Validation of in vivo Mouse PK Assay by Mixed Effects Modelling: Estimation of between-study variability.
James Yates, Rebecca Watson and Jason Cheung
AstraZeneca R&D.
Objectives: A capillary bleed sampling technique was evaluated in-house. This technique allowed the sampling of multiple time points from a single mouse. The reproducibility of the data produced by this technique was investigated by carrying out a number of in vivo studies. The variability in the data may then be partitioned into inter-study and inter-animal. A particular compound was chosen for its well-defined 2-compartmental kinetics. Reproducibility, at least for this compound, would be assessed if the majority of variability could be assigned to inter-animal variability rather than inter-study variability. To achieve these objectives the data were analysed in NONMEM VI, the NLME toolbox in R and WinBUGS for a comparison of results.
Methods: The NONMEM analysis was performed using the FOCE method. The NLME tool in R uses the Lindstrom-Bates method [1] which is similar to FOCE because it linearises about the current individual estimate. WinBUGS was used with uninformative priors and five independent chains. The latter two methods allow the construction of a multi-level mixed effects model. The implementation of inter-study and inter-individual variability in NONMEM was as described in the literature[2].
Results: The data set contained profiles of 48 animals over 7 studies; each individual profile contained a maximum of 5 time points. The results from NONMEM suggested that the inter-animal variability was large (~70% CV on each parameter) whereas there was very little evidence of inter-study variability except for on the volume of the peripheral compartment. The NLME results were qualitatively very similar with the majority of variance at the inter-animal level, except for a moderate variability at the study level on the inter-compartmental flow. This is possibly due to drug being detectable in some animals at 24hrs and not in others. The NLME results did show some sensitivity to initial parameter values. The WinBUGS output allocated variance more evenly across the inter-animal and inter-study levels, with more variance at the inter-animal level.
Conclusions: The analysis is readily applicable, though there are a number of pitfalls - especially with respect to the NONMEM implementation of the statistical models. The results also show that with this quantity of data different conclusions may be reached by using different methods. The results however point towards a more sophisticated use of data when planning drug discovery life-phase activities.
References:
[1]. M.J. Lindstrom and D.M. Bates. 1990. Nonlinear Mixed Effects Models for Repeated Measures Data. Biometrics. 46: 673-687
[2]. S.Laport-Simitsidis et al. 2000. Inter-study variability in population-pharmacokinetic meta-analysis: When and how to estimate it? Journal of Pharmaceutical Sciences. 89:155-167.
