川崎病丙种球蛋白无反应评分模型的建立与研究

Establishment of a new scoring model for IVIG non-response of Kawasaki disease

目的 基于数字科研(E-Science)模式下,通过上海市儿童医院和上海俊泽软件有限公司共同研发的临床数字科研平台分析川崎病(Kawasaki disease,KD)患者初始静脉注射丙种球蛋白(intravenous immunoglobulin,IVIG)治疗无反应的可能危险因素,并通过对危险因素加权积分建立数学模型,探讨有效预测MG无反应的方法,为及时有效治疗及判断疾病预后提供临床依据.方法 收集2013年1月至2016年11月于上海市儿童医院住院的595例川崎病患儿数据,采集指标包括患儿性别、年龄、发热至第一次使用IVIG时间以及相关实验室检查结果.采用多因素逻辑回归进行IVIG无反应的影响因素分析,预测模型中的指标根据逻辑回归方程的自变量系数进行赋分,对新的预测模型做ROC曲线,并计算曲线下面积,截断值的选取采用约登指数最高的点,并根据截断值计算新的预测模型的灵敏度和特异度,最后将新建模型与Kobayashi和Egami评分模型进行比较.结果 本研究共纳入川崎病患儿595例,其中CRP、NLR、LDH、ALB、FDP水平是川崎病患儿IVIG无反应的影响因素(P＜0.05).根据逻辑回归方程系数建立评分模型,该模型预测丙球无反应型川崎病的灵敏度和特异度分别为69.7％和80.4％,ROC曲线下面积为0.825(95％ CI:0.769-0.882),将得分为6分设置为截断点,大于或等于6分为阳性,小于6分为阴性.用Kobayashi和Egami评分模型进行测试,发现新的评分系统ROC曲线下面积以及灵敏度和特异度都较高.结论 本研究发现新的评分模型对于预测川崎病患儿IVIG无反应具有良好的效果,值得临床推广并以大规模数据进行验证和调整.

Objective To analyze the possible risk factors of IVIG non-response of Kawasaki disease (KD),Shanghai Children's Hospital and Shanghai Junze Software develop an research platform,which is based on E-Science model.Through the mathematical model by integrating the risk factors to explore the method of effective prediction for IVIG non-response,and to provide the clinical basis for timely and effective treatment and prognosis of the disease.Methods The data of KD children who were hospitalized in Shanghai Children's Hospital from January 2013 to November 2016 were included.The indexes included gender,age,time of IVIG treatment,and laboratory examinations.The multivariate logistic regression was used to analyze the influencing factors of IVIG non-response.The indexes in the model were deduced according to the independent variables of the logistic regression equation.The ROC curve and the area under the curve were calculated for the new prediction model.The sensitivity and specificity of the new prediction model were calculated according to the cutoff value.Finally,the new model was compared with the Kobayashi and Egami scoring model.Results The levels of CRP,NLR,LDH,ALB and FDP in children with KD were influencing factors for IVIG nonresponse (P ＜ 0.05).According to the logistic regression equation,the sensitivity and specificity of the model used to predict IVIG non-response were 69.7％ and 80.4％,respectively,and the area under the ROC curve was 0.825 (95％ CI:0.769-0.882).Kobayashi and Egami scoring models were tested,the sensitivity and specificity of the new scoring system were better than previous ones.Condusion The scoring model established in this study has a good effect in predicting IVIG non-response in KD patients and could be used in clinical practice,and it is worthy to be validated and adjusted by large-scale data.