Lower educational attainment is associated with an increased risk of dementia in older adults, but the underlying brain changes remain unclear. This study addresses this gap by examining longitudinal imaging-based measures of brain structure and function in a diverse cohort of adults. The researchers hypothesize that individual differences in large-scale brain network organization reflect functional integrity and that its degradation may predict cognitive impairment beyond brain atrophy and pathological burden. The study aims to determine if a link exists between educational attainment and longitudinal changes in brain system segregation and if these changes predict impending clinical decline. Understanding this link is crucial for elucidating environmental determinants of brain disease and developing new brain health biomarkers. The urgent need to understand the causes of Alzheimer's disease (AD), a leading cause of dementia, further motivates this research, as it is characterized by complex interactions between genetic and environmental factors. While biomarkers of AD pathology (amyloid and tau) and neurodegeneration are well-studied, their relationship with cognitive decline is not fully understood, suggesting other crucial factors, such as brain function, might be at play. Characterizing brain function in older adults is challenging, but resting-state functional connectivity (RSFC) offers a promising avenue. RSFC, representing the correlation structure of brain region signals during rest, reflects large-scale brain network organization. Prior research has shown age-related declines in brain system segregation (a measure of network modularity) and SES-related differences in this segregation. This study leverages a rich dataset with multiple longitudinal MRI sessions and clinical visits (up to 10 years post-scan) to investigate the relationship between education, brain network changes, and cognitive decline. The researchers account for demographic factors, health indicators, and measures of AD-related genetic risk and pathology to ascertain the unique prognostic value of brain network changes.

Existing literature links lower education to increased risks of mental health disorders and dementia in older age. This association is likely mediated by socioeconomic factors influencing resource access, environmental stimulation, health habits, and stress levels. However, studies examining the link between education, brain structure, and brain pathology have yielded inconsistent results. While some studies suggest a relationship between lower socioeconomic status (SES) and reduced brain system segregation in middle-aged adults, others find no such relationship. This inconsistent evidence underscores the need for longitudinal studies that consider a wider range of factors influencing brain aging. The existing literature on Alzheimer's disease focuses heavily on the role of amyloid plaques and tau tangles, as well as neurodegeneration, utilizing the amyloid, tau, and neurodegeneration (A/T/N) framework for staging and classification. While progress has been made in understanding the trajectory of these biomarkers and their relationship to cognitive decline, the presence of individuals with similar biomarker profiles but differing clinical profiles suggests the influence of other, yet unidentified, moderators. The incorporation of brain function, specifically resting-state functional connectivity (RSFC), into models of aging and AD has been limited due to challenges in characterizing brain signals in older and cognitively impaired populations using task-related functional imaging. However, RSFC offers a valuable approach for studying brain network organization and its potential role in aging and dementia.

This longitudinal study analyzed data from 265 participants (age 45-86 at baseline) enrolled in studies at the Charles F. and Joanne Knight Alzheimer Disease Research Center. Each participant underwent two to five MRI sessions (including resting-state scans) over 0.88-9.24 years. The sample included participants from diverse zip codes within the greater St. Louis metropolitan area, exhibiting reasonable geographic diversity. Educational attainment was categorized as 'college+' (16+ years of education) and 'below college' (<16 years). Participant characteristics were compared across education groups, considering socioeconomic index (SEI), neighborhood median household income, area deprivation index (ADI), clinical status, AD-related pathology (CSF pTau, PiB amyloid), cardiovascular health, and history of traumatic brain injury (TBI). Resting-state functional brain networks were constructed using a surface-based node set, with nodes labeled according to their functional system assignment. Brain system segregation, a measure of network modularity, was calculated for each session. Longitudinal changes in brain system segregation were analyzed as a function of educational attainment and age using linear mixed-effects models, accounting for covariates such as sex, head motion, and various health indicators. The prognostic utility of brain network changes was examined by relating them to trajectories of clinical decline (CDR-SB scores) measured up to 10 years post-last scan. Multiple models were built to test the independence of the observed relationship between brain network changes and cognitive decline from AD-related genetic risk (APOE status), AD-related pathology, and general measures of health. Additionally, the relationship between brain system segregation and measures of cortical thinning was examined to determine whether structural changes captured the observed education-related network changes. Finally, system-specific changes in resting-state correlations were compared between education groups in older adults. Permutation testing was employed for multiple comparison correction. All analyses used linear mixed-effects models to account for the longitudinal nature of the data and participant-level variability.

The study revealed several key findings. First, older adults (≥65 years) without a college degree exhibited significantly greater declines in brain system segregation over time compared to their college-educated peers. This difference was observed even after controlling for other factors such as sex, head motion, and various health measures. This effect was particularly evident among older adults, while younger adults in both groups exhibited relatively less pronounced segregation changes. Second, declining brain system segregation was a significant predictor of impending cognitive and functional impairment (CDR-SB scores), indicating increased dementia severity, even up to 10 years after the last MRI scan. This predictive relationship was independent of AD-related genetic risk (APOE status), the presence of AD-associated pathology (CSF phosphorylated tau, cortical amyloid), and cortical thinning. This demonstrates that brain network changes offer unique prognostic value beyond established AD biomarkers. Third, while older adults without a college degree demonstrated more prominent brain network decline, the association between brain network decline and future cognitive impairment held true regardless of educational attainment. In other words, once network decline occurred, the subsequent cognitive decline was comparable across education levels. Finally, analysis of system-specific changes in resting-state correlations revealed education-related differences in how integrative processing systems (association systems) changed over time, with 'below college' older adults showing distinct patterns of change involving the default mode, memory-retrieval, and frontal-parietal systems. These changes were not fully captured by overall changes in cortical thinning.

The study's findings provide strong evidence that declines in resting-state brain system segregation are associated with an increased risk of future cognitive impairment in older adults, independent of known AD biomarkers. The observation that this decline is more pronounced in older adults without a college degree highlights the significant role of educational attainment in shaping brain network trajectories. However, the lack of moderation by educational attainment on the link between brain network decline and cognitive outcomes suggests that the consequences of network degradation are equally detrimental regardless of educational background, once the changes occur. These findings strongly support the hypothesis that brain system segregation, reflecting functional brain network integrity, is a crucial indicator of brain health and a strong predictor of impending cognitive decline. This measure may capture aspects of 'cognitive reserve' in a way that is not fully captured by traditional measures of brain structure or AD pathology. Further, the study highlights the importance of considering functional network changes alongside structural and pathological changes when assessing AD risk and prognosis. The observed differences in system-specific network changes between education groups hint at possible mechanisms underlying the impact of educational attainment on brain network integrity and aging, suggesting specific circuits involved in integrative processing are particularly susceptible to environmental influences. Future research should explore these findings further to clarify the pathways linking socioeconomic factors to brain network organization and cognitive decline, and to identify potentially modifiable targets for interventions to mitigate the risk of age-related cognitive decline.

This study demonstrates that declining resting-state brain system segregation is a significant predictor of impending cognitive decline in older adults, independent of established AD biomarkers and educational attainment. This highlights the importance of functional brain network organization as a key aspect of brain health in older age, and its potential use as a preclinical warning sign for cognitive impairment. Future research should investigate the underlying mechanisms that link socioeconomic factors, brain network changes, and cognitive outcomes, and explore the potential of brain network interventions to improve cognitive resilience in aging populations.

The study's sample, while diverse, had limited representation of individuals with very low education and non-white participants compared to the broader Knight ADRC dataset, potentially leading to underestimation of the effects of lower education and health disparities. Longitudinal studies are also susceptible to attrition bias, where participants with poorer health are less likely to complete the study. The reliance on global signal regression in RSFC processing introduces potential biases, although methods were used to mitigate this. The lack of direct measurement of cognitive engagement and lifestyle factors limits a complete understanding of the factors contributing to brain network changes and their link to educational attainment. Finally, while the CDR-SB measures both cognitive and functional impairment, it’s not exclusively AD-related; future studies could benefit from more specific cognitive assessments.