婴幼儿视网膜母细胞瘤生存率预测列线图模型的构建

A nomogram model for predicting survival rate in infants with retinoblastoma

目的:构建预测婴幼儿视网膜母细胞瘤(RB)生存率的列线图。方法:收集2004年至2019年SEER数据库中929例RB患儿(0~3岁)的临床资料作为训练集,采用Cox回归分析筛选出婴幼儿RB生存的影响因素,并基于回归结果构建预测RB患儿3、5和10 a生存率的列线图。收集2010年至2019年郑州大学第一附属医院诊断的RB患儿111例为外部验证集。利用C指数、ROC曲线、校准曲线和决策曲线分析(DCA)评估列线图模型的预测价值。结果:以因RB死亡为终点事件,Cox回归分析结果显示,肿瘤基底直径≥15 mm、T3期、T4期、M1期和放化疗联合治疗是生存预后的危险因素,HR(95%CI)分别为3.049(1.459~7.782)、6.341(1.811~12.372)、9.554(6.054~13.989)、5.087(1.602~16.155)、10.452(1.180~19.598)。在训练集和外部验证集中,列线图模型预测的C指数(95%CI)分别为0.948(0.923~0.973)、0.858(0.728~0.987);在训练集中,列线图模型3、5和10 a生存率预测的AUC(95%CI)分别为0.966(0.946~0.986)、0.960(0.930~0.990)、0.934(0.906~0.962),在外部验证集中分别为0.825(0.769~0.881)、0.843(0.781~0.905)、0.868(0.796~0.940);校准曲线与标准曲线重合度较高,同时DCA显示模型具有理想的临床收益。结论:构建的列线图可用于婴幼儿RB生存率的预测,为临床个性化诊疗提供参考。

Aim:To construct a nomogram model for predicting survival rate in infants with retinoblastoma(RB).Methods:A total of 929 cases of RB in infancy(0-3 years old)from 2004 to 2019 in SEER datebase were collected as training set,Cox regression analysis was used to screen the influencing factors of survival rate,and nomogram for predicting 3-,5-,and 10-year survival rate was constructed based on the regression results.A total of 111 cases of infants with RB diagnosed at the First Affiliated Hospital of Zhengzhou University from 2010 to 2019 were collected as the external validation set.The predictive value of nomogram model was evaluated using C-index,ROC curve,calibration curve,and decision curve analysis(DCA).Results:With death due to RB as the endpoint event,tumor basal diameter≥15 mm,T3 stage,T4 stage,M1 stage,and combination of radiotherapy and chemotherapy were the risk factors,with HR(95%CI)of 3.049(1.459-7.782),6.341(1.811-12.372),9.554(6.054~13.989),5.087(1.602-16.155),10.452(1.180-19.598).The C-index(95%CI)of the nomogram model in the training set and in the external validation set,was 0.948(0.923-0.973)and 0.858(0.728-0.987),respectively.In the training set,AUC(95%CI)for the prediction of 3-,5-,and 10-year survival rate was 0.966(0.946-0.986),0.960(0.930-0.990),0.934(0.906-0.962),respectively,in the external validation set was 0.825(0.769-0.881),0.843(0.781-0.905),0.868(0.796-0.940),respectively.The calibration curve overlapped well with the standard curve in the training set and the external validation set,and DCA showed that the predictive model had the desired clinical effect.Conclusion:The nomogram constructed in this study can be used for the prediction of survival rate of infants with RB,and can provide a reference for individualized clinical treatment.