基于临床特征的决策树模型预测胃癌隐匿性腹膜转移风险  

A prediction model of occult peritoneal metastasis in gastric cancer created using the decision tree model based on clinical characteristics

作　　者：邢继尧 李浙民 王胤奎[1] 李子禹[1] Xing Jiyao;Li Zhemin;Wang Yinkui;Li Ziyu(Gastrointestinal Cancer Center,Peking University Cancer Hospital and Beijing Institute for Cancer Research,Key laboratory of Carcinogenesis and Translational Research(Ministry of Education),Beijing 100142,China)

机构地区：[1]北京大学肿瘤医院暨北京市肿瘤防治研究所胃肠肿瘤中心,恶性肿瘤发病机制及转化研究教育部重点实验室,北京100142

出　　处：《腹部外科》2025年第2期112-117,共6页Journal of Abdominal Surgery

基　　金：北京市属医学科研院所公益发展改革试点项目(JYY2023-3);科技部重点研发计划青年计划(2023YFC2415200);北京大学肿瘤医院科学研究基金(PY202335)。

摘　　要：目的腹膜转移是胃癌最常见的远处转移部位,预后较差。隐匿性腹膜转移病人缺乏影像学征象,术前诊断困难,给临床诊疗带来挑战。该研究通过决策树模型,建立一种基于临床特征的胃癌隐匿性腹膜转移预测工具。方法回顾性地纳入2015年1月至2019年12月期间在北京大学肿瘤医院胃肠肿瘤中心接受增强CT检查及腹腔镜探查的胃癌病人。通过单因素及多因素分析筛选与腹膜转移相关的临床病理特征,利用决策树方法构建诊断模型,并在训练集和验证集中评估模型效能。结果研究共纳入414例病人,其中63例为腹膜转移病人(15.22%)。多因素logistic回归分析显示,肿瘤长径(OR=1.023,95%CI:1.003~1.044,P=0.026)、cN3分期(OR=6.587,95%CI:1.362~31.846,P=0.019)、BorrmannⅣ型(OR=47.012,95%CI:4.903~1521.541,P=0.005)和糖类抗原(CA)125水平(OR=1.014,95%CI:1.002~1.027,P=0.021)是腹膜转移的独立危险因素。决策树模型包含6个分类变量,包括肿瘤长径、cN分期、Borrmann分型、CA125水平、CA19-9水平和Lauren分型。在验证集中腹膜转移诊断准确性为87.2%,敏感度为68.3%,特异度为90.6%,曲线下面积值为0.807(95%CI:0.766~0.847;P<0.001),阴性预测值为0.941。结论该研究建立了一种基于临床特征的决策树模型,对于CT阴性的胃癌隐匿性腹膜转移具有较好的诊断效能。Objective Peritoneal metastasis is the most common site of distant metastasis in gastric cancer and associated with poor prognosis.Occult peritoneal metastasis(OPM),characterized by the absence of imaging signs,is challenging to diagnose preoperatively,posing difficulties for clinical decision-making.This study aims to develop a prediction tool for OPM in gastric cancer based on clinical features using a decision tree model.Methods This was a retrospective study involving gastric cancer patients who underwent contrast-enhanced CT and laparoscopic exploration at the Gastrointestinal Oncology Center of Peking University Cancer Hospital from 2015 to 2019.Univariate and multivariate analyses were conducted to identify clinicopathological features associated with OPM.A decision tree model was constructed and its diagnostic performance was evaluated in both the training and validation sets.Results A total of 414 patients were included,of whom 63(15.22%)had peritoneal metastasis.Multivariate logistic regression analysis identified that tumor long diameter(OR=1.023,95%CI:1.003–1.044;P=0.026),cN3 staging(OR=6.587,95%CI:1.362–31.846;P=0.019),Borrmann typeⅣ(OR=47.012,95%CI:4.903–1521.541;P=0.005),and CA125 level(OR=1.014,95%CI:1.002–1.027;P=0.021)were independent risk factors for OPM in gastric cancer.The decision tree model incorporated six classification variables,including the tumor long diameter,cN stage,Borrmann classification,CA125 level,CA19-9 level,and Lauren classification.In the validation set,the model achieved a diagnostic accuracy of 87.2%,sensitivity of 68.3%,specificity of 90.6%,an area under the curve(AUC)of 0.807(95%CI:0.766–0.847;P<0.001),and a negative predictive value of 0.941.Conclusion This study developed a decision tree model based on clinical features,demonstrating good diagnostic performance for OPM in gastric cancer with negative CT findings.

关 键 词：胃癌 腹膜转移 决策树 危险因素 

分 类 号：R656.6[医药卫生—外科学]