Incorporating uncertainty and variability into PBPK based predictions of human PK
H. Jones(1), R. Gieschke(2)
(1)Preclinical and (2)Clinical Modeling & Simulation, F. Hoffmann-La Roche AG, Basel, Switzerland
Background: PBPK (physiologically based pharmacokinetic) models are used in preclinical research to predict drug concentrations in plasma and various organs based on physiological (volumes, blood flows) and compound specific (partition coefficients, intrinsic clearance) parameters. In general, PBPK model parameters represent averages (often from a small number of measurements), literature data, or even educated guesses. Some PBPK model parameters might be very well characterized in terms of central tendency and dispersion (e.g. blood flow in human). When predicting drug concentrations such knowledge should be incorporated, thus addressing variability. Other PBPK model parameters might not be known enough to assign a firm statistical distribution. When predicting drug concentrations such (mis-)knowledge should be incorporated, thus addressing uncertainty.
Objectives: The objective was to characterize the impact of variability in physiological and compound specific parameters on PK predictions using PBPK models.
Methods: A PBPK model [1,2] was realized in ACSL (advanced continuous simulation language, AEgis Inc). Variability on physiological parameters was incorporated using the P3M database . After selection of relevant records (e.g. male, 18-45 y), multivariate (log)normal random vectors were generated preserving the predefined variance-covariance structure. Compound specific variability was incorporated based on experimental results (in vitro incubations) and standard PK equations. The ACSL program generated the random parameter vectors for a specified number of subjects and summarized predicted plasma-concentration versus time profiles into median, 5% and 95% percentiles.
Results/Conclusion: So far, preliminary results from one iv administered compound suggest that physiologically related variability (as obtained from the P3M database) may not contribute much in addition to compound specific variability. However, the relationship between physiological and compound specific parameters and PK predictions needs to be investigated in more detail to better understand the sensitivity of the system.
 Poulin P. JPS (91) 2002, 129-156
 Poulin P. JPS (91) 2002, 1358-1370
 Price P. Crit Rev Tox (33) 2003, 469-503