Gao Xiangyu, Liu Wenjun
Department of Pediatrics, Laboratory of Hematologic Tumors and Birth Defects in Children, Affiliated Hospital of Southwest Medical University, Birth Defects Clinical Medical Research Center of Sichuan Province, Luzhou, Sichuan, China.
Medicine (Baltimore). 2020 May;99(19):e20115. doi: 10.1097/MD.0000000000020115.
B lymphocytic leukemia (B-ALL) is a hematopoietic malignant disease characterized by an accumulation of early B cells. This study aimed to construct a children B-ALL Nomogram prediction model based on Therapeutically Applicable Research to Generate Effective Treatments database, so as to further guide clinical diagnose and treatment.Clinical data related to children B-ALL were collected from the TARGET database, among which, the stage II clinical data were used as the prediction model, while the stage I clinical data were utilized as the external verification model. The stage II clinical factors were analyzed through Lasso regression analysis to screen the risk factors for the construction of Nomogram prediction model. In addition, the model prediction capacity and accuracy were verified internally and externally using the ROC curve, C-index and calibration curve, respectively.A total of 1316 B-ALL children were enrolled in this study. Lasso regression analysis revealed that, Age, Gender, WBC, CNSL, MRD29, BMR, CNS R, BCR-ABL1, BMA29, DS, and DI were the important prognostic risk factors. The C-index values of internal and external verification models were 0.870 and 0.827, respectively, revealing the ideal model discriminating capacity. Besides, the calibration curve had high contact ratio, which suggested favorable consistency between the incidence predicted by the model and the actual incidence. Moreover, the AUC values of the ROC curve were 0.858, 0.787, 0.898, and 0.867, respectively, indicating high model prediction accuracy in predicting the 3- and 5-year survival rates of children with B-ALL.The Nomogram prediction model plotted in this study exhibits favorable prediction capacity and clinical practicability for the survival rate of B-ALL children, which contributes to patients screening and clinical intervention.
B淋巴细胞白血病（B-ALL）是一种以早期B细胞积聚为特征的造血系统恶性疾病。本研究旨在基于“生成有效治疗方案的治疗应用研究”（TARGET）数据库构建儿童B-ALL列线图预测模型，以进一步指导临床诊断和治疗。从TARGET数据库收集与儿童B-ALL相关的临床数据，其中，II期临床数据用作预测模型，而I期临床数据用作外部验证模型。通过Lasso回归分析对II期临床因素进行分析，以筛选用于构建列线图预测模型的危险因素。此外，分别使用ROC曲线、C指数和校准曲线对模型预测能力和准确性进行内部和外部验证。本研究共纳入1316例B-ALL儿童。Lasso回归分析显示，年龄、性别、白细胞计数、中枢神经系统白血病（CNSL）、微小残留病29（MRD29）、骨髓缓解率（BMR）、中枢神经系统复发（CNS R）、BCR-ABL1、骨髓侵犯29（BMA29）、危险度（DS）和诱导分化指数（DI）是重要的预后危险因素。内部和外部验证模型的C指数值分别为0.870和0.827，显示出理想的模型区分能力。此外，校准曲线的拟合度高，表明模型预测的发病率与实际发病率之间具有良好的一致性。此外，ROC曲线的AUC值分别为0.858、0.787、0.898和0.867，表明该模型在预测B-ALL儿童3年和5年生存率方面具有较高的预测准确性。本研究绘制的列线图预测模型对B-ALL儿童的生存率具有良好的预测能力和临床实用性，有助于患者筛选和临床干预。
