Zhenjiang Guo ¹ Ning Wang ^2* Guangyuan Zhao ^1* Liqiang Du ^1* Zhaobo Cui ^2* Fangzhen Liu ^1*

• ¹ Department of Gastrointestinal Surgery, Hengshui People’s Hospital,, Hengshui, Hebei Province, China
• ² Department of Respiratory and Critical Care Medicine, Hengshui People's Hospital, Hengshui, China

Objective: To develop and validate a model for preoperative prediction of positive proximal margins for AEG by transabdominal approach, and to analyze the safe margin distances for patients with different risks of positive proximal margins.A retrospective analysis was performed on 284 AEG patients who underwent surgery via the transabdominal approach between January 2017 and December 2023. Patients were divided into a training set (n=201, first five years) and a test set (n=83, last two years). The synthetic minority oversampling technique (SMOTE) was applied to address class imbalance in the training set. Two nomogram models were developed: one based on the original training set and the other using the SMOTE dataset. The model's performance was compared using the test set, with the area under the curve (AUC) used to evaluate discrimination and the Hosmer-Lemeshow test used for model fit. The best-performing model was used to calculate total scores for the entire cohort, and the optimal cutoff value was determined via the ROC curve. Patients were classified into low-and high-risk groups based on the total score, and optimal margin distances were determined using Youden's index.The model developed using the SMOTE dataset showed superior AUC for predicting positive proximal margins in the test set compared to the model based on the original training set (0.814 vs. 0.780). Independent predictors of positive proximal margins included Borrmann classification, Lauren classification, cT stage, tumor differentiation, and Siewert classification. The Hosmer-Lemeshow test showed a good model fit (χ² = 5.397, P = 0.612). Using a cutoff total score of 206.811, patients were divided into low-risk (score < 206.811) and high-risk (score ≥ 206.811) groups, with an AUC of 0.788. For the low-risk group, a proximal margin distance of 2.75 cm yielded an AUC of 0.824, with a sensitivity of 54.5%, specificity of 97.9%, and a Youden's index of 0.524. For the high-risk group, a margin distance of 3.85 cm provided an AUC of 0.813, sensitivity of 73.1%, specificity of 80.0%, and a Youden's index of 0.531.The nomogram may offer a valuable preoperative tool for assessing the risk of positive proximal margins in AEG patients.