with erdafitinib may be primed for clinical application through a tumor-agnostic, biomarker-driven treatment approach.

Supplementary Material

PV - Appendix Figure 2

Supplemental Figure 2. Duration of treatment for 33-patient cohort including subjects with tumor FGFR alterations not centrally confirmed who were not included in the primary analysis

PV - Appendix Figure 1

Supplemental Figure 1. Best lesion size change from baseline for 33-patient cohort including subjects with tumor FGFR alterations not centrally confirmed who were not included in the primary analysis

PV - Appendix Figure 3 BOTTOM

Supplemental Figure 3. Median PFS and OS for 33-patient cohort including subjects with tumor FGFR alterations not centrally confirmed who were not included in the primary analysis

Context Summary.

Key Objective:

Subprotocol K2 of the NCI-MATCH trial was a phase 2 single-arm trial demonstrating the efficacy of the oral FGFR inhibitor erdafitinib in solid tumors with FGFR1–3 mutations or gene fusions exclusive of urothelial carcinoma.

Knowledge generated:

Erdafitinib demonstrated a centrally confirmed ORR of 16% and clinical benefit rate of 32% across a spectrum of solid tumors with objective responses and prolonged SD in tumor types including cholangiocarcinoma, adenoid cystic head and neck cancer, gynecologic cancer, and IDH1/2-wildtype brain tumors. Subjects were predominantly heavily pretreated, and erdafitinib was safe and well tolerated.

Relevance:

Positive results from this study support the potential advancement of erdafitinib to treat FGFR-altered malignancies, particularly those with FGFR2/3 fusions, beyond current FDA-approved indications in a potentially tumor-agnostic fashion.