Integration of objective nutritional status and tumor burden predicts 5-year survival after gastric cancer surgery: a...

This study aimed to develop an objective prognostic score by integrating nutritional indices and tumor burden to improve long-term survival prediction after gastrectomy for gastric cancer.

This multicenter study included 1395 patients with gastric cancer who underwent curative gastrectomy and were divided into a training (n = 1006) cohort and an external validation (n = 389) cohort. In the training cohort, the optimal objective nutritional index was identified among the five candidates. Random forest analysis was used to screen key preoperative biomarkers, and the Objective Nutrition and Tumor Burden Score (ONTBS) was developed using Cox regression. Model performance was assessed using AUC, C-index, AIC, BIC, and decision curve analysis.

Among the five nutritional scores, CONUT showed the best survival prediction (AUC = 0.580).Random forest analysis identified CA19-9 (p < 0.001) and CEA (p = 0.01) as the most prognostically relevant biomarkers. Multivariate Cox analysis confirmed CONUT (HR 1.130, 95% CI 1.032-1.237, p = 0.008), CA19-9 (HR 1.0021, 95% CI 1.001-1.002, p < 0.001), pT stage (pT3: HR 4.311, 95% CI 2.553 -7.279, p < 0.001; pT4: HR 7.020, 95% CI 4.010 -12.291, p < 0.001), and PLNR (HR 1.040, 95% CI 1.028 -1.053, p < 0.001) as independent risk factors, forming the ONTBS. In the training cohort, ONTBS showed a higher C-index (0.83) than TNM (0.81) and CONUT (0.57), with lower AIC (3399.37) and BIC (3403.02) values than TNM (3441.86 and 3445.51) and CONUT (3798.61 and 3802.26), respectively. Time-dependent ROC curve and DCA demonstrated consistently superior performance and clinical benefits, as confirmed by the external validation cohort.

ONTBS integrates objective nutritional status and tumor burden and outperforms TNM staging in predicting long-term survival after curative gastric cancer surgery.