Considerations and Caveats when Applying Global Sensitivity Analysis Methods to Physiologically Based Pharmacokinetic...

Area of Science:

• Pharmacokinetics and Drug Metabolism
• Computational Biology
• Systems Pharmacology

Background:

• Physiologically based pharmacokinetic (PBPK) models are crucial for predicting drug behavior.
• Global sensitivity analysis (GSA) identifies key model parameters influencing predictions.
• Understanding parameter influence is vital for accurate pharmacokinetic (PK) assessments.

Purpose of the Study:

• To compare three GSA methods (Morris, Sobol, extended Sobol) for a minimal PBPK (mPBPK) model.
• To investigate the impact of input parameter correlations on GSA results.
• To identify key parameters affecting maximal plasma concentration (Cmax), time to Cmax (Tmax), and area under the curve (AUC) for quinidine, alprazolam, and midazolam.

Main Methods:

• Application of Morris, Sobol, and extended Sobol GSA methods to an mPBPK model.
• Analysis of PK outputs: Cmax, Tmax, and AUC24h.
• Comparison of influential parameters identified by methods that do and do not account for parameter correlations.

Main Results:

• Morris and Sobol methods showed comparable informativeness for independent parameters.
• Extended Sobol method, accounting for correlations, identified different influential parameters.
• Overestimation of volume of distribution at steady state (Vss) on AUC24h and underestimation of liver volume (Vliver) were observed.
• Specific enzyme clearance and abundance parameters were also misjudged by certain methods.

Conclusions:

• Interpreting GSA results requires careful consideration of model assumptions and method limitations.
• Ignoring parameter correlations in GSA can lead to biased identification of key PK parameters.
• Decisions on parameter influence should integrate knowledge of model structure, GSA limitations, and parameter inter-correlations.