Brain activity at rest may flag Alzheimer's risk years before memory slips, study finds

A pattern of brain activity recorded while people lie quietly in an MRI scanner predicted who would develop memory and thinking problems years later, according to a Stanford Medicine-led study. In healthy adults ages 55 to 70, the pattern forecast decline at least as well as the amyloid and tau markers that currently anchor Alzheimer's research, and in some comparisons better.

The work, published Nov. 4 in Science Advances, rests on a measure the team calls the Alzheimer's resilience-vulnerability connectome, or ARC score. It tracks how different brain regions communicate, a property scientists call functional connectivity. The researchers built the score by contrasting two groups at opposite extremes of aging: people with Alzheimer's disease and so-called super agers, older adults whose memories hold up like those of people decades younger.

The appeal is timing. Amyloid plaques and tau tangles are reliable signs of Alzheimer's, but they show up relatively late in the disease process. A marker that shifts before those proteins accumulate could widen the window for monitoring and, eventually, for treatment.

"What we really care about is defining early markers that can detect earlier signs of cognitive decline," said Yu Zhang, an assistant professor of psychiatry and behavioral sciences at Stanford and the study's senior author.

How the score was built and tested

The design choice that sets this study apart is its comparison group. Earlier super-ager research typically compared these unusually sharp older adults with peers who scored normally on cognitive tests. Zhang argued that approach muddies the picture, because people with similar test scores can have very different brains underneath, with one already building up amyloid and another showing no such changes.

So the team compared super agers with people of roughly the same age who already had Alzheimer's. For this study, a super ager was someone at least 75 years old who performed on memory tests like a typical 45-year-old while keeping normal executive function for their age. Depending on the definition used, roughly 10% of older adults may qualify.

Drawing on imaging collected by the Open Access Series of Imaging Studies, the Harvard Aging Brain Study and the Alzheimer's Disease Neuroimaging Initiative, the researchers analyzed resting-state functional MRI scans from 180 super agers and 110 people with Alzheimer's. Activity rose in some regions and fell in others, producing signatures distinct enough that the resulting ARC score separated the two groups with about 85% accuracy.

Then came the harder test. The team applied the score to more than 1,300 adults ages 55 to 70 who had no cognitive problems. At the start, people with high and low scores performed similarly on cognitive tests. Over the follow-up years, they diverged. Those whose brain activity looked more like the Alzheimer's group declined faster on standard tests of memory, attention and executive function, and were more likely to accumulate amyloid-beta plaques and receive a diagnosis of mild cognitive impairment or Alzheimer's. Those whose patterns resembled super agers generally declined the slowest.

Follow-up varied across the pooled datasets. A few participants were tracked for as long as 12 years, but most for about five to six.

Against demographic factors, genetic risk and amyloid and tau biomarkers, the ARC score held up well. "Our model shows comparable or better prediction ability, and it can do so at an earlier time point," said Kanhao (Alex) Zhao, a Stanford postdoctoral scholar and the study's lead author. The researchers also found that the score and conventional markers worked better together. "And if we combine our score with the existing biomarkers, the prediction performance improves," Zhao said.

That second finding may be the more practical one. Pitched as a replacement for amyloid and tau testing, the ARC score would face a long, skeptical road. As an added layer that sharpens risk estimates, it has a clearer job.

What the findings do and don't show

The study shows an association between a brain-activity pattern and later decline. It does not show that the pattern causes decline or protection, and the researchers are explicit that the score is not a diagnostic tool. For now they describe it as a possible way to track cognitive health and estimate future risk.

The score also isn't fixed. Zhang said brain activity changes over time, and someone with a resilient pattern today may not keep it. The share of people with super-ager-like scores generally falls with age, and brain injuries and neuropsychiatric conditions can affect the pattern as well.

Several other limits deserve weight. The study relied on existing research datasets rather than a single cohort designed for this question, and most follow-up ran only five to six years, short against a disease that develops over decades. Volunteers in large imaging studies also tend to differ from the general population. And no independent assessment of the work was reported alongside it.

Practicality is another hurdle. Functional MRI is expensive, time-consuming and not part of routine checkups, and resting-state measurements have historically been more dependable for comparing groups than for judging any one person. An 85% accuracy figure for separating two extreme groups is encouraging, but sorting healthy 60-year-olds whose brains sit somewhere in the middle is a tougher task. Until the score is validated in new, more diverse populations, it looks best suited to research settings, such as choosing participants for prevention trials, rather than doctors' offices.

The team also measured how ARC scores related to brain age, an estimate of how old a brain appears based on functional MRI from healthy people. The two correlated somewhat, but Zhang said ARC speaks more directly to whether a person is vulnerable or resistant to decline.

The bet on changing brain activity

The most ambitious idea in the study is that this kind of marker might be changeable in a way genes and protein buildup are not. Zhang pictures treatments that nudge a person's brain activity toward the resilient pattern, through cognitive training or transcranial magnetic stimulation, which uses magnetic pulses to influence brain activity and has shown promise in depression treatment.

Neither approach has been tested for preventing cognitive decline, and the researchers do not yet know whether shifting brain activity would alter the course of Alzheimer's at all. That gap is large. A pattern can predict an outcome without being the lever that controls it, and plenty of biomarkers have proved better at forecasting disease than at guiding treatment.

Still, the super-ager comparison gives the idea a concrete target. Zhang said treatment design depends on knowing which brain region to aim at and how to change it, and the resilient pattern defines the goal, "so we know what a good pattern should be."

The next step is already in motion. Zhang's team is preparing a trial that will measure ARC scores before and after cognitive training. Whether training moves the score is the first question; whether moving it changes anyone's memory years later is the one that will decide if this becomes more than a sophisticated forecast.