影响替考拉宁谷浓度的多因素分析及预测模型研究

Discussion on multi-factor prediction model of Teicoplanin not reaching the target trough concentration

目的探讨影响替考拉宁谷浓度达标(Cmin≥10 mg/L)的因素,建立预测模型并进行验证,为临床用药提供参考。方法采用多因素logistic回归分析筛选影响替考拉宁谷浓度达标的因素,利用回归系数及常数项建立预测模型1,参考OR值建立评分预测模型2。通过受试者工作曲线(ROC曲线)评估模型1、2的预测效果。结果共纳入189例患者,149例建模组,40例验证组。建模组36.24%的患者替考拉宁谷浓度<10 mg/L,logistic回归分析结果发现前3 d平均给药剂量(MID)、尿素氮(BUN)、内生肌酐清除率(Ccr)和合并抗休克血管活性药物是影响替考拉宁达标的重要因素。比较建立的两个模型,模型1预测效果显著优于模型2,ROC曲线下面积(AUC)分别为0.806和0.685(P=0.003)。选用模型1进行外部验证,AUC=0.840,实际运用效果灵敏度、特异度及准确度分别为84.00%、73.33%和80.00%。结论MID、BUN、Ccr和合并抗休克血管药物是影响替考拉宁达标的因素,建立的多因素logistic预测模型(模型1)对患者替考拉宁血药浓度是否达标有较好的预测效果。

Objective To explore the factors that affect the patient’s Teicoplanin trough concentration to reach the target(Cmin≥10 mg/L),and to establish a predictive model and verify it in order to provide reference for clinical medication.Method Multivariable logistic regression was used to screen independent factors of Teicoplanin trough concentration reaching to target.Regression coefficients and constant terms were used to establish prediction model 1,OR value was refered to establish scoring prediction model 2,and areas under the ROC curve(AUC)was used for evaluating the two predictive models. Results A total of 189 patients were enrolled(modeling group n=149,verification group n=40). Among the modeling group,the non-compliance rate was 36. 24%. logistic regression analysis showed that mean administration dose(MID),urea nitrogen(BUN),endogenous creatinine clearance rate(Ccr)and the combination of antishock vasoactive drugsin the first three days were the important factors affecting Teicoplanin trough concentration(P<0. 05). The prediction effect of Model 1 was significantly better than Model 2(AUC:0. 806 vs. 0. 685,P=0. 003). Model 1 was selected for external verification,AUC=0. 840,the actual application effect sensitivity,specificity and accuracy were 84. 00%,73. 33%,and 80. 00%,respectively. Conclusion MID,BUN,Ccr and combined the vasoactive drugs were important factors that affected patient’s Teicoplanin trough concentration reaching to the target level. The established multivariate logistic prediction model(model 1)had a good prediction effect on whether the blood concentration of Teicoplanin reached the target level.