Linking amyloid to clinical outcome: A quantitative semi‐mechanistic model based on the A/T/N biomarker framework to simulate the natural history of Alzheimer’s disease and the effects of anti‐amyloid treatment

Abstract Background According to FDA’s literature review of clinical trial data from a number of anti‐amyloid antibodies, a “clear relationship” was demonstrated between the extent of amyloid plaque removal measured by amyloid PET and the magnitude of clinical benefit measured by the placebo‐corrected difference in the CDR sum of boxes (CDR‐SB). To understand this relationship from a theoretical perspective we developed a quantitative semi‐mechanistic representation of the A (amyloid)/T (tau)/N (neurodegeneration) framework (the Q‐ATN model) in order to simulate the natural history of disease progression and the effects of anti‐amyloid therapy. Method Published data and biologically plausible mechanisms were used to construct, calibrate and validate the components of the Q‐ATN model (anti‐amyloid effect, amyloid PET, tau PET, cortical thickness [CT] and clinical outcome [CDR‐SB]). A parabolic input function was used to describe the annual rate of increase of the amyloid PET level in placebo patients (in Centiloids/Yr) as a function of the baseline level. Clinical trial simulations corresponding to mean treatment responses were made for different anti‐amyloid antibodies using individualized pharmacokinetic (PK) and pharmacodynamic (PD) models of the anti‐amyloid effect. A 5‐Yr simulation of gantenerumab treatment, based on the dosing regimen employed in the ongoing phase 3 GRADUATE trials (NCT03444870; NCT03443973), was performed to illustrate the predicted long‐term effects of gantenerumab maintained at the target dose. Result The simulated time‐course of biomarkers and clinical outcomes were consistent with natural history studies and described the effects of several anti‐amyloid antibodies observed in trials with positive and negative outcomes. Figure 1A‐D shows the simulated 5‐Yr time‐course of gantenerumab and placebo treatment on A. amyloid PET; B. tau PET; C. medial temporal cortical thickness and D. CDR‐SB. The model predicts that the difference in CDR‐SB between active and placebo groups increases markedly with the time of treatment, reaching ‐5.2 units at the 5‐Yr endpoint of the simulation. Conclusion The Q‐ATN model provides a novel theoretical approach for linking amyloid to clinical outcome based on the A/T/N framework. It illustrates the predicted disease‐modifying potential of gantenerumab – with increasing treatment benefits over time. The predictions of the Q‐ATN model will be compared with forthcoming clinical trial data.