Active Stat3 and Her-2 as combined survival predictors show superiority to TNM staging system for postoperative patients with gastric cancer

Objectives: TNM staging of gastric cancer (GC) is useful in predicting prognosis, but its definition is only possible after surgery. It is therefore desirable to develop a method that can predict prognosis and assist management options before surgery. Methods: This study investigated 110 GC patients after radical gastrectomy and followed-up for 136 months. Patients' complete clinicopathological data were collected and gastroscopically biopsied or surgically resected tissues were examined for the expression of Her-2, nm-23, CEA and phosphorylated Stat3 (p-Stat3) using immunohistochemistry (IHC). Univariate and multivariate ROC curves, Kaplan-Meier survival curves, and SPSS Version 22.0 and R (version 3.6.1) statistical software were used to analyze the data. Results: Three major findings were observed: (1) Tissue levels of p-Stat3, Her-2, CEA and nm-23 were correlated with GC patients' survival probability termed as survival prediction power (SPP). (2) Using 5-year survival as an end-point, the SPP of the p-Stat3+Her-2 combination was stronger (AUC=0.867) than that of TNM staging (AUC=0.755). (3) Using cut-off values derived from ROC curves, Kaplan-Meier analyses showed that the p-Stat3+Her-2 molecular combination could clearly predict overall survival rates between the predictive low-risk patients (69.2%) and the predictive high-risk patients (13.2%) with a discriminative difference as high as 56.0%. Conclusions: We conclude that area under the ROC curve (AUC) can be used to quantify SPP powers for biomarkers, making cross-comparisons possible among different survival predictors. This study has first established a multi-factor survival prediction model by which the p-Stat3+Her-2 combination has the best discriminative capability to differentiate low-risk patients from high-risk patients in terms of survival prognosis.