摘要
目的评价Smart Dose、PharmVan及JPKD 3款个体化给药辅助决策系统对万古霉素血药浓度的预测能力,分析影响Smart Dose、PharmVan及JPKD预测能力的因素。方法回顾性分析2018年1月至2020年7月在温州医科大学附属第一医院使用万古霉素治疗并进行治疗药物监测(TDM)的成年住院患者的临床资料。使用Smart Dose和PharmVan软件预测万古霉素初始方案的稳态谷浓度,对于因稳态谷浓度不达标而需要调整万古霉素给药方案的患者,使用Smart Dose、PharmVan、JPKD软件分别预测调整方案后的万古霉素稳态谷浓度。计算实测浓度与预测浓度之间的相对预测偏差(PE),绘制各软件PE的箱式图;比较各软件预测万古霉素血药浓度PE绝对值(APE)的差异。根据APE将各软件的预测数据分为预测准确组(APE<30%)和预测不准确组(APE≥30%),比较两组患者性别、年龄、体重、合并症等指标和万古霉素用药情况以及TDM结果的差异;采用单因素和多因素Logistic回归分析影响Smart Dose、PharmVan、JPKD预测万古霉素血药浓度准确性的危险因素;并绘制受试者工作特征曲线(ROC曲线),评价影响因素对软件预测准确性的判断价值。结果共纳入185例患者,收集万古霉素TDM数据258例次,其中初始方案185例次,调整方案73例次。Smart Dose与PharmVan对初始方案万古霉素血药浓度的预测准确性差异无统计学意义,Smart Dose、PharmVan和JPKD对调整方案的预测准确性差异亦无统计学意义。Smart Dose预测调整方案血药浓度的准确性优于初始方案〔22.94%(10.50%,36.24%)比29.33%(13.07%,47.99%),P<0.05〕。影响Smart Dose预测初始方案万古霉素血药浓度准确性的单因素分析显示:预测准确组高血压患者的比例明显高于预测不准确组〔43.3%(42/97)比27.3%(24/88),P<0.05〕,影响SmartDose预测调整方案万古霉素血药浓度准确性的单因素分析显示:预测准确组合并心脏瓣膜病患者的比例明显低于预测不准确组〔23.4%(11/47)比46.2%(12/26),P<0.05〕,影响JPKD软件预测调整方案万古霉素血药浓度准确性的单因素分析显示:预测准确组体重明显高于预测不准确组(kg:62.8±14.9比54.8±12.8,P<0.05)。多因素Logistic回归分析显示,高血压是Smart Dose预测万古霉素初始方案血药浓度的优势因素〔优势比(OR)=0.526,95%可信区间(95%CI)为0.281~0.983,P=0.044〕;低体重是JPKD预测调整方案血药浓度不准确的独立危险因素(OR=1.042,95%CI为1.001~1.085,P=0.043)。ROC曲线分析显示,体重评估JPKD预测万古霉素血药浓度准确性的ROC曲线下面积(AUC)=0.663,95%CI为0.529~0.796,P=0.023;当体重低于55.95 kg时,JPKD预测万古霉素稳态谷浓度不准确的风险增加,其敏感度为75%,特异度为60%。结论个体化给药辅助决策系统Smart Dose、Pharm Van及JPKD对万古霉素血药浓度的预测能力无明显差异。Smart Dose对调整方案血药浓度的预测能力较初始方案高。Smart Dose的初始方案模块对合并有高血压的患者预测能力较好;JPKD对低体重的患者预测能力较差,尤其是体重低于55.95 kg时,预测不准确的风险增加。
Objective To evaluate the predictive performance of the individualized drug delivery decision-making system including Smart Dose,PharmVan and JPKD on predicting the Vancomycin plasma concentration and to analyze the related factors affecting the predictive performance.Methods The clinical data of patients who were treated with Vancomycin and received therapeutic drug monitoring(TDM)admitted to the First Affiliated Hospital of Wenzhou Medical University from January 2018 to July 2020 were retrospectively collected.Smart Dose and PharmVan were used to predict the plasma concentration of Vancomycin of the initial regimen.Smart Dose,PharmVan and JPKD were used to predict the plasma concentration of Vancomycin of the adjustment regimen for patients whose initial steady-state trough concentration were not qualified.The relative predictive error(PE)between the measured plasma concentration and predicted plasma concentration was calculated and box plotted.Mann-Whitney U test was used to evaluate the difference of the absolute value of PE(APE)predicted by each software for Vancomycin plasma concentration.The TDM results were divided into accurate prediction group(APE<30%)and the inaccurate prediction group(APE≥30%)according to the APE value.Patients and disease characteristics including gender,age,body weight complication,Vancomycin medication and TDM results were collected from electronic medical records.Univariate analysis and multivariate Logistic regression analysis were used to screen the related factors that influence the predictive performance of Smart Dose,PharmVan and JPKD;and receiver operating characteristic curve(ROC curve)was drawn to evaluate its predictive value.Results A total of 185 patients were enrolled,and 258 plasma concentration of Vancomycin were collected,including 185 concentrations of initial regimen and 73 concentration of adjustment regimen.There was no significant difference in the APE of the initial regimen of plasma concentration between Smart Dose and PharmVan.No significant difference in the APE of the adjustment regimen of plasma concentration was found among Smart Dose,PharmVan and JPKD.The accuracy of Smart Dose in predicting the plasma concentration of the adjustment regimen was better than that of the initial regimen[22.94%(10.50%,36.24%)vs.29.33%(13.07%,47.99%),P<0.05].The univariate analysis of factors affecting the performance of Smart Dose in predicting the concentration of initial regimen showed that the proportion of patients with hypertension in the accurate prediction group was significantly higher than that in the inaccurate prediction group[43.3%(42/97)vs.27.3%(24/88),P<0.05].The univariate analysis of factors affecting the performance of Smart Dose in predicting the concentration of adjustment regimen showed that the proportion of patients with valvular heart disease in the accurate prediction group was significantly lower than that in the inaccurate prediction group[23.4%(11/47)vs.46.2%(12/26),P<0.05].The univariate analysis of factors affecting the performance of JPKD in predicting the concentration of adjustment regimen showed that the body weight of patients in the accurate prediction group was significantly higher than that in the inaccurate prediction group(kg:62.8±14.9 vs.54.8±12.8,P<0.05).Multivariate Logistic regression analysis indicated that hypertension was a beneficial factor for Smart Dose to predict the initial plasma concentration of Vancomycin[odds ratio(OR)=0.526,95%confidence interval(95%CI)was 0.281-0.983,P=0.044],and low body weight was an independent risk factor for the inaccurate prediction of JPKD for adjustment regimen(OR=1.042,95%CI was 1.001-1.085,P=0.043).ROC curve analysis indicated that the area underROC curve(AUC)of the body weight for evaluating the accuracy of JPKD in predicting Vancomycin plasma concentration was 0.663,and 95%CI was 0.529-0.796(P=0.023).When the body weight was less than 55.95 kg,the risk of inaccurate prediction of JPKD in predicting Vancomycin plasma concentration was increased,and the predictive sensitivityand specificity were 75%and 60%respectively.Conclusions There is no significant difference in the predictive performance of Smart Dose,PharmVan or JPKD on Vancomycin plasma concentration.Smart Dose had a better predictive performance for the Vancomycin plasma concentration of adjustment regimen than initial regimen.Smart Dose had a better predictive performance when the patient was concomitant with hypertension.JPKD had a poor predictive performance for low-body weight patients.The predictive performance of JPKD was decreased when the body weight was lower than 55.95 kg.
作者
韩璐
徐方敏
张晓珊
王誉臻
林观样
余旭奔
Han Lu;Xu Fangmin;Zhang Xiaoshan;Wang Yuzhen;Lin Guanyang;Yu Xuben(Department of Pharmacy,the First Affiliated Hospital of Wenzhou Medical University,Wenzhou 325000,Zhejiang,China;Department of Pharmacy,Wenzhou Medical University,Wenzhou 325000,Zhejiang,China)
出处
《中华危重病急救医学》
CAS
CSCD
北大核心
2021年第3期263-268,共6页
Chinese Critical Care Medicine
基金
浙江省温州市科研项目(Y20190654)。
