Brain scan score linked to later Alzheimer’s-related decline

Those patterns became the basis for ARC. Zhao explained that the score measures whether an individual’s connectivity more closely resembles SuperAgers or people with Alzheimer’s disease. Lower scores were more SuperAger-like, while higher scores were more Alzheimer’s-like.

How well did ARC work?

In a separate test set of 143 people who already belonged to one of those groups, ARC put about 7 in 10 in the correct category. That amounted to about 71% accuracy.

The researchers then calculated ARC from the first scans of 1,315 adults with no cognitive impairment. They compared those baseline results with later memory tests and clinical assessments. Most datasets covered about five to six years, and the longest extended to 12.

The practical question was whether people with similar current memory scores would diverge later. People with lower and higher ARC scores performed similarly on most tests at first. Over time, however, those with more Alzheimer’s-like scans declined faster on assessments of memory and everyday functioning.

Those with more Alzheimer’s-like scans also were more likely to develop Alzheimer’s disease or mild cognitive impairment. Mild cognitive impairment can make memory or thinking problems noticeable while a person still manages most daily activities independently.

People with more SuperAger-like scans were younger on average, so the researchers checked whether age explained the results. In an additional analysis of one dataset, differences in clinical ratings and test performance persisted after they accounted for baseline age.

What did the follow-up scans show?

Follow-up scans showed faster accumulation of amyloid beta, a protein that can build into plaques in the brain. People with more Alzheimer’s-like scans accumulated it faster, even though the groups started at similar levels.

The researchers found no clear difference in the buildup of tau, another protein linked to Alzheimer’s. Adding ARC improved predictions of who would develop mild cognitive impairment or Alzheimer’s beyond three existing measures:

• age;
• scans measuring amyloid beta and tau;
• genetic risk linked to the APOE gene.

If the findings hold up, ARC could help recruit higher-risk people into Alzheimer’s prevention trials before memory problems begin. Researchers still need to test that use in new groups of participants.

What must happen before clinical use?

Zhao said: “We would need larger prospective studies following cognitively healthy individuals over several years, with validation across diverse populations and MRI scanners.” A prospective study follows people forward in time, rather than reviewing outcomes that have already happened.

He also said such studies must establish reliable individual-level predictions and show value beyond existing biomarkers, meaning biological measures. They must also determine whether ARC-informed assessments improve clinical decisions.

The current results do not yet demonstrate reliable predictions for each person. They also do not show that using ARC would improve real-world clinical decisions.