Interspecies and in vitro‐in vivo scaling for quantitative modeling of whole‐body drug pharmacokinetics in patients: Application to the anticancer drug oxaliplatin

• Main Authors: Simona Catozzi (Author), Roger Hill (Author), Xiao‐Mei Li (Author), Sandrine Dulong (Author), Elodie Collard (Author), Annabelle Ballesta (Author)
• Format: Book
• Published: Wiley, 2023-02-01T00:00:00Z.
• Subjects: article
• Online Access: Connect to this object online.
• Notes: 2163-8306; 10.1002/psp4.12895
• Keywords: Therapeutics. Pharmacology; RM1-950
• Source: CPT: Pharmacometrics & Systems Pharmacology, Vol 12, Iss 2, Pp 221-235 (2023)

Abstract Quantitative systems pharmacology holds the promises of integrating results from laboratory animals or in vitro human systems into the design of human pharmacokinetic/pharmacodynamic (PK/PD) models allowing for precision and personalized medicine. However, reliable and general in vitro‐to‐in vivo extrapolation and interspecies scaling methods are still lacking. Here, we developed a translational strategy for the anticancer drug oxaliplatin. Using ex vivo PK data in the whole blood of the mouse, rat, and human, a model representing the amount of platinum (Pt) in the plasma and in the red blood cells was designed and could faithfully fit each dataset independently. A "purely physiologically‐based (PB)" scaling approach solely based on preclinical data failed to reproduce human observations, which were then included in the calibration. Investigating approaches in which one parameter was set as species‐specific, whereas the others were computed by PB scaling laws, we concluded that allowing the Pt binding rate to plasma proteins to be species‐specific permitted to closely fit all data, and guaranteed parameter identifiability. Such a strategy presenting the drawback of including all clinical datasets, we further identified a minimal subset of human data ensuring accurate model calibration. Next, a "whole body" model of oxaliplatin human PK was inferred from the ex vivo study. Its three remaining parameters were estimated, using one third of the available patient data. Remarkably, the model achieved a good fit to the training dataset and successfully reproduced the unseen observations. Such validation endorsed the legitimacy of our scaling methodology calling for its testing with other drugs.