Haeseker Michiel, Croes Sander, Neef Cees, Bruggeman Cathrien, Stolk Leo, Verbon Annelies
*Medical Microbiology, Maastricht University Medical Centre; †Care and Public Health Research Institute (CAPHRI); ‡Clinical Pharmacy and Toxicology, Maastricht University Medical Centre; and §Department of Internal Medicine, Erasmus Medical Centre, Rotterdam, the Netherlands.
Ther Drug Monit. 2016 Feb;38(1):120-6. doi: 10.1097/FTD.0000000000000250.
The aim of this study was to investigate whether vancomycin clearance (CLva) can be adequately predicted with CLva prediction methods. Additionally, other covariates influencing the CLva were investigated and predictivity of monitoring of only trough levels to 24-hour area under the curve (AUC24) was evaluated.
Routine vancomycin plasma levels were measured with a fluorescence polarization immunoassay. Pharmacokinetic (PK) parameters of individual patients, that is, CLva and volume of distribution, were determined with maximum a posteriori Bayesian estimation. CLva was calculated with the 3 prediction methods, which are solely based on creatinine clearance (CLcr) estimated with Cockcroft and Gault formula and was compared with the calculated CLva with maximum a posteriori Bayesian estimation. Prediction errors were calculated. Correlations between CLva and CLcr, creatinine, age, weight, sex, and neutropenia were made. Furthermore, correlations between trough levels and AUC24 were evaluated.
A total of 171 patients were included. Prediction errors and absolute prediction errors of the 3 methods ranged from 28% to 80% and 39% to 83%, respectively. In the multivariate analysis, CLva was significantly associated with CLcr, creatinine, age, weight, sex, and neutropenia. Linear correlation between AUC24 and trough levels was R(2) 0.38.
Large prediction errors make the CLva algorithms based on estimated plasma CLcr unsuitable for use in patient care. Additionally, other factors, which are not accounted for in the current algorithms, influence the CLva individually. Owing to low association of AUC24 and trough levels, the AUC24 cannot be predicted with through levels. For a reliable AUC24 guided vancomycin dosing, therapeutic drug monitoring is necessary.
本研究旨在调查万古霉素清除率（CLva）预测方法能否充分预测CLva。此外，还研究了影响CLva的其他协变量，并评估了仅监测谷浓度对24小时曲线下面积（AUC24）的预测性。
采用荧光偏振免疫分析法测定万古霉素常规血浆浓度。通过最大后验贝叶斯估计确定个体患者的药代动力学（PK）参数，即CLva和分布容积。使用3种预测方法计算CLva，这3种方法仅基于用Cockcroft和Gault公式估算的肌酐清除率（CLcr），并与通过最大后验贝叶斯估计计算的CLva进行比较。计算预测误差。分析CLva与CLcr、肌酐、年龄、体重、性别和中性粒细胞减少之间的相关性。此外，评估谷浓度与AUC24之间的相关性。
共纳入171例患者。3种方法的预测误差和绝对预测误差分别为28%至80%和39%至83%。多变量分析中，CLva与CLcr、肌酐、年龄、体重、性别和中性粒细胞减少显著相关。AUC24与谷浓度之间的线性相关系数R²为0.38。
较大的预测误差使得基于估算血浆CLcr的CLva算法不适用于患者护理。此外，当前算法未考虑的其他因素会分别影响CLva。由于AUC24与谷浓度之间的关联性较低，无法通过谷浓度预测AUC24。为实现可靠的AUC24指导下的万古霉素给药，有必要进行治疗药物监测。
