A novel AI system can predict Alzheimer's disease up to seven years prior to first symptoms, possibly opening the door for earlier treatments, according to research recently featured by the U.S. National Institute of Health.
Using patients’ past medical records as data inputs to train the machine learning system, the AI model was 70% accurate at predicting Alzheimer’s seven years out and 80% the day before diagnosis, the study found. In fact, its predictive accuracy improved to as much as 90% when researchers added basic demographic details such as birth year, gender, ethnicity, and race.
“In the past few decades, electronic health records have become a source of rich... data that can be leveraged to understand and predict complex diseases, particularly Alzheimer's disease,” the study noted. The researchers tapped into prior studies using health records to track the development of Alzheimer's, as well as models that classified or predicted a dementia diagnosis from clinical data.
“Neurodegenerative disorders are devastating, heterogeneous, and challenging to diagnose, and their burden in aging populations is expected to continue to grow. Among these, Alzheimer's disease is the most common form of dementia after age 65, and its hallmark memory loss and other cognitive symptoms are costly and onerous to both patients and caregivers,” the researchers wrote.
To run the study, researchers from the University of California–San Francisco compiled clinical data for more than 250,000 individuals from its vast medical records database of millions of people collected from 1980 to 2021. Almost 3,000 of those patients had been diagnosed with Alzheimer's.
The AI models were trained on 70% of the patient records, which included both Alzheimer's patients and controls—people who had not been diagnosed with the disease. The remaining 30% of the total patient records were “saved” to be used for the evaluative portion of the study.
The AI was able to predict the onset of Alzheimer's with a high degree of accuracy.
"These findings potentially support hypotheses suggesting Alzheimer's disease can be associated with general aging or frailty, which might present in non-neurological body systems either before or concurrent with Alzheimer's,” the researchers wrote. “Furthermore, interpretation of these models allows for the identification of higher-order groups of predictors that may contribute to disease heterogeneity or together, contribute to Alzheimer's disease risk.”
Specifically, some of those early predictors named in the research that contribute to Alzheimer's risk were high levels of cholesterol and other fats in the blood, congestive heart failure, dizziness, cataracts, and deteriorating cartilage between bone joints.
Perhaps one of the most surprising findings was identifying osteoporosis as a female-specific predictor of Alzheimer's risk.
"In the University of California Data Discovery Platform, osteoporosis-exposed individuals… showed a quicker progression to Alzheimer's disease compared to matched unexposed individuals,” the study noted, “When stratified by sex, this progression was significant when comparing female individuals with osteoporosis... versus female controls.”
This level of predictive power could be a game-changer in the fight against Alzheimer's, which currently has no cure. Having years of lead time for prospective Alzheimer's patients could result in new ways to slow or halt the disease before it causes irreversible damage.
The lead researcher did not respond to a request for comment from Decrypt.
Edited by Ryan Ozawa.