Biomarker prevalence in US patients with gastric/gastroesophageal junction adenocarcinoma (GEA) from real-world data.

Abstract

444 Background: The therapeutic landscape for gastric (GA)/ gastroesophageal junction (GEJ) adenocarcinoma (GEA) has advanced with biomarker (BM)-targeted therapies. c-Met protein (also known as MET protein) is a potential target for antibody-drug conjugates; however, the prevalence and overlap of MET protein with key BMs in GEA are not well understood. Detailed molecular profiling from patients in real-world (RW) settings can inform patient selection and treatment strategies. Methods: We conducted a retrospective analysis of 1819 patient samples (1081 GEJ and 738 GA), with either treatment-naïve (70%) or treatment-experienced (30%) samples, using the Caris and linked ConcertAI Translational360 database to assess BM prevalence. Leveraging high concordance between c-Met immunohistochemistry (IHC) and RNA, we defined MET mRNA enriched tissue samples as the top quartile (Q1) of samples with the highest expression. Using IHC or transcriptomics, we assessed Jaccard overlap of MET Q1 with human epidermal growth factor receptor 2 (HER2), programmed death-ligand 1 (PD-L1) with combined positive scores of ≥1 or ≥5 (CPS1+ or CPS5+), claudin 18 (CLDN18), carcinoembryonic antigen-related cell adhesion molecule 5 (CEACAM5) , and fibroblast growth factor receptor 2 (FGFR2) . Due to high concordance between CLDN18 IHC and RNA, we used Youden index analyses to predict enriched samples. Results: Consistent with previous literature, HER2 positivity (+), PD-L1 CPS1+, PD-L1 CPS5+ and CLDN18+ were detected in 149 (11%), 790 (76%), 347 (33%), and 752 (41.3%) of BM-evaluable samples, respectively, by IHC. There was no substantial over- or under-representation of MET Q1 across clinical variables. Overall, MET Q1 had relatively low concurrent expressions with HER2 and CLDN18. Among GEJ samples, MET Q1 overlapped with CLDN18+ (15%), HER2+ (8%), PD-L1 CPS1+ (29%), and PD-L1 CPS5+ (25%); in GA, MET Q1 overlapped with CLDN18+ (15%), HER2+ (4%), PD-L1 CPS1+ (19%), and PD-L1 CPS5+ (18%). Treatment in general reduced overlap between MET Q1 and PD-L1 (from 27% to 20%), which may be due to chemotherapy (27% to 18%), but not anti-PD1 therapy (27% to 36%). In HER2-negative populations, MET Q1 had a consistently low (11%–12%) overlap with FGFR2, whereas MET - CEACAM5 overlaps were numerically higher (21% in GEJ and 16% in GA). Additional GEA subtype analyses are ongoing. Conclusions: MET Q1 exhibits relatively low overlap with HER2, CLDN18, and FGFR2 , indicating different disease biology in these GEA subtypes. MET/ PD-L1 overlap decreased in treatment-experienced patients, thus favoring using anti-PD1 in earlier line of treatment. RW analysis of biomarker prevalence and overlap may support c-MET targeted clinical strategies.