A population pharmacokinetic model for Gd-DTPA in DCE-MRI
A. Steingoetter (1), D. Menne (2), R. Braren (3)
(1) Division of Gastroenterology and Hepatology, Dept. of Internal Medicine, University of Zurich, Zurich, Switzerland; (2) Menne Biomed Consulting, Tuebingen, Germany; (3) Institute of Radiology, Klinikum rechts der Isar der Technischen Universität München, Munich, Germany
Objectives: In dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), the kinetics of an injected contrast agent (Gd-DTPA) can be simultaneously observed within multiple tissues. Due to MRI specific obstacles, however, kinetics in the plasma cannot directly be measured. In 2 compartment models commonly applied in DCE-MRI [1,2], these data are approximated from previous data and assigned to all subjects. These models disregard the individual variations in circulation time and those by systemic drug effects, and cannot model the multiple in normal and tumor tissue. The integration of multiple simultaneously acquired tissue concentration curves within a study population allows for the development of more complex models. Based on population nonlinear mixed effects modeling (popPK), this study was aimed at developing and evaluating a robust multi-compartment model for Gd-DTPA in a rat tumor model.
Methods: 33 data sets with concentration curves of tumor, muscle and liver were included for popPK modeling. Each curve consisted of 150 samples (every 6 sec). Tumors were classified into four levels of necrosis as determined by histology. All popPK analyses were performed with full MCMC Bayesian analysis method using NONMEM® 7.1. Initial parameter estimates were calculated by SAEM. The burn-in phase consisted of 3000 Bayes samples. Model parameters were estimated using 3000 Bayes samples. Structural model building was based on physiological and histological considerations, standard numerical criteria and supported by anatomical evidence. Bayesian chain plots (CP), CWRES, OFV, standard error (SE) of parameters and DIC were applied as selection criteria.
Results: With 3 observed compartments, the final model consisted of 3 serial muscle and liver compartments branching off the central compartment and 2 serial tumor compartments branching off the first liver compartment. CP, CWRES, OFV, SE and DIC confirmed the superiority of this final model in comparison to all reduced models. The covariate model building highlighted an effect of tumor necrosis as well as heart rate on the Gd-DTPA kinetics.
Discussion: A multi-compartment model for Gd-DTPA is presented. Based on standard numerical model building and selection criteria the presented model is not over-determined. Covariate analysis demonstrated that this model can detect changes in tissue structure and circulation.
 Tofts P. et al JMRI (1999)10:223-232  Brix G. et al Biomed Tech (2006)51:325–330