Evaluation of a real-world clinico-genomics database of patients with upper gastrointestinal (GI) malignancies in Japan.

Abstract

400 Background: The genomic profile of gastrointestinal (GI) cancers provides essential information for precision medicine, however, there is limited availability of real-world data in Asian populations as they are not well-represented in most publicly accessible clinico-genomic databases. In this retrospective study, we evaluated the clinical and genomic characteristics of upper GI cancer patients in the database from an academia-industry collaborative nationwide cancer genome screening project in Japan. Methods: The Screening Projectfor Individualized Medicine in Japan (J-SCRUM) GI-SCREEN enrolled advanced GI cancer patients from 26 sites before planned or during systemic therapy from 2015 to 2019. Archived formalin-fixed paraffin-embedded tumor tissue was analyzed by next-generation sequencing (NGS) using Oncomine Cancer Research Panel and Oncomine Comprehensive Assay. Patients with gastric (GC), esophageal (EC) and pancreatic (PC) cancers were identified in the database. The frequency of key actionable genomic alterations in each cancer cohort was explored and compared to similar upper GI cohorts in The Cancer Genome Atlas (TCGA), a publicly available genomics database initiated in 2005 and concluded sample collection in 2013 with most samples from White cancer patients. Results: A total of 872 GC (mean age at registration [SD]: 64y [12]; percent males: 65%), 265 EC (64y [9]; 78%) and 495 PC (64y [10]; 60%) patients were identified. The majority (>70%) of samples were obtained from primary sites. Adenocarcinoma was the most common histological type for GC (85%) ad PC (76%), while squamous cell (76%) was the most common among EC patients. In the GC cohort, the prevalence of key actionable genomic alterations in J-SCRUM GI-SCREEN vs TCGA (N = 380) was, TP53 mutation (49% vs 49%), ERBB2 amplification (14% vs 13%), PIK3CA mutation (11% vs 14%), MYC amplification (9% vs 12%) and MET amplification (2% vs 3%). In the EC cohort, the prevalence in J-SCRUM vs TCGA (N = 182) was TP53 mutation (77% vs 87%), NFE2L2 mutation (21% vs 10%), PIK3CA mutation (9% vs 9%) and CDKN2A mutation (8% vs 8%). In the PC cohort, the prevalence in J-SCRUM vs TCGA (N = 150) was KRAS hotspot mutation (90% vs 93%), KRAS G12D (43% vs 41%), KRAS G12V (32% vs27%), TP53 mutation (58% vs 64%), BRCA1/2 deleterious mutation (5% vs 5%), ATM deleterious mutation (6% vs 5%) and MYC amplification (5% vs 5%). Conclusions: As part of a nationwide cancer genome screening project, J-SCRUM GI SCREEN database provides comprehensive genomic information for Asian population (Japan). The prevalence of key actionable gene alterations in each tumor type was highly comparable between J-SCRUM GI-SCREEN and TCGA data.