神经母细胞瘤患儿生存预测模型的构建和验证  

Establishment and validation of a survival prediction model for children with neuroblastoma

作　　者：陈伟明[1] 白建喜 张炳 吴典明[1] 方一凡 Chen Weiming;Bai Jianxi;Zhang Bing;Wu Dianming;Fang Yifan(Department of General Surgery,Fujian Children’s Hospital,College of Clinical Medicine for Obstetrics&Gynecology and Pediatrics,Fujian Medical University,Fuzhou 350011,China)

机构地区：[1]福建省儿童医院普外科福建医科大学妇儿临床医学院,福州350011

出　　处：《临床小儿外科杂志》2025年第3期243-249,共7页Journal of Clinical Pediatric Surgery

基　　金：福建省自然科学基金(2023J011300);福建医科大学启航基金(2021QH1168)。

摘　　要：目的构建用于评估神经母细胞瘤患儿生存预后的临床预测模型。方法从美国国立癌症研究所的“监测、流行病学和最终结果”(Surveillance,Epidemiology,and End Results,SEER)数据库中收集2000—2019年间公开发布的神经母细胞瘤患儿临床资料,将筛选后的数据按2∶1的比例随机划分为训练集和验证集。采用单因素和多因素Cox比例风险回归模型确定预后相关因素,构建神经母细胞瘤预后预测模型并绘制列线图。采用一致性指数(C指数)、受试者操作特征(receiver operating characteristic,ROC)曲线下面积(area under curve,AUC)和校准曲线评估列线图的辨别和校准能力,采用决策曲线分析(decision curve analysis,DCA)评价列线图的临床实用性。结果多因素Cox回归分析结果显示:年龄1~5岁(HR=3.595,95%CI:2.122~6.091)或≥5岁(HR=4.556,95%CI:2.566~8.091)、肿瘤长径≥7.5 cm(HR=1.934,95%CI:1.347~2.777)、肿瘤位于肾上腺和腹膜后(HR=1.520,95%CI:1.029~2.247)、肿瘤转移(HR=5.263,95%CI:2.146~12.907)和未接受手术治疗(HR=1.866,95%CI:1.225~2.833)为神经母细胞瘤患儿预后不良的独立危险因素。将上述变量纳入预测模型并绘制列线图,进行1、3、5年的总生存率预测。预测模型在训练集和验证集中的C指数分别为0.803和0.785,训练集中ROC曲线在1、3、5年的总生存率预测AUC值分别为0.782、0.829、0.842,验证集中分别为0.748、0.807、0.855。DCA曲线表明,列线图的预测概率阈值在0.1~0.6之间时,使用列线图预测神经母细胞瘤患儿的预后更准确。结论年龄≥1岁、肿瘤长径≥7.5 cm、肿瘤原发于肾上腺和腹膜后、存在远处转移和未接受手术治疗是神经母细胞瘤患儿预后不良的独立危险因素,据此构建的预测模型能较为准确地预测神经母细胞瘤患儿预后,为临床决策提供参考。Objective To establish and validate a clinical prediction model assessing the survival of children with neuroblastoma(NB).Methods The relevant clinical data were collected from NB children between 2000 and 2019 from the database of Surveillance,Epidemiology and End Results(SEER).The primary cohort was randomized into two sets of training(n=987)and validation(n=493).The prognostic effects of variables were examined with a Cox proportional hazard model.Nomograms were established for estimating 1/3/5-year overall survival(OS)based upon Cox regression model.Concordance index(C-index),area under time-dependent receiver operating characteristics curve(AUC-ROC),calibration plotting and decision-curve analysis(DCA)were utilized for evaluating the model.Results The results of the multivariate Cox regression analysis indicated that 1-5 years(HR=3.595,95%CI:2.122-6.091)or≥5 years(HR=4.556,95%CI:2.566-8.091),tumour length≥7.5 cm(HR=1.934,95%CI:1.347-2.777),tumour located in the adrenal and retroperitoneal(HR=1.520,95%CI:1.029-2.247),tumour metastasis(HR=5.263,95%CI:2.146-12.907),and not receiving surgical treatment(HR=1.866,95%CI:1.225-2.833)were identified as independent risk factors for poor prognosis in children with neuroblastoma.A nomogram was constructed based on the aforementioned independent prognostic factors for the purpose of overall survival prediction at 1/3/5-year in children with neuroblastoma.C-index of prediction model in training and validation sets were 0.803 and 0.785 respectively.AUC of ROC curves in training set at 1/3/5-year overall survival were 0.782,0.829 and 0.842 respectively.Validation set revealed that AUC of ROC curve of 1/3/5-year overall survival were 0.748,0.807 and 0.855 respectively.DCA indicated that using the nomogram for predicting the prognosis of NB children was more beneficial at a prediction probability threshold of(0.1-0.6).Conclusions Nomogram is a reliable and powerful tool for distinguishing and predicting the survival of NB children,thus assisting clinicians in making clini

关 键 词：神经母细胞瘤 列线图 预后 外科手术 儿童 

分 类 号：R739.4[医药卫生—肿瘤]