Bayesian Hierarchical Modeling of Receptor Occupancy in PET Trials

Abstract

Receptor occupancy (RO) PET is a non-invasive way to determine drug on target. Given the complexity of procedures, long acquisition times, and high cost, ligand displacement imaging trials often have a limited size and produce sparse RO results over the time course of the blocking drug. To take the best advantage of the available data, we propose a Bayesian hierarchical model to analyze RO as a function of the displacing drug. The model has three components: the first estimates RO using brain regional time-radioactivity concentrations, the second shapes the pharmacokinetic profile of the blocking drug, and the last relates PK to RO. Compared to standard 2-steps RO estimation methods, our Bayesian approach quantifies the variability of the individual RO measures. The model has also useful prediction capabilities: to quantify brain RO for dosage regimens of the drug that were not tested in the experiment. This permits the optimal dose selection of neuroscience drugs at a limited cost. We illustrate the method in the prediction of RO after multiple dosing from a single-dose trial.

Key Words:

• Bayesian analysis
• Brain imaging
• Heirarchical model
• PET
• Receptor occupancy