This work stands somewhat apart: its closest neighbors in the archive study links between genes, environment, and brain structure, cognitive reserve, and dementia, but none of them breaks down the socioeconomic gradient of brain aging into specific modifiable risk factors. Previous studies either used a single status measure or treated risk factors as confounders without conducting a formal mediation analysis. Here, for the first time, three socioeconomic status indicators are systematically compared and the contribution of each of the eleven factors is quantified.
The sample included 36,878 UK Biobank participants (mean age 65.1 years, 55.2% women) with structural brain MRI data. Brain age was estimated using the NeuroFM neural network on T1- and T2-weighted images, and the brain age gap (BAG) was calculated as the difference between predicted and chronological age. Socioeconomic status was measured with three indicators: household income, education, and the Townsend Deprivation Index (TDI). Risk factors included physical ones (cholesterol, HbA1c, BMI, blood pressure), lifestyle ones (smoking, alcohol, sleep, diet, physical activity), and social-emotional ones (stress, frequency of social contacts), as well as the composite Life's Essential 8 score. Mediation was assessed using the Baron and Kenny framework with bootstrap confidence intervals, separately for a single mediator (LE8) and for parallel models with all factors simultaneously.
Low income and high area deprivation were associated with a larger BAG (an older-looking brain) in a graded manner: the difference between extreme groups was 0.35 years for income and 0.33 years for TDI. Education showed associations in opposite directions depending on the MRI modality. LE8 mediated 36% of the income-BAG association and 17.9% of the TDI-BAG association. In parallel models, smoking was the largest mediator for all indicators: an indirect effect of 0.022 years for income and 0.016 years for TDI. HbA1c and systolic blood pressure followed smoking in effect size. Alcohol acted in the opposite direction: it suppressed the income-BAG association (indirect effect -0.013) and strengthened the TDI-BAG association (0.016), reflecting the different direction of alcohol's association with income and deprivation.
The study is the first to break down the link between socioeconomic status and brain aging into the contributions of individual modifiable factors, showing that status indicators are not interchangeable. Unlike neighboring works that either treated status as a single measure or did not perform formal mediation, this one uses three indicators and parallel models with eleven factors. This makes it possible to single out smoking as a priority target for interventions aimed at reducing inequalities in brain health.
The topic could move toward longitudinal studies with repeated measurements of BAG and risk factors to test temporal sequence and get closer to causal conclusions. Regional measures of brain aging derived from segmentation and surface reconstruction could add anatomical specificity. Studying social-emotional factors with more reliable instruments than single-item self-reports is also promising.
The results could influence the epidemiology of brain aging, preventive medicine, and public health policy by pointing to smoking as a key modifiable factor of inequality.
Conduct a longitudinal analysis with repeated MRI and risk factor measurements to establish temporal sequence. Include regional measures of brain structure and more precise measures of social-emotional factors.
The work links social determinants of health to the neurobiology of aging, answering the question of which specific mechanisms translate social inequality into biological brain aging. This contributes to solving the problem of explaining inequalities in dementia and cognitive decline.
The average brain age gap among participants was about 3.4 years on T1-weighted images, comparable to the effect of a decade of smoking.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.