小于2岁神经母细胞瘤儿童的特异性生存列线图构建与验证：一项基于SEER数据库的研究

A Nomogram for Predicting Cancer-Specific Survival of Neuroblastoma in Children < 2 Years Old: A SEER Database Analysis

摘要: 背景：神经母细胞瘤(NB)是儿童最常见的非中枢神经系统实体肿瘤，占所有儿童恶性肿瘤的8~10%。我们的目的是探讨儿童神经母细胞瘤的预后因素，并构建一个列线图(nomogram)来预测2岁以下儿童神经母细胞瘤的癌症特异性生存(CSS)。方法：从监测、流行病学和最终结果(SEER)数据库下载2000年至2018年2岁以下神经母细胞瘤儿童的临床病理信息。所有患者被随机分配到训练队列(70%)或验证队列(30%)。采用最小绝对收缩和选择算子(LASSO)回归和逐步回归多变量Cox回归分析筛选2岁以下神经母细胞瘤患儿癌症特异性生存(CSS)的独立危险因素。选取的危险因素构建列线图(nomogram)来预测神经母细胞瘤患儿(<2岁)1年、3年和5年时的CSS。列线图的预测精度和判别能力由一致性指数(C-index)、受者工作曲线下面积(AUC)和校正曲线决定。采用决策曲线分析(Decision curve analysis, DCA)和Kaplan-Meier估计量验证预测模型的临床应用效果。结果：研究对象包括631名年龄小于2岁的神经母细胞瘤患儿。LASSO回归和逐步回归多因素Cox回归分析显示，年龄、肿瘤大小、肿瘤分期和远处转移手术是2岁以下神经母细胞瘤患儿发生CSS的独立危险因素。我们构建了一个列线图(nomogram)来预测神经母细胞瘤患者(<2岁)的CSS。训练组和验证组的c指数分别为0.888 (95% CI: 0.849~0.927)和0.885 (95% CI: 0.818~0.952)。训练队列和验证队列的AUC也表明该预测模型具有较好的准确性。修正曲线表明，预测模型的观测值与预测值高度吻合。DCA具有良好的临床应用价值。结论：在这项研究中，我们探讨了小于2岁神经母细胞瘤患者的预后因素。我们发现年龄、肿瘤大小、肿瘤分期和远处转移手术是CSS的独立危险因素。我们构建了一个列线图(nomogram)来预测2岁以下NB患儿的CSS。该预测模型可为医生和患者的临床决策提供参考。

Abstract: Background: Neuroblastoma(NB) is the most common non-central nervous system solid tumor in children, accounting for 8~10% of all childhood malignancies. We aim to explore prognostic factors of NB in the children and construct a nomogram to predict cancer-specific survival (CSS) in children less than 2 years old with NB. Methods: The clinicopathological information of children (<2 years old) with NB from 2000 to 2018 was downloaded from the Surveillance, Epidemiology, and End Results (SEER) database. All patients were randomly assigned to a training cohort (70%) or a validation cohort (30%). Least absolute shrinkage and selection operator (LASSO) regression and stepwise backward multivariate Cox regression analysis were used to screen the independent risk factors of cancer-specific survival (CSS) in children (<2 years old) with NB. The selected risk factors were used to construct a nomogram to predict the CSS at 1-, 3-, and 5 years in children (<2 years old) with NB. The prediction accuracy and discriminant ability of nomogram are determined by consistency index (C-index), the area under the receiver operating curve (AUC) and correction curve. Decision curve analysis (DCA) and Kaplan-Meier estimator were used to verify the clinical application effect of the predictive model. Results: The subjects included 631 children aged less than 2 years with NB. LASSO regression and stepwise backward multivariate Cox regression analysis showed that age, tumor size, tumor stage and distant surgery were independent risk factors for CSS in children (<2 years old) with NB. We constructed a nomogram to predict CSS in patients(<2 years old) with NB. The C-index of the training cohort and validation cohort was 0.888 (95% CI: 0.849~0.927) and 0.885 (95% CI: 0.818~0.952), respectively. The AUC of training queue and verification queue also shows that the prediction model has good accuracy. The correction curve shows that the observed value of the prediction model is highly consistent with the predicted value. DCA shows good clinical application value. Conclusions: We found that age, tumor size, tumor stage, and distant surgery were independent risk factors for patient with NB. We constructed a new nomogram to predict CSS in children < 2 years old with NB.