ABSTRACT Background Early esophageal cancer is prone to lymphovascular invasion (LVI) and ductal involvement (DI), which seriously affects the prognosis of patients. Therefore, identifying the potential risk factors for LVI/DI is crucial. This study aims to clarify endoscopic appearances which are predictive for LVI/DI in patients with node‐negative superficial esophageal squamous cell carcinoma (SESCC). Methods A total of 401 lesions were included for model development. Endoscopic image features were selected through multivariate logistic regression and LASSO (least absolute shrinkage and selection operator) regression analysis, and the optimal model was determined by calculating cNRI (continuous net reclassification improvement) and IDI (integrated discrimination improvement). Additionally, internal and external validation were performed using data from 173 and 133 lesions, respectively. Results Five endoscopic image features, including lesion length, macroscopic type, surface granularity, surface nodularity, and surface erosion, were identified as predictive factors and were incorporated into the nomogram. The nomogram demonstrated substantial predictive efficacy, as evidenced by the AUC (area under curve) values of 0.854 (95% CI: 0.809–0.899) for the training set, 0.821 (95% CI: 0.749–0.894) for the internal validation set, and 0.860 (95% CI: 0.765–0.955) for the external validation set. Calibration curves showed good agreement between the nomogram predictions and actual observations. DCA (decision curve analysis) and CICs (clinical impact curves) confirmed the nomogram's clinical applicability. Conclusions We constructed a nomogram based on preoperative endoscopic image features for predicting the risks of LVI/DI in patients with node‐negative SESCC, which may aid in clinical decision‐making.
Yue et al. (Thu,) studied this question.