P. Kellie Turner, Brian Moser, Jay Tuttle, Volker Wacheck
Eli Lilly and Company Ltd
Objectives: The aim of the analysis was to build a semi-mechanistic model to describe the PK of a monoclonal antibody administered by IV infusion.
Methods: A population pharmacokinetic (PK) analysis was conducted using NONMEM version 7.2 to describe the IV-infusion data. Estimation methods included first order conditional estimation with interaction (FOCEI), Monte Carlo importance sampling expectation maximization (IMP), and IMP assisted by mode a posteriori estimation (IMPMAP). The final model was selected based upon change in objective function value (OFV), precision estimates, diagnostic plots, and visual predictive checks (VPCs).
Results: PK data were available from 23 individuals. Samples for determination of total antibody concentrations were taken prior to infusion, mid-infusion, end of infusion, and at 2, 4, 24, 72-120, 144-192, and 336 hours after the end of the infusion. Clearance decreased with increasing dose from the first to the third of six escalating dose levels before levelling, consistent with TMDD. The model-building process began with a TMDD model using FOCEI. A two-compartment TMDD model using quasi-equilibrium approximation (1) was tested, but was not adequate because the antibody plasma concentrations continued to rise for 2-4 hours after the end of the IV infusion in a majority of individuals. Therefore, a transit model was added to the TMDD model and tested with FOCEI. Due to problems with model minimization, covariance step failure, and protracted run times with FOCEI, the IMP method was tested, but the OFV increased with successive iterations. Therefore, the IMPMAP method was implemented to test a series of 1 to 5 transit compartments. Compared to FOCEI, run times were shorter with IMPMAP, and the model successfully minimized with covariance step completion. Finally, a model with infusion into a depot compartment and 4 transit compartments followed by a 2-compartment TMDD was selected that best fit the data.
Conclusions: The population PK model for this monoclonal antibody adequately described the data, and can be used for clinical trial dosing scenario simulations. This model, or sub-sections of the model, may also apply to other compounds administered by IV infusion that show a delayed tmax. Furthermore, the IMPMAP estimation method offers a useful alternative to FOCEI for complex PK/PD models such as this TMDD model with depot and transit compartments.
References:
[1] Gibiansky L et al. J Pharmacokinet Pharmacodyn 2008; 35:573–591.
Reference: PAGE 21 () Abstr 2563 [www.page-meeting.org/?abstract=2563]
Poster: Estimation methods