On the borderline: GEJ coding as a driver for stage migration and survival difference—A SEER analysis.

Abstract

e16151 Background: The classification of gastro-esophageal junction (GEJ) tumors as esophageal (ICD-O-3 C15.5, C15.2) versus gastric cardia (C16.0) differs by institutional practice. We postulated that this coding heterogeneity leads to stage migration, which could impact trial eligibility and survival comparisons by assigning different stages to physiologically comparable tumors. Methods: Using SEER Research Data (17 Registries, 2000-2022), we identified 23,892 patients individuals with adenocarcinoma at the GEJ border (C15.2/C15.5 "esophagus-coded" vs. C16.0 "stomach-coded").Overall survival (OS) and stage distribution differences (chi-square test) were the main results. Overall and stratified by AJCC 7th edition stage, we conducted Kaplan-Meier and Cox regression analyses. Log-rank tests were used to evaluate within-stage survival differences in order to identify stage migration effects. Results: Esophagus-coded tumors (n = 10,705) versus stomach-coded tumors (n = 13,187) showed significantly different stage distributions (p = 2.1×10⁻³⁷). Stomach-coded tumors had higher stage IV prevalence (43.0% vs 37.0%) and lower stage III (22.1% vs 26.5%). Advanced-stage disease (III-IV) was more common in stomach-coded tumors (65.1% vs 63.5%, p = 0.01). Overall median survival was similar (13 vs 12 months, log-rank p = 0.62), but within-stage survival differed significantly in stages II, III, and IV (p = 0.0003, p = 0.01, p = 0.004, respectively), with no difference in stage I (p = 0.11). In stage II, esophagus-coded tumors showed worse survival despite earlier nominal stage. Results persisted in sensitivity analyses restricted to specified anatomic sites excluding NOS codes (n = 19,030). Conclusions: GEJ adenocarcinomas that are otherwise the same but coded as esophageal in some settings and gastric in others show consistent stage migration and survival differences even within the same stage, indicating that coding conventions can bias stage-based comparisons. This directly affects how we design clinical trials, write treatment guidelines, and interpret comparative effectiveness studies. To make results comparable across datasets and institutions, we need either standardized GEJ classification rules or a GEJ-specific staging framework.