Novel, noninvasive testing is able to predict dementia onset with 80% accuracy up to 9 years before clinical diagnosis.
The results suggest resting-state functional MRI (rs-fMRI) could be used to identify a neural network signature of dementia risk early in the pathological course of the disease, an important advance as disease-modifying drugs such as those targeting amyloid beta are now becoming available.
“The brain has been changing for a long time before people get symptoms of dementia, and if we’re very precise about how we do it, we can actually, in principle, detect those changes, which could be really exciting,” study investigator Charles R. Marshall, PhD, professor of clinical neurology, Centre for Preventive Neurology, Wolfson Institute of Population Health, Queen Mary University of London, London, England, told this news organization.
“This could become a platform for screening people for risk status in the future, and it could one day make all the difference in terms of being able to prevent dementia,” he added.
The findings were published online in Nature Mental Health.
The rs-fMRI measures fluctuations in blood oxygen level–dependent signals across the brain, which reflect functional connectivity.
Brain regions commonly implicated in altered functional connectivity in Alzheimer’s disease (AD) are within the default-mode network (DMN). This is the group of regions “connecting with each other and communicating with each other when someone is just lying in an MRI scanner doing nothing, which is how it came to be called the default-mode network,” explained Dr. Marshall.
The DMN encompasses the medial prefrontal cortex, posterior cingulate cortex or precuneus, and bilateral inferior parietal cortices, as well as supplementary brain regions including the medial temporal lobes and temporal poles.
This network is believed to be selectively vulnerable to AD neuropathology. “Something about that network starts to be disrupted in the very earliest stages of Alzheimer’s disease,” said Dr. Marshall.
While this has been known for some time, “what we’ve not been able to do before is build a precise enough model of how the network is connected to be able to tell whether individual participants were going to get dementia or not,” he added.
The investigators used data from the UK Biobank, a large-scale biomedical database and research resource containing genetic and health information from about a half a million UK volunteer participants.
The analysis included 103 individuals with dementia (22 with prevalent dementia and 81 later diagnosed with dementia over a median of 3.7 years) and 1030 matched participants without dementia. All participants had MRI imaging between 2006 and 2010.
The total sample had a mean age of 70.4 years at the time of MRI data acquisition. For each participant, researchers extracted relevant data from 10 predefined regions of interest in the brain, which together defined their DMN. This included two midline regions and four regions in each hemisphere.
Greater Predictive Power
Researchers built a model using an approach related to how brain regions communicate with each other. “The model sort of incorporates what we know about how the changes that you see on a functional MRI scan relate to changes in the firing of brain cells, in a very precise way,” said Dr. Marshall.
The researchers then used a machine learning approach to develop a model for effective connectivity, which describes the causal influence of one brain region over another. “We trained a machine learning tool to recognize what a dementia-like pattern of connectivity looks like,” said Dr. Marshall.
Investigators controlled for potential confounders, including age, sex, handedness, in-scanner head motion, and geographical location of data acquisition.
The model was able to determine the difference in brain connectivity patterns between those who would go on to develop dementia and those who would not, with an accuracy of 82% up to 9 years before an official diagnosis was made.
When the researchers trained a model to use brain connections to predict time to diagnosis, the predicted time to diagnosis and actual time to diagnosis were within about 2 years.
This effective connectivity approach has much more predictive power than memory test scores or brain structural measures, said Dr. Marshall. “We looked at brain volumes and they performed very poorly, only just better than tossing a coin, and the same with cognitive test scores, which were only just better than chance.”
As for markers of amyloid beta and tau in the brain, these are “very useful diagnostically” but only when someone has symptoms, said Dr. Marshall. He noted people live for years with these proteins without developing dementia symptoms.
“We wouldn’t necessarily want to expose somebody who has a brain full of amyloid but was not going to get symptoms for the next 20 years to a treatment, but if we knew that person was highly likely to develop symptoms of dementia in the next 5 years, then we probably would,” he said.
Dr. Marshall believes the predictive power of all these diagnostic tools could be boosted if they were used together.
Potential for Early Detection, Treatment
Researchers examined a number of modifiable dementia risk factors, including hearing loss, depression, hypertension, and physical inactivity. They found self-reported social isolation was the only variable that showed a significant association with effective connectivity, meaning those who are socially isolated were more likely to have a “dementia-like” pattern of DMN effective connectivity. This finding suggests social isolation is a cause, rather than a consequence, of dementia.
The study also revealed associations between DMN effective connectivity and AD polygenic risk score, derived from meta-analysis of multiple external genome-wide association study sources.
A predictive tool that uses rs-fMRI could also help select participants at a high risk for dementia to investigate potential treatments. “There’s good reason to think that if we could go in earlier with, for example, anti-amyloid treatments, they’re more likely to be effective,” said Dr. Marshall.
The new test might eventually have value as a population screening tool, something akin to colon cancer screening, he added. “We don’t send everyone for a colonoscopy; you do a kind of pre-screening test at home, and if that’s positive, then you get called in for a colonoscopy.”
The researchers looked at all-cause dementia and not just AD because dementia subtype diagnoses in the UK Biobank “are not at all reliable,” said Dr. Marshall.
Study limitations included the fact that UK Biobank participants are healthier and less socioeconomically deprived than the general population and are predominantly White. Another study limitation was that labeling of cases and controls depended on clinician coding rather than on standardized diagnostic criteria.