A prediction model for lymph node metastasis in T1 esophageal squamous cell carcinoma.

BACKGROUND: Endoscopic resection is widely used for the treatment of T1 esophageal cancer, but it cannot be used to treat lymph node metastasis (LNM). This study aimed to develop a prediction model for LNM in patients with T1 esophageal squamous cell carcinoma.

METHODS: A prospectively maintained database of all patients who underwent surgery for esophageal cancer between January 2002 and June 2010 was retrospectively reviewed, and patients with T1 squamous cell carcinoma were included in this study. Correlations between LNM and clinicopathological variables were evaluated using univariable and multivariable logistic regression analyses. The penalized maximum likelihood method was used to estimate regression coefficients. A prediction model was developed and internally validated using a bootstrap resampling method. Model performance was evaluated in terms of calibration, discrimination, and clinical usefulness.

RESULTS: A total of 240 patients (197 male, 43 female) with a mean age of 57.9 years (standard deviation ± 8.3 years) were included in the analysis. The incidence of LNM was 16.3%. The prediction model consisted of four variables: grade, T1 stage, tumor location and tumor length. The model showed good calibration and good discrimination with a C-index of 0.787 (95% confidence interval [CI], 0.711-0.863). After internal validation, the optimism-corrected C-index was 0.762 (95% CI, 0.686-0.838). Decision curve analysis demonstrated that the prediction model was clinically useful.

CONCLUSIONS: Our prediction model can facilitate individualized prediction of LNM in patients with T1 esophageal squamous cell carcinoma. This model can aid surgical decision making in patients who have undergone endoscopic resection.