Revisiting MET: Clinical Characteristics and Treatment Outcomes of Patients with Locally Advanced or Metastatic, MET‐Amplified Esophagogastric Cancers

Abstract

Background

Metastatic esophagogastric cancers (EGCs) have a poor prognosis with an approximately 5% 5‐year survival. Additional treatment approaches are needed. c‐MET gene‐amplified tumors are an uncommon but potentially targetable subset of EGC. Clinical characteristics and outcomes were evaluated in patients with MET‐amplified EGC and compared with those without MET amplification to facilitate identification of these patients and possible treatment approaches.

Patients and Methods

Patients with locally advanced or metastatic MET‐amplified EGC at Massachusetts General Hospital (MGH) were identified using fluorescent in situ hybridization analysis, with a gene‐to‐control ratio of ≥2.2 defined as positive. Non–MET‐amplified patients identified during the same time period who had undergone tumor genotyping and treatment at MGH were evaluated as a comparison group.

Results

We identified 233 patients evaluated for MET amplification from 2002 to 2019. MET amplification was seen in 28 (12%) patients versus 205 (88%) patients without amplification. Most MET‐amplified tumors occurred in either the distal esophagus (n = 9; 32%) or gastroesophageal junction (n = 10; 36%). Of MET‐amplified patients, 16 (57%) had a TP53 mutation, 5(18%) had HER2 co‐amplification, 2 (7.0%) had EGFR co‐amplification, and 1 (3.5%) had FGFR2 co‐amplification. MET‐amplified tumors more frequently had poorly differentiated histology (19/28, 68.0% vs. 66/205, 32%; p = .02). Progression‐free survival to initial treatment was substantially shorter for all MET‐amplified patients (5.6 vs. 8.8 months, p = .026) and for those with metastatic disease at presentation (4.0 vs. 7.6 months, p = .01). Overall, patients with MET amplification had shorter overall survival (19.3 vs. 24.6 months, p = .049). No difference in survival was seen between low MET‐amplified tumors (≥2.2 and <25 MET copy number) compared with highly amplified tumors (≥25 MET copy number).

Conclusion

MET‐amplified EGC represents a distinct clinical entity characterized by rapid progression and short survival. Ideally, the identification of these patients will provide opportunities to participate in clinical trials in an attempt to improve outcomes.

Implications for Practice

This article describes 233 patients who received MET amplification testing and reports (a) a positivity rate of 12%, similar to the rate of HER2 positivity in this data set; (b) the clinical characteristics of poorly differentiated tumors and nodal metastases; and (c) markedly shorter progression‐free survival and overall survival in MET‐amplified tumors. Favorable outcomes are reported for patients treated with MET inhibitors. Given the lack of published data in MET‐amplified esophagogastric cancers and the urgent clinical importance of identifying patients with MET amplification for MET‐directed therapy, this large series is a valuable addition to the literature and will have an impact on future practice.

Introduction

Esophageal, gastric, and gastroesophageal (GE) junction cancers, collectively referred to as esophagogastric cancers (EGCs), have been increasing rapidly in incidence over the past 3 decades [1]. EGC is the third leading cause of cancer‐related death worldwide [2]. Survival with metastatic disease remains limited, with a median survival of approximately 12–14 months and a 5‐year overall survival rate of approximately 5% [3]. The backbone of frontline therapy for advanced disease remains 5‐fluorouracil in combination with a platinum agent, most commonly oxaliplatin in U.S. patients [4]. Taxane‐based cytotoxic chemotherapy, either alone or in combination with the antiangiogenic agent ramucirumab as second‐line therapy, is additionally supported by level 1 evidence, although treatment heterogeneity is common in Western patients [4–6].

Following a common paradigm in multiple tumor types, large molecular profiling efforts such as The Cancer Genome Atlas and the Asian Cancer Research Group have identified recurrent molecular alterations in EGC [7, 8]. Among molecular biomarkers, HER2 overexpression and/or amplification (15%–20%), microsatellite instability (3%–5%), and PD‐L1 positivity (50%–70%) are now linked with U.S. Food and Drug Administration approvals in patients with advanced disease [8, 9]. Other potentially actionable recurrent alterations include EGFR amplification, FGFR2b overexpression, and CLDN18.2 overexpression [10]. Despite preclinical rationale, none have yet had sufficient evidence of clinical activity to be approved by the U.S. Food and Drug Administration.

The cell surface receptor MET and its ligand, hepatocyte growth factor (HGF), play important roles in a number of important biological processes including normal cell migration, development, and growth [11] (Fig. 1). The HGF and MET receptor tyrosine kinase (RTK) pathway is normally tightly regulated. Physiologic MET tyrosine kinase receptor activation is mediated by HGF binding, which leads to MET α and β chain dimerization, in turn causing autophosphorylation at intracellular amino acid residues Y1234 and Y1235 within the enzymatic kinase site. Subsequent receptor phosphorylation mediates the recruitment of signal transducers and adaptors that lead to multiple potential downstream pathways, including those involving RAS, phosphatidylinositol (3,4,5)‐triphosphate–mTOR, signal transducer and activator of transcription 3, and nuclear factor‐κB [11, 12]. MET signaling, however, can be subverted by inappropriate autocrine loops, ligand/receptor overexpression/gene amplification, or activating mutations in MET, which are found in a number of human cancers [13, 14]. A number of cancers have been found to have MET gene amplification or mutations, including hereditary and sporadic papillary renal carcinoma, non‐small cell lung cancer (NSCLC), esophagogastric cancer, hepatocellular cancer, head and neck cancer, ovarian carcinoma, small cell lung cancer, and glioma [13, 15, 16]. MET gene amplification is a driver in a subset of esophagogastric cancers and NSCLCs [17–20]. Increased MET protein expression has been correlated with advanced disease stage and poor prognosis in several cancers, including colorectal cancer [21–23]. In addition to direct effects of HGF and MET on multiple biological processes in tumor cells, the HGF/MET axis can affect other cells such as those involved in immune function and potential response to cancers as well as tumor endothelial cells, thereby affecting the tumor microenvironment and promoting angiogenesis [13, 24] (Fig. 1). The HGF/MET pathway is also increasingly implicated in drug resistance, particularly in EGFR‐mutant NSCLCs that possess de novo or acquired resistance to small‐molecule inhibitors of the epidermal growth factor receptor (EGFR) [18, 25, 26].

Because of the widespread role that dysregulated MET plays in a variety of cancers, MET has been an attractive target for the development of inhibitors. In vitro, gastric cancer cell lines possessing MET amplification demonstrated dramatic sensitivity to the effects of a preclinical inhibitor of MET [17]. This in vitro observation was recapitulated in a phase I clinical trial of crizotinib, a potent inhibitor of MET as well as ALK and ROS1 tyrosine kinases, when several patients with MET‐amplified EGC experienced disease shrinkage upon treatment with crizotinib [3, 27].

The optimal methods to assess MET in EGC continue to evolve, and immunohistochemistry and/or fluorescent in situ hybridization (FISH) remain the most common approaches [28, 29]. We previously published a small heterogenous series of MET‐amplified EGC highlighting the frequency and feasibility of FISH testing [3]. Subsequent clinical efforts to target MET in EGC have been met with largely negative results, although factors including patient selection, MET overexpression cutoffs, and optimal therapeutic agent are likely confounders [28–32]. We sought to build upon existing data and look more granularly at the MET‐amplified population at our tertiary center where routine MET testing began in 2002.

Materials and Methods

Study Population

We retrospectively identified cases of locally advanced or metastatic esophagogastric adenocarcinoma or squamous cell carcinoma at the Massachusetts General Hospital (MGH) from 2002 to 2019 for analysis. Key inclusion was prior MET testing performed by FISH at MGH. All work was conducted under an institutional review board–approved protocol.

Data Collection and Survival Analysis

Medical records for patients whose tumors were MET amplified and not amplified on the basis of MGH testing were reviewed to extract data on clinicopathologic characteristics and outcomes. Patients were deidentified, and baseline clinicopathologic characteristics, concurrent molecular abnormalities, sites of disease, first‐line chemotherapy regimens, and survival outcomes were recorded.

Pathology and Molecular Analyses

Hematoxylin and eosin–stained, formalin‐fixed, paraffin‐embedded tumor tissues obtained from surgical specimens and/or diagnostic biopsies of either metastatic or primary cancer sites were evaluated by gastrointestinal pathologists at MGH in order to establish the diagnosis of esophagogastric adenocarcinoma or squamous cell carcinoma.

MET gene copy number (CN) FISH was performed on formalin‐fixed paraffin‐embedded tissue by hybridization for MET and centromere (CEP) 7, as previously reported [33]. Where possible, amplification of HER2 and/or EGFR was performed on the same tissue sample using two separate hybridizations for MET/EGFR/CEP7 and HER2/CEP17. Details of hybridization have been described previously [3]. MET amplification (MET positive) was defined as a gene‐to‐CN control probe ratio (G:CN) of greater than or equal to 2.2 scored in 50 tumor nuclei as used, which was extrapolated from established HER2 criteria and has been used previously. Specifically, polysomy, high polysomy, and equivocal (G:CN) ratio (i.e., 1.8 to less than 2.2) were scored as negative for amplification (MET negative) [3]. MET‐amplified tumors were defined as those with MET/CEP7 ratios ranging from ≥2.2 to <25 (low amplification) and ≥25 (high amplification).

For most patients, additional tumor gene analyses were available using the same tissue specimens that had been subjected to MET FISH. The methodology for genotyping at MGH has been previously described by Dias‐Santagata et al. [33].

Statistical Analysis

The distribution of continuous variables was assessed using the Shapiro‐Wilk test and, because of nonnormality, summarized using medians and interquartile ranges. Comparisons were made using the Wilcoxon rank‐sum test. Categorical variables are summarized with frequencies and proportions and compared using either the chi‐square test or Fisher's exact test. Regression analysis was performed to find independent predictors. In the first stage, the univariate Cox regression model was used to identify the eligible predictors, which had a marginal association of 10% with the outcome (progression and death). In the second stage, a backward stepwise multiple Cox proportional hazard model was used to identify the independent predictors of mortality with entry probability of 0.1 and exit probability of 0.05.

Survival estimates were generated using Kaplan‐Meier methodology and compared among the various group using a log‐rank test. Overall survival (OS) was defined as the time from initial diagnosis to death. Patients who were alive at last contact were censored at this date. Progression‐free survival (PFS) was defined as the time from initial treatment (surgery, chemotherapy, and/or radiotherapy) to the earliest date of progression or death. Patients who were alive and progression‐free at last contact were censored at this date.

Two‐tailed p values less than .05 were considered statistically significant. All analyses were conducted using Stata 15.1 (StataCorp, College Station, TX).

Results

We identified 28 patients with MET‐amplified tumors and 205 patients with non–MET‐amplified tumors from the years 2002 to 2019. Baseline demographics and clinical characteristics are shown in Table 1. For the entire cohort, the majority of patients were male (n = 175, 75.0%). Mean age at diagnosis was 59.9 years. There was no difference seen in baseline characteristics between patients with or without MET amplification except anemia, which was more common in MET‐negative (nonamplified) patients (43% vs. 18.5%, p = .013). The most common tumor site evaluated for MET amplification was the primary disease site 211 (90.5%). This reflects the tendency to establish diagnosis through endoscopic biopsy of the primary site of disease, and therefore this tissue is more available for MET testing, rather than reflecting the incidence of MET amplification at primary versus metastatic disease sites. The location of the primary tumor was most commonly in the gastroesophageal junction or distal esophagus (Table 2). MET‐amplified tumors were more likely to be intestinal type (95.5% vs. 68.0%, p = .029) and poorly differentiated compared with non–MET‐amplified tumors (68.0% vs. 32.0%, p = .02; Table 2). Patients with MET amplification were more likely to present with bulky metastasis measuring >5 cm (n = 7, 25% vs. n = 22, 11%) than non–MET‐amplified patients (p = .05; supplemental online Table 1). There were no differences in sites of metastatic disease or stage at presentation. At the time of diagnosis, 53 (23%) patients presented with locally advance disease, whereas the majority (180, 77%) presented with metastatic disease (supplemental online Table 1). Two of the six MET‐positive patients with localized disease underwent resection, but both subsequently developed metastases.

There was no significant difference in the percentage of patients receiving adjuvant or neoadjuvant therapy for locally advanced disease. Five of 6 MET‐amplified patients received neoadjuvant chemotherapy, whereas 1 patient received neoadjuvant radiation (he refused concurrent chemotherapy); 44 of 47 non–MET‐amplified patients received adjuvant or neoadjuvant chemotherapy (supplemental online Table 2). Chemoradiation (with carboplatin/paclitaxel) was the most common regimen given in the neoadjuvant setting to patients with esophageal and gastroesophageal cancers who did not present with metastatic disease. Similarly, there was no significant difference in the percentage of patients who received at least one line of therapy for metastatic disease (21 of 22 MET‐amplified patients compared with 154 of 158 non–MET‐amplified patients). The most commonly used first‐line regimen for patients with metastatic disease was FOLFOX chemotherapy, used in over 50% of both MET‐amplified and non–MET‐amplified patients. A higher percentage of non–MET‐amplified patients with metastatic disease received second‐line chemotherapy (71%) compared with 50% of MET‐amplified patients with metastatic disease (supplemental online Table 2). Eighteen percent of MET‐amplified patients with metastatic disease versus 14% of non–MET‐amplified patients received trastuzumab, most commonly with FOLFOX in the first line. Compared with 17% of non–MET‐amplified patients, only 7% of MET‐amplified patient received immunotherapy, due at least in part to the fact that only six of the MET‐amplified patients were diagnosed in 2017 or later when data for the potential benefit of treatment with immune checkpoint inhibitor therapy became available.

Of patients who presented with metastatic disease, overall the primary site was the most common site of radiation (13%), followed by bone (1%), chest (1%), lymph node (1%), and other sites (1%; supplemental online Table 2).

Common Genetic Alterations in Signaling Pathways Associated with MET Amplification

Of the 233 patients who were tested for gene amplification, HER2 was the most common (17%) amplification found, followed by MET (12%) and EGFR (5%). Other genetic alterations in signaling pathways included TP53 (42%), CDKN2A (8%), PIK3CA (7%), KRAS (6%), and FGFR (3%; Table 3).

HER2 (n = 5, 18.0%) was the most common gene co‐amplified with MET. Four of the 5 patients had co‐amplification based on initial biopsy, whereas the fifth had amplification of MET seen on a biopsy after trastuzumab therapy for HER2‐amplified disease, which may have represented a resistance mechanism (insufficient tissue was available from the initial biopsy for testing to confirm this). One additional MET amplification was seen in a patient with a FGFR2 amplification after anti‐FGFR therapy, also likely representing a resistance mechanism as the pretreatment sample was negative for MET amplification. Mutations and/or amplifications of TP53, CDKN2A, PIK3CA, EGFR, KRAS, and FGFR2 were found in patients with MET amplification in 16 (57%), 3 (11.0%), 3 (11%), 2 (7.0%), 2 (7%), and 1 (3.5%) of patients, respectively. PD‐1/PD‐L1 expression was seen in 4 of 6 MET‐amplified patients and 19 of 34 non–MET‐amplified patients of the 40 total patients who were tested (routine testing for this did not start until 2017 when data suggesting pembrolizumab's benefit for patients with PD‐L1–positive tumors became available; Table 3).

Univariate and Multivariate Analyses

Univariate Cox proportional hazard analysis showed that Eastern Cooperative Oncology Group performance status (ECOG PS; hazard ratio [HR], 1.3; p = .05); history of gastroesophageal reflux disease (HR, 0.69; p = .034); metastasis at presentation (HR, 1.8; p = .004); metastases to bone (HR, 1.5; p = .005), liver (HR, 1.2; p = .002), and lymph nodes (HR, 1.1; p = .02); and family history of gastrointestinal (GI) cancer (HR, 1.5; p = .045) were significant prognostic factors for OS. MET amplification and MET copy number ≥ 25 were also poor prognostic factors for OS (HR, 1.8; p = .016 and HR, 1.4; p = .024, respectively).

On multivariate Cox proportional hazard analysis, ECOG PS ≥2 (HR, 1.5; 95% confidence interval [CI], 1.1–2.0; p = .002), presentation with pain (HR, 1.45; 95% CI, 1.0–2.0; p = .03), MET amplification (HR, 1.6; 95% CI, 1.0–2.6; p = .047), metastasis at presentation (HR, 2.0; 95% CI, 1.3–3.1; p = .001), and a family history of GI cancer (HR, 1.5; 95% CI, 1.0–2.3, p = .027) were all significantly associated with poor OS. Patients who were MET‐amplified (HR, 5.1; 95% CI, 2.0–12; p ≤ .001) and received radiation for primary or metastatic disease (HR, 1.6; 95% CI, 1.1–2.4; p = .008) were at higher risk of disease progression (supplemental online Table 3).

Genotype‐Specific Treatment Outcomes

Median PFS was substantially shorter for patients with MET amplification compared with no MET amplification for all patients (n = 233; 5.6 vs. 8.8 months; HR, 1.6; 95% CI, 1.1–2.3; p = .026; Fig. 2) and for those with metastatic disease at presentation (4.0 vs. 7.6 months; HR, 1.8; 95% CI 1.0–2.8; p = .010; supplemental online Fig. 1). When all patients were analyzed, MET‐amplified patients had a shorter median OS than non–MET‐amplified patients (19.3 vs. 24.6 months; HR, 1.6; 95% CI 1.0–2.5; p = .049), whereas for those who presented with metastatic disease, there was a trend toward shorter survival in the MET‐amplified patients (15.3 vs. 22.1 months; HR, 1.3; 95% CI, 0.7–2.4; p = .25; Fig. 3 and supplemental online Fig. 2).

There was no difference seen in the comparisons of OS between MET‐amplified and non–MET‐amplified patients when all patients (n = 233) or only those with adenocarcinoma (n = 227) after excluding squamous cell carcinoma (n = 5) and a patient with adenocarcinoma with neuroendocrine differentiation (n = 1) were analyzed. Among the whole cohort, OS was 19.3 versus 24.6 months (HR, 1.6; 95% CI, 1.0–2.5; p = .049), and for patients with adenocarcinoma, the results were similar, 19.3 versus 24.6 months (HR, 1.58; 95% CI, 0.9–2.5; p = .06), although because of the smaller numbers of patients the p value was no longer significant (Fig. 3 and supplemental online Fig. 3). Similarly, there was not a significant difference when only patients with metastatic disease overall versus metastatic patients with adenocarcinoma only were compared (supplemental online Figs. 3 and 4 ). The absence of a difference in survival seen is probably due to the small number of patients with small cell cancer (n = 5).

In order to determine whether the 11 patients who received MET‐directed therapy (AMG337 or crizotinib) affected the OS in MET‐amplified patients with metastatic adenocarcinoma, we determined the OS after censoring those patients at the time of starting MET‐directed therapy. Even though this produced a 1.9 month difference in OS (15.3 vs. 13.4 months), this was not statistically significant (15.3 vs. 22.1 months; HR, 1.4; 95% CI, 0.8–2.4; p = .22) compared with 13.4 versus 22.1 months (HR, 1.6; 95% CI, 0.8–3.2; p = .17) in the MET‐amplified and non–MET‐amplified groups after censoring (supplemental online Figs. 4 and 5).

There was no significant difference in PFS (4.0 months vs. 5.8 months; HR, 0.9; 95% CI, 0.41–2.0; p = .85) or OS (19.3 months vs. 15.3 months; HR, 0.9; 95% CI, 0.4–2.2; p = .86) between patients with lower level MET‐amplified tumors (≥2.2 and <25) versus higher amplified tumors (≥25 MET copy number; supplemental online Figs. 6 and 7). MET‐amplified tumors with co‐expression of EGFR (n = 2) amplification had a shorter mean OS (12.8 months vs. 25.3 months, numbers too small for statistical comparison) compared with non–MET‐amplified tumors with this amplification (n = 10). MET‐amplified tumors with initial co‐expression of HER2 (n = 5) also had a trend toward worse OS (19.3 vs. 26.2 months; HR, 2.3; 95% CI, 0.7–6.9; p = .13) compared with non–MET‐amplified HER2‐expressing tumors (n = 35), although this did not reach significance. However, these numbers should be interpreted cautiously because of the small sample size of patients with HER2 and EGFR co‐amplification.

Survival Among Patients Receiving MET‐Directed Therapy

Among the 28 patients with MET amplification, 11 patients received MET inhibitors at some point in their treatment. Six patients received the MET inhibitor AMG337, and five patients received crizotinib. A statistically significant improvement in median overall survival was seen among those patients receiving MET inhibitors compared with those who did not receive MET‐directed therapy (33.0 months vs. 10.9 months; HR, 0.3; 95% CI, 0.1–0.9; p = .021; Fig. 4). Because anti‐MET therapy was given in different lines of treatment, it was not possible to do a PFS comparison for patients who received MET‐directed therapy versus those who did not. Median duration of MET‐directed therapy was 9 months with a range from 2 to 82 months. Of patients who received MET‐directed therapy, the following mutations in genes involved in signaling pathways were seen: TP53 (7), CDKN2A (2), PIK3CA (1), KRAS (1), and ERBB2 (1). Amplifications were seen in EGFR (2), MYC (2), FGFR (1), and HER2 (1). The patients with co‐amplifications in a growth factor receptor did not respond to anti‐MET therapy. In two cases, the patients with the FGFR and HER2 amplifications, the MET overexpression was likely a resistance mechanism after prior treatment with an anti‐FGFR agent and trastuzumab, respectively. The latter patient also had a co‐amplification of EGFR at the time MET amplification was found, consistent with tumor heterogeneity and evidence that resistance development often involves more than one potential mechanism. The other patient with a concomitant EGFR amplification on the initially tested specimen also did not respond. The longest responding patient who had high MET amplification with a PIK3CA mutation was treated only with anti‐MET therapy from diagnosis and lived slightly over 7 years from start of therapy.

Response to Anti‐HER2 Therapy and Immunotherapy

HER2 expression was seen in 40 patients, and 33 of them received anti‐HER2 therapy (trastuzumab). Median OS was 27.3 months and 19.3 months (HR, 2.8; 95% CI, 0.9–9.2; p = .07) among non–MET‐amplified HER2+ and MET‐amplified HER2+ patients, respectively (supplemental online Fig. 8).

PD‐L1 was tested in the six MET‐amplified patients diagnosed in 2017 or later when the data for a role for immunotherapy in patients with PD‐L1–expressing tumors was reported (Table 3). Of the four patients who were positive (three with a score of 1 and one with a score of 13), two ultimately received immunotherapy, whereas the other two were treated with MET‐directed therapy after progressing on standard therapy. One of the two treated with immunotherapy had evidence of minor response but ultimately progressed, and one progressed rapidly after 9 weeks of treatment. There was no difference in OS (24.6 months vs. 27.0 months, respectively) between the non–MET‐amplified and MET‐amplified patients who received immunotherapy.

Discussion

In this retrospective study, we evaluated the clinicopathological features and outcomes of patients with MET‐amplified and non–MET‐amplified EGCs.

The difference in first‐line median PFS between patients with non–MET‐amplified versus MET‐amplified EGCs was statistically significant, confirming previous observations that MET‐amplified EGCs represent a particularly aggressive subtype of an inherently aggressive disease [3, 34, 35]. Interestingly, the difference in OS between the groups is less striking than the difference in PFS. This may reflect the inclusion in this analysis of several patients with MET‐amplified cancers who participated in a phase I clinical trial of AMG337, a highly selective and potent small‐molecule inhibitor of MET receptor signaling. As previously reported, the response rate of MET‐amplified patients on this early phase trial was 29.6%, and several of these patients received AMG337 on protocol for over 2 years, which may have skewed the OS survival analysis for the MET‐amplified cohort [36]. No difference in OS was seen between those with high MET amplification, ≥25, versus lower level amplification, ≥2.2 to <25. This is consistent with the findings from the RILOMET‐1 study [37]. There are a number of potential explanations for this, including that there is a threshold level for MET amplification that determines overall biology or possibly that the highly amplified tumors were the ones more likely to have been treated on one of the clinical trials (crizotinib or AMG337) or to have responded to anti‐MET therapy. The numbers of patients are too small to distinguish between these possibilities.

MET gene amplification in EGC is a relatively uncommon event. Most series report MET amplification in 2% to 10% of tumors, although the frequency depends on the testing methodology, specific population tested, histology, and the proximity of the primary tumor to the GE junction (3,35,36). In the current study, MET amplification was observed in 12% (28 of 233) of patients. This slightly higher percentage may be due to a number of factors, including defining the lower cutoff of MET amplification at a MET/CEP7 ratio of 2.2 or higher and the fact that there likely was some selection bias in which patients were screened for MET amplification (such as those with tumors near to the GE junction), which was a requirement for inclusion in this analysis, especially in the earlier years of testing. In addition, two of the MET‐amplified patients were tested at time of progression on targeted therapy (one after trastuzumab therapy for HER2 amplification and one after anti‐FGFR therapy for FGFR2 amplification), and in both cases the MET amplification may have been a resistance mechanism to the previous therapies. In fact, the pretreatment sample for the patient with the FGFR amplification did not show MET amplification. Activation of the MET pathway has been identified as a reasonably common mechanism for development of resistance to therapies targeting growth factor pathways in cancer [38]. If these two are excluded, then 11.2% (26 of 233) of the patients had de novo MET amplification.

We observed co‐occurrence of a number of other amplified or mutated genes important for cell signaling or cell cycle regulation (e.g., TP53, CDKN2A, EGFR, HER2, and PIK3CA) with MET amplification, with HER2 and/or EGFR co‐amplification seen in approximately 20% of patients on pretreatment biopsies. Co‐amplified patients tended to have aggressive disease with shorter survival times when compared with patients without MET amplification with these amplifications, although these did not reach statistical significance. Because tumors with co‐amplified or mutant protein (whether these are present in the pretreated tumor or develop as resistant mechanisms to targeting one of the proteins) may be resistant to inhibition of signaling through just one of the pathways, combinations of anti‐MET–directed therapy and therapy targeting the other amplified (e.g. HER2, EGFR, or FGFR) or mutant proteins should continue to be investigated for this subgroup of patients in future clinical trials [39, 40].

The finding that a family history of gastrointestinal malignancy was associated with decreased OS on the multivariate analysis is consistent with some but not all previous studies that have addressed this question, although all of these addressed the question of family history of upper gastrointestinal malignancies specifically (as well as in one case also family history of any cancer) and not a family history of all gastrointestinal cancers [41–44] .The reason for this is not certain. When we analyzed them separately, there was not a statistically significant difference in OS for patients with family history versus those without for either those patients with locally advanced disease or those with metastatic disease at presentation (data not shown). Given that this was a retrospective database, there was no specific information on germline analysis for most patients with positive family history of gastrointestinal cancer to help address this on the genetic level. However, because the majority of these patients were treated before immunotherapy was available, any benefit from that for those with Lynch syndrome would not have been seen. Given the variability in findings from different studies, the question of whether family history of gastroesophageal cancer, gastrointestinal cancer, or any cancer is associated with prognosis of patients with gastroesophageal cancers with locally advanced or metastatic disease needs to be addressed in a larger prospective study in the current era when immunotherapy is used for those patients with microsatellite instability–high tumors.

An important observation is the median OS of 22.1 months in the MET‐negative population with metastatic disease observed in our cohort. Although this may be influenced by several factors, including selection bias of only those patients well enough to be considered for tumor genetic testing, it is also likely affected by the availability of ancillary services (nutrition, palliative care, etc.) and multidisciplinary management in dedicated EGC programs at tertiary centers [45].

Conclusion

Here we further refine the understanding of MET‐amplified EGC and highlight treatment implications. Therapeutically exploiting molecular biomarkers, including MET, in EGC has been plagued by multiple barriers, among the most important of which is tumoral heterogeneity. Despite, encouraging phase II studies of c‐MET antibodies (rilotumumab, onartuzumab), larger phase III EGC studies using the same screening criteria for c‐MET positivity revealed no survival benefits, regardless of c‐MET staining [46]. Larger efforts to assess for concordance between tissue (including both primary and metastatic sites) and blood (circulating tumor DNA) in order to evaluate the effect of tumor heterogeneity on response, as well as the impact of other molecular features, such as the presence of RTK co‐amplification, may aid in refining the population who are most likely to benefit from MET‐directed therapies, either alone or in combination. Small trials with novel MET inhibitors and combination strategies with robust correlative work are likely the best path forward in revisiting MET as a therapeutic target in EGC [47].

Acknowledgments

We thank all our team and especially Carissa Desilus, Ashley Lester, and Ryan Frazier for excellent technical assistance. J.W.C. acknowledges partial funding by Benz (P30CA06516, 9/1/2009 to 8/31/2011, Role: Investigator) and the NCI‐ASCO Clinical Investigator Team Leadership Supplemental Award.

Author Contributions

Surendra Pal Chaudhary, Eunice L. Kwak, Katie L. Hwang, Jochen K. Lennerz, Ryan B. Corcoran, Rebecca S. Heist, Andrea L. Russo, Aparna Parikh, Darrell R. Borger, Lawrence S. Blaszkowsky, Jason E. Faris, Janet E. Murphy, Christopher G. Azzoli, Eric J. Roeland, Lipika Goyal, Jill Allen, John T. Mullen, David P. Ryan, A. John Iafrate, Samuel J. Klempner, Jeffrey W. Clark, Theodore S. Hong

Conception/design: Surendra Pal Chaudhary, Eunice L. Kwak, Jeffrey W. Clark, Theodore S. Hong

Provision of study material or patients: Eunice L. Kwak, Katie L. Hwang, Jochen K. Lennerz, Ryan B. Corcoran, Rebecca S. Heist, Aparna Parikh, Darrell R. Borger, Lawrence S. Blaszkowsky, Jason E. Faris, Janet E. Murphy, Christopher G. Azzoli, Eric J. Roeland, Lipika Goyal, Jill Allen, John T. Mullen, David P. Ryan, A. John Iafrate, Samuel J. Klempner, Jeffrey W. Clark, Theodore S. Hong

Collection and/or assembly of data: Surendra Pal Chaudhary, Eunice L. Kwak, Katie L. Hwang, Jochen K. Lennerz, Darrell R. Borger, A. John Iafrate, Jeffrey W. Clark, Theodore S. Hong

Data analysis and interpretation: Surendra Pal Chaudhary, Eunice L. Kwak, Katie L. Hwang, Samuel J. Klempner, Jeffrey W. Clark, Theodore S. Hong

Manuscript writing: Surendra Pal Chaudhary, Eunice L. Kwak, Katie L. Hwang, Jochen K. Lennerz, Ryan B. Corcoran, Rebecca S. Heist, Andrea L. Russo, Aparna Parikh, Darrell R. Borger, Lawrence S. Blaszkowsky, Jason E. Faris, Janet E. Murphy, Christopher G. Azzoli, Eric J. Roeland, Lipika Goyal, Jill Allen, John T. Mullen, David P. Ryan, A. John Iafrate, Samuel J. Klempner, Jeffrey W. Clark, Theodore S. Hong

Final approval of manuscript: Surendra Pal Chaudhary, Eunice L. Kwak, Katie L. Hwang, Jochen K. Lennerz, Ryan B. Corcoran, Rebecca S. Heist, Andrea L. Russo, Aparna Parikh, Darrell R. Borger, Lawrence S. Blaszkowsky, Jason E. Faris, Janet E. Murphy, Christopher G. Azzoli, Eric J. Roeland, Lipika Goyal, Jill Allen, John T. Mullen, David P. Ryan, A. John Iafrate, Samuel J. Klempner, Jeffrey W. Clark, Theodore S. Hong

Disclosures

Eunice L. Kwak: Novartis (E); Ryan B. Corcoran: Abbvie, Amgen, Array Biopharma/Pfizer, Astex Pharmaceuticals, AstraZeneca, Avidity Biosciences, Bristol‐Myers Squibb, Chugai, Elicio, Fog Pharma, Genentech, Guardant Health, Ipsen, LOXO, Merrimack, Natera, N‐of‐one/Qiagen, Novartis, Revolution Medicines, Roche, Roivant, Shionogi, Shire, Spectrum Pharmaceuticals, Symphogen, Taiho, Warp Drive Bio, Zikani Therapeutics (C/A), C4 Therapeutics, Fount Therapeutics/Kinnate Biopharma, nRichDx (SAB), Avidity Biosciences, C4 Therapeutics, Fount Therapeutics/Kinnate Biopharma, nRichDx, Revolution Medicines (OI), Asana, AstraZeneca, Eli Lilly & Co. (RF); Rebecca S. Heist: Boehringer Ingelheim, Tarveda, Apollomics, Novartis (C/A), Novartis, Abbvie, Daichii Sankyo, Agios, Corvus, Genentech Roche, Mirati, Exelixis, Eli Lilly & Co. (RF—institution); Aparna Parikh: Eli Lilly & Co. (C/A), Puretech, Natera (other—ad board), Eli Lilly & Co., Array, Guardant, Bristol‐Myers Squibb, Tesaro (RF—institution); Darrell R. Borger: Takeda Pharmaceuticals (E); Lawrence S. Blaszkowsky: Pfizer (OI); Jason E. Faris: Novartis (E, OI), Immunitas (C/A), H3B Therapeutics, Incyte (RF); Eric J. Roeland: Asahi Kasei Pharmaceuticals, DRG Consulting, Napo Pharmaceuticals, American Imaging Management, Immuneering Corporation, Prime Oncology (C/A), Heron Pharmaceuticals, Vector Oncology (SAB), Oragenics, Inc, Galera Pharmaceuticals, Enzychem Lifesciences Pharmaceutical Company (other—data safety monitoring board); Lipika Goyal: Taiho, Incyte, Debiopharm, Agios, Klus, Alentis, Pieres Pharmaceuticals, QED, H3Biomedicine, SIRTEX (SAB), QED, Alentis (C/A), AstraZeneca (other—data monitoring committee); David P. Ryan: MPM Capital, Acworth Pharmaceuticals, Thrive Earlier Detection (OI), MPM Capital, Oncorus, Gritstone Oncology, Maverick Therapeutics, 28/7 Therapeutics (C/A); A. John Iafrate: Archer Dx (OI), Repare Bio (C/A); Samuel J. Klempner: Eli Lilly & Co., Merck, Bristol‐Myers Squibb, Boston Biomedical, Foundation Medicine Inc, Pieris (C/A), Turning Point Therapeutics (OI); Theodore S. Hong: Synthetic Biologics, Novocure, Merck (C/A), Taiho, Ipsen, Bristol‐Myers Squibb, Tesaro, AstraZeneca (RF). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board

References

Author notes