Advances in the application of nomograms for patients with gastric cancer associated with peritoneal metastasis

Abstract This review elucidates advancements in nomogram applications for predicting peritoneal metastasis (PM) and prognostication in gastric cancer (GC). PM accounts for 40–50% of GC-related mortalities globally, with Asian populations disproportionately affected. Nomograms integrate multifactorial clinical, radiological, and molecular data to enhance individualized risk assessment, addressing limitations of conventional TNM staging. Clinical nomograms incorporating TNM staging, serum biomarkers (e.g., CA19-9), and peritoneal carcinomatosis index (PCI) demonstrate superior predictive accuracy. For instance, Ji et al. constructed a nomogram integrating PCI, preoperative tumor markers, and peritoneal metastasis duration, achieving an area under the ROC curve (AUC) of 0.985 for overall survival prediction. Imaging-based models leverage CT radiomics and deep learning algorithms to detect occult PM, with Huang et al.’s deep learning model attaining an AUC of 0.900. Hybrid approaches combining radiomic signatures with clinical variables further improve diagnostic sensitivity (e.g., Dong et al.’s tri-modal nomogram, AUC = 0.958). Molecular biomarkers, including collagen deposition patterns, Legumain (LGMN) expression, and circulating tumor cells (CTCs), have been incorporated into nomograms to refine risk stratification. For example, a collagen signature-based model demonstrated superior discrimination compared to traditional staging systems (C-index = 0.68 vs. 0.61). Challenges persist, including data heterogeneity, limited multicenter validation, and algorithmic complexity. Future directions involve AI-driven optimization (e.g., federated learning), standardization of variable selection, and exploration of mechanistic targets (e.g., FMO family genes, TME remodeling). In conclusion, nomograms represent a promising tool for personalized GC management. Overcoming translational barriers through interdisciplinary collaboration will be critical to bridging the gap between predictive analytics and therapeutic intervention.