论文部分内容阅读
本研究旨在比较不同实验设计下正态化预测分布误差(normalised prediction distribution errors,NPDE)和可视化预测检验(visual predictive check,VPC)对模型的评价效能。本研究通过仿真方法,分别在采血时间相同的单剂量、多剂量给药以及采血时间不同的多剂量给药3种条件下,比较考察NPDE和VPC对正确模型、参数群体典型值偏差或参数个体间变异(inter-individual variability)偏差造成的错误模型的评价能力。结果显示,VPC没有明确的判断标准并且会受到实验设计的影响,采血时间不同的多剂量给药实验设计下,VPC结果已很难辨别并且对模型的辨识能力也明显下降;而NPDE具有相应的统计学检验,其模型评价能力不受实验设计因素的影响。结果提示,临床研究中VPC不适用的数据及模型,NPDE依然可以进行合理的评价。
The purpose of this study was to compare the effectiveness of normalized prediction distribution errors (NPDEs) and visual predictive tests (VPCs) on the model under different experimental designs. In this study, the simulation methods were used to study the effects of NPDE and VPC on the correct model, the deviation of the typical values of the parameter groups, or the parameter individuals under the conditions of single dose, multiple dose administration and multi-dose administration with the same blood sampling time Evaluation of the error model caused by inter-individual variability deviations. The results showed that VPC did not have a clear criterion and would be influenced by the experimental design. Under the multi-dose experimental design with different blood sampling times, the VPC results were hard to distinguish and the model’s ability to recognize was significantly decreased. However, NPDE had corresponding Statistical test, its model evaluation ability from experimental design factors. The results suggest that, in clinical studies, VPC does not apply data and models, NPDE can still be a reasonable assessment.