中国癫痫患儿丙戊酸群体药动学模型的建立

目的通过非线性混合效应模型（NONMEM）法建立癫痫患儿丙戊酸群体药动学模型，为制定个体化给药方案提供依据 。方法收集2004至2008年在复旦大学附属儿科医院住院并应用丙戊酸常规治疗的癫痫患儿的临床资料，包括血药浓度数据、生理指标及与其他抗癫痫药联合用药情况。采用NONMEM法建立丙戊酸群体药动学模型，采用一级吸收和消除的一房室模型拟合丙戊酸的药动学过程, 吸收速率常数固定为1.9·h-1，个体间变异用对数模型，采用个体间变异和残差变异有交互作用的一级速率条件算法计算。考察生理和联合用药对表观分布容积（V/F）和相对清除率（CL/F）的影响。模型验证采用200次bootstrap法。结果纳入癫患儿321例，收集血药浓度数据点390个，每例患儿采样1～5个。丙戊酸单药治疗115/321例（35.8％），与其他抗癫药合用206/321例（64.2％）。年龄中位数为1.8（0.1～13.6）岁，体重中位数为11.5（2.6～84）kg，丙戊酸剂量为30.8（4.8～88.9）mg·d-1·kg-1。患儿年龄，与氯硝西泮、卡马西平或苯巴比妥合用均是丙戊酸 CL/F的影响因素。模型的参数为：CL/F＝0.223·（Age/1.8）0.353·1.22CLO·1.36LEI,V/F＝13.0·（Total body weight/11.5）；与氯硝西泮合用，CLO＝1，否则为0；与卡马西平或苯巴比妥合用，LEI＝1，否则为0。与氯硝西泮合用可增加丙戊酸 CL/F约22％，与卡马西平或苯巴比妥合用可增加丙戊酸 CL/F约36％。最终模型较基本模型可降低CL/F的个体间变异（34.3％ vs 51.5％）。Bootstrap法的验证结果与模型计算值相符。结论建立的丙戊酸群体药动学模型有一定代表性,可为采用Bayesian法进行个体化给药奠定基础。

Establishment of population phamacokinetic model of valproic acid in Chinese epileptic children

ObjectiveTo determine the population pharmacokinetics of valproic acid(VPA) and identify the factors that explained PK variability in Chinese epileptic children. Methods390 steady-state serum concentration samples were collected totally. The plasma samples for VPA concentration were analyzed using automated fluorescence polarization method (FPIA) with monoclonal antibodies. In total, 321 patients aged 1.8(0.1-13.6) years and weighted 11.5(2.6-84) kg with 390 trough concentrations were extracted. Total body weight, VPA daily dose, gender and comedication with clonazepam, topiramate, carbamazepine, phenobarbital and oxcarbazepine were considered as covariates. A decrease of OFV (from basic model to covariate model) by more than 6.63 was considered to be statistically significant (P<0.01, df=1). An increase of OFV from the full model of at least 10.83 (P<0.001, df = 1) was used as the chosen criterion for retaining the covariate in the model. Pharmacokinetic modeling was performed using NONMEM program (version 6, level 1.1) complied with G77 Fortran. A one-compartment pharmacokinetic model with first-order absorption and elimination (subroutine ADVAN2 and TRANS2) was used to describe the concentration-time data. The first-order conditional estimation method with η-ε interaction (FOCE-I) was used throughout the model-building procedure. Various statistical models describing the random residual variability were considered. The reliability and stability of the population pharmacokinetic model developed was further assessed by a nonparametric bootstrap procedure. During this process, each patient was randomly sampled with replacement from the original data set to form new data set containing the same number of patients as the original data set. ResultsThe final model for VPA apparent clearance(CL/F),including age (years), CLO comedication and liver enzyme inducer(carbamazepine or phenobarbital) covariates with a significant influence on this parameter, was as follows: CL/F＝0.223·（Age/1.8）0.353·1.22CLO·1.36LEI,V＝13.0·（Total body weight/11.5） where CLO was 1 when co-medicated with clonazepam preparation, otherwise CLO was 0; LEI was 1 when co-medicated with carbamazepine or phenobarbital preparation.The typical values for CL/F and apparent volume of distribution (V/F) were 0.223 L·h-1 and 13.0 L, respectively. Interindividual variability (coefficient of variation) for CL/F was 34.3% for final model and 51.5% for basic model; The residual unexplained variability was 25.6 %.In a single attempt of 200 bootstrap analysis, 6(3.0%) runs terminated due to rounding errors and were excluded in the analysis. The parameter(except V/F) estimates from nonparametric bootstrap procedure were comparable and within 5% of the estimates from NONMEM. ConclusionsA population pharmacokinetic model was proposed to estimate the individual CL/F for paediatric patients receiving VPA in terms of patient′s age, and concomitant clonazepam, carbamazepine or phenobarbital, in order to establish a prior dosage regimens.