Optimal Design to Estimate the Time Varying Receptor Occupancy Relationship in a PET Experiment
Stefano Zamuner, Roberto Gomeni
GlaxoSmithkline - CPK/MS Verona - Italy
Objectives: Positron emission tomography (PET) is one of the most effective imaging techniques used to investigate ligand-receptor binding in living brain. In a PET experiment the total number of subjects and the number of PET scans per individual are limited either for cost or for ethical reasons (no more that 3 scan per subject). On this basis, the definition of the experimental design may become a critical issue for the assessment of the population PK/RO relationship especially when a slow dissociation model characterizes the drug-receptor interaction. The purpose of this work is to define an optimal PET experiment strategy and to present a case study where this methodology is applied to evaluate the PK/ RO relationship for a novel 5-HT1A antagonist.
Methods: A mechanistic model was used to estimate the time-varying relationship between RO and PK. A receptor association-dissociation model was established from pre-clinical data:
Designs were optimized using a D-optimality criterion as implemented in the WinPOPT software (1).
A comparison of population optimal designs with an empirical approach for designing a PET experiment aimed to estimate the PK/RO relationship is presented. PK parameters obtained from a population PK model were considered as a known independent variable in the exploration of parsimonious PET experiment. A series of alternative designs were considered to explore the influence of: a) the PET scan time allocation, b) the number of subjects to elementary design and, c) the number of dose levels.
Results: All designs explored have a total of 32 samples (N=2 PET scans x 16 subjects). In the empirical design PET scans were performed at Tmax and trough levels (four doses with same PET scan time for each dose group).
The efficiency criterion (2) for any given design showed that all the optimized designs improved the empirical ones (efficiency increased of at least 500%). Allocation of the appropriate time appeared to be the most critical factor to improve efficiency compared with number of groups/doses.
Finally, a Monte Carlo simulation was used to assess the performance of optimal designs by estimating the kon and koff parameters (fixed/random effects) from simulated data. Performances were measured as bias, precision and accuracy. The optimized design provided more accurate and reliable model parameter estimates.
Conclusions: The results show that population D-optimal design provided more accurate and reliable model parameter estimates.
 Retout S, at al. Comput. Meth. Prog. Biomed. 2001; 65(2):141-51.
 Duffull S, et al. J. Pharmacokinet. Pharmacodyn. 2005; 32:441-57.