Central obesity is selectively associated with cerebral gray matter atrophy in 15,634 subjects in the UK Biobank

The study investigates whether obesity—particularly central/abdominal adiposity—is associated with alterations in brain structure and connectivity. Prior research links obesity to type 2 diabetes, cardiovascular disease, and increased risk of cognitive impairment and dementia, with suggested structural brain changes. However, findings for white matter (WM) have been inconsistent, and most studies rely on body mass index (BMI), which does not capture fat distribution. The authors hypothesized that central obesity, measured by WHR, abdominal MRI, DXA, and impedance, would be more strongly associated with reduced gray matter (GM) volume than general obesity and examined whether WM volume or network efficiency (from diffusion MRI connectomics) are affected. They also assessed potential mediation by cardiometabolic risk factors and explored region-specific GM associations (basal ganglia, thalamus, hippocampus).

Previous work has reported global GM atrophy in obesity, but WM findings are mixed, with reports of both decreases and increases in WM volume and microstructure. Network metrics from diffusion MRI often correlate better with cognition than macrostructural WM measures and may be sensitive markers of WM injury. Central (visceral) obesity is more strongly linked than peripheral (subcutaneous) obesity to metabolic syndrome, type 2 diabetes, myocardial infarction, and Alzheimer’s disease. Most large-scale studies have used BMI, which does not reflect fat distribution; abdominal MRI and DXA can more precisely characterize visceral and subcutaneous fat. Prior studies have suggested associations between obesity and subcortical GM (e.g., thalamus, caudate, putamen, hippocampus), but comprehensive, multimodal examinations of central obesity versus general adiposity, and their relationships with GM, WM, and network efficiency, have been limited, particularly at large scale.

Design and cohort: Cross-sectional analysis within the UK Biobank cohort. Approximately 500,000 participants aged 40–69 were recruited (2006–2010). A subset underwent neuroimaging ~7.7 years later. This analysis included 15,634 participants with brain MRI and relevant covariates. Ethical approval: 11/NW/0382; informed consent; application 36509. Measures of obesity/body composition: (a) Anthropometry: BMI (kg/m^2) and WHR (waist/hip circumferences) from the imaging visit. (b) Whole-body bioelectrical impedance (Tanita BC418MA): segmental fat mass and fat-free mass. (c) Abdominal MRI (Siemens 1.5T MAGNETOM Aera): single-slice MOLLI and additional sequences; AMRA-derived visceral adipose tissue (VAT), abdominal subcutaneous adipose tissue (ASAT), total adipose tissue volume, total lean tissue volume. (d) Dual-energy X-ray absorptiometry (DXA; GE Lunar iDXA): android-to-gynoid fat mass ratio, VAT mass, trunk-to-leg fat mass, trunk-to-leg lean mass, fat mass index (FMI), and lean body mass index (LBMI); daily QC and phantom calibration. Brain imaging: Siemens Skyra 3T MRI. T1-weighted images processed by UK Biobank pipeline for brain volumes: total brain, GM, WM, plus regional volumes (caudate, putamen, pallidum, thalamus, hippocampus) via FAST. WM hyperintensities (WMH) quantified using T1 and T2-FLAIR with BIANCA. Diffusion MRI preprocessing with FSL (eddy, motion, gradient distortions); tensors from b=1000 s/mm^2; FA images registered to standard space. Deterministic tractography (MRtrix3) from FA images; streamline termination: 20–250 mm length, turning angle >45°, FA <0.15. Structural connectome constructed using the 90-region AAL atlas (cerebellum excluded). Edges defined by streamlines, weighted by number of streamlines × inverse mean length; weights <1 zeroed. Global and local network efficiencies computed in R (brainGraph/igraph). Cognitive testing: Touchscreen tasks at MRI visit: pairs-matching visual memory (errors), reaction time (mean ms), prospective memory (binary), Trail Making Test (Part A and B; attention/task switching). Blood biochemistry and blood pressure: From baseline and repeat visits: blood glucose, HDL, LDL, HbA1c; CRP referenced as inflammation marker. Manual blood pressure (systolic and diastolic) from first visit used in analyses. Statistical analysis: Participants stratified into six BMI/WHR groups to distinguish general vs central obesity (sex-specific WHR cutoffs; BMI: normal 18.5–25, overweight 25–30, obese >30; underweight BMI <18.5 excluded). Primary outcomes: normalized total brain volume, GM volume, WM volume, log-transformed WMH volume, network efficiencies, and cognitive scores. Analysis of covariance (ANCOVA) tested group differences with covariates: sex, age, Townsend Deprivation Index, alcohol intake, current smoking, diabetes mellitus, systolic and diastolic blood pressure, HbA1c. Partial correlations assessed associations between continuous body composition indicators and neuroimaging outcomes with the same covariates. ROI analyses examined subcortical regions (caudate, putamen, pallidum, thalamus, amygdala, nucleus accumbens, hippocampus), correcting for total GM volume; Bonferroni correction applied (p<0.004). Mediation analysis (IBM SPSS Amos SEM) tested whether blood glucose, HbA1c, HDL cholesterol, systolic/diastolic blood pressure, and CRP mediated associations between central obesity indicators (e.g., WHR, VAT) and GM volume. Interaction analyses assessed sex × obesity indicators on GM, WM, WMH, total brain volumes. Sample sizes: Brain volumes N=15,634; WMH N=14,662; DTI/network N=14,368; obesity assessments: impedance N≈15,437; abdominal MRI N≈5,155; DXA N≈4,212; cognitive measure availability varied (e.g., visual memory N=15,631; Trail Making N=7,519).

- Central obesity relates to lower GM volume: Across BMI/WHR groups, higher WHR (central obesity) was associated with progressively reduced cerebral GM volume; WHR was a stronger driver than BMI (GM: p=6.7×10^-16, ηp^2=0.004). The combined BMI/WHR group also associated with lower normalized total brain volume (p=2.2×10^-16, ηp^2=0.011). - No WM or network efficiency association: No significant association with WM volume (p=0.135, ηp^2≈0.001) or with brain network efficiency (global p=0.594; local p=0.607). A weak association with WMH was observed (p=2.4×10^-32, ηp^2=0.011). - Regional GM associations: After adjusting for total GM volume and multiple comparisons, significant associations with central obesity remained for bilateral thalamus, caudate, pallidum, and nucleus accumbens (all p<0.004). Associations with putamen, amygdala, and hippocampus were not significant after correction. - Body-fat mass indicators: Bioelectrical impedance—higher whole-body fat mass correlated with lower GM volume (partial r=-0.11, 95% CI [-0.13, -0.10], p=2.93×10^-47) and lower total brain volume (r=-0.07, p=1.77×10^-19); no WM volume or network metric associations. Abdominal MRI—greater total adipose tissue volume (r=-0.08, p=1.16×10^-11), visceral adipose tissue volume (r=-0.096, p=4.28×10^-12), and abdominal subcutaneous adipose tissue volume (r=-0.094, p=1.18×10^-11) each correlated with lower GM volume; no WM volume or network efficiency associations. DXA—higher fat mass index associated negatively with GM and total brain volume (not WM); - Lean mass indicators: Bioelectrical impedance—higher fat-free mass associated with lower GM (r=-0.17, p=5.77×10^-101), lower total brain (r=-0.12, p=1.44×10^-53) and slightly lower WM (r=-0.03, p=3.7×10^-5); no network associations. Abdominal MRI—lean tissue volume normalized by body weight showed small positive correlations with GM, WM, and total brain volumes; DXA—LBMI negatively associated with GM and total brain volumes (not WM). - Fat distribution (central vs peripheral): Abdominal MRI VAT volume showed a significant negative correlation with GM volume (p≈10^-12) and total brain volume, but not WM; similar for ASAT and total adipose tissue. DXA VAT mass negatively associated with GM and total brain volume. Android-to-gynoid and trunk-to-leg fat ratios showed strong negative unadjusted correlations with GM that were attenuated to non-significance after adjusting for covariates. - Cognitive outcomes: No significant ANCOVA effects of BMI/WHR group on visual memory errors, Trail Making (B–A), prospective memory, or symbol digit matching after multiple-comparison correction. - Mediation: Associations of WHR and VAT (MRI/DXA) with GM volume were not mediated by blood glucose, HbA1c, HDL cholesterol, systolic/diastolic blood pressure, or CRP (all p>0.05). - Sex interactions: Significant interactions between sex and some obesity indicators (e.g., body fat mass, fat-free mass by impedance; ASAT and total adipose tissue by abdominal MRI) on brain volumes; no interactions for android-to-gynoid fat mass ratio, trunk-to-leg fat or lean mass ratios.

The findings support the hypothesis that central, rather than general, obesity is selectively associated with reduced gray matter volume. Using multiple complementary measures of adiposity and fat distribution (WHR, abdominal MRI, DXA, impedance) in a large population sample, the study shows consistent negative associations of central obesity with GM volume, not with WM volume or connectome efficiency. Regionally, central obesity relates to smaller volumes in subcortical nuclei involved in reward and inhibitory control (thalamus, caudate, pallidum, nucleus accumbens), aligning with known neurocircuits of energy balance and eating behavior. The lack of mediation by glycemia, lipids (HDL), blood pressure, or CRP suggests that commonly measured components of metabolic syndrome and systemic inflammation do not directly explain the GM atrophy, pointing to other mechanisms (e.g., neuroinflammation, adipokines, hormonal or vascular pathways) or reverse causality. Absence of WM and network alterations suggests that GM changes may occur earlier, with potential downstream WM effects later in life; longitudinal data are needed. Overall, central adiposity provides more informative risk stratification for brain structural compromise than BMI alone, advancing understanding of obesity-related brain changes.

Central body-fat distribution is associated with global and regionally specific gray matter atrophy, whereas white matter volume and network efficiency appear unaffected in this large population-based cohort. Being overweight by BMI alone showed limited adverse associations, underscoring the importance of central adiposity measures. The study’s multimodal assessment of adiposity strengthens the evidence base. Future work should focus on longitudinal trajectories to determine causality and temporality, identify biological mediators (beyond standard metabolic and inflammatory markers), and evaluate whether modifying central obesity mitigates GM loss and related cognitive risks.

- Cross-sectional design precludes causal inference. - Potential selection bias in UK Biobank toward healthier volunteers, limiting generalizability. - Some effect sizes are small, though highly significant due to large sample size. - Body composition measurement constraints: abdominal MRI covered only the abdomen (no whole-body MRI), DXA can be less accurate in individuals with obesity, and impedance depends on hydration status. - Variability in available sample sizes across modalities and cognitive tests may affect power for specific analyses.