Fransson Martin, Gréen Henrik
Programming Environments Laboratory, Department of Computer and Information Science, Linköping University, SE-581 83 Linköping, Sweden.
Eur J Pharm Sci. 2008 Feb 5;33(2):128-37. doi: 10.1016/j.ejps.2007.10.005. Epub 2007 Nov 6.
Two main types of model structures have been proposed for the pharmacokinetics of paclitaxel; an empirical model structure based on total plasma concentrations of paclitaxel, and a mechanism-based model structure derived from both total and unbound paclitaxel concentrations and concentrations of the formulation vehicle Cremophor EL. The purpose was to compare the two pharmacokinetic model structures when only total paclitaxel concentrations are available. To support the mechanism-based model structure with Cremophor EL concentrations, in silico concentrations were obtained from simulations of a pharmacokinetic model available in the literature. Local algebraic observability was tested on both model structures; the mechanism-based model structure was found, with high probability, not to be algebraically observable if total paclitaxel concentration is considered to be the only model output, and if no kind of prior information is used. Sensitivity analysis was performed to reveal which parameter should be fixed in order to make it locally observable. Parameter estimation was then performed on both model structures using nonlinear mixed effects and data from a clinical study. The estimated mechanism-based model turned out to have a somewhat better fit to data than the corresponding empirical model, delta AIC=-31, where AIC is the Akaike Information Criterion. Hold-out validation was performed on three patients, but did not favour any of the models. In conclusion, since the mechanism-based model structure behaved at least as good as the empirical model structure, it is suggested that the former model structure should be used since it offers a more accurate description of the disposition.
对于紫杉醇的药代动力学，已经提出了两种主要类型的模型结构；一种是基于紫杉醇总血浆浓度的经验模型结构，另一种是基于紫杉醇总浓度和游离浓度以及制剂载体聚氧乙烯蓖麻油EL浓度的基于机制的模型结构。目的是在仅可获得紫杉醇总浓度时比较这两种药代动力学模型结构。为了用聚氧乙烯蓖麻油EL浓度支持基于机制的模型结构，通过对文献中可用的药代动力学模型进行模拟获得了计算机模拟浓度。对两种模型结构都进行了局部代数可观测性测试；如果将紫杉醇总浓度视为唯一的模型输出且不使用任何类型的先验信息，则发现基于机制的模型结构很可能不是代数可观测的。进行敏感性分析以揭示应固定哪个参数以使其局部可观测。然后使用非线性混合效应和来自一项临床研究的数据对两种模型结构进行参数估计。结果表明，估计的基于机制的模型比相应的经验模型对数据的拟合稍好，AIC差值为-31，其中AIC是赤池信息准则。对三名患者进行了留一法验证，但没有支持任何一个模型。总之，由于基于机制的模型结构表现至少与经验模型结构一样好，因此建议使用前一种模型结构，因为它能更准确地描述药物处置情况。
