川崎病患儿丙种球蛋白耐药列线图模型的构建与验证

Construction and validation of a nomogram model of intravenous immunoglobulin resistance in Kawasaki disease based on Lasso regression

摘要:

目的 基于Lasso回归和列线图构建并验证川崎病患儿对丙种球蛋白耐药的预测模型，以期为临床诊疗提供帮助。方法 回顾性收集2014年7月—2020年7月西南医科大学附属医院收治的474例川崎病患儿的临床资料，采用Lasso回归分析筛选重要的临床因素构建Nomogram模型，通过绘制受试者工作特征（ROC）曲线、Calibration校准曲线及DCA曲线验证模型的区分度、校准度及临床有效性。结果 共纳入474例患儿资料，其中339例作为训练集，135例作为验证集。Lasso回归分析显示，心脏表现、心外并发症、首剂静脉注射免疫球蛋白使用时间、中性粒细胞比例、红细胞分布宽度-标准差、血小板压积、白蛋白、系统性免疫-炎症指数及C反应蛋白/白蛋白是川崎病患儿丙种球蛋白耐药的预测因素。基于上述预测因素构建Nomogram模型，并分别在训练集与验证集人群中进行验证。训练集ROC曲线下面积（AUC）为0.784（95% CI：0.701，0.867），当最佳阈值取0.045时，相应的特异性和敏感性分别为0.490（95% CI：0.434，0.546）和0.935（95% CI：0.849，1.000）；验证集的AUC为0.784（95% CI：0.643，0.925），当最佳阈值取0.142时，相应的特异性和敏感性分别为0.851（95% CI：0.788，0.915）和0.714（95% CI：0.478，0.951）。Calibration校准曲线在训练集和验证集人群的C值分别为0.784和0.784，P 值分别为0.953和0.251。DCA曲线显示，Pt值在0.01～0.58时，训练集存在临床净获益。结论 川崎病丙种球蛋白耐药列线图Nomogram预测模型方便临床使用，有助于早期识别丙种球蛋白耐药的高危患儿。

Abstract:

Objective To construct and validate a predictive model for intravenous immunoglobulin (IVIG) resistance in children with Kawasaki disease (KD) based on Lasso regression and nomogram, aiming to assist clinical diagnosis and treatment.Methods Clinical data from 474 children with KD treated at the Affiliated Hospital of Southwest Medical University from July 2014 to July 2020 were retrospectively collected. Lasso regression analysis was used to select important clinical factors to build the Nomogram model. The model's discrimination, calibration, and clinical effectiveness were verified through the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA).Results A total of 474 cases were included, with 339 cases in the training set and 135 cases in the validation set. Lasso regression analysis identified cardiac manifestations, extracardiac complications, time of first intravenous immunoglobulin use, neutrophil ratio, red cell distribution width-standard deviation, platelet crit, albumin, systemic immune-inflammatory index, and C-reactive protein/albumin as predictive factors for IVIG resistance in children with KD. A Nomogram model was constructed based on these predictive factors and validated in both the training and validation sets. The area under the ROC curve (AUC) in the training set was 0.784 (95% CI: 0.701, 0.867), with a specificity of 0.490 (95% CI: 0.434, 0.546) and sensitivity of 0.935 (95% CI: 0.849, 1.000) at the optimal threshold of 0.045. The AUC in the validation set was 0.784 (95% CI: 0.643, 0.925), with a specificity of 0.851 (95% CI: 0.788, 0.915) and sensitivity of 0.714 (95% CI: 0.478, 0.951) at the optimal threshold of 0.142. The C-values in the calibration curve for the training and validation sets were 0.784 and 0.784, with P-values of 0.953 and 0.251, respectively. The DCA curve showed a clinical net benefit in the training set when the threshold probability (Pt) ranged from 0.01 to 0.58.Conclusion The Lasso regression Nomogram model for predicting IVIG resistance in KD is convenient for clinical use and helps identify high-risk children for IVIG resistance early.