基于seer数据库的食管癌骨转移临床特征分析及食管癌患者生存期预测模型的构建与验证.pdf

中文摘要

基于seer数据库的食管癌骨转移临床特征分析及食管癌患者生

存期预测模型的构建与验证

摘要

目的：本研究拟对食管癌骨转移相关因素进行分析并构建与验证食管

癌患者生存期的预测模型。

方法：从2010年至2016年的SEER数据库中提取了7596例诊断为

食管癌的患者数据（年龄，性别，种族，婚姻状况，肿瘤部位，肿瘤大小，

病理类型，组织学分级，T分期，N分期，手术情况，放疗情况，化疗情

况，骨转移情况）。用Logistic回归分析食管癌骨转移的相关因素，采用

Cox回归选择影响食管癌患者的预后因素并构建列线图预测模型。采用一

致性指数(Concordanceindex,C-index)、校准曲线和决策曲线分析(Decision

curveanalysis,DCAs)对列线图进行验证。

结果：206/7596（2.7%）食管癌患者发生骨转移。通过单因素Logistic

分析，食管癌骨转移与年龄、性别、肿瘤部位、肿瘤大小、T分期、N分

期、手术情况、放疗情况等因素相关。而多因素Logistic回归分析表明，

患病年龄越小、肿瘤体积越大、男性、食管下1/3部位的原发肿瘤、N3

分期是食管癌发生骨转移的独立危险因素；食管癌手术以及放射治疗可以

降低发生骨转移的风险。此外，多因素COX分析结果显示了影响食管癌

患者预后的潜在因素，包括年龄、性别、种族、婚姻状况、肿瘤部位、肿

瘤大小、病理、T分期、N分期、组织学分级、手术情况和化疗情况。多

因素COX回归分析的结果纳入列线图预测模型的构建，预测模型的

C-index为0.739（0.731~0.747），1年、2年、3年生存率的ROC曲线AUC

分别为0.81、0.81、0.82。

结论：年龄、性别、肿瘤部位、肿瘤大小、N分期、手术情况、放疗

情况被证明是影响食管癌骨转移的主要因素。通过基于SEER数据库构建

的食管癌患者的列线图，我们建立了一个预测模型，该模型展现出了良好

的区分度和校准度。这表明该模型具有良好的准确性和临床适用性。

关键词：食管癌，骨，肿瘤转移，SEER数据库，列线图

1

目录

中文摘要············································································································································1

英文摘要············································································································································2

英文缩写············································································································································4

研究论文基于seer数据库的食管癌骨转移临床特征分析及食管癌患者生

存期预测模型的构建与验证

前言············································································································································5

资料与方法····························································································································6