Prediction of lymph node status in stage T1 esophageal squamous cell carcinoma.

Abstract

263 Background: With the development of endoscopic therapy and the introduction of immunotherapies, it is more important for preserving the esophagus and uninvaded lymph nodes to avoid the damages involving the long-term quality of life and normal immune function caused by esophagectomy with radical lymphadenectomy. Methods: We collected patients with T1 ESCC who underwent endoscopic resection or esophagectomy from 2014 to 2021 in three cancer centers. The primary end point was presence of lymph node metastasis (LNM) in surgical resection specimens or detection of metastases during follow-up. Univariate and multivariate logistic regression were used to screen the clinicopathological features related to LNM. Using the function 'lrm' constructed the prediction model. The area under the curve (AUC) of receiver operating characteristic and calibration graphs were used to assess the performance of the model. An external validation was conducted to evaluate the generalization of the model. We then prospectively collected 28 surgically resected tissue samples of T1 ESCC to further supplement the model from the cellular and molecular level, and used 6 markers (CK, CD34, D2-40, CD4, CD8, FOXP3) and DAPI dye to process multiple immunofluorescence (mIF) staining. Afterwards, we analyzed the density, spatial distribution of microvessels, lymphatic capillaries and tumor-associated immune cells and the interaction between them and malignant cells in the tumor microenvironment of T1 ESCC to explore the relationship between tumor microenvironment and risk of LNM. Results: 1109 patients with T1 ESCC were included. 653 patients from two centers constituted the training cohort, and 456 patients from the third center served as the validation cohort. Multivariate logistic regression based on the training cohort showed that tumor invasion into the submucosa, poor histological grade, and lymphovascular invasion were risk factors for LNM in T1 ESCC. Subsequently, we constructed a model based on three variables to predict the risk of LNM in T1 ESCC. The model demonstrated good discriminative ability (AUC 0.858, 95%CI 0.826-0.890). After mIF staining of 28 prospectively collected T1 ESCC tissue samples, we observed that microvessels and lymphatic capillaries were abundant in the mucosa in lymph node-positive samples but not in lymph node-negative samples. In addition, the average density of FOXP3+ T cells in samples with and without metastasis was 253/mm2 and 26/mm2 (P < 0.001), and the average density of CD8+ T cells was 112/mm2 and 432/mm2 (P < 0.001) in quantitative analysis, respectively. Conclusions: This study (RENMIN-237) developed a well-performing LNM risk stratification model of T1 ESCC to reduce both overtreatment and undertreatment, and we found that FOXP3+ T cells in the tumor microenvironment are a risk factor for LNM, while CD8+ T cells can serve as a protective factor against LNM.