关键词: 胃肿瘤, 体层摄影术, 螺旋计算机, 腔内超声检查, 肿瘤分期, Logistic模型, 超声内镜, 临床T分期

Abstract: Objective　To establish a new clinical T staging model for patients with gastric cancer (GC) and to evaluate its predictive effect, so as to provide the basis for improving the predictive value of clinical T staging. Methods　A total of 227 GC patients underwent radical surgery in our hospital were enrolled in this study. Among them, 102 cases were pT1-pT2 gastric cancer, 125 cases were pT3-pT4 gastric cancer. All patients underwent endoscopic ultrasonography (EUS) and multi-slice spiral computed tomography (CT) examination before operation. Univariate analysis was used to compare the clinical and pathological data, including gender, age, tumor location, Borrmann classification, CT based T staging, the layers of tumor invading the gastric wall under EUS and the maximum short diameter of longitudinal section of tumor under EUS, between pT1-pT2 and pT3-pT4 patients. According to the clinical experience, CT-based T staging and the layers of tumor invading the gastric wall under EUS were included to establish the traditional conventional clinical T staging model (CCTSM). Multivariate Logistic regression analysis was used to further evaluate the risk factors of pT3-pT4 after univariate analysis, and the significant variables were included in the revised clinical T staging model (RCTSM). The receiver operating characteristic (ROC) curve was constructed to assess the performance of two prediction models. Results　The corresponding Logistic regression equation was Logit (P)= -2.599+2.409× CT based T staging + 2.553× the layers of tumor invading the gastric wall under EUS. The results of univariate analysis and multivariate Logistic regression analysis showed that CT based T3-T4 staging (OR=12.528, 95%CI: 4.347-36.109), the 5th layer of tumor invading the gastric wall under EUS (OR=7.533, 95%CI: 2.539-22.353), the longer maximum of the short diameter of tumor longitudinal section under EUS (OR=31.084, 95%CI: 8.681-111.307) were independent risk factors of pT3-pT4 stage in the GC patients. The Logistic regression equation of the revised clinical T staging model was established with these three variables: Logit (P)= -7.884+2.528× CT based T staging + 2.019× the layers of tumor invading the gastric wall under EUS + 3.437×the maximum short diameter of longitudinal section of tumor under EUS. The clinical value of the RCTSM in predicting pT3~pT4 was better than that of the CCTSM (AUC: 0.952 vs. 0.891, Z=3.870, P＜0.01). In the lymph node positive subgroup, the predictive value of the RCTSM was also better than that of the CCTSM (AUC: 0.916 vs. 0.864, Z=2.058，P＜0.05). Conclusion　The RCTSM can better predict the pathological T staging in patients with gastric cancer and provide reliable basis for individualized treatment of GC patients.

Key words: stomach neoplasms, tomography, spiral computed, endosonography, neoplasm staging, Logistic models, endoscopic ultrasonography, clinical T stage