用群体药代动力学预测癫痫儿童丙戊酸钠血药浓度

目的 用群体药代动力(PPK)模型和贝叶斯法预测血药浓度。方法 用癫痫儿童丙戊酸钠PPK模型和USC~*PACK软件中的贝叶斯程序,对100例新癫痫患儿丙戊酸钠的血药浓度进行预测。将预测值与实测值做配对t检验,相关和回归分析,计算平均预测误差、预测误差的百分比、不同预测误差百分比的符合率及其95％可信区间、构成比和评价预测的准确程度。结果 预测值与实测值的相关系数为0.99,P<0.001,线性回归Y_(OBS)=0.99)Y_(PRED),决定系数为0.98,P<0.001;平均预测误差为-0.43μg·mL~(-1),预测误差百分比分别为5％,10％,15％,20％,25％,30％的符合率为62％,74％,82％,85％,89％,93％。结论 PPK模型和贝叶斯法可准确预测丙戊酸钠稳态血药浓度。

Prediction of valproate serum concentrations in children with epilepsy by population pharmacokinetics

Objective To predict valproate serum concentrations of pediatric patients with epilepsy, using the population pharmacokinetics (PPK) model of valproate and Bayes approach. Methods Based on the PPK model of valproate, valproate serum concentrations of 100 patients were predicted by Bayesian fitting program of USC*PACK software. Paired-samples t test, correlation analysis, regression analysis were run between predicted and observed concentrations. To assess the accuracy of prediction, mean prediction error (predicted concentration minus observed concentration), percentage of prediction error (prediction error divided by corresponding observed concentration), coincidence rates and their 95% confidence intervals (CI) for the different percentages of prediction error were calculated. Results Correlation coefficient was 0.99,P<0.001; regression function was yOBS = 0.99*YPRED, P<0.001; mean prediction error was -0.43 ug.mL-1. The coincidence rates for different percentages of prediction error by 5%, 10%, 15%, 20%, 25% and 30% were 62%, 74%, 82%, 85%, 89%, and 93%, respectively. Conclusion Based on the PPK model of valproate, steady state concentrations of valproate in children with epilepsy were accurately predicted by Bayes approach.