Logistic模型和ROC曲线对参麦注射液致不良反应的预测分析

Predictive analysis on Shenmai injection-induced adverse reactions with Logistic model and ROC curve

应用Logistic模型和ROC曲线探讨参麦注射液致不良反应/事件的相关危险因素,对其相关不良反应/事件发生的风险进行预测。采用前瞻性、多中心、大样本、巢式病例对照研究的方法收集使用参麦注射液患者的病历资料,分析接受参麦注射液治疗患者致不良反应/事件的危险因素,建立Logistic模型,绘制风险因素的ROC曲线。研究期间有7 632例(男3 477名,女4 155名)患者纳入研究,8例患者出现药品不良反应/事件。多因素Logistic模型分析显示,年龄分组(≥50)(OR=5.061,95%CI:2.197~7.924,P=0.001)、用药总天数(OR=-1.020,95%CI:-1.652^-0.388,P=0.002)、单次用药剂量(OR=0.245,95%CI:0.127~0.364,P=0.000)等是发生药物不良反应/事件的独立危险因素。根据Logistic模型分析结果,分别用年龄分组(≥50)、用药总天数、单次用药剂量及联合预测因子构建ROC曲线,联合预测因子的ROC曲线下面积(0.975 3,95%CI:0.944 3~1.000,P<0.005)大于其他3个单一指标,风险预测价值优先。与参麦注射液不良反应的独立危险因素包括年龄分组(≥50)、用药总天数和单次用药剂量。在临床实际工作中,可将患者的年龄分组(≥50)用药总天数和用药剂量带入联合预测因子计算公式(P_(预测)=1/[1+e^(-(-21.588+5.061×X_(年龄)-1.020×X_d+0.245×X_(mL))]),预测患者可能发生不良反应,以便调整给药方案。

To study relevant risk factors of Shenmai injection induced adverse reactions by using Logistic model and ROC curve, and made the prediction for the occurrence of relevant adverse reactions/events. Case data of patients treated with Shenmai injection were collected by using the prospective, multi-center, large-sample, nested-case control method, in order to analyze the risk factors of Shenmai injection-induced adverse reactions/events, establish the logistic model and draw the receiver operating characteristic （ROC） curve for risk factors. During the study, 7 632 patients （including 3 477 males and 4 155 females） were included, and eight of them suffered adverse reactions/events. Based on a multi-factor Logistic model analysis, the age （≥50 years） （OR=5.061, 95% CI：2.197-7.924;P=0.001）, the total number of medication days （OR=-1.020,95% CI：-1.652--0.388;P=0.002） and the single dose （OR=0.245, 95% CI：0.127-0.364;P=0.000） were significant independent risk factors for Shenmai injection-induced adverse reactions/events. According to the results, ROC curves were drawn with age （≥50 years）, the total number of days of medication and single dose;The area under ROC curves the joint predictor （0.975 3, 95% CI：0.944 3-1.000, P〈0.005） was larger than that of the other three single indexes, with a higher risk prediction value. The independent risk factors for Shenmai injection-induced adverse reactions/events included the age （≥50 years）, the total number of days of medication and single dose. In clinical practice, the age （≥50 years）, the total number of days of medication and the medication dose can be substituted in the joint predictor calculation formula （P=1/[1＋e-（-21.588＋5.061×X年龄-1.020×Xd＋0.245×XmL）] to predict the potential adverse reactions of patients and adjust the dosage regimen.