Using VT (total volume of distribution from PET) in estimating the PK-Receptor Occupancy relationship in the absence of reference regions.
Alienor Berges(1), Stefano Zamuner(2), Bart Laurijssens(1), Roger Gunn(3), Vincent Cunningham(3), Chao Chen(1)
(1)GlaxoSmithkline, CPKMS, Greenford, UK; (2)GlaxoSmithkline, CPKMS, Verona, Italy; (3)GlaxoSmithkline, Imaging, London, UK
Introduction: In a PET (Positron Emission Tomography) study, receptor occupancy (RO) data can be obtained from the fractional decrease in the volume of distribution of the specifically bound radioligand (VS) [1]. This VS can be derived from VT (total volume of distribution) when VND (non-displaceable radioligand) can be measured directly from a region of reference or estimated by linear regression using volume of distribution at baseline (VT0) and after drug administration (VT) [2].
Objective: We propose to estimate the PK/RO relationship with a population approach using VT values, instead of the derived RO. In general, this approach allows estimating inter-subject variability on both VT0 and the VND and uses a more appropriate residual error model. This method was applied in a PET study and was compared to the more conventional approach of performing the modelling directly with RO estimates.
Methods: PET data was obtained from an ongoing neuroreceptor drug occupancy study. Measurements were performed at baseline, at tmax and 24h post dose. A range of doses were tested in order to characterise the exposure-occupancy relationship. The following equation based on the Emax model [2] was used to describe VT:
VTij = VT0ij - (CPi / (EC50i + CPi) * (VT0ij - VNDi)
where i is a subject, j is a brain region, and EC50i, VT0ij, VNDi are parameters that were estimated using a population approach including all the brain regions. Receptor Occupancy was calculated as 100%*CPi/(EC50i+CPi) For comparison, an Emax model was applied to the derived RO values. In this case, different residual error models were tested (proportional, additive and a model derived from the propagation of the error applied to VT observations). Results were assessed in terms of the accuracy of the parameter estimates and using simulation-based diagnostics.
Results: The point estimates of EC50 were similar between the different methods and well estimated (SEM <40%). The inter-individual variability on the EC50 could only be estimated fitting either VT values or RO values using the error propagation model. From the VPC, the proportional and additive error models appeared to inflate the overall variability.
Conclusions: The model using VT values provided additional information on VT0 and VND, as well as robust estimates of the EC50, inter-subject and residual variability. A population modelling approach is able to successfully characterize the PK-RO relationship in PET studies where the total distribution volume is the outcome measure of interest and no reference region exists.
References:
[1] Lammertsma et al. JCBFM 11:545-556, 1991.
[2] Lassen et al. JCBFM 15:152-65, 1995.
