Development of Lymph Node Metastasis Prediction Model on Each Station in Gastric Cancer Patient

Abstract

Background and Aim The prediction of lymph node metastasis (LNM) on each lymph node (LN) station is important for tailored surgery. The aim of this study was to develop a prediction program which can calculate the probability of LNM according to LN stations in gastric cancer patients. Methods We retrospectively analyzed 4,660 patients who underwent gastrectomy for primary gastric cancer from 2003 to 2013 and the LN status was well examined in according to the LN stations, at Seoul National University Hospital. We reviewed preoperative endoscopic findings and/or gross pathologic findings, and reclassified the tumor locations by the endoscopic terms. All of the involved locations were included into the analysis. The variables which can get preoperatively were evaluated. Multiple logistic regression analysis was used to develop a LNM prediction model using whole data for each LN station. The performance of the prediction model was validated in terms of discrimination and calibration using a total of 200 bootstrap samples. Results The multiple analysis identified depth of tumor, gross type, involved locations as covariates associated with LNM. But the significant factors were somewhat different according to the LN stations. In the validation, the prediction equation exhibited good discrimination. Calibration plot of the prediction equation predicted LNM rate corresponding closely with the actual rate. Conclusions We developed a LNM prediction program on each LN stations. Validation revealed good discrimination and calibration, suggesting good clinical utility. The LNM prediction program improved individualized predictions of LNM of each LN stations.