A physiological-based model of ICG for the evaluation of liver function based on COMBINE standards

COMBINE2021 talk with slides available from: https://docs.google.com/presentation/d/e/2PACX-1vSrgq_6yhIwWlqXTI_d6C-sXq4kDs35r6CdXUVbKT4MzzzbUdvdtPSQNFVjFFzX8FgShExuTHqeSd_u/pub?start=false&loop=false&delayms=3000

Background
Determining liver function is a crucial task in hepatology, e.g., for liver disease diagnosis or evaluation of pre- and postoperative functional capacity of the liver. An accurate assessment is especially relevant in the context of liver surgery as postoperative complications are often associated with reduced functional capacity of the liver. An important method for quantitative evaluation of liver function are pharmacokinetic measurements of test compounds specifically metabolized by the liver, often called dynamical liver function tests. Test substances such as indocyanine green (ICG) are routinely applied in the clinics. Key challenges are hereby the large interindividual variability of such tests and that test values such as ICG-R15 or LiMAx often poorly correlate with clinical outcome.

Methods
The physiological based pharmacokinetic (PBPK) models were parameterized and validated based on extensive literature curation. All data was made open and FAIR using our pharmacokinetics database (https://pk-db.com). All models were encoded in the Systems Biology Markup Language (SBML). Based on the model predictions, classification models were developed to predict outcome after hepatectomy.


Results
We developed for the evaluation of dynamical liver function tests and applied them in the context of liver surgery. We applied our approach among others to the ICG based tests such as ICG-R15 and ICG-PDR and methacetin based tests such as LiMAx and MBT. By combining extensive data curation, physiological based pharmacokinetics models and classification models we could demonstrate that our approach allows us to predict survival in hepatectomy based on model predicted postoperative ICG-R15 values (see figure). An important advantage of the approach is that model parameters correspond to physiological parameters (e.g. cardiac output, hematocrit, bilirubin, body weight, liver volume, liver perfusion) which can readily be determined in the clinics and allow directly for model individualization.

Conclusions
Evaluation of liver function based on PBPK computational models can provide important insights in physiological factors affecting liver function. By combining the approach with classification methods survival after hepatectomy could be predicted. By using COMBINE standards such as SBML, OMEX Metadata, COMBINE archives and SED-ML high-quality and reproducible models and simulation workflows could be established.