Doctors have assumed amyloid plaques are the first sign of Alzheimer’s. A new study found the brain is already changing 7 years before those plaques appear on scans.
The medical model of Alzheimer’s disease has a clear sequence. Amyloid-beta plaques accumulate in the brain over years. At some point they reach a threshold detectable by positron emission tomography imaging. That moment of amyloid positivity is considered the starting gun: the earliest biological sign that Alzheimer’s disease has begun. Everything that follows, tau tangles, neuroinflammation, synaptic loss, cognitive decline, builds on that foundation. The clinical field has organized early detection, prevention trials, and therapeutic windows around this model for more than a decade.
A new study published in Nature Neuroscience has examined what was happening in people’s brains before that starting gun fired. What it found suggests the race had already been underway for years.
Researchers at the Center for Lifespan Changes in Brain and Cognition at the University of Oslo, led by James M. Roe and senior author Anders M. Fjell in collaboration with William J. Jagust at the University of California Berkeley, combined longitudinal data from three separate cohorts of cognitively healthy adults: 4,570 MRI brain scans and 1,684 amyloid PET scans. The key methodological feature is what they excluded. Every MRI scan used in the primary analysis was acquired before the participant developed elevated amyloid. These were images of brains that would later show disease, captured at a time when the best available test said nothing was wrong.
When the researchers compared the brains of participants who later converted to amyloid positivity against those who remained amyloid negative, the pre-conversion brains were already measurably different. Cortical thickness showed distinctive patterns up to 7 years before any amyloid became detectable on PET scanning.
“We found that structural changes in the brain occur many years before high levels of plaque are seen on PET scans, which is the brain scan currently used to identify the earliest signs of Alzheimer’s disease,” said Roe.
What the brain scans showed before any amyloid appeared
The specific pattern the researchers found was not what intuition would predict. People developing Alzheimer’s did not show a thinner cortex before amyloid appeared. They showed a thicker cortex in certain regions, combined with less cortical thinning over time than people who stayed amyloid negative.
This counterintuitive finding has a biological explanation that has been proposed in prior research but never confirmed at this scale or this early a time point. Before neurons die and the cortex thins, the affected regions may undergo a period of compensatory swelling. Neurons attempt to maintain their function in the face of early molecular stress by enlarging their cell bodies and dendrites. The cortex, paradoxically, appears slightly thicker in the regions that will later show the most severe degeneration.
This early thickening phase represents a window in which the disease process has begun at the molecular level but has not yet caused the cellular death that produces the thinning traditionally associated with Alzheimer’s. The amyloid PET scan was, in effect, detecting a later stage of a process that had already been underway for nearly a decade.
The brain regions showing these pre-amyloid changes were not random. They corresponded spatially to the regions where amyloid preferentially accumulates in early Alzheimer’s, providing biological coherence to the finding. The timing of the thickness changes also tracked the temporal progression of amyloid accumulation, meaning the early MRI signal and the later amyloid signal were part of the same underlying process, with standard imaging catching only the second act.
Why this overturns the current detection model
The amyloid cascade hypothesis, which places amyloid plaques at the origin of the Alzheimer’s disease process, has guided research and drug development for more than 30 years. Recent FDA approvals of amyloid-targeting therapies have been built on this framework, with clinical trials selecting participants based on confirmed amyloid positivity.
If meaningful brain changes are occurring 7 years before amyloid reaches the PET detection threshold, several consequences follow. First, the therapeutic window assumed by current trials may be too late. Drugs that clear amyloid after it has become detectable are entering a brain that has already been changing for nearly a decade. Second, the participants selected for those trials, people who have just crossed the amyloid positivity threshold, may have already passed the stage where intervention is most likely to prevent damage.
Third and most practically significant for people worried about their own risk: a clean amyloid PET scan does not mean Alzheimer’s has not begun. It means amyloid has not yet reached the detection threshold. The brain may already be in the early compensatory phase that, in some individuals, precedes clinical disease by a decade or more.
“Structural changes in the brain can be traced at least 7 years before pathological amyloid becomes visible on PET scans,” Fjell said. “This challenges the current view that amyloid-PET represents the earliest detectable change in Alzheimer’s disease.”
What MRI might reveal that amyloid PET cannot
Amyloid PET scanning has significant practical limitations beyond its apparent temporal limitations. The scans are expensive, typically costing several thousand dollars. They involve radiation exposure from radioactive tracers. They are not widely available outside major research and clinical centers. And they require specialized equipment and expertise to interpret.
Standard structural MRI, by contrast, is available at virtually every hospital. It is less expensive, involves no radiation, and is already used routinely for neurological evaluation across many conditions. The finding that cortical thickness changes on standard MRI precede amyloid PET positivity by at least 7 years raises the possibility that the most widely available brain imaging technology may contain early Alzheimer’s information that has not previously been recognized or systematically extracted.
The researchers do not claim that cortical thickness on a single MRI can diagnose Alzheimer’s or predict with certainty who will develop the disease. The changes they detected are statistical patterns across groups of people, not individual predictors. Many people who show early cortical thickness changes of this type may never develop clinical Alzheimer’s disease, and many who develop the disease may not show the pattern. The findings are about population-level biology, not about individual clinical decision-making.
What they establish is that the biological substrate of Alzheimer’s disease is not sitting quietly in the brain, doing nothing, until amyloid reaches a detectable threshold. Something is happening years earlier. The brain is responding to molecular processes that current clinical tools cannot yet see, and standard MRI may be capturing an echo of that response.
What comes next
The finding raises an immediate practical question: can the pre-amyloid MRI signal be refined into a reliable individual-level marker? Current cortical thickness measures vary between individuals for reasons unrelated to Alzheimer’s disease risk, including genetics, education, body size, and scan acquisition parameters. Extracting a meaningful early Alzheimer’s signal from that background variation requires more sophisticated analytical approaches than standard clinical interpretation.
The researchers call for longitudinal studies that follow larger cohorts across the full timeline from early adulthood through amyloid conversion and eventual cognitive decline. They also point to the potential value of machine learning approaches that can identify subtle, spatially distributed patterns in MRI data that human readers cannot reliably detect.
The clinical implications also require careful development before they can guide individual patient decisions. Knowing that someone’s brain has an MRI pattern associated with future amyloid conversion creates an ethical and practical challenge: what action should follow? Effective interventions for the pre-amyloid stage of Alzheimer’s disease do not yet exist in proven form, which means early detection in an individual patient does not automatically translate into better outcomes.
What the study changes is the conceptual framework. Alzheimer’s disease, even by its strictest current biological definition, appears to begin earlier than the gold standard test can measure. The gap is at least 7 years. Whatever is happening in that window, whether it is a treatment target, a diagnostic opportunity, or simply a biological reality that researchers need to account for, it is no longer invisible. The MRI was recording it all along.
The study, “Cortical thickness changes precede high levels of amyloid by at least 7 years”, was authored by James M. Roe, William J. Jagust, Susan M. Landau, Theresa M. Harrison, and colleagues at the University of Oslo and the University of California Berkeley, and published August 19, 2026 in Nature Neuroscience.
Source: University of Oslo / University of California Berkeley. DOI: 10.1038/s41593-026-02363-4
