Teresa Collins (1, 2), James WT Yates (1), Amin Rostami-Hodjegan (2)
(1) Innovative Medicines, AstraZeneca, Alderley Park, UK. (2) School of Pharmacy and Pharmaceutical Sciences, University of Manchester, UK
Objectives: To assess the performance of sequential methods simulated under typical pre-clinical-type study designs and to make a recommendation of a quick and robust sequential method for performing population PKPD analysis.
Methods: Using NONMEM stochastic simulation and estimation, 200 datasets were simulated and fitted using a simultaneous pharmacokinetic/pharmacodynamic fit. For the sequential methods: IPP, IPPSE and PPP&D, which are described elsewhere [1, 2, 3], a pharmacokinetic model was fitted to the concentration data from the simulated datasets then PK information included in the subsequent pharmacodynamic model fit. Structural models explored included a one compartment oral pharmacokinetic model, direct EMAX and indirect response IMAX model. A maximum of 5 concentration samples per individual per dose was explored under single and multiple dosing scenarios. Each method was compared in terms of success rate, bias and time saved to the simultaneous fit in a 3 dimensional plot.
Results: The criteria set for bias was exceeded particularly with precision on the EC50 or IC50 at lower numbers of PK observations per individual. With the direct PD model PPP&D was slower but had less bias, but success rates were similar across sequential methods. Success rates for simultaneous fit were lower with the indirect model fit making this a more important factor. IPP and IPPSE showed much higher success rates, with bias and time saved being less distinguishing between sequential methods. In contrast to discussion of the propagation of PK parameter estimates between methods in earlier studies [2], IPPSE and PPP&D are quite different as IPPSE is largely constrained to the PK parameter estimated from the PK step; whilst for PPP&D, because the PD estimation step has concentration data available it is most similar to a simultaneous fit.
Conclusions: Despite differences in the simulation settings, results with previously investigated structural models agreed well with previous studies in terms of bias, time saved and success rates. However when other structural models were explored the ranking of sequential methods and importance of performance criteria was different. Therefore it seems not possible to recommend a sequential method based on a single set of simulation conditions.
References:
[1] Lacroix, B. D., L. E. Friberg, and M. O. Karlsson. “Evaluating the IPPSE Method for PKPD Analysis”. PAGE 19 (2010) Abstr 1843 [www.page-meeting.org/?abstract=1843].
[2] Lacroix, B. D., L. E. Friberg, and M. O. Karlsson. “Evaluation of IPPSE, an Alternative Method for Sequential Population PKPD Analysis.” Journal of pharmacokinetics and pharmacodynamics (2012b).
[3] Zhang, L., S. L. Beal, and L. B. Sheiner. “Simultaneous Vs. Sequential Analysis for Population PK/PD Data I: Best-Case Performance.” Journal of pharmacokinetics and pharmacodynamics 30.6 (2003): 387-404.
Reference: PAGE 22 () Abstr 2697 [www.page-meeting.org/?abstract=2697]
Poster: New Modelling Approaches