Laboratory of Medical Biochemistry and Clinical Analysis, Department of Bioanalysis, Ghent University, Ottergemsesteenweg 460, 9000, Ghent, Belgium.
J Pharmacokinet Pharmacodyn. 2024 Dec;51(6):639-651. doi: 10.1007/s10928-023-09894-4. Epub 2023 Nov 12.
Physiologically based pharmacokinetic (PBPK) models can be used to leverage physiological and in vitro data to predict monoclonal antibody (mAb) concentrations in serum and tissues. However, it is currently not known how consistent predictions of mAb disposition are across PBPK modelling platforms. In this work PBPK simulations of IgG, adalimumab and infliximab were compared between three platforms (Simcyp, PK-Sim, and GastroPlus). Accuracy of predicted serum and tissue concentrations was assessed using observed data collected from the literature. Physiological and mAb related input parameters were also compared and sensitivity analyses were carried out to evaluate model behavior when input values were altered. Differences in serum kinetics of IgG between platforms were minimal for a dose of 1 mg/kg, but became more noticeable at higher dosages (> 100 mg/kg) and when reference (healthy) physiological input values were altered. Predicted serum concentrations of both adalimumab and infliximab were comparable across platforms, but were noticeably higher than observed values. Tissue concentrations differed remarkably between the platforms, both for total- and interstitial fluid (ISF) concentrations. The accuracy of total tissue concentrations was within a three-fold of observed values for all tissues, except for brain tissue concentrations, which were overpredicted. Predictions of tissue ISF concentrations were less accurate and were best captured by GastroPlus. Overall, these simulations show that the different PBPK platforms generally predict similar mAb serum concentrations, but variable tissue concentrations. Caution is therefore warranted when PBPK models are used to simulate effect site tissue concentrations of mAbs without data to verify the predictions.
基于生理学的药代动力学（PBPK）模型可用于利用生理学和体外数据来预测单克隆抗体（mAb）在血清和组织中的浓度。然而，目前尚不清楚 PBPK 建模平台之间对 mAb 处置的预测结果是否一致。在这项工作中，使用三种平台（Simcyp、PK-Sim 和 GastroPlus）比较了 IgG、阿达木单抗和英夫利昔单抗的 PBPK 模拟。使用从文献中收集的观察数据评估预测的血清和组织浓度的准确性。还比较了生理和 mAb 相关的输入参数，并进行了敏感性分析，以评估当输入值发生变化时模型的行为。对于 1mg/kg 的剂量，平台之间 IgG 的血清动力学差异最小，但在更高剂量（>100mg/kg）和参考（健康）生理输入值发生变化时，差异变得更加明显。三种平台预测的阿达木单抗和英夫利昔单抗的血清浓度相似，但明显高于观察值。平台之间的组织浓度差异显著，包括总液和间质液（ISF）浓度。除脑组织浓度外，所有组织的总组织浓度的准确性均在观察值的三倍以内，而脑组织浓度则被高估。组织 ISF 浓度的预测准确性较低，GastroPlus 能够更好地捕捉到。总体而言，这些模拟表明，不同的 PBPK 平台通常预测相似的 mAb 血清浓度，但组织浓度存在差异。因此，在没有数据验证预测的情况下，使用 PBPK 模型来模拟 mAb 的效应部位组织浓度时应谨慎。
