(M-025) A quantitative systems pharmacology (QSP) model to enable prediction of ARIA-E incidence with anti-Aβ monoclonal antibody therapies for Alzheimer’s disease

Sarah DiBartolo, PhD – Principal Scientist, Certara; Fei Hua, PhD – VP Modeling and Simulation, Certara; Sven Mensing, PhD – Head of Pharmacometrics, AbbVie; Mohamad Shebley, PhD, FCP – Head of Clinical Pharmacology Neuroscience, AbbVie; Sven Stodtmann, PhD – Director, Pharmacometrics, AbbVie; Hao Xiong, PhD – Director and Research Fellow, Clinical Pharmacology, AbbVie

Objectives: Monoclonal antibodies (mAbs) targeting amyloid beta (Aβ) have shown promising efficacy in treating Alzheimer’s Disease (AD). However, amyloid-related imaging abnormalities (ARIA) are a common side effect and remain a challenge for anti-Aβ drug development. Here, we have expanded an existing quantitative systems pharmacology (QSP) model of AD [1], describing Aβ biology and the activity of anti-Aβ mAbs, to enable prediction of ARIA-E incidence for APOEe4 carrier and non-carrier patients.

Methods: The model expansion included describing both Aβ40 and Aβ42 species in plasma, cerebrospinal fluid, and brain interstitial fluid (ISF), as well as the addition of a cerebrovascular compartment, which is associated with cerebral amyloid angiopathy (CAA). CAA concentration is directly related to the plaque concentration in ISF. The model was calibrated and validated against clinical PK and biomarker data for four anti-Aβ mAbs: aducanumab, lecanemab, donanemab, and gantenerumab. The model was parameterized separately for APOEe4 carrier and non-carrier patients to match differences in biomarker and CAA levels between these populations. Proposed mechanisms for ARIA-E were explored by looking at correlations between various model outputs and overall reported ARIA-E incidence for these 4 molecules at different dose levels and with different routes of administration. Model outputs that correlated well with ARIA-E incidence were then further explored by considering temporal dynamics.

Results: ARIA-E incidence over time was most highly correlated with cumulative mAb-CAA binding out of several hypotheses tested. The correlation analysis supported a mAb concentration in the cerebrovascular space lower than in plasma but similar to or higher than in brain ISF. ARIA-E incidence for the four molecules was generally well-predicted using reported in vitro CAA binding potencies for these molecules, although gantenerumab appeared to be an outlier. The model was able to capture clinical observations such as the time course of ARIA-E occurrence with treatment, decreased incidence with titration vs. flat dosing, and similar incidence with IV and SC administration. Sensitivity analysis revealed that predicted ARIA-E incidence is most sensitive to parameters affecting CAA concentration, including the ADCP clearance rate. However, ARIA-E does not correlate well with CAA clearance itself. Finally, the model predicts that increasing mAb brain penetration leads to increased plaque reduction with minimal impact on ARIA-E incidence.

Conclusions: Our modeling suggests that ARIA-E is mechanistically driven by mAb binding to CAA, and that CAA may be more directly accessible to the blood than parenchymal plaque. The model is able to predict ARIA-E incidence with different dose levels and dosing regimens and among different molecules based on their in vitro CAA binding (with the exception of gantenerumab), enabling model-informed decision-making regarding safety.

Citations: [1] Madrasi, K. et al. Alzheimers Dement 17, 1487–1498 (2021).