Gastroesophageal adenocarcinoma (GEA) is a lethal disease where targeted therapies, even when guided by genomic biomarkers, have had limited effi cacy. A potential reason for the failure of such therapies is that genomic profi ling results could commonly differ between the primary and metastatic tumors. To evaluate genomic heterogeneity, we sequenced paired primary GEA and synchronous metastatic lesions across multiple cohorts, fi nding extensive differences in genomic alterations, including discrepancies in potentially clinically relevant alterations. Multiregion sequencing showed signifi cant discrepancy within the primary tumor (PT) and between the PT and disseminated disease, with oncogene amplifi cation profi les commonly discordant. In addition, a pilot analysis of cell-free DNA (cfDNA) sequencing demonstrated the feasibility of detecting genomic amplifi cations not detected in PT sampling. Lastly, we profi led paired primary tumors, metastatic tumors, and cfDNA from patients enrolled in the personalized antibodies for GEA (PANGEA) trial of targeted therapies in GEA and found that genomic biomarkers were recurrently discrepant between the PT and untreated metastases. Divergent primary and metastatic tissue profi ling led to treatment reassignment in 32% (9/28) of patients. In discordant primary and metastatic lesions, we found 87.5% concordance for targetable alterations in metastatic tissue and cfDNA, suggesting the potential for cfDNA profi ling to enhance selection of therapy. SIGNIFICANCE:We demonstrate frequent baseline heterogeneity in targetable genomic alterations in GEA, indicating that current tissue sampling practices for biomarker testing do not effectively guide precision medicine in this disease and that routine profi ling of metastatic lesions and/or cfDNA should be systematically evaluated. Cancer Discov; 8(1);[37][38][39][40][41][42][43][44][45][46][47][48]

Purpose: Gastroesophageal adenocarcinoma (GEA) has a poor prognosis and few therapeutic options. Utilizing a 73gene plasma-based next-generation sequencing (NGS) cellfree circulating tumor DNA (ctDNA-NGS) test, we sought to evaluate the role of ctDNA-NGS in guiding clinical decisionmaking in GEA. Experimental Design: We evaluated a large cohort (n ¼ 2,140 tests; 1,630 patients) of ctDNA-NGS results (including 369 clinically annotated patients). Patients were assessed for genomic alteration (GA) distribution and correlation with clinicopathologic characteristics and outcomes. Results: Treatment history, tumor site, and disease burden dictated tumor-DNA shedding and consequent ctDNA-NGS maximum somatic variant allele frequency. Patients with locally advanced disease having detectable ctDNA postoperatively experienced inferior median disease-free survival (P ¼ 0.03). The genomic landscape was similar but not identical to tissue-NGS, reflecting temporospatial molecular heterogeneity, with some targetable GAs identified at higher frequency via ctDNA-NGS compared with previous primary tumor-NGS cohorts. Patients with known microsatellite instabilityhigh (MSI-High) tumors were robustly detected with ctDNA-NGS. Predictive biomarker assessment was optimized by incorporating tissue-NGS and ctDNA-NGS assessment in a complementary manner. HER2 inhibition demonstrated a profound survival benefit in HER2-amplified patients by ctDNA-NGS and/or tissue-NGS (median overall survival, 26.3 vs. 7.4 months; P ¼ 0.002), as did EGFR inhibition in EGFR-amplified patients (median overall survival, 21.1 vs. 14.4 months; P ¼ 0.01). Conclusions: ctDNA-NGS characterized GEA molecular heterogeneity and rendered important prognostic and predictive information, complementary to tissue-NGS. See related commentary by Frankell and Smyth, p. 6893

Previous anti-EGFR trials in unselected gastroesophageal adenocarcinoma (GEA) patients were resoundingly negative. We identified EGFR amplification in 5% (19/363) of patients at the University of Chicago, including 6% (8/140) who were prospectively screened with intention-to-treat using anti-EGFR therapy. Seven pts received >1 dose of treatment: three first line FOLFOX plus ABT-806, one second line FOLFIRI plus cetuximab, and three third/fourth line cetuximab alone. Treatment achieved objective response in 58% (4/7) and disease control in 100% (7/7) with a median progression-free survival of 10 months. Pre and post-treatment tumor NGS, serial plasma ctDNA NGS, and tumor IHC/FISH for EGFR revealed pre-existing and/or acquired genomic events including EGFR negative clones, PTEN deletion, KRAS amplification/mutation, NRAS, MYC and HER2 amplification, and GNAS mutations serving as mechanisms of resistance. Two evaluable patients demonstrated interval increase of CD3+ infiltrate, including one who demonstrated increased NKp46+, and PD-L1 IHC expression from baseline, suggesting an immune therapeutic mechanism of action. EGFR amplification predicted benefit from anti-EGFR therapy, albeit until various resistance mechanisms emerged. Statement of Significance: This paper highlights the role of EGFR inhibitorsin EGFR amplified GEA -despite negative results in prior unselected phase III trials. Using serial ctDNA and tissue NGS, we identified mechanisms of primary and acquired resistance in all patients, as well as potential contribution of antibody-dependent cell-mediated cytotoxicity (ADCC) to their clinical benefit.

The 1-year and median overall survivals (mOS) of advanced gastroesophageal adenocarcinomas (GEA) are ~50% and <12 months. Baseline spatial and temporal molecular heterogeneity of targetable alterations may be a cause of failure of targeted/immunooncologic therapies. This heterogeneity, coupled with infrequent incidence of some biomarkers, has resulted in stalled therapeutic progress. We hypothesized that a personalized treatment strategy, applied at first diagnosis then serially over up to three treatment lines using monoclonal antibodies combined with optimally-sequenced chemotherapy, could contend with these hurdles. This was tested using a novel clinical expansion platform type-II design with a survival primary endpoint. Of 68 patients by intention-to-treat, 1-year survival was 66% and mOS was 15.7 months, meeting the primary efficacy endpoint (one-sided p=0.0024). First-line response rate (74%), disease control rate (99%), and median progression free survival (8.2 months) were superior to historical controls. The PANGEA strategy led to improved outcomes warranting a larger randomized study. Significance This study highlights excellent outcomes achieved by individually optimizing chemotherapy, biomarker profiling, and matching of targeted therapies at baseline and over time for GEA. Testing a predefined treatment strategy resulted in improved outcomes versus historical controls. Therapeutic resistance observed in correlative analyses suggests that dual targeted inhibition may be beneficial. Research.

IMPORTANCE Patients with locally advanced gastroesophageal adenocarcinoma (ie, stage ՆT3 and/or node positive) have high rates of recurrence despite surgery and adjunctive perioperative therapies, which also have high toxicity profiles. Evaluation of pharmacogenomically dosed perioperative gFOLFIRINOX (fluorouracil, leucovorin, oxaliplatin, and UGT1A1 genotype-directed irinotecan) to optimize efficacy while limiting toxic effects may have value. OBJECTIVE To evaluate the coprimary end points of margin-negative (R0) resection rates and pathologic response grades (PRGs) of gFOLFIRINOX therapy among patients with locally advanced gastroesophageal adenocarcinoma. DESIGN, SETTING, AND PARTICIPANTS This single-group phase 2 trial, conducted at 2 academic medical centers from February 2014 to March 2019, enrolled 36 evaluable patients with locally advanced adenocarcinoma of the esophagus, gastroesophageal junction, and gastric body. Data analysis was conducted in May 2019. INTERVENTIONS Patients received biweekly gFOLFIRINOX (fluorouracil, 2400 mg/m 2 over 46 hours; oxaliplatin, 85 mg/m 2 ; irinotecan, 180 mg/m 2 for UGT1A1 genotype 6/6, 135 mg/m 2 for UGT1A1 genotype 6/7, or 90 mg/m 2 for UGT1A1 genotype 7/7; and prophylactic peg-filgrastim, 6 mg) for 4 cycles before and after surgery. Patients with tumors positive for ERBB2 also received trastuzumab (6-mg/kg loading dose, then 4 mg/kg). MAIN OUTCOMES AND MEASURES Margin-negative resection rate and PRG. RESULTS A total of 36 evaluable patients (27 [78%] men; median [range] age, 66 [27-85] years; 10

Background The randomized phase III RAINBOW trial established paclitaxel (pac) plus ramucirumab (ram) as a global standard for second‐line (2L) therapy in advanced gastric and gastroesophageal junction adenocarcinoma, together gastroesophageal adenocarcinoma (GEA). Patients (pts) receiving first‐line (1L) FOLFOX often develop neuropathy that renders continued neurotoxic agents in the 2L setting unappealing and other regimens more desirable. As such, FOLFIRI‐ram has become an option for patients with 2L GEA. FOLFIRI‐ramucirumab (ram) has demonstrated safety and activity in 2L colorectal cancer, but efficacy/safety data in GEA are lacking. Subjects, Materials, and Methods Patients with GEA treated with 2L FOLFIRI‐ram between August 2014 and April 2018 were identified. Clinicopathologic data including oxaliplatin neurotoxicity rates/grades (G), 2L treatment response, progression‐free survival (PFS), overall survival (OS), safety, and molecular features were abstracted from three U.S. academic institutions. Kaplan‐Meier survival analysis was used to generate PFS/OS; the likelihood ratio test was used to determine statistical significance. Results We identified 29 pts who received 2L FOLFIRI‐ram. All pts received 1L platinum + fluoropyrimidine, and 23 of 29 (79%) had post‐1L neuropathy; 12 (41%) had G1, and 11 (38%) had G2. Patients were evenly split between esophagus/gastroesophageal junction (12; 41%) and gastric cancer (17; 59%). Among evaluable pts (26/29), the overall response rate was 23% (all partial response) with a disease control rate of 79%. Median PFS was 6.0 months and median OS was 13.4 months among all evaluable pts. Six‐ and 12‐month OS were 90% (n = 18/20) and 41% (n = 7/17). There were no new safety signals. Conclusion We provide the first data suggesting FOLFIRI‐ram is a safe, non‐neurotoxic regimen comparing favorably with the combination of pac + ram used in the seminal RAINBOW trial. Implications for Practice Results of this study provide initial support for the safety and efficacy of second‐line (2L) FOLFIRI‐ramucirumab (ram) after progression on first‐line platinum/fluoropyrimidine in patients with gastroesophageal adenocarcinoma (GEA). The overall response, progression‐free survival, overall survival, and toxicity profile compare favorably with paclitaxel (pac) + ram and highlight the importance of the ongoing phase II RAMIRIS trial examining FOLFIRI‐ram versus pac + ram in 2L GEA (NCT03081143). FOLFIRI‐ram may warrant consideration for inclusion as an alternate regimen in consensus guidelines for GEA.

PURPOSE Robust institutional tumor banks depend on continuous sample curation or else subsequent biopsy or resection specimens are overlooked after initial enrollment. Curation automation is hindered by semistructured free-text clinical pathology notes, which complicate data abstraction. Our motivation is to develop a natural language processing method that dynamically identifies existing pathology specimen elements necessary for locating specimens for future use in a manner that can be re-implemented by other institutions. PATIENTS AND METHODS Pathology reports from patients with gastroesophageal cancer enrolled in The University of Chicago GI oncology tumor bank were used to train and validate a novel composite natural language processing-based pipeline with a supervised machine learning classification step to separate notes into internal (primary review) and external (consultation) reports; a named-entity recognition step to obtain label (accession number), location, date, and sublabels (block identifiers); and a results proofreading step. RESULTS We analyzed 188 pathology reports, including 82 internal reports and 106 external consult reports, and successfully extracted named entities grouped as sample information (label, date, location). Our approach identified up to 24 additional unique samples in external consult notes that could have been overlooked. Our classification model obtained 100% accuracy on the basis of 10-fold cross-validation. Precision, recall, and F1 for class-specific named-entity recognition models show strong performance. CONCLUSION Through a combination of natural language processing and machine learning, we devised a re-implementable and automated approach that can accurately extract specimen attributes from semistructured pathology notes to dynamically populate a tumor registry.

TPS198 Background: Targeted therapies (tx) in GEA have had limited efficacy despite recognition of numerous ‘targetable’ molecular events. This may be due to the molecular heterogeneity (MH) that exists between patients (pts), within the primary tumor (PT), between PT and synchronous metastatic lesions (MLs), and in lesions over time. Current biomarker profiling (BP) is performed on one site, usually the PT, yet this fails to capture the MH of GEA, with likely major clinical implications. Classic trial designs are challenged by MH, low frequency oncogenic drivers, and scarcity of tissue. There is need for novel trial designs and BP technologies that address these concerns, provide tx algorithms for pts with multiple aberrations, and have access to several txs. Methods: This phase IIa, open-label, non-randomized ‘platform trial’ enrolls pts with newly diagnosed metastatic GEA or recurrent disease after curative-intent surgery. Baseline tumor BP is performed on PT/ML along with circulating free (cf)DNA. Pts receive first line (1L) mFOLFOX6 + biologic tx based on BP of the ML using a prioritized tx algorithm (HER2+: trastuzumab; MET+: none; FGFR2+: none; EGFR+: ABT806; MSI-H: nivolumab; ‘RAS-like’: ramucirumab). MET/FGFR2 arms (~10% of all pts) are tx’d with cytotoxics only and followed for natural outcome until/if tx becomes available on study. At first progression (PD1), pts undergo biopsy of growing ML and change to 2L FOLFIRI + biologic agent as assigned in 1L tx. Upon results of PD1 biopsy, pts change to a new biologic tx if the molecular category evolves. At PD2, pts change to 3L FOLTAX + biologic as determined after PD1, and switch biologic tx from PD2 biopsy result. All PD1/PD2 tumor/cfDNA samples undergo BP to assess evolution and resistance mechanisms. Co-primary endpoints: safety, feasibility, and overall survival (OS) of this personalized treatment strategy (excluding MET/FGFR2) compared to historical controls (HR 0.66). Secondary endpoints include rate of baseline MH between PT and ML leading to new treatment assignment; utility of cfDNA; overall progression-free survival (PFS)/response rate (RR); OS/PFS/RR in each targetable group. Since 8/2015, 38 of 68 planned pts have been accrued. Clinical trial information: NCT02213289.