胰腺导管腺癌腹腔隐匿性转移术前预测模型的构建

Construction of preoperative predictive model for occult abdominal metastatic disease of pancreatic ductal adenocarcinoma

目的建立能有效预测胰腺导管腺癌腹腔隐匿性转移发生风险的术前预测模型。方法回顾性分析2018年9月至2020年12月我科收治的986例行手术治疗的胰腺导管腺癌患者的临床资料,将单因素分析中P≤0.2的变量纳入二元logistic回归模型,筛选出腹腔隐匿性转移的独立预测因素,并据此建立列线图预测模型。使用ROC曲线评估该预测模型预测腹腔隐匿性转移的能力。结果986例胰腺导管腺癌患者中腹腔隐匿性转移的发生率为8.42%(83/986);男593例(60.14%)、女393例(39.86%);年龄为(62.40±9.43)岁。单因素分析显示,腹腔隐匿性转移与腹痛、腹痛时间、总胆红素、前白蛋白、丙氨酸转氨酶、天冬氨酸转氨酶、碱性磷酸酶、乳酸脱氢酶、癌胚抗原、甲胎蛋白、糖类抗原(CA)125、CA19-9、CA724、腹水、肿瘤大小、肿瘤部位、突破被膜、侵犯周围脏器、腹腔干侵犯程度、肠系膜上动脉侵犯程度、脾动脉侵犯程度、脾静脉侵犯程度、第9组可疑淋巴结转移、第13组可疑淋巴结转移、第14组可疑淋巴结转移、第17组可疑淋巴结转移、腹膜后存在可疑淋巴结转移、肝可疑转移瘤、肝可疑转移瘤位置、腹腔不确定性病灶、腹腔手术史有关(P均<0.05)。多因素二元logistic回归分析显示,年龄、腹水、肿瘤短径、侵犯周围脏器、肠系膜上动脉侵犯程度、第13组可疑淋巴结转移、肝可疑转移瘤、腹腔不确定性病灶、腹腔手术史是胰腺导管腺癌发生腹腔隐匿性转移的独立预测因素。基于上述独立预测因素构建列线图预测模型,该模型预测胰腺导管腺癌腹腔隐匿性转移的ROC AUC值为0.783(P=0.001),最佳临界风险评分为77.68,灵敏度为0.650,特异度为0.787。结论本研究建立的列线图预测模型有助于提高胰腺导管腺癌腹腔隐匿性转移的术前诊出率。

Objective To establish a preoperative prediction model which can effectively predict the risk of occult abdominal metastatic disease of pancreatic ductal adenocarcinoma.Methods The clinical data of 986 patients with pancreatic ductal adenocarcinoma who underwent surgery in our department from Sep.2018 to Dec.2020 were retrospectively analyzed.The variables with P≤0.2 in univariate analysis were included in binary logistic regression model,and independent predictors of occult abdominal metastatic disease were screened out;finally a nomogram prediction model was established.Then receiver operating characteristic(ROC)curve was used to evaluate the prediction efficiency.Results The incidence of occult abdominal metastatic disease was 8.42%(83/986)in the pancreatic ductal adenocarcinoma patients(593[60.14%]males and 393[39.86%]females).The average age of the patients was(62.40±9.43)years old.The results of univariate analysis showed that occult abdominal metastatic disease was associated with abdominal pain,duration of abdominal pain,total bilirubin,prealbumin,alanine aminotransferase,aspartate aminotransferase,alkaline phosphatase,lactate dehydrogenase,carcinoembryonic antigen,α-fetoprotein,carbohydrate antigen(CA)125,CA19-9,CA724,ascites,tumor size,tumor site,breakthrough of capsule,the invasion of surrounding organs,the degree of invasion of celiac trunk,superior mesenteric artery,splenic artery and splenic vein,suspected metastasis in the No.9,No.13,No.14 and No.17 lymph nodes,suspected metastasis of retroperitoneal lymph nodes,suspected metastatic tumor and the sites in the liver,abdominal lesions of unknown nature,and history of abdominal surgery(all P<0.05).Multivariate analysis showed that age,ascites,short cross-sectional diameter of tumor,the invasion of surrounding organs,the degree of invasion of superior mesenteric artery,suspected metastasis in the No.13 lymph nodes,suspected liver metastases,uncertain lesions in abdominal cavity and history of abdominal surgery were independent predictors of occult abdominal metastatic disease.Based on the above independent predictors,a nomogram prediction model has been constructed,and the area under the ROC curve was 0.783(P=0.001).The optimal risk score,sensitivity and specificity of the model were 77.68,0.650 and 0.787,respectively.Conclusion The nomogram prediction model can help to improve the preoperative diagnosis rate of occult abdominal metastatic disease of pancreatic ductal adenocarcinoma.