Treatment Effects of Different Therapeutic Strategies and Development of a Prognostic Prediction Model for Early Gastric Cancer: A Population-Based and Multicenter Study

Background: Early gastric cancer (EGC) lacks a predictive tool integrating endoscopic therapy (ET), gastrectomy (G), and chemoradiotherapy (CRT) to support individualized risk stratification. This study compared cancer-specific survival (CSS) and overall survival (OS) across the four treatment strategies and developed and externally validated a CSS prediction model incorporating oncologic characteristics and treatment strategy (presented as a nomogram and online tool). Methods: In a U.S. cohort (2017–2021), 1766 patients with EGC were classified into four treatments groups (ET, ET + CRT, G, G + CRT); an independent multicenter cohort (n = 1333) provided external validation. Kaplan–Meier curves, log-rank tests, and Cox regression assessed unadjusted differences, with confounding balanced by stabilized inverse probability of treatment weighting (sIPTW). Predictors were selected by univariable Cox, LASSO, Boruta, and best-subset regression; performance was evaluated by C-index, calibration curves, and decision curve analysis. Results: After sIPTW adjustment, CSS did not differ significantly between ET and G (p = 0.856), whereas G was associated with better CSS than G + CRT (p < 0.05). In exploratory analyses of node-positive patients, G + CRT was associated with lower cancer-specific mortality than G alone (p < 0.05). The CSS prediction model was validated internally and externally (AUCs for 1-, 3-, and 5-year CSS: 0.769, 0.737, 0.717 internally; 0.943, 0.890, 0.872 externally). Conclusions: This is the first externally validated EGC CSS model integrating all four treatments. For eligible patients, ET was associated with survival comparable to G and superior to G + CRT; among node-positive patients, G + CRT was associated with lower mortality than G alone. The model supports individualized prediction and treatment decisions.