Background: Gastroesophageal cancer (GEC) is a global health problem. Multiple phase III anti-EGFR trials of unselected metastatic GEC patients (pts) were negative, yet subset analyses suggested potential benefit in pts with EGFR amplification (amp). We sought to quantify the incidence of EGFR amp in metastatic/recurrent GEC pts compared to the locally advanced TCGA rate, treat them with EGFR antagonists, quantify response/duration, and assess for mechanisms of action and resistance (MOA, MOR). Methods: 106 stage IV GEC pts in any therapy (tx) line (L) were prospectively screened for EGFR amp between 9/1/14-8/31/16 at a single site using tumor next generation sequencing (NGS). EGFR amp samples were then assessed by FISH, immunohistochemistry (IHC) and mass spectroscopy (MS) to confirm. Pts received 1L ABT-806 (1/4) or 2L+ cetux (3/4) monoclonal anti-EGFR antibodies, in combination with chemotherapy if 1L or 2L (2/4), or as monotherapy for 3L+ (2/4). Pts were followed for response (ORR) and disease control (DCR). Pre and post-treatment NGS, serial ctDNA, and EGFR/PD-L1 FISH/IHC/MS were obtained to monitor for MOA and MOR to tx. Results: Six (5.7%) pts harbored EGFR amp tumors with copies/cell ranging 54-167. Four eligible pts received >1 dose of tx with monoclonal anti-EGFR therapy: 1 FOLFOX-ABT-806 (1L), 1 FOLFIRI-cetux (2L), and 2 cetux alone (3L, 4L); ORR was 100% (4/4) by RECIST, respectively (-70%, -65%, -77%, -43%). All 3 cetux treated pts developed classic rash, which disappeared when/if progression (2/3). The 4-month DCR was 75% (3/4). No pts showed PD-L1 staining pre/post therapy. Upon progression, the 2L pt acquired de novo PTEN exon 6 deletion after 10 months, while basal NRAS mutated and HER2 amp ctDNA clones had expanded in the 4L pt after 4 months of monotherapy. Conclusions:EGFR amp incidence was similar to TCGA. Despite large negative trials with EGFR antagonists for GEC, EGFR amp predicted benefit (100%) from anti-EGFR tx, including monotherapy in 50%, albeit until various resistance mechanisms emerged, including HER2, RAS, and PTEN/AKT pathway activation. No PD-L1 expression was observed pre/post treatment, consistent with our previous data showing that EGFR-amplified cases trend towards a ‘non-T cell-inflamed’ phenotype. Clinical trial information: NCT02213289