Establishment Of A Scoring System Predicting The Lymph Node Metastasis In Undifferentiated Early Gastric Cancer And A Preliminary Exploration Of ESD Feasibility

ObjectiveThis study aims to establish an evaluation system for patients with undifferentiated early gastric cancer so we can used it to evaluate risk of lymph node metastasis (LNM), as well as to expand the possibility of endoscopic submucosal dissection (ESD) for treating undifferentiated early gastric cancer.MethodsFirst of all, Meta-analysis was performed to compare the effectiveness and safety in early gastric cancer between surgery and endoscopic submucosal dissection. As a result,we found that ESD was not only effective and safe but also economically viable. Secondly, we collected data of 116 patients from our hospital with undifferentiated early gastric cancer who received surgical treatment. Univariate and multivariate logistic regression analysis was used to retrospectively analyze the relation between clinicopathological factors and lymph node metastasis. We then enlarged the number of cases into 192 to build the predictively scoring system. Finally, we conducted prospective clinical trial to evaluate the accuracy of this model.Results1. The tumor size, depth of invasion and lymphatic vessel involvement (LVI) that were significantly associated with LNM by univariate analysis, were found to be significant and independent risk factors for LNM by multivariate analysis. LNM was observed in 81% of patients with both risk factors, but in 1% of 36 patients without risk factors in intramucosal undifferentiated EGC. The 5-year survival rates were 88.9%, 72.4% and 33.3%,respectively in cases with none, one and two of the risk factors respectively in intramucosal undifferentiated EGC (P < 0.05).2. Based on logistic regression results, we set up the equation as:Z= -7.720 +2.063*(tumor size) + 1.403*(depth of invasion) + 1.858 (lymphatic vessel involvement).We got the cut-off point at the highest score of YI index, which is 34 with a sensitivity of 94.3% and a specificity of 6.2. Logistic regression determined AUC to be 0.885 and our network model determined it to be 0.888.3. In our prospective study, the sensitivity, specificity and coincidence rate of the prediction model is 100%, 68.8% and 80.4%, respectively. The high-risk group has lymph node positive rate of 65.5%. While the lymph node positive rate of low-risk group is 0.Conclusions1. The tumor size, LVI, and depth of invasion were independently associated with the presence of LNM in undifferentiated EGC.2. We established a prediction model to predict the positive lymph node rate in patients with undifferentiated early gastric cancer.3. Prospective study proved the stability, accuracy and feasibility of this model.