Centre for Human Drug Research, Leiden, the Netherlands.
Clin Pharmacokinet. 2010 Sep;49(9):619-32. doi: 10.2165/11533760-000000000-00000.
Prediction of pharmacokinetics in humans is essential for translating preclinical data to humans and planning safe and efficient clinical studies. The performance of various methods in extrapolation of preclinical pharmacokinetic data to humans is usually benchmarked by the fraction of predictions falling within a predefined interval that is centred on the value observed clinically. Recently, such an approach was used to compare physiologically based pharmacokinetic (PBPK) modelling and allometry in predicting the pharmacokinetics of a set of compounds in humans. Here, we present an analysis of the same dataset, focusing on predictions falling outside such a relatively narrow and centrally located interval. These are the main risk determinants in extrapolation of preclinical pharmacokinetic data to humans and should therefore be thoroughly understood in a risk mitigation approach to the design of early-phase human studies.
Values that had been previously predicted by allometry and by PBPK modelling in terms of the apparent total clearance after oral administration, apparent volume of distribution, area under the plasma concentration-time curve, maximum plasma drug concentration, time to reach the maximum plasma concentration and terminal elimination half-life in humans were used to generate a log-transformed dataset of predicted/observed ratios. The probabilities of mispredicting the values of these pharmacokinetic parameters using PBPK modelling and allometry were estimated by a bootstrap procedure on this set of ratios.
Our results, albeit from a limited dataset, indicated that although PBPK modelling yielded higher fractions of satisfactory predictions than allometry, both methodologies were associated with a significant and occasionally high probability of obtaining mispredictions of pharmacokinetic parameters by factors of >2, >3 and >10. In line with recent proposals to extend the goals of early-phase human studies beyond safety and tolerability, and considering the need to mitigate risks in studies dealing with novel and highly potent drug candidates, we discuss these results in a pharmacological context.
Concise recommendations are given regarding the use of allometric and PBPK extrapolation methodologies in the translation process. The results presented here should alert clinical investigators to the limitations inherent in all approaches to prediction of human pharmacokinetics from preclinical data. We propose an adaptive approach to the design of early-phase clinical studies, particularly when dealing with compounds that are characterized by novel and only partially understood pharmacological profiles.
预测人体的药代动力学对于将临床前数据转化为人体数据以及规划安全有效的临床研究至关重要。各种方法在将临床前药代动力学数据外推至人体时的性能通常通过落在以临床观察值为中心的预定义区间内的预测分数来衡量。最近，这种方法用于比较生理相关药代动力学（PBPK）建模和比例法预测一组化合物在人体中的药代动力学。在这里，我们分析了相同的数据集，重点关注落在相对较窄且居中的区间之外的预测。这些是将临床前药代动力学数据外推至人体的主要风险决定因素，因此，在早期人体研究设计的风险缓解方法中，应彻底理解这些因素。
根据口服后表观总清除率、表观分布容积、血浆浓度-时间曲线下面积、最大血浆药物浓度、达到最大血浆浓度的时间和末端消除半衰期，使用先前通过比例法和 PBPK 建模预测的数值，生成预测/观察比值的对数转换数据集。通过对该比值集进行自举程序，估计 PBPK 建模和比例法预测这些药代动力学参数值的错误概率。
尽管来自有限的数据集，但我们的结果表明，尽管 PBPK 建模产生的满意预测分数高于比例法，但两种方法都与药代动力学参数获得 >2、>3 和 >10 倍错误预测的概率显著且偶尔很高有关。考虑到新药和高活性候选药物研究中需要降低风险，并且考虑到将早期人体研究的目标扩展到安全性和耐受性之外的提议，我们在药理学背景下讨论了这些结果。
在转化过程中，提出了关于使用比例法和 PBPK 外推方法的简明建议。这里呈现的结果应该提醒临床研究人员注意从临床前数据预测人体药代动力学中所有方法固有的局限性。我们提出了一种早期临床研究设计的自适应方法，特别是在处理具有新颖且部分理解的药理学特征的化合物时。
