Development, validation and comparison of the nomograms for predicting postoperative survival and recurrence in gastric cancer

Abstract

Gastric cancer (GC) is one of the most common malignant tumors worldwide, with an overall five-year survival rate of less than 50%. The aim of this study is to construct and validate the new nomogram model for predicting recurrence and survival of GC patients after gastrectomy. We conducted a long-term, multicenter retrospective analysis of 562 GC patients who underwent gastrectomy from two medical centers. Multiple risk factors affecting prognosis were screened from preoperative and postoperative clinical data of patients. Then we constructed postoperative and preoperative models separately. The predictive accuracy of each model was evaluated by calibration curves and receiver operating characteristic (ROC) curves and DeLong’s test. The predictive performance of preoperative and postoperative models at different time points was assessed by plotting the time-dependent ROC (t-ROC) curves. Four nomograms were proposed to predict postoperative recurrence and survival based on preoperative and postoperative factors, respectively. The calibration curves indicated well consistency between the predicted and actual observed results. In addition, the area under the ROC curve (AUC) value, t-ROC curves and DeLong’s test showed that the accuracy and predictive performance of the postoperative nomograms are significantly superior to those of the preoperative nomograms and the TNM staging system. The postoperative nomogram can better predict postoperative recurrence and survival, and provide reliable guidance for clinical treatment and follow-up decisions of GC patients.