Jennifer Dong(1), Wei Gao(1), Gianluca Nucci(1), Mats Karlsson(2)
(1) Pfizer Worldwide R&D, Groton, CT, USA; (2) Uppsala University, Uppsala, Sweden
Objectives: An integrated glucose-insulin (IGI) model describing glucose and insulin regulation during oral glucose tolerance tests (OGTT) has been proposed in both healthy and type 2 diabetes (T2DM) subjects [1-3], which could be integrated to simulate the outcome of longer duration trials for anti-diabetic agents [4-5]. The object of this work was 1) to understand if the IGI model can be used to gain insight in the main mechanism of action (MOA) of a test anti-diabetic agent and 2) to assess the precision and bias of the IGI model in estimating quantitatively the main drug effects.
Methods: A typical early signal of efficacy trial was simulated in which a cohort of 12 T2DM subjects received a 75 g oral glucose load at baseline and at the steady state of drug effect. This sample size allowed to detect a clinically meaningful drug effect of 20% reduction in glucose excursions after an OGTT. Four categories of MOA were simulated to represent: hepatic glucose production lowering (HGP 65%↓), glucose dependent insulin secretion increase (IPRG 300%↑), overall insulin secretion increase (ISEC 130%↑), and insulin sensitivity increase (ISEN 200%↑). For each MOA, 50 realizations of the study design were simulated with intensive glucose and insulin sampling over 3 hours. During Part 1 (identifying correct MOA) of the model estimation step, inclusion of different MOA factor in the model was applied to the baseline and post drug data simultaneously in a one-by-one manner. Goodness-of-fit plots and the Objective Function Value (OFV) provided by NONMEM (v7.1.2) were used for selecting the correct MOA. For part 2 (quantifying drug effects), percent Relative Estimation Error (REE%) for the specific MOA factor was determined.
Results: The IGI model without incorporation of drug effect was used as the reference run for the calculation of delta OFV. For each of the 4 MOA categories, the model incorporating the correct drug effect generated the lowest OFV (typically a difference in OFV of 750-1300). The difference to the second best MOA model was generally 300-500. Goodness-of-fits plots also supported correct model selection. In addition, the correct model consistently provided accurate estimation of the drug effect with the median REE < 5% (90% CI < 10%) for each MOA factor.
Conclusions: The IGI model was able to both correctly categorize and quantify the MOA responsible for the glucose and insulin response following simulated standard OGTT protocols in T2DM patients.
References:
[1] Silber HE, Frey N, Karlsson MO. J Clin Pharmacol. 2010; 50(3):246-256.
[2] Silber HE, Jauslin PM, Frey N, Karlsson MO. Basic Clin Pharmacol Toxicol. 2010; 106(3):189-194
[3] Jauslin PM, Frey N, Karlsson MO. J Clin Pharmacol. 2011; 51(2):153-164.
[4] de Winter W, DeJongh J, Post T, Ploeger B, Urquhart R, Moules I, Eckland D, Danhof M. J Pharmacokinet Pharmacodyn. 2006; 33(3):313-343.
[5] Hamrén B, Bjork E, Sunzel M, Karlsson MO. Clin Pharmacol Ther. 2008; 84:228-235
Reference: PAGE 21 () Abstr 2584 [www.page-meeting.org/?abstract=2584]
Poster: Model evaluation