Efficacy of Combined VEGFR1-3, PDGFα/β, and FGFR1-3 Blockade Using Nintedanib for Esophagogastric Cancer

All tumors depend on angiogenesis for growth and metastatic progression, and increased VEGFR2 signaling is associated with poorer outcomes in gastric cancer (1, 2). This rationale motivated trials of the VEGFR2-directed mAb, ramucirumab, in patients with metastatic esophagogastric cancer, which demonstrated improved progression-free (PFS) and overall survival (OS; refs. 3, 4). Ramucirumab is now FDA-approved for use alone or in combination with paclitaxel in patients with esophagogastric cancer following disease progression on first-line chemotherapy. Nonetheless, the vast majority of patients treated with ramucirumab ultimately progress, and novel therapeutic options are urgently needed for this population.

Large-scale sequencing initiatives (5–9) have revealed that amplification and simultaneous activation of multiple receptor tyrosine kinases (RTK) is one of the hallmarks of esophagogastric cancer. Upregulation of the proangiogenic FGFR and platelet-derived growth factor receptor (PDGFR) families of RTKs provide escape mechanisms that can mediate therapeutic resistance to VEGFR2 inhibition in preclinical models (10). Furthermore, genomic profiling studies of esophagogastric tumors indicate that FGFR2 alterations are present in 5% to 9% of esophagogastric tumors, and it has been postulated that this genomically defined subset of patients may be particularly sensitive to dual FGFR and VEGFR inhibition (8, 11).

Nintedanib is a multikinase inhibitor that potently inhibits VEGFR1-3, FGFR1-3, and PDGFRα/β (IC₅₀, 20–100 nmol/L). Nintedanib leads to sustained (>30 hours) blockade of VEGFR2 in vitro, which, in mice bearing solid tumor xenografts, translates to reduced vessel density and vessel integrity after 5 days, as well as profound growth inhibition (12). Nintedanib was approved in Europe for use in combination with docetaxel for the treatment of patients with metastatic lung adenocarcinoma (13). As the progression of cancer has been shown to be biologically dependent on angiogenesis (1), and as nintedanib can inhibit both VEGFR signaling and putative angiogenic bypass mechanisms such as FGFR signaling, we conducted a phase II study of nintedanib in patients with previously treated metastatic esophagogastric adenocarcinoma to test the hypothesis that inhibition of multiple angiokinases may be more effective than selective VEGFR2 inhibition.

This was a single-arm, open-label, nonrandomized phase II study of nintedanib at a dose of 200 mg administered twice daily by mouth until intolerable adverse events, progressive disease, or death (ClinicalTrials.gov study NCT02234596). The primary objective was to define the proportion of patients who were progression-free at 6 months (6-month PFS). The study had an exact binomial single-stage design (14), in which 32 patients were treated to differentiate between 6-month PFS of sufficient and insufficient drug activity at 10% (based on historical controls; ref. 15) and 28%, respectively, with type I and II error rates of 10% each. On the basis of this study design, nintedanib would be considered worthy of further study if at least 6 patients were alive and progression-free at 6 months. Secondary objectives included determining the objective response rate (as defined by RECIST 1.1; ref. 16) and identification of predictive or prognostic molecular biomarkers by tumor sequencing.

Eligible patients were at least 18 years old and had a diagnosis of metastatic or recurrent esophageal or gastroesophageal junction (GEJ) adenocarcinoma with radiographically measurable or evaluable lesions by RECIST 1.1 criteria (16). Patients may have received up to one prior chemotherapy regimen for metastatic disease or up to two prior regimens if they had previously received curative-intent chemotherapy or chemoradiotherapy. Other eligibility criteria included adequate performance status and organ function. Exclusion criteria included ERBB2-amplified disease, prior treatment with a VEGFR2 inhibitor, or a history of an arterial thromboembolic or hemorrhagic event. During the study enrollment period, there was an active study for second-line patients with HER2-positive metastatic esophagogastric cancer; therefore, patients with HER2-positive disease were excluded from this study. Patients with a history of deep vein thrombosis or pulmonary embolism and stable on an anticoagulation regimen were eligible.

Twenty-seven samples were of adequate quality for molecular analysis; 21 samples were obtained prior to first-line therapy, five prior to second-line therapy, and one after nintedanib therapy (third-line).The MSK-IMPACT next-generation sequencing (NGS) assay was performed in a Clinical Laboratory Improvement Amendments–certified laboratory as described previously, with results reported in the electronic medical record (17, 18). MSK-IMPACT detects mutations, small insertions and deletions, copy-number alterations (CNA), and select structural rearrangements in cancer-associated genes. Several versions of the assay were used, depending upon the date of tumor sequencing (Supplementary Table S1). Only the 341 genes common to all three versions of the MSK-IMPACT assay were analyzed. Alterations in 10 canonical oncogenic signaling pathways were assessed as described previously (19). A pathway was classified as activated or inactivated in an individual tumor sample when at least one member gene was affected by a known or likely driver alteration, as defined by the OncoKB knowledge base (20).

All patients who received nintedanib were included in the analysis. PFS and OS were calculated from the date of treatment initiation to the date of radiographic disease progression, death, or last evaluation. PFS and OS were estimated using Kaplan–Meier methods and compared between primary tumor locations using the log-rank test. Response rate was summarized using binomial proportions, and exact 95% confidence intervals (CI) were calculated.

In patients with available molecular profiling, the genomic alteration status of 10 signaling pathways was assessed as described by Sanchez-Vega and colleagues (19). Only pathways altered in 10% or more of samples were considered in the analysis. Global CNAs, tumor mutation burden (TMB), and tumor purity were compared using the Wilcoxon rank-sum test after grouping samples by anatomical site of the primary tumor, and for CNA comparisons, affected signaling pathway. Pathway alteration status between primary tumor sites was compared using Fisher exact test. Distributions of PFS between patients whose tumors carried alterations in each pathway were compared with patients whose tumors did not using permutation log-rank test (21). All statistical, biomarker, and MSK-IMPACT analyses were performed using R version 3.5.0 (R Foundation for Statistical Computing) using the “survminer” package. All P values were two-sided. P values less than 0.05 were considered to indicate statistical significance.

The study was conducted in accordance with the Declaration of Helsinki, International Ethical Guidelines for Biomedical Research Involving Human Subjects, the Belmont Report, and the U.S. Common Rule. All patients provided a written informed consent approved by the Memorial Sloan Kettering Cancer Center Institutional Review Board. Nintedanib was provided by Boehringer Ingelheim GmbH. The senior academic authors had full access to all clinical and molecular data collected during the study and had final responsibility for the decision to submit the manuscript.

All genomic and clinical data from this study are publicly available through the cBioPortal for Cancer Genomics (www.cbioportal.org; ref. 22).

From October 2, 2014 to June 16, 2017, 32 patients were enrolled (Table 1; Supplementary Table S1). One patient withdrew consent during the third month of treatment. This patient was included in the final PFS and OS analyses. The study population consisted exclusively of patients with esophageal/GEJ (17 patients, 53%) or gastric (15, 47%) adenocarcinomas. The majority of patients (23, 72%) had suffered disease progression on one prior systemic chemotherapy regimen for metastatic disease. Eight patients (25%) had peritoneal carcinomatosis, and 12 (38%) had multiple sites of metastases.

The median PFS was 1.9 months (95% CI, 1.6–3.6 months; Fig. 1A). Median OS was 14.2 months (95% CI, 10.8 months–NR; Fig. 1B). No patient achieved a complete or partial response by RECIST 1.1 criteria; 14 (44%) achieved stable disease (Fig. 1C). Two patients had progression of nontarget lesions despite shrinkage of their primary lesion (Fig. 1C). Of the 27 (84%) patients with measurable lesions, 10 (37%) achieved stable disease. Six of the 32 patients (19% CI, 9%–38%) were progression-free at 6 months, and thus the study met its prespecified primary endpoint. Prolonged disease stabilization of 12 months or greater was seen in 2 patients, and another 2 achieved stable disease for at least 8 months.

Of the 32 patients, none required dose reduction for toxicity. One patient discontinued the study drug as a result of grade 2 fatigue. Common adverse events are shown in Table 2. The majority were grade 1 or 2 on the basis of Common Terminology Criteria for Adverse Events criteria (23). Observed grade 3 adverse events included hypertension (5 patients), liver enzyme elevation (2 patients), lymphopenia (1 patient), hypertriglyceridemia (1 patient), and fatigue (1 patient). There was a single grade 3 thromboembolic event, but no bleeding or perforation events were observed. The only grade 4 event was hypertriglyceridemia (1 patient). There were no grade 5 events.

Tumor molecular profiling was performed using a targeted NGS platform (MSK-IMPACT; refs. 17, 18). Tumor samples of 27 patients were adequate for molecular analysis (18 primary tumors and 9 metastatic samples). We achieved a mean sequencing coverage of 668.7× and identified an average of 3.76 nonsynonymous mutations per Mb per tumor sample (range, 0–11 mutations; Supplementary Table S2). The most commonly altered genes, affected by somatic mutations, amplifications, or homozygous deletions, were TP53 (16 patients, 59%), ARID1A (7, 26%), KRAS (6, 22%), FGFR2 (5, 19%), CDKN2A (4, 15%), and CCND1 (3, 11%).

There was a high incidence of FGFR2 alteration (19% of all amplifications defined by NGS only) in tumors collected from our study population as the trial was enriched for this group given our hypothesis that FGFR2-amplified tumors may be more sensitive to FGFR inhibition. In the 5 patients with FGFR2 amplification, 2 had amplifications determined from a primary tumor sample, and 3 were from metastatic samples. Four of 5 patients had measurable disease and low-level FGFR2 amplification. Only one of 5 patients had peritoneal metastases. Although patients with FGFR2 amplifications had longer PFS compared with patients without such alterations, the difference was not statistically significant (median PFS, 3.5 vs. 1.9 months; P = 0.92). Of note, the only patient with high FGFR2 amplification did not have evaluable lesions, but experienced the longest PFS (6.6 months) among patients with FGFR2-amplified tumors. Tumors from the 3 patients with the longest PFS (>8 months) did not have FGFR amplification.

Pathway level analysis accounting for tumor anatomic location revealed that alterations in the cell cycle and TGFβ pathways were significantly associated with differences in PFS (Fig. 2A; Supplementary Table S3). However, the TGFβ pathway was not explored further because it was altered in fewer than 10% of patients. The median PFS in patients with cell-cycle pathway alterations was significantly worse compared with patients without such alterations (1.6 vs. 2.7 months, P = 0.02; Fig. 2B). Cell-cycle pathway alterations were associated with more DNA CNAs (fraction of the genome altered 0.31 vs. 0.10 in samples without cell-cycle alterations, P = 0.005; Fig. 2C). Consistent with this finding, cell-cycle pathway alterations were more common in the chromosomal instability subtype (defined as tumors with >5% fraction genome altered) than in the genomically stable subtype (8/16 = 50% vs. 1/11 = 9%, P = 0.042, Fisher's test). TMB was significantly higher in esophageal/GEJ adenocarcinomas compared with stomach adenocarcinomas (5.23 vs. 0.94 mutations/MB, P = 0.0005), whereas cell-cycle pathway alteration status (P = 0.42), CNA (P = 0.12), and tumor purity (P = 0.2) did not vary significantly by tumor anatomical location.

This phase II trial of the multitargeted kinase inhibitor, nintedanib, in patients with esophagogastric adenocarcinoma met its primary PFS endpoint; 19% of patients were progression-free at 6 months. Although minor tumor regressions were noted in some patients with stable disease as best response by RECIST 1.1 criteria, the effects of nintedanib were primarily cytostatic. This result is similar to that reported in the LUME-Lung 1 trial, in which nintedanib plus chemotherapy increased the proportion of patients with stable disease (49.6% vs. 37.9% in the chemotherapy-alone group; ref. 13). In that trial, only 4.4% patients achieved a partial response and, as observed in this study, no patient achieved a complete response.

The median PFS we observed, 1.9 months (95% CI, 1.6–3.6 months), is similar to that reported for other angiokinase-targeted therapies in advanced esophagogastric cancer. PFS with the VEGFR2 mAb, ramucirumab, as monotherapy, as determined in the phase III REGARD trial, was 2.1 months (vs. 1.3 months with placebo, P < 0.0001); the trial also found improved median survival (median OS 5.2 vs. 3.8 months, P = 0·047; ref. 4). The PFS benefit of regorafenib, a multikinase inhibitor targeting the proangiogenic kinases (VEGFR2/3) and prooncogenic RTKs (RET, KIT, and PDGFR), was similarly modest (3.1 vs. 0.9 months in the placebo arm, P < 0.001) in a phase II trial (24).

As nintedanib inhibits multiple compensatory proangiogenic pathways, including those activated by PDGF and FGF family receptors, we hypothesized that patients with FGFR alterations may be particularly sensitive to this drug. We examined the correlation of FGFR alteration status with clinical benefit to nintedanib, but found that FGFR2 alterations were not a predictive biomarker of clinical benefit (median PFS, 3.5 vs. 1.9 months, P = 0.81). Similar results were seen in the SHINE trial, where patients with FGFR2 amplifications or polysomy gastric cancers were randomized to the FGFR 1/2/3 inhibitor, AZD4547, or paclitaxel. AZD4547 did not improve PFS in the FGFR2 amplified/polysomy patients compared with paclitaxel (25). In contrast, in the MATCH-W study of AZD4547, among the 52 patients whose tumors were evaluated by NGS, only the 2 patients whose tumors harbored FGFR fusions (FGFR3-TACC3) achieved partial responses (26). Thus, although early results with FGFR inhibitors in patients with FGFR fusions are promising, the current body of data suggest that the activity of these agents in the broader population of patients with other potentially oncogenic alterations in the FGFR pathway (receptor amplification, ligand activation) is likely modest at best. One possible explanation is that the presence of an FGFR fusion may be more indicative of pathway addiction and thus drug sensitivity than other potentially oncogenic FGFR alterations. In addition, intrapatient heterogeneity (27) and the presence of concurrent genomic drivers may also contribute. More potent and selective FGFR inhibitors, including those that are isoform-selective, combinations of targeted agents, or mAbs may be required to achieve maximal antitumor effects.

Although our study met its primary PFS endpoint, the overall activity of nintedanib was similar to the antitumor activity reported for other VEGFR2 inhibitors in patients with esophagogastric cancer. Specifically, the clinical benefit of nintedanib was modest at best and thus, we do not believe that the agent warrants further development in esophagogastric cancer. Our exploratory genomic biomarker analyses suggest that FGFR alterations are not predictive of clinical outcome in patients treated with multi-targeted kinase inhibitors such as nintedanib. We acknowledge that a clear limitation in our study is the small sample size of 5 patients with FGFR2 alterations to evaluate benefits specifically in FGFR2-amplified tumors. The study was also not designed to obtain multiple tumor biopsies or to capture circulating tumor DNA correlative data to evaluate for tumor heterogeneity over time. Combining nintedanib with chemotherapy may be a better approach to address our hypothesis.

Alterations in cell-cycle pathway genes were associated with poor outcome and may be prognostic, although we cannot exclude the possibility that they are predictive of therapeutic resistance to nintedanib. As prospective molecular tumor profiling is becoming increasingly routine in esophagogastric cancer (8), future development of targeted novel agents should be based upon compelling preclinical data and trials thereof should incorporate prospective molecular tumor profiling to ensure that all patients have adequate tumor mutational data for correlative analyses.

G.Y. Ku reports receiving commercial research grants from Arog, AstraZeneca, Bristol-Myers Squibb, and Merck. M.F. Berger is a consultant/advisory board member for Roche, and reports receiving commercial research support from Illumina. M. Schattner is a consultant/advisory board member for Boston Scientific. D.B. Solit reports receiving speakers bureau honoraria from Loxo Oncology, Pfizer, Vivideon Therapeutics, Lilly Oncology, and Illumina, and is a consultant/advisory board member for Loxo Oncology and Pfizer. Y.Y. Janjigian is a consultant/advisory board member for Bristol-Myers Squibb, Merck, Amgen, Boehringer Ingelheim, and Lilly, and reports receiving commercial research grants from Bristol-Myers Squibb, Eli-Lilly, Daiichi-Sankyo, and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Won, D.H. Ilson, Y.Y. Janjigian

Development of methodology: J.F. Chou, D.H. Ilson, N. Schultz, Y.Y. Janjigian

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Won, J.F. Hechtman, G.Y. Ku, S.B. Chalasani, M.S. Boyar, Z. Goldberg, A.M. Desai, L. Tang, D.P. Kelsen, M. Schattner, D.B. Solit, Y.Y. Janjigian

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Won, A. Basunia, W.K. Chatila, J.F. Chou, M.F. Berger, M. Capanu, D.B. Solit, N. Schultz, Y.Y. Janjigian

Writing, review, and/or revision of the manuscript: E. Won, A. Basunia, W.K. Chatila, J.F. Hechtman, J.F. Chou, G.Y. Ku, L. Tang, D.P. Kelsen, M. Schattner, D.H. Ilson, M. Capanu, D.B. Solit, N. Schultz, Y.Y. Janjigian

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Won, A. Basunia, Y. Tuvy, D.B. Solit, Y.Y. Janjigian

Study supervision: Y.Y. Janjigian

This work was funded in part by Boehringer Ingelheim Inc., the NCI Cancer Center Core grant P30 CA008748, the Robertson Foundation (to N. Schultz), and the Marie-Josée and Henry R. Kravis Center for Molecular Oncology. The authors gratefully acknowledge the members of the Molecular Diagnostics Service in the Department of Pathology at Memorial Sloan Kettering Cancer Center (New York, NY) for generating the MSK-IMPACT genomic profiling data. The authors thank Jessica Moore and Erin Patterson for outstanding editorial support. Afatinib was provided by Boehringer Ingelheim.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.