RESUMEN
OBJECTIVES: The latest version of the National Comprehensive Cancer Network recommends neoadjuvant therapy followed by surgical treatment or radical chemoradiotherapy for patients with cT3N0M0. Neoadjuvant therapy can improve the prognosis of patients with locally advanced esophageal cancer. Therefore, the evaluation or prediction of T stage is particularly important because the treatment could differently affect the prognosis. Here, we establish a model to predict the T stage of patients with T2-3N0M0 to help choose the best treatment strategy. METHODS: From 1637 patents with esophageal cancer, we enrolled 48 patients and performed least absolute shrinkage and selection operator regression to screen for independent factors influencing pathological T stage. We, then, trained the decision tree to obtain the decision tree diagram and divided the T stages obtained by different methods into two categories, T2 and T3, for survival analysis. RESULTS: A total of 21 and 27 cases were predicted to be T2 and T3, respectively, under ultrasonic gastroscopy, 19 and 29 under magnetic resonance imaging, and 22 and 26 under pathological examination. Multivariate logistic regression analysis revealed that the muscularis propria thickness (MPT) (p = 0.0097) and the muscularis propria + mucosa thickness (MPMT) in the largest tumor cross-section (p = 0.0239) were independent influencing factors. We plotted a decision tree diagram with these two factors. MPT in the largest tumor cross-section >1.3 mm could be judged as pT3; if ≤1.3 mm, MPMT should be considered a thickness ≥1.7 mm could be judged as pT2 (otherwise pT3). Corresponding survival analysis was performed according to the T stage under different examination modalities. CONCLUSION: MPT in the largest tumor cross-section and MPMT in the largest tumor cross-section are independent predicting factors of pathological T stage.