We developed a disease model to describe natural progression in Alzheimer’s Disease (AD) using data from multiple interventional clinical trials in AD and the Alzheimer’s Disease Neuroimaging Initiative (ADNI), which spanned a range of disease severity. Specifically, the model predicts longitudinal trajectory of the Clinical Dementia Rating Scale – Sum of Boxes (CDR-SB) score and can be utilized to assess drug effects for molecules in development.

Disease progression models for AD (describing progression of scores that assess cognitive impairment in AD [eg, CDR-SB]) have evolved in complexity over time, from simple linear growth models to using non-linear relationships to account for variation in rate of change of cognitive impairment as the disease advances.

We used data from placebo arms of the Roche/Genentech ABBY, BLAZE, SCarlet RoAD, and Marguerite RoAD AD clinical trials and from ADNI. Using nonlinear mixed-effects modeling, we estimated a disease onset time (DOT) for each subject. Additionally, the CDR-SB score was described by a disease progression rate (RATE) and an exponential growth term (ALPHA). To assess the predictive performance of the model, data from CREAD, CREAD II, and TAURIEL were used for external validation.

We identified baseline CDR-SB and baseline Mini-Mental State Examination (MMSE) scores as significant covariates to help explain between-subject variability. All parameters were very well estimated. DOT was estimated to be on average 3.3 years with large variability before entering the trial (or study start [ADNI]); the RATE was estimated at 0.305(/y). External validation confirmed that the model’s predicted progression agreed with observed progression in the trials.

We developed a disease progression model for AD that predicted the trajectory of CDR-SB over time. By predicting individual trajectories of disease progression in the absence of treatment, the model can reliably assess drug effects for molecules in development and can be leveraged for clinical trial design.