多发性骨髓瘤患者肾损害预测模型的建立与验证

DEVELOPMENT AND VALIDATION OF A PREDICTION MODEL FOR RENAL IMPAIRMENT IN PATIENTS WITH MULTIPLE MYELOMA

目的建立多发性骨髓瘤(MM)并发肾损害(RI)的预测模型并进行验证。方法收集2012年6月1日—2019年10月31日于我院确诊的958例MM患者的临床资料,其中RI患者197例,非RI患者761例,按病例纳入时间分为训练组(RI患者134例,非RI患者538例)和验证组(RI患者63例,非RI患者223例)。应用单因素和多因素Logistic回归分析训练组MM患者并发肾损害的危险因素。根据赤池信息准则(AIC)建立MM并发RI模型,并将该模型用于验证组人群进行验证。采用受试者工作特征曲线(ROC)、Hosmer-Lemeshow检验及决策曲线分析(DCA)验证与评估该模型的区分度、校准度和临床实用价值。结果多因素Logistic回归分析显示,血尿酸、尿蛋白、血红蛋白、球蛋白、血免疫球蛋白轻链κ/λ比值、校正血清钙、尿隐血为MM患者RI的独立危险因素。由上述指标构建MM并发RI的预测模型。训练组和验证组ROC曲线下面积分别为0.882和0.928,Hosmer-Lemeshow检验P=0.374(P>0.05),显示该模型具有良好的区分度和校准度,DCA结果表明该模型的安全性及患者的临床净获益较高。结论MM并发RI预测模型的准确度和临床实用性较高,利于MM并发RI早期防治。

Objective To develop and validate a model for predicting renal impairment(RI)in patients with multiple myeloma(MM).Methods We collected the clinical data of 958 patients with MM who were diagnosed in the Affiliated Hospital of Qingdao University from June 1,2012 to October 31,2019,including 197 patients with RI and 761 patients without RI.According to the time of inclusion,the patients were divided into training group(134 cases of RI and 538 cases of non-RI)and validation group(63 cases of RI and 223 cases of non-RI).Univariate and multivariate logistic regression analyses were used to determine risk factors for the presence of RI in MM using the data of the training group.A model for predicting RI in patients with MM was established according to the Akaike information criteria,and was validated in the validation group.The discriminatory ability,calibration,and clinical practicability of the model were evaluated by a receiver operating characteristic(ROC)curve,the Hosmer-Lemeshow test,and decision curve analysis(DCA),respectively.Results The multivariate logistic regression analysis showed that serum uric acid,urinary protein,hemoglobin,globulin,serum immunoglobulin free light chain ratio(κ/λ),corrected serum cal-cium,and hematuria were independent risk factors for the presence of RI in MM.These indicators were used to construct a prediction model for RI in MM.The area under the ROC curve was 0.882 in the training group and 0.928 in the validation group.The P value for the Hosmer-Lemeshow test was 0.374(>0.05),suggesting good discrimination and calibration of the model.DCA showed that this model had good safety and high net clinical benefit.Conclusion This prediction model for RI in MM has high accuracy and good clinical practicality,and can facilitate early prevention and treatment of RI in MM.