Machine learning finds four behavior symptom profiles tied to early dementia processes

Credit: Pixabay/CC0 Public Domain

Behavioral changes—such as anxiety, depression, irritability, apathy or agitation, collectively known as neuropsychiatric symptoms—may appear long before a dementia diagnosis. A new study from Karolinska Institutet and the University of Perugia, published in Alzheimer’s & Dementia, shows that these symptoms form recognizable patterns across older adults ranging from cognitively unimpaired to those with dementia.

The study included 1,234 individuals aged 65 and older who had visited a memory clinic in Perugia, Italy. Neuropsychiatric symptoms were assessed using a standardized interview tool. Strikingly, 42% of participants without dementia already showed such symptoms.

Using machine learning, the researchers identified four distinct neuropsychiatric symptoms profiles: no or very few symptoms; a combination of depression–apathy–anxiety; a profile characterized by depression–anxiety; and, a pattern dominated by delusions–agitation–irritability. These patterns appeared in both individuals with and without dementia, although more pronounced in those with dementia.

“Our results show that neuropsychiatric symptoms—an important dimension of brain health—can appear early and are not just something that emerges in established dementia,” says Anna Marseglia, last author and assistant professor at the Department of Neurobiology, Care Sciences and Society, Karolinska Institutet.

“This may give us valuable clues about who is at higher risk of progressing to dementia. Also, our study suggests that certain symptom clusters co-occur with treatable health conditions, highlighting opportunities for preventive strategies and earlier support.”

Indeed, the study identified links across the different neuropsychiatric symptoms profiles and modifiable health factors, including high LDL cholesterol, low HDL, poorly regulated blood sugar, thyroid disorders, and underweight.

“Because several of these health conditions are treatable, this opens possibilities for better monitoring and supporting older adults who show early behavioral changes,” says Anna Giulia Guazzarini, visiting Ph.D. student from the University of Perugia within the Westman Neuroimaging research group at KI and first author of the study. “However, since our study is cross-sectional, we cannot determine causality.”

The next step is to follow participants over time to understand how the four symptom profiles progress clinically, and to examine in detail the role of lipids and glucose alterations, thyroid-related and underweight factors in that progression. The researchers also plan to link these behavioral patterns to neuroimaging and blood-based biomarkers of Alzheimer’s disease and other dementias to understand better the underlying mechanisms.