Target mediated drug disposition model for the DPP-4 inhibitor BI1356 (proposed trade name ONDERO): Is the structure identifiable?
S. Dittberner (1), V. Duval (2), A. Staab (2), U. Jaehde (1)
(1) Dept. Clinical Pharmacy, Institute of Pharmacy, Rheinische Friedrich-Wilhelms-Universität Bonn, Bonn, Germany; (2) Boehringer Ingelheim Pharma GmbH & Co. KG, Biberach a.d.R., Germany
Objectives: BI 1356 is a DPP-4 inhibitor with non-linear pharmacokinetics (PK) due to saturable binding to plasma and tissue DPP-4. A PK model taking both binding sites into account was developed to analyse 2 multiple oral dose studies in patients (1, 2.5, 5 and 10 mg BI 1356 q.d.). The aim was to test whether both binding sites were identifiable using the given study designs and the total plasma concentration.
Methods: Three methods were used to investigate the identifiability of both saturable binding sites. First, their influences on the PK profiles were evaluated by simulating PK profiles with varying amounts of binding sites and by log-likelihood profiling of the parameters reflecting the amount of central and peripheral binding sites. Secondly, the PK data was simulated based on 2 scenarios: binding in only the central compartment (S1) or binding in both central and peripheral compartment (S2) for the given study design. Each dataset was then re-estimated assuming either binding only in the central compartment (M1) or in both compartments (M2). The objective function was used to compare models. Finally, the Fisher information matrix (FIM) for the given model and study design was determined with WinPOPT and used to calculate the relative standard errors (RSE) of the parameter estimates.
Results: The first test showed that changes in the concentration of central binding sites (BMAX) affected the PK profiles of all dose groups. In contrast, changes in the amount of peripheral binding partners (AMAX2) affected the 1 mg dose group predominantly. The log-likelihood profiles showed BMAX to be estimated more precisely than AMAX2. In the second test, the re-estimations for S1 showed that for 69 out of 70 simulated datasets M2 was not superior to M1. In contrast, the re-estimations for S2 showed that in 100% of the simulated datasets M2 was superior to M1. Hence the correct model was chosen for both scenarios. In the third test, the RSE of the parameter estimates determined by FIM were 4 % for BMAX and 26 % for AMAX2. Omitting the 1 mg dose group resulted in an increased RSE for AMAX2.
Conclusions: The identifiability of a PK model that takes binding of the DPP-4 inhibitor BI 1356 to its target in plasma and tissue into account was investigated using different tests. All tests supported the identifiability of the structure using the available clinical data.