Incidence, risk factors, and a predictive model for lymph node metastasis of submucosal (T1) colon cancer: A population‐based study

Abstract

OBJECTIVE: This study aimed to assess the incidence, identify independent factors, and develop a lymph node metastasis (LNM) prediction model for patients with T1 colon cancer. METHODS: Statistics were drawn from the Surveillance, Epidemiology, and End Results database between 2004 and 2014. A multivariate logistic regression analysis was performed to determine independent predictors of LNM. A nomogram for predicting the possibility of LNM was developed based on those factors. RESULTS: A total of 5397 patients with T1 colon cancer were identified. The overall LNM rate was 15.0% (808/5397). A multivariate analysis showed that age (odds ratio [OR] 0.97, P < 0.001), tumor size (OR 1.01, P < 0.001), moderate (OR 1.77, P = 0.001) or poorly differentiated/undifferentiated tumor (OR 5.60, P < 0.001), right colon cancer (OR 1.39, P = 0.008), and a positive carcinoembryonic antigen level (OR 1.51, P = 0.004) were independent predictive factors for LNM. The area under the receiver operating characteristic curve was 0.68 (95% confidence interval [CI] 0.65-0.71) in the training set and 0.65 (95% CI 0.61-0.67) in the validation set. A calibration plot showed good consistency between the bias-corrected prediction and the ideal reference line with 1000 additional bootstraps (mean absolute error = 0.007). CONCLUSIONS: The incidence of LNM was high in patients with T1 colon cancer. A nomogram for predicting the probability of LNM for T1 colon cancer may be used to help determine the optimal treatment for these patients.