Predicting late-phase outcome from early-phase findings using a Model-Based Approach – Application to Type 2 Diabetes Mellitus
Maria C. Kjellsson (1), Valerie Cosson (2), Nicolas Frey (2), Norman A Mazer (2), Mats O. Karlsson (1)
(1) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden, (2) Modeling and Simulation, Translational Research Sciences, Pharma Research and Early Development, F Hoffmann-La Roche AG, Basel, Switzerland
Objectives: Predicting late-phase outcome from early-phase findings is increasingly being used to inform decisions in drug development. However, if the biomarker in early-phase is different from late-phase this bridging is more challenging. In this work, we present a model-based bridging approach for type 2 diabetes mellitus (T2DM) as an example of drug development programs where different biomarkers are used in phase 1 (meal tolerance tests with glucose measurements) and phase 2 (HbA1c measurements) for efficacy assessment.
Methods: Two previously developed semi-mechanistic models were used; an integrated glucose and insulin (IGI) model [1-2] predicting glucose and insulin concentration after meal test provocation experiments and an integrated glucose-red blood cell-HbA1c (IGRH) model [3] predicting the time-dependent change in HbA1c levels from the average glucose concentration (Cg,av) and life-spans of red blood cells. Study and drug specific parameters of the IGI model were estimated using data from a resampled phase1 study in 59 diabetic patients receiving placebo or a glucokinase activator (GKA) for one week with repeated meal test challenges. Phase1 data was resampled to mimic the more severely diseased patients tested in phase2. From this adapted IGI model, Cg,av was simulated according to a phase2 study design and used in the IGRH model to predict the HbA1c response. This bridging approach was validated by comparing the predicted relative change in HbA1c to the actual outcome of the phase 2 study.
Results: The re-estimated parameters of the IGI model and the GKA drug effect parameters were in good agreement with previously reported parameters [2]. The main trend in relative change in HbA1c over time was well captured except for the last observations at week 12. This last observations was not included in the uncertainty either. If instead of looking at change from placebo, change from the lowest drug arm was assessed the predictions were good.
Conclusions: Using a model based approach allowed to predict reasonably well Phase 2 HbA1c response from effect on glucose and insulin observed in Phase I
References:
[1] Jauslin PM, Frey N, Karlsson MO. Modeling of 24-hour glucose and insulin profiles of patients with type 2 diabetes. J Clin Pharmacol. 2011;51(2):153-164.
[2] Jauslin PM, Karlsson MO, Frey N. Identification of the Mechanism of Action of a Glucokinase Activator From Oral Glucose Tolerance Test Data in Type 2 Diabetic Patients Based on an Integrated Glucose-Insulin Model. J Clin Pharmacol. 2011; Dec: Epub.
[3] Lledo R, Mazer NA, Karlsson MO. A mechanistic model for the steady-state relationship between HbA1c and average glucose levels in a mixed population of healthy volunteers and diabetic patients. PAGE. 2010;19:1783.
