A New Statistical Model Identified Two-thirds of Clinical T1 Gastric Cancers as Possible Candidates for Endoscopic Treatment.

BACKGROUND: Clinical T1 gastric cancer has low metastatic potential to lymph nodes and is generally curable by local treatment. Endoscopic resection can preserve the whole stomach and does not impair the patient's quality of life; however, its indication is strictly limited to the subset of patients without nodal metastasis. The study was designed to predict reliably the patients without nodal metastasis based only on the clinical information. METHODS: We examined patients with clinical T1 disease who were treated with surgery. The clinically available information was evaluated for its ability to predict nodal metastasis by logistic regression model. Then, the predictive ability of the logistic regression model using the risk factors for nodal metastasis was evaluated by a receiver operating characteristic curve. RESULTS: A total of 511 patients were entered into this study. The clinical depth (cT1a or cT1b), maximal tumor diameter, and pathological type were confirmed to be significantly different between patients with and without nodal metastasis. The cutoff value of the tumor diameter differed depending on the histology and clinical depth: 79 mm for differentiated type and 48 mm for undifferentiated type in cT1a tumors, and 43 mm for differentiated type and 11 mm for undifferentiated type in cT1b tumors. According to these criteria, 348 of the 511 patients (68.1 %) were classified to have predictive N0 status. The negative predictive value was 95.7 % (95 % confidence interval 94.0-97.5 %). CONCLUSIONS: The predictive criteria based on the multivariate logistic model identified that almost two-thirds of the patients with clinical T1 gastric cancer were possible candidates for endoscopic treatment.