Observational study finds dementia risk increases with accelerated brain aging in adults — Evidence Review
Published in JAMA Network Open, by researchers from UC San Francisco, Beth Israel Deaconess Medical Center
Researched by
— the AI search engine for science
Table of Contents
A new study shows that analyzing sleep brainwaves with machine learning can estimate "brain age," and a higher brain age than actual age is linked to increased dementia risk. Most related research supports the idea that accelerated brain aging, especially when measured by objective biomarkers, predicts cognitive decline and dementia (1, 4, 6). Read the full study in JAMA Network Open.
- Existing studies using MRI-based and neuroimaging biomarkers also find that a discrepancy between biological brain age and chronological age is a strong predictor of dementia, supporting the validity of this EEG-based approach (4, 6, 8).
- Reviews indicate that targeting biological aging processes may reduce cognitive decline, and that machine learning can reliably estimate brain age for early detection of neurodegenerative disorders (1, 6, 8).
- The new study adds to a growing body of work suggesting non-invasive, accessible tools (like sleep EEG) could help identify individuals at risk for dementia before symptoms develop, complementing other biomarker strategies (5, 6, 10).
Study Overview and Key Findings
Dementia remains a major public health concern, with early detection crucial for improving outcomes. The search for accessible, non-invasive biomarkers has intensified, given the limitations of current diagnostic approaches. This new study stands out for leveraging sleep EEG—a widely available, minimally disruptive tool—to estimate brain age and identify individuals at elevated risk for dementia, offering a potential advance in both research and clinical practice.
| Property | Value |
|---|---|
| Organization | UC San Francisco, Beth Israel Deaconess Medical Center |
| Journal Name | JAMA Network Open |
| Authors | Yue Leng, Haoqi Sun, Robert J. Thomas, M. Brandon Westover |
| Population | People aged 40 to 94 without dementia |
| Sample Size | approximately 7000 people |
| Methods | Observational Study |
| Outcome | Dementia risk associated with estimated brain age |
| Results | For every 10-year increase in brain age, dementia risk rose by nearly 40% |
The study introduced a machine-learning system that estimates "brain age" from 13 features of sleep EEG data. Researchers applied this model to EEG recordings from about 7,000 adults without dementia at baseline, following them for up to 17 years. They found that individuals whose brain activity appeared older than their chronological age had a significantly increased risk of developing dementia; each 10-year gap between brain age and real age raised dementia risk by nearly 40%. Notably, this association remained significant even after accounting for lifestyle, medical, and genetic risk factors.
Literature Review: Related Studies
To contextualize these findings, we searched the Consensus database of over 200 million research papers. The following search queries were used:
- brain aging dementia risk
- artificial intelligence brain age assessment
- cognitive decline predictors neurodegenerative diseases
Summary Table of Related Topics and Findings
| Topic | Key Findings |
|---|---|
| How does biological brain aging relate to dementia risk? | - Accelerated brain aging, assessed with neuroimaging or EEG, is a significant predictor of cognitive decline and dementia (1, 4, 6, 8, 9, 10). - Multiple biological pathways—including neuropathological and molecular changes—drive the link between aging and dementia (1, 3, 13). |
| Can machine learning accurately estimate brain age and aid prediction? | - Machine learning models (especially deep learning) can accurately estimate brain age from MRI and EEG data, with brain-age gaps correlating with dementia risk (4, 6, 7, 8, 9, 10). - Combining global and local features using advanced neural networks improves prediction accuracy (7, 8). |
| What factors modulate or predict cognitive decline? | - Modifiable risk factors (sleep quality, vascular health, physical activity) and genetic factors (APOE, GBA mutations) influence cognitive decline risk (2, 5, 11, 14). - Biomarkers such as plasma tau, amyloid, and synaptic proteins predict cognitive decline in preclinical stages (12, 13, 14). |
| How can early detection and risk stratification be improved? | - Non-invasive, scalable biomarkers like sleep EEG or routine MRI may enable earlier risk detection and intervention (5, 10). - Multi-domain interventions and risk stratification in memory clinics are recommended for dementia prevention (5). |
How does biological brain aging relate to dementia risk?
Research consistently links accelerated biological brain aging with increased dementia risk. The new study expands this concept by using EEG-derived brain age, aligning with earlier work using MRI-based models and neuropathological analyses (1, 4, 8, 9, 10).
- Reviews highlight that both normal cognitive decline and dementia share core aging pathways such as mitochondrial dysfunction, inflammation, and protein accumulation (1).
- Large cohort and autopsy studies show that the accumulation of multiple neuropathologies with age accounts for most of the increased dementia risk in older adults (3).
- MRI and EEG-based estimates of brain age that exceed chronological age ("brain-age gap") are repeatedly associated with elevated dementia risk and cognitive decline (4, 9, 10).
- Loss of specific synaptic proteins and molecular markers correlates closely with cognitive decline in neurodegenerative diseases (13).
Can machine learning accurately estimate brain age and aid prediction?
The use of machine learning in predicting brain age has become increasingly reliable, with growing evidence that these models can support early diagnosis and risk stratification (4, 6, 7, 8, 9, 10).
- Deep learning and advanced machine learning workflows achieve high accuracy in predicting age from brain imaging, with mean absolute errors as low as 2-4 years (7, 8, 10).
- Brain-age models built on MRI and now EEG data have shown that a greater brain-age gap correlates with higher dementia risk, supporting their use as clinical biomarkers (4, 6, 8, 10).
- The combination of global and local feature extraction improves the precision of brain age estimation, as shown in transformer-based models (7).
- The current study's use of sleep EEG introduces a non-invasive, widely accessible data source that complements these imaging-based approaches (6, 10).
What factors modulate or predict cognitive decline?
Beyond brain age, several modifiable and genetic factors contribute significantly to cognitive decline, suggesting intervention points for prevention (2, 5, 11, 12, 13, 14).
- Chronic diseases, poor sleep, vascular risk, and limited physical activity all contribute to accelerated brain aging and dementia risk (2, 5).
- Genetic factors such as APOE ε4 and GBA mutations are linked to faster cognitive decline in both Alzheimer's and Parkinson’s disease (11, 14).
- Biomarkers like plasma P-tau217, amyloid, and synaptic protein loss are strong predictors of future cognitive decline, even before clinical symptoms appear (12, 13).
- Early identification and modification of these risk factors could slow or prevent progression to dementia (5, 12).
How can early detection and risk stratification be improved?
The integration of scalable, non-invasive biomarkers and multi-domain risk assessment is a central theme for improving early detection and prevention (5, 10).
- Routine MRI-based brain-age models and now EEG-based approaches offer promising, accessible screening tools for identifying individuals at risk during standard clinical care (10).
- New memory clinic models advocate for comprehensive risk assessment—including genetic, lifestyle, and brain pathology measures—to enable early, targeted interventions (5).
- Machine learning-derived biomarkers could facilitate more efficient clinical trials by selecting high-risk individuals, as shown with plasma tau markers (12).
- The accessibility of sleep EEG for brain-age estimation, as in the new study, could help reach larger populations outside traditional clinical settings (5, 10).
Future Research Questions
While the current study demonstrates the promise of EEG-based brain age as a risk marker for dementia, several questions remain. Further research is needed to validate these findings across diverse populations, refine machine learning models, and determine how such biomarkers can be implemented in clinical practice to prevent or delay cognitive decline.
| Research Question | Relevance |
|---|---|
| How does EEG-derived brain age compare to MRI-based brain age in predicting dementia risk? | Comparing the predictive value, accessibility, and cost-effectiveness of these modalities will clarify their complementary or substitutive roles in early dementia detection (4, 6, 10). |
| Can improving sleep quality alter brain age estimates and reduce dementia risk? | Understanding if interventions targeting sleep can actually reverse or slow brain aging would inform prevention strategies and public health recommendations (1, 5, 6). |
| What machine learning features from sleep EEG are most predictive of future cognitive decline? | Identifying which EEG patterns or features drive risk prediction will help refine models and potentially guide targeted interventions (6, 8). |
| How generalizable are EEG-based brain age models across different ethnic, socioeconomic, and clinical populations? | Ensuring these models work across diverse groups is critical for equitable application and avoiding biases in dementia risk prediction (2, 9). |
| Can multi-modal biomarkers (combining EEG, MRI, and blood-based measures) improve early detection of dementia? | Combining different types of biomarkers may enhance predictive power and enable more personalized risk stratification (5, 12, 13). |