脐尿管癌生存的预测模型和改良分期系统的构建:基于SEER数据库的真实世界研究  

Establishment of prediction models and modified staging systems for the survival of urachal carcinoma:a real-world study based on SEER database

作　　者：肖卓慧 刘奕婷[1] 张阁钧[2] XIAO Zhuohui;LIU Yiting;ZHANG Gejun(Department of Healthcare Management,First Hospital of China Medical University,Shenyang,llo0o0,China;Department of Urology,First Hospital of China Medical University)

机构地区：[1]中国医科大学附属第一医院健康管理科,沈阳110000 [2]中国医科大学附属第一医院泌尿外科

出　　处：《临床泌尿外科杂志》2024年第3期184-191,共8页Journal of Clinical Urology

摘　　要：目的:基于SEER数据库构建并验证脐尿管癌患者的预后列线图模型并建立改良的分期系统。方法:回顾性检索2010—2018年脐尿管癌患者的临床信息,将患者队列随机以6:4比例分配为训练队列和验证队列。在单因素、多因素Cox回归分析以及双向逐步回归分析的基础上,筛选出独立危险因素并建立列线图。分别采用C指数、受试者工作特征(receiver operating characteristic,ROC)曲线和曲线下面积(area under curve,AUC)值、校准曲线、决策曲线分析(decision curve analysis,DCA)来验证列线图预测准确性。最后使用患者新列线图的评分结合生存信息建立新的分期系统。结果:共筛选出413例患者,分为训练队列247例,验证队列166例。经单因素、多因素Cox回归分析和双向逐步回归分析后,确定年龄、N分期、M分期、Mayo分期及Sheldon分期为脐尿管癌的独立预后因素(P<0.05),其中Mayo分期与Sheldon分期二者呈线性相关,因此分别以二者为基础建立列线图预测模型,C指数分别为0.71±0.02、0.71±0.02。对于3年、5年ROC曲线的AUC值,在以Mayo分期为基础的模型在训练集分别为75.1%,74.9%,在验证集分别为82.2%、72.2%;在以Sheldon分期为基础的模型在训练集分别为75.5%、75.5%,在验证集分别为80.5%、73.3%。校正曲线显示2个模型对患者预后的预测值和实际值一致性较好。DCA曲线显示列线图获益良好。基于列线图总分得到的新风险分期的生存曲线显示新的分期系统分层效果显著。结论:本研究分别基于Mayo分期和Sheldon分期构建了可准确预测脐尿管癌患者预后的列线图模型,并建立了新的分期系统。Objective: To establish and validate the prognosis nomogram models and modified staging systems of urachal carcinoma patients based on the SEER database. Methods: We retrospectively retrieved the clinical information of urachal carcinoma patients from the SEER database diagnosed from 2010 to 2018, then randomly assigned the patients cohort into training cohort and testing cohort in a ratio of 6∶4. The independent risk factors were screened and the nomograms were established based on the univariate, multivariate Cox regression analysis and bidirectional stepwise regression analysis. The C-index, receiver operating characteristic(ROC)curve, area under curve(AUC)value, calibration curve and decision curve analysis(DCA)curve were used to validate the prediction ability of the nomogram models. Lastly we constructed the new staging systems based on the nomogram scores combine with survival information. Results: A total of 413 patients were screened, 247 in the training cohort and 166 in the testing cohort. After the univariate, multivariate Cox regression and bidirectional stepwise regression analysis, the age, N stage, M stage, Mayo stage and Sheldon stage were determined as independent risk factors of the prognosis of urachal carcinoma patients(P<0.05). Mayo stage and Sheldon stage were linearly correlated with each other. Thus, we constructed the prognosis prediction nomogram models based on Mayo stage and Sheldon stage seperately, and the C-index were 0.71±0.02, 0.71±0.02. For the AUC values of 3-year and 5-year ROC curves, the Mayo-based model were 75.1% and 74.9% in the training cohort, 82.2% and 72.2% in the testing cohort;the Sheldon-based model were 75.5% and 75.5% in the training cohort, 80.5% and 73.3% in the testing cohort. The calibration curves showed satisfactory consistency between the predictive value and real value of prognosis. The DCA curves showed the new models benefited well. The survival curves of new staging systems established based on the nomogram points showed significant stratification

关 键 词：脐尿管癌 预后 列线图 SEER数据库 

分 类 号：R737.1[医药卫生—肿瘤]