Abstract
Most patients with gastric cancer (GC) are first diagnosed at stage III-IV and surgery resection remains the primary therapeutic modality for these patients. However, clinical staging used for prediction of those patients provides limited information. We collected clinicopathological data and disease-progression information from 508 patients with stage III-IV GC at three Chinese hospitals and 1298 patients from the Surveillance, Epidemiology, and End Results database. Based on the stepwise multivariate regression model, we constructed a novel nomogram to predict overall survival (OS). The performance of discrimination for this model was measured using Harrell's concordance index (C-index) and receiver-operating characteristic curve (ROC), and was validated using calibration plots. Multivariate Cox regression analyses showed that tumor size, age at diagnosis, N stage, tumor grade, and distant metastases were outstanding independent prognostic factors of stage III-IV GC. We developed a nomogram based on these five prognostic predictors. In the training set, the C-index of the nomogram was 0.645 (95% CI: 0.611-0.679), which was higher than that of the American Joint Committee on Cancer TNM system alone (sixth TNM: 0.544; seventh TNM: 0.575; eighth TNM: 0.568). Similar results were observed in validation cohort. Moreover, calibration blots demonstrated good consistency between the actual and predicted OS probabilities. According to the nomogram, GC individuals could be classified into three groups (low-, middle-, and high-risk) (P < .001). Our nomogram complements the current staging system for prediction of individual prognosis with stage III-IV GC, and may be helpful for making individualized treatment decisions.