Stratification and Personalization of Computational Models for Liver Function Tests

Introduction: The field of pharmacokinetics describe the kinetics of substances administered to the body, consisting of absorption, distribution, metabolization and elimination (ADME) of the substance. An important use case is the testing of liver function via test substances like caffeine, methacetin or galactose in dynamical liver function tests. Understanding the kinetics of test substances and drugs is crucial to evaluate therapeutic outcome and diagnostic value.

Results: Physiologically based pharmacokinetic models (PBPK) allow to simulate the kinetics of test substances by modeling ADME, and analyze the effect of parameter changes. An important challenge is the personalization and stratification of such models, which can enhance the predictive performance by accounting for lifestyle and pharmacological modifiers like smoking and oral contraceptive use. Models for various test substances (caffeine, paracetamol, codeine) are currently established in our group. A key requirement for the development of these models are reproducible and standardized workflows from the patient data to the stratified/personalized models. An important outcome of our work is a semi-automatic “workflow” for the rapid development of such models. A key component is the curation and standardization of unstructured datasets from literature and clinical cooperation partners in a newly designed pharmacokinetic database (PKDB). The database is accessible via a REST API and provides a web interface for curation, validation and representation of the data. Hereby, the data can be queried for a variety of subgroups and across multiple clinical trials. The standardized PKDB output builds the foundation for parameterization, personalization and stratification of our pharmacokinetic models.