滤泡状甲状腺癌术前预测模型的建立与验证  

Establishment and validation of preoperative prediction model for follicular thyroid carcinoma

作　　者：王晓娜 汪威[1] 刘倩[2] 程元甲 张云 朱立东[1] WANG Xiao-na;WANG Wei;LIU Qian;CHENG Yuan-jia;ZHANG Yun;ZHU Li-dong(Department of General Surgery,Beijing Puren Hospital,Beijing 100062,China;Department of No.Four Wards of General Surgery,First Affiliated Hospital of Peking University,Beijing,100034,China;Department of Special Emergency Surgery,Characteristic Medical Center of PAPF,Tianjin 300162,China)

机构地区：[1]北京市普仁医院普外科,北京100062 [2]北京大学附属第一医院外科四病区,北京100034 [3]武警特色医学中心特勤急救外科,天津300162

出　　处：《中国现代普通外科进展》2022年第9期702-706,共5页Chinese Journal of Current Advances in General Surgery

摘　　要：目的:构建术前滤泡状甲状腺癌(FTC)预测模型,并验证预测效果。方法:回顾性分析2015年1月—2020年12月模型队列(北京大学第一医院患者)和验证队列(北京市普仁医院患者)接受手术治疗的甲状腺结节患者的术前资料。根据术后病理分别将两队列患者分为FTC组和非FTC组。在模型队列使用Logistic回归分析,探索FTC发生的危险因素,根据危险因素构建术前FTC预测模型。使用受试者工作特征(ROC)曲线确定最佳诊断切点。分别在模型队列和验证队列中验证预测模型预测FTC的准确度。结果:模型队列患者术前血清甲状腺球蛋白(Tg)水平升高(P<0.001)、结节直径大(P<0.001)、无钙化(P=0.003)、结节呈实性(P<0.001)、结节边缘平滑(P=0.035)和结节可见血供(P<0.001)是甲状腺结节患者最终被诊断为FTC的危险因素。模型队列中根据Logistic回归因素和系数构建预测FTC模型:Y=0.013×Tg(ng/mL)+0.7×结节直径(cm)+0.919×(钙化状态)+2.355×(结节性质)+1.073×(结节边缘)+1.424×(结节血供)-10.681。使用ROC曲线分析得出最佳诊断切点为Y=-3.96,此时诊断FTC的准确度为91.53%,敏感度为87.16%。在验证队列中,预测模型Y的诊断敏感度为90.91%、特异度为90.99%。模型队列和验证队列中,预测模型判断FTC准确率均高于术中冷冻切片(P<0.001)。结论:利用术前血清Tg水平和甲状腺结节超声特征,所构建的FTC术前预测模型预测效果可靠。Objective:To construct a predictive model for preoperative follicular thyroid carcinoma(FTC)and verify the predictive effect.Methods:From January 2015 to December 2020,the preoperative data of patients with thyroid nodules in the model cohort(Peking University First Hospital)and the validation cohort(Beijing Puren Hospital)were retrospectively analyzed.According to the postoperative pathology,the patients were divided into FTC group and non-FTC group.Logistic regression analysis was used in the model cohort to explore the risk factors of FTC,and the preoperative FTC prediction model was constructed according to the risk factors.Receiver operating characteristic(ROC)was used to determine the diagnostic cut-off point.The accuracy of FTC prediction model is verified in model queue and validation queue.Results:The risk factors of FTC in patients with thyroid nodules were increased serum thyroglobulin(Tg)level(P<0.001),large nodule diameter(P<0.001),without calcification(P=0.003),solid nodule(P<0.001),smooth nodule edge(P=0.035)and visible blood supply(P<0.001).In the model cohort,the predictive FTC model based on logistic regression factors and coefficients was Y=0.013×Tg(ng/ml)+0.7×nodule diameter(cm)+0.919×(calcification)+2.355×(nodule nature)+1.073×(nodule margin)+1.424×(nodule blood supply)-10.681.ROC curve analysis showed that the best cut-off point was Y=-3.96,which the accuracy and sensitivity of FTC diagnosis were 91.53%and 87.16%.In the validation cohort,the diagnostic sensitivity and specificity of prediction model y were 90.91%and 90.99%.In the model cohort and validation cohort,the accuracy of predicting FTC was higher than that of intraoperative frozen section(P=0.001).Conclusion:Based on the preoperative serum Tg level and ultrasonic characteristics of thyroid nodules,the prediction model of FTC is reliable.

关 键 词：预测模型 滤泡性甲状腺癌 术前预测 甲状腺球蛋白 

分 类 号：R736.1[医药卫生—肿瘤]