Abstract:

Objective To explore the preoperative predictive factors influencing microscopic positive proximal margin in upper gastric cancer, and to establish a nomogram prediction model and to validate it internally. Methods Retrospective analysis of 187 patients with upper gastric cancer operated in the Department of Gastrointestinal Surgery of Hengshui People's Hospital from January 2018 to October 2022 were included in this study. Patients were divided into the microscopic positive proximal margin (the R0 group, n=15) and the negative microscopic proximal margin group (the R1 group, n=172) according to histopathological diagnosis. Preoperative factors that may influence positive upper margin of proximal gastric cancer were collected, including patient age, gender, tumor size, tumor location, Borrmann staging, tumor differentiation, Lauren staging, cT stage and cN stage. Receiver operating characteristic (ROC) curve was used to figure out the optimal cut-off value for predicting positive margin of proximal gastric cancer by tumor length. Multivariate Logistic regression was used to analyze the variables with statistical difference between the two groups, and independent risk factors were screened out, and prediction mode was constructed. The prediction accuracy of the model was verified internally using Bootstrap method. Results The best threshold for predicting positive margin of proximal gastric cancer by tumor length was 4.85 cm. Univariate analysis showed that there were significant differences in tumor length, tumor location, Borrmann staging, Lauren staging, cT staging and cN staging between the two groups (all P<0.05). Multivariate Logistic regression analysis showed that tumor length >4.85 cm (OR=4.000, 95%CI: 1.039-15.399), tumor located in esophagogastric junction (OR=7.108, 95%CI: 1.604-31.494), Borrmann staging Ⅲ—Ⅳ(OR=6.991, 95%CI: 1.538-31.782), Lauren staging as diffuse or mixed (OR=7.583, 95%CI: 1.814-31.701) and cT staging as cT4 (OR=8.249, 95%CI: 1.890-36.007) were independent predictors of microscopic positive proximal margin of advanced upper gastric cancer before surgery, and a prediction model was established based on results of multivariate analysis. The area under ROC curve (AUC) value for subjects with the model was 0.862 after internal validation. The calibration curve showed that the model predicted the probability of microscopic positive proximal margin occurrence in good agreement with the probability of actual microscopic positive proximal margin occurrence （Hosmer-Lemeshow χ²=6.145，P=0.523）. Conclusion The established nomogram prediction model can predict the probability of positive upper incisal margin of proximal gastric cancer before operation, and provide clinical guidance for formulating surgical strategy.

Key words: stomach neoplasms, margins of excision, root cause analysis, Logistic models, nomograms, forecasting