Predictors of lymph node metastasis in patients with non-curative endoscopic resection of early gastric cancer

Although surgery is recommend for non-curative endoscopic resection of early gastric cancer (EGC), only a part of patients are found to have lymph node (LN) metastasis. This study aimed to identify the predictors of LN metastasis in patients with non-curative endoscopic resection. Between April 2005 and July 2013, consecutive patients who received non-curative endoscopic resection and then underwent gastrectomy with lymphadenectomy or followed at least 1 year with abdominal computed tomography were retrospectively enrolled at a single tertiary hospital. Non-curative resection was defined as a resection beyond the expanded criteria in pathologic mapping. The predictors for LN metastasis were identified by fitting a multivariate logistic regression model. Among the 1783 consecutive patients who received endoscopic resection of EGC, non-curative resection was performed in 323 (18.1 %) patients. Of these patients, a total of 267 patients were enrolled, and the rate of LN metastasis was 6.7 % (18/267). In multivariate analysis, venous invasion [odds ratio (OR), 7.83; 95 % confidence interval (CI) 2.20-27.86; p = 0.001], sm2 invasion (tumor invasion a parts per thousand yen500 A mu m into submucosa; OR 4.98; 95 % CI 1.34-18.47; p = 0.016), or antral tumor location (OR 12.65; 95 % CI 1.57-102.00; p = 0.017) were independent predictors for LN metastasis. The rates of LN metastasis were 1.1 % (95 % CI 0-2.7) for patients with one or no predictor and 17.8 % (95 % CI 9.7-25.8) for those with two or more predictors. Additional gastrectomy with lymphadenectomy after non-curative endoscopic resection of EGC is recommended for the patients with two or more identified predictors. However, close follow-up without immediate surgery might be considered cautiously for those with only one or no predictor.