基于SEER数据库的滤泡状甲状腺癌发生远处转移风险预测模型  

A Risk Prediction Model for Distant Metastasis of Follicular Thyroid Cancer Based on SEER Database

作　　者：沈刘海 章诗琪[2] SHEN Liuhai;ZHANG Shiqi(Department of Nuclear Medicine,Anhui No.2 Provincial People’s Hospital,Hefei 230041,China;Department of Endocrinology,First Affiliated Hospital of Anhui Medical University,Hefei 230032,China)

机构地区：[1]安徽省第二人民医院核医学科,安徽合肥230041 [2]安徽医科大学第一附属医院内分泌科,安徽合肥230032

出　　处：《标记免疫分析与临床》2024年第4期692-697,共6页Labeled Immunoassays and Clinical Medicine

基　　金：国家自然科学基金青年基金资助项目(编号:81900746)。

摘　　要：目的探讨滤泡状甲状腺癌(FTC)发生远处转移的风险预测的模型。方法筛选2004~2015年SEER数据库中被诊断为滤泡状甲状腺癌的患者信息。符合纳入和排除标准的1275例患者为研究对象,分析各项预测因素,利用ROC获取年龄及诊断至治疗间隔时间(TDT)的最佳截断值。选择单因素分析中有统计学意义的变量采用Lasso回归及十折交叉验证初步筛选变量,再用多因素Logistic回归构建预测模型,并用ROC曲线及AUC评价模型诊断效能。结果4.6%的FTC(59例)发生远处转移。其中性别、年龄、是否微浸润癌、肿瘤分期(T分期)、淋巴结分期(N分期)、TDT、淋巴结转移比率(LNR)、肿瘤大小、原发灶浸润情况及是否多灶性10个变量单因素分析中差异有统计学意义,Lasso回归及十折交叉验证筛选出其中6个变量纳入多因素Logistic回归,采用逐步法向前向后法筛选年龄、T分期、TDT、LNR4个变量纳入最终模型。模型ROC的曲线下面积(AUC)为0.8815,95%CI为0.8353~0.9276。结论年龄、T分期、TDT、LNR 4个变量构建的预测模型有很高的预测价值,可以为临床诊疗决策提供帮助,其中LNR及TDT对FTC发生远处转移的预测,在以往的研究中不曾有过,亦是本文的亮点。Objective This study aims to delineate the prediction of distant metastasis in follicular thyroid cancer.Methods A total of 1,275 patients diagnosed as follicular thyroid cancer in the SEER database from 2004 to 2015 were selected for the study.Various predictive factors were evaluated.Furthermore,the optimal cut-off of age and the duration from diagnosis to treatment(TDT)were obtained by ROC analysis.Statistically significant variables in univariate analysis was preliminarily screened using Lasso regression and ten-fold cross validation.A predictive model was then constructed using multivariate logistic regression.The diagnostic effectiveness of the model using ROC curves and AUC was finally assessed.Results 4.6%of FTC(59 cases)experienced distant metastasis in our study.Among them,10 variables in univariate analysis reached statistical significance,including gender,age,whether there was micro invasive cancer,T stage,N stage,TDT,LNR,tumor size,primary lesion infiltration and multifocal.Lasso regression and ten-fold cross validation identified 6 variables and the stepwise forward backward method was subsequently used.Age,T stage,TDT,and LNR were selected for the final model.AUCs of this model was 0.8815,with a 95%CI of 0.8353-0.9276.Conclusion The predictive model constructed with age,T-stage,TDT,and LNR variables is highly predictive which might favor clinical diagnosis and treatment for FTC.The prediction of distant metastasis of FTC by LNR and TDT has not been reported in previous studies,which is a highlight of this article.

关 键 词：滤泡状甲状腺癌 SEER数据库 预测模型 

分 类 号：R736.1[医药卫生—肿瘤]