Abstract
BACKGROUND: The relationship between regional adiposity and dementia remains poorly understood. METHODS: This study included 440 861 UK Biobank participants initially free of dementia, stroke and cancer. Hazard ratios (HRs) and 95% confidence intervals (CIs) of dementia across quartiles of waist circumference (WC) or hip circumference (HC) were calculated, after multivariable adjustment with mutual adjustment for WC and HC. The potential mediating roles of 21 biomarkers covering distinct metabolic pathways were assessed. In a subsample with body composition quantified by dual-energy X-ray absorptiometry, the associations of android and gynoid fatness with magnetic resonance imaging-derived volumetric brain measures were further assessed. RESULTS: Over a median follow-up of 12.7 years, there were 2899 incident cases of dementia in females and 3306 cases in males. After multivariable adjustment, WC was positively associated with the risk of dementia (predominantly vascular dementia) in both sexes, but the associations disappeared after further adjusting for the treatments for metabolic disorders. HC was inversely associated with dementia both in females (HR(Q4 vs. Q1) = 0.75; 95% CI: 0.64, 0.86; P-trend < 0.001) and in males (HR(Q4 vs. Q1) = 0.83; 95% CI: 0.72, 0.95; P-trend = 0.018), with the leading mediator being apolipoprotein B (11.15%) and vitamin D (9.03%) for female and male associations, respectively. In both sexes, gynoid fat percent was related to larger grey matter volumes (β (female) = 1.21; 95% CI: 0.74, 1.68; β (male) = 0.96; 95% CI: 0.28, 1.64) and smaller volumes of white matter hyperintensities (β (female) = -1.96; 95% CI: -2.50, -1.43; β (male) = -2.57; 95% CI: -3.39, -1.76). CONCLUSIONS: Android and gynoid adipose tissues may exert divergent influences on the development of dementia, with evidence for shared and distinct mechanisms between sexes.