A step-wise analysis of multiple biomarkers drives the development of a Semi-Mechanistic Comprehensive Model. Application to modulation of Amyloid-β.
Eline van Maanen(1), Julie A. Stone(2), Tamara van Steeg(1), Maurice Ahsman(1), Mary J. Savage(2), Maria S. Michener(2), Huub Jan Kleijn(2)
(1) LAP&P, Leiden, Netherlands; (2) Merck Research Laboratories, Whitehouse Station, NJ
Objectives: Integrating different biomarkers, to model a biological cascade of responses, results in technical challenges in NONMEM. An example of such is discussed for the multi-step production of amyloid protein (Aβ) and clearance from brain. Altered Aβ processing is hypothesized to lead to development of Alzheimer's disease (AD). The objective was to develop an approach to build a comprehensive model describing biomarker inter-relationships and their responses to inhibitors acting on different sequence in the Aβ processing pathway and clearance to CSF as an aid to drug development targeting AD.
Methods: Data on 4 biomarkers (sAPPβ, Aβ40, Aβ42, sAPPα) measured in CSF in cisterna-magna-ported rhesus monkeys receiving BACE (1 study) or GS (2 studies) inhibitors were available. BACE acts earlier in the cascade, affecting all 4 biomarkers; GS affects Aβ40 and Aβ42. A step-wise modeling approach was followed: I) Each biomarker-inhibitor combination was evaluated. II) For BACE a comprehensive biomarker model was developed by sequentially adding each biomarker, using one drug effect term to account for response of all 4 biomarkers. III) Aβ40 and Aβ42 response to GS was integrated into the model.
Results: I) For each biomarker-inhibitor combination comparable model structures and IC50 estimates were obtained. II) An indirect response model described sAPPβ response to BACE. The sAPPβ pool was then used as moderator to describe the baseline and behavior of Aβ40 and Aβ42 in presence of BACE. Next, sAPPα (informing an alternate pathway) was incorporated in the model. A precursor pool, shared by sAPPα and sAPPβ, was introduced to describe all 4 biomarkers with one common drug effect. The effect of BACE was built-in the model as inhibition of loss of this precursor pool. Different transit rates for transit compartments from site-of-action (brain) to measurement-site (CSF) allowed the rate of onset of response to differ for each biomarker. III) The response of Aβ40 and Aβ42 to GS was built-in the model, as inhibition of the exit-rate constant of the sAPPβ pool.
Conclusions: A step-wise modeling approach was developed to facilitate the build of a comprehensive model for describing 4 biomarkers measured in response to 2 inhibitors. The model characterized the response and inter-relationships of the biomarkers and gives insight into the mechanism and rate-limiting processes of the system. It is expected that the model will aid further development of drugs targeting AD.
