An individualized systems model to optimize Alzheimer’s disease prevention strategies

Background: A large number of biopsychosocial factors are implicated in the prevention of Alzheimer’s Disease (AD). These factors are not independent causes but part of a complex causal network that underlies the condition. Computational models that would capture this system-wide multicausality could help identify causal pathways and inform multifactorial prevention strategies.
Method: We developed a system dynamics model (SDM) from a causal loop diagram that was parameterized using empirical data from multiple cohorts (including the Alzheimer’s Disease Neuroimaging Initiative). The SDM contains over 20 known risk factors and pathophysiological processes, including blood pressure, smoking, neuronal dysfunction, and amyloid-beta and phosphorylated tau burden. We simulated 5-year cognitive decline trajectories for individuals and explored several “what if” scenarios regarding the effect of changes in modifiable risk factors on cognitive decline.
Result: Our SDM was able to simulate the cognitive decline trajectories of individuals with good accuracy (< 20% mean absolute percentage error). These predictions also generalized well to an independent test sample from the same data set (<2% error increase). The effect of changes in modifiable risk factors on cognitive decline in the SDM were checked against literature reported ranges. We also developed a workflow to further calibrate and validate the SDM.
Conclusion: Our SDM demonstrates the feasibility of system-wide modelling approaches for AD prevention. Such a simulation model could eventually be used to better understand the interactive effects of modifiable risk factors on AD pathophysiology and help optimize individualized prevention strategies.