Developing and validating nomograms for predicting the survival in patients with clinical local-advanced gastric cancer

BACKGROUND: Many patients with gastric cancer are at a locally advanced stage during initial diagnosis. TNM staging is inaccurate in predicting survival. This study aims to develop two more accurate survival prediction models for patients with locally advanced gastric cancer (LAGC) and guide clinical decision-making. METHODS: We recruited 2794 patients diagnosed with LAGC (2010–2015) from the Surveillance, Epidemiology, and End Results (SEER) database and performed external validation using data from 115 patients with LAGC at Yantai Affiliated Hospital of Binzhou Medical University. Univariate and multifactorial survival analyses were screened for meaningful independent prognostic factors and were used to build survival prediction models. Concordance index (C-index), receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) were evaluated for nomograms. Finally, the differences and relationships of survival and prognosis between the three different risk groups were described using the Kaplan–Meier method. RESULTS: Cox proportional risk regression model analysis identified independent prognostic factors for patients with LAGC, and variables associated with overall survival (OS) included age, race, marital status, T-stage, N-stage, grade, histologic type, surgery, and chemotherapy. Variables associated with cancer-specific survival (CSS) included age, race, T-stage, N-stage, grade, histological type, surgery, and chemotherapy. In the training cohort, C-index of nomogram for predicting OS was 0.722 (95% confidence interval [95% CI]: 0.708–0.736] and CSS was 0.728 (95% CI: 0.713–0.743). In the external validation cohort, C-index of nomogram for predicted OS was 0.728 (95% CI:0.672–0.784) and CSS was 0.727 (95% CI:0.668–0.786). The calibration curves showed good concordance between the predicted and actual results. C-index, ROC, and DCA results indicated that our nomograms could more accurately predict OS and CSS than TNM staging and had a higher clinical benefit. Finally, to facilitate clinical use, we set up two web servers based on nomograms. CONCLUSION: The nomograms established in this study have better risk assessment ability than the clinical staging system, which can help clinicians predict the individual survival of LAGC patients more accurately and thus develop appropriate treatment strategies.