Risk stratification of gastric neuroendocrine tumors in autoimmune gastritis: Evaluating the clinical value of an integrated clinical-endoscopic model

Abstract

Risk stratification of gastric neuroendocrine tumors (G-NETs) arising in the context of autoimmune gastritis (AIG) remains a significant clinical challenge, as current approaches based on isolated biomarkers or endoscopic findings fail to fully account for the multifactorial nature of tumor development. Li et al[13] recently published a study in World Journal of Gastroenterology, and this article synthesizes existing evidence on the pathophysiological mechanisms linking AIG to G-NETs, and systematically evaluates the development and validation of integrated clinical-endoscopic models, demonstrating the enhanced performance of machine learning techniques in identifying robust predictors such as age, Helicobacter pylori status, vitamin B12 levels, severity of corpus atrophy, and serum gastrin concentration for accurate risk stratification. The clinical implications of these models are examined across several domains: Enabling risk-adapted endoscopic surveillance schedules, guiding chemopreventive interventions including vitamin B12 supplementation, improving prognostic precision through incorporation of proliferative indices, and enhancing diagnostic consistency across diverse healthcare settings. Although current models show promising discriminative ability (area under the curve: 0.830), they are constrained by reliance on single-center cohorts and limited integration of molecular data. Future efforts should prioritize multicenter validation, incorporation of genomic markers, and the creation of multimodal frameworks that integrate endoscopic imaging with serological and genetic profiles to support personalized management of AIG-induced G-NETs.

Keywords: Gastric neuroendocrine tumors; Autoimmune gastritis; Integrated model; Endoscope; Integrated model

Core Tip: Autoimmune gastritis induces gastric neuroendocrine tumors through a multistep process involving oxyntic mucosal destruction, hypergastrinemia, and enterochromaffin-like cell proliferation. Integrated clinical-endoscopic models utilizing machine learning (e.g., Boruta algorithm) outperform single-parameter approaches, with key predictors including atrophy severity, gastrin levels, and Helicobacter pylori history. These models enable personalized risk-adapted surveillance, inform vitamin B12 chemoprevention, and improve prognostic stratification by incorporating endoscopic and proliferative indices. Current limitations include single-center bias and incomplete molecular integration; future development requires multicenter validation and multimodal biomarkers.