基于SEER数据库的下咽鳞状细胞癌患者预后列线图模型构建及验证  

Construction and validation of prognostic nomogram model for patients with hypopharyngeal squamous cell carcinoma based on SEER database

作　　者：胡晶[1] 廖红明 何本超[1] Hu Jing;Liao Hongming;He Benchao(Department of Otorhinolaryngology Head and Neck Surgery,Tianmen First People's Hospital,Tianmen 431700,China)

机构地区：[1]天门市第一人民医院耳鼻咽喉头颈外科,天门431700

出　　处：《国际耳鼻咽喉头颈外科杂志》2025年第1期6-13,共8页International Journal of Otolaryngology-Head and Neck Surgery

摘　　要：目的构建及验证预测下咽癌患者预后的列线图。方法从SEER数据库下载和提取2010年1月至2015年12月下咽癌患者的临床资料,随机分为建模组和验证组,最小绝对值选择与收缩算子(lasso)回归对自变量进行筛选,最小误差处自变量个数为7,进行多因素Cox回归,选择向后LR法筛选自变量,最终纳入5个自变量,包括:T分级、M分级、年龄、骨转移、AJCC分期,构建预后模型A;建模组通过lasso回归筛选自变量,选择1se处自变量个数为3,包括:T分级、M分级、年龄,构建预后模型B;构建T分级、N分级、M分级模型定义为模型C,通过不同模型间相互比较,选择最佳模型。构建列线图预测下咽癌患者1年、3年、5年的生存概率,采用一致性指数(c-index)、校准曲线、受试者工作特征(receiver operating characteristic,ROC)曲线下面积(AUC)评估模型的区分度及准确度,临床决策曲线评估模型的临床适用价值。结果共纳入1941例患者,包括建模组(n=1370)和验证组(n=571)。将模型A与模型B进行对比,P>0.05,模型A赤池信息准则(AIC=9844.6406)大于模型B赤池信息准则(AIC=9801.332),故选择模型B为下咽癌预后模型。经lasso回归和Cox回归分析表明,T分级、M分级、年龄是影响下咽癌的独立危险因素。模型B建模组和验证组的c-index分别为0.66(95%CI:0.63~0.68)和0.66(95%CI:0.62~0.69),建模组1年、3年、5年对应的ROC曲线AUC分别为0.699、0.692、0.693;验证组1年、3年、5年对应的ROC曲线AUC分别为0.698、0.667、0.677,建模组和验证组的AUC结果表明,该列线图具有良好的准确性。校准曲线表明观察值和预测值一致。结论T分级、M分级、年龄是下咽癌患者预后的独立影响因素,构建的列线图比传统的TNM分期系统具有更好的临床应用价值。Objective Construction and validation of nomogram for predicting the prognosis of patients with hypopharyngeal cancer.Methods The clinical data of patients diagnosed with pharyngeal cancer between January 2010 and December 2015 are from the SEER database.The dataset was then divided into a modeling group and a validation group using random sampling.The least absolute shrinkage and selection operator(lasso)regression technique was employed to identify independent variables.At the point of minimum error,the number of independent variables identified was seven.Subsequently,multivariate cox regression was conducted,utilizing the backward LR method to screen the independent variables.Ultimately,a total of five independent variables were included in the analysis,namely T stage,M stage,age,bone metastasis,AJCC staging,and the construction of prognostic model A.The modeling group employed lasso regression to identify independent variables,ultimately selecting three variables at 1se,namely T stage,M stage,and age.The T stage,N stage,and M stage models were defined as Model C,and a comparative analysis was conducted to determine the optimal model.Develop a nomogram to estimate the likelihood of survival for individuals with hypopharyngeal cancer at 1,3,and 5 years.Subsequently,assess the model's discriminatory power and accuracy through the utilization of the c-index,calibration curve,and area under the receiver operating characteristic curve(AUC).Additionally,employ clinical decision curve analysis(DCA)to evaluate the practicality of the model in a clinical setting.Results A total of 1941 patients were included in the study,divided into the modeling group(n=1370)and the validation group(n=571).Comparing Model A with Model B,the statistical analysis indicated that the Akashi information criterion(AIC=9844.6406)of Model A was significantly higher than that of Model B(AIC=9801.332)with a p-value greater than 0.05.Consequently,Model B was selected as the preferred prognosis model for hypopharyngeal cancer.The results of the l

关 键 词：下咽肿瘤 癌 鳞状细胞 预后 SEER数据库 总生存率 Nomogram模型 

分 类 号：R73[医药卫生—肿瘤]