Why we need a more dynamic view of Alzheimer's disease.
Cognition does not fail because protein accumulates or tissue shrinks. It fails when neuronal activity across distributed networks loses its timing. That layer sits between amyloid and amnesia, and current disease models leave it out.
Alzheimer's disease is not pathology alone. It is pathology expressed through a brain network.
The dominant narrative is linear. Pathology accumulates, neurons are lost, structure deteriorates, cognition declines. It has given the field biomarkers, staging and the first disease-modifying drugs. It also takes a shortcut: neuronal activity, the process that actually implements cognition, is treated as a byproduct.
The physio in pathophysiology is missing. The model cannot say how much each pathology weighs on function, how mixed pathologies interact, or whether removing one of them rescues the network.
Structure and dynamics shape each other. Pathology destabilises neuronal communication; neuronal activity in turn influences amyloid and tau aggregation, clearance and spread. Decline emerges from the loop.
In a nonlinear system, small changes in timing or coupling have disproportionate effects, and severe dysfunction can exist with the structure intact. Epilepsy, encephalopathy and coma show this every day. Four familiar observations in Alzheimer's disease follow the same logic.
Equal pathological burden, markedly different cognition. The difference lies in how the network responds to damage.
Symptoms shift within a day while pathology evolves over years. Performance tracks the current functional state of the network.
Some brains keep functioning under a heavy load. Network organisation and adaptability set the margin.
Highly connected hub regions run at high activity and metabolic cost. They fail early, and integration fails with them.
The brain computes in milliseconds. Alzheimer pathology builds over decades. Molecular and structural biomarkers observe the slow end of this axis. Neurophysiology observes the rest.
Cardiology would not manage the heart without the ECG. EEG and MEG offer the same kind of access to the brain: reproducible, robust changes in oscillatory activity and network coordination that track disease severity. A few minutes of resting-state EEG is cheap, low-burden and repeatable.
Less specific than a molecular marker, closer to the clinical phenotype.
The EEG deteriorates along the course of Alzheimer's disease and carries information on what comes next.
Target engagement is not benefit. Function responds within seconds to months, long before clinical scales move.
Across neurology and psychiatry, the largest clinical effects come from therapies that modulate brain activity directly, with the underlying pathology left in place. The established Alzheimer drugs belong to this family.
Whether non-invasive stimulation alters the course of Alzheimer's disease is an open empirical question. A phase 2 trial of precuneus rTMS reported slower decline over 24 weeks. If pathology impairs cognition by disrupting neuronal communication, then restoring coordinated activity acts at a mechanistically relevant level.
The tools of network neuroscience exist. What has been missing is the framework that gives them a place in disease models and in the clinic. Four objections deserve a serious answer.
Functional markers do not identify a molecular disease. They answer a different question: what state the system is in.
Dynamic signals are context-dependent and harder to capture in a single number. Harmonised acquisition and analysis are a precondition for clinical use.
When activity is both consequence and cause, causal claims need careful design: interventions on dynamics with pathology as an outcome, and the reverse.
Neurophysiology sits at the margin of dementia research, trial design and guidelines. Integration is the problem, more than missing knowledge.
A dynamic view does not replace pathology-based frameworks. It places molecular and structural change inside the functional processes through which they reach cognition. As the field moves from linear chains to complex system failure, brain dynamics become a practical necessity for diagnosis, prognosis, trial design and treatment.