A Risk-Scoring Model for Predicting Lymph Node Metastasis in Early Gastric Cancer Patients: a Retrospective Study and External Validation.

BACKGROUND: The possibility of lymph node metastasis (LNM) is critical to the assessment of the indication for endoscopic submucosal dissection. Thus, the aim of this study is to identify the risk factors for LNM and construct a risk-scoring model for patients with early gastric cancer to guide treatment. METHODS: A retrospective examination of reports and studies carried out January 2000 and December 2014 was conducted. A risk-scoring model for predicting LNM was developed based on the data thus collected. In addition, the model is subject to verification and validation by three institutions. RESULTS: Of the 1029 patients, 228 patients (22%) had LNM. Multivariate analysis showed that female, depressed type, undifferentiated type, submucosa, tumor size, and lymphovascular invasion were significantly associated with LNM. An 11-point risk-scoring model was used to predict LNM risk. An area under the receiver operating characteristic (AUROC) of the risk-scoring model was plotted using the development set and the AUROC of the model [0.76 (95% CI 0.73-0.80)] to predict LNM risk. After internal and external validation, the AUROC curve for predicting LNM was 0.77 (95% CI 0.68-0.86), 0.82 (95% CI 0.72-0.91), and 0.82 (95% CI 0.70-0.94), respectively. CONCLUSIONS: A risk-scoring model for predicting LNM was developed and validated. It could help with personalized care for patients with EGC.