Contour-like model for precision risk stratification in gastric cancer patients underwent neoadjuvant therapy: A multicenter retrospective study

Background

Lymph node metastasis (LNM) is a critical determinant of prognosis in gastric cancer (GC). Accurate evaluation of lymph node involvement enhances prognostic accuracy and informs postoperative strategies.

Methods

This retrospective study included 649 GC patients who received neoadjuvant chemotherapy followed by curative surgery at two centers between 2009 and 2019. An additional cohort of 292 patients was selected from the SEER database using matching criteria. Collected variables included the number of retrieved lymph nodes (rLNs), positive lymph nodes (pLNs), pathological T stage after treatment (ypT), and Tumor Regression Grade. A novel contour-like ypTN (Con-ypTN) model was constructed using a Gaussian process-augmented Cox regression approach to predict the overall prognosis. Model performance was evaluated through receiver operating characteristic curve analysis, the Delong test, calibration plots, and decision curve analysis.

Results

The Con-ypTN model demonstrated strong prognostic discrimination. AUC values were 0.853 (95 % CI: 0.807–0.900) in the training cohort. Calibration plots and Delong test results showed good agreement between predicted and actual outcomes across all datasets. Notably, the Con-ypTN model significantly outperformed all comparator staging systems (P < 0.05). Patients classified as high risk by the Con-ypTN model had significantly worse survival outcomes than those in the low-risk group (P < 0.05).

Conclusion

The Con-ypTN model provides a robust and clinically relevant tool for prognostic stratification of GC patients treated with neoadjuvant chemotherapy. The model enables precise identification of high-risk individuals, offering improved guidance for postoperative clinical decision-making.