Abstract
BACKGROUND: The detection rate of early-stage gastric cancer remains relatively low. Early detection is beneficial for improving prognosis. We aim to explore the risk factors for lymph node metastasis(LNM) in elderly patients with early gastric cancer (EGC), and to construct a nomogram prediction model for validation. METHODS: 136 patients diagnosed with elderly EGC accompanied by LNM in Cangzhou Central Hospital and Cangzhou People's Hospital from January 2021 to June 2024 were selected as the training cohort. The new gastric cancer screening score (GCSS), regenerated-protein 4(REG4), tumor abnormal protein(TAP) levels and clinicopathologic characteristics were compared in elderly EGC, gastric precancerous disease and control groups. Lasso-Logistic regression was used to identify risk factors affecting LNM, and the nomogram model was established and verified. The clinical decision curve analysis (DCA) and clinical impact curve (CIC) were performed to evaluate the model. In addition, the clinical data of 72 patients with elderly EGC who met the inclusion criteria in two hospitals from July 2024 to June 2025 were selected as the validation cohort to verify the nomogram model. RESULTS: There was no statistically significant difference in general clinical data between the training cohort and the validation cohort (P > 0.05). The incidence of LNM in the training cohort was 22.79% (31/136), while that in the validation cohort was 23.61% (17/72). GCSS, REG4, and TAP levels in EGC group were higher than gastric precancerous disease group and control group. Lasso-Logistic regression showed that REG4, TAP, vascular invasion, undifferentiated type, infiltrate into the submucosa, and diameter ≥ 2 cm were independent risk factors for LNM in elderly EGC. The Nomogram model was constructed based on the independent risk factors screened out, which showed that the predicted values were in good agreement with the measured values, The area under the receiver operating characteristic curve (ROC) (AUC) of the training cohort and the validation cohort was 0.853 and 0.878, respectively. The goodness-of-fit test was conducted using the Hosmer-Lemeshow method, and the result was P = 0.751. The DCA and CIC of the training and validation cohort both indicate that the model had good clinical application value. CONCLUSION: The results of this study indicate that REG4, TAP, vascular invasion, undifferentiated type, infiltrate into the submucosa and diameter ≥ 2 cm are at a higher risk for LNM in elderly EGC. A nomogram model can be helpful for early prediction of LNM in elderly EGC, It provides certain references for the treatment strategies of elderly EGC.