Associations between obesity indices and cognitive function: Evidence from a middle‐aged observational and Mendelian randomisation study

AIMS: Clinical trials investigating the relationship between obesity and cognitive decline have yielded inconsistent results across various age groups. This study explored the associations of body mass index (BMI), waist-hip ratio (WHR) and trunk fat mass (TFM) with cognitive function using a middle-aged cross-sectional and Mendelian randomisation (MR) approach. MATERIALS AND METHODS: A total of 72 899 middle-aged participants from the UK Biobank (UKB) were used to examine the association between obesity indices (BMI, WHR and TFM) and cognitive function through multivariable linear regression. MR analyses were subsequently performed to explore the potential causal relationships. For BMI, 449 single-nucleotide polymorphisms (SNPs) were considered (n = 532 396), for WHR, 36 SNPs (n = 210 082) and for TFM, 277 SNPs (n = 331 093), as exposure variables. Cognitive function outcomes, derived from large genome-wide association studies of European ancestry, included fluid intelligence score (FIS, n = 149 051), memory (n = 48 080) and complex processing speed (n = 114 724). MR analysis was primarily conducted using inverse-variance weighted methods, supplemented by sensitivity analyses to assess pleiotropy and exclude variants with potential pleiotropic effects. RESULTS: In a cross-sectional analysis of 72 899 middle-aged UKB participants, higher obesity indices (BMI and WHR) were consistently associated with poorer cognitive performance across multiple domains. After full covariate adjustment, participants in the highest BMI tertile (T3) demonstrated significant adverse cognitive declines relative to the lowest tertile (T1). Specifically, T3 exhibited inverse associations with FIS (β = -0.024, 95% confidence interval [CI]: -0.042 to -0.006, p-value adjusted using the Benjamini-Hochberg false discovery rate procedure [p_BH] = 0.032), memory (β = -0.111, 95% CI: -0.130 to -0.092, p_BH <0.001). Similarly, WHR tertiles 2 and 3 (both higher than T1) showed consistent adverse effects across cognitive domains, with the negative correlation observed in T3 for FIS (β = -0.072, 95% CI: -0.096 to -0.048, p_BH <0.001), memory (β = -0.116, 95% CI: -0.141 to -0.090, p_BH <0.001) and processing (β = -0.091, 95% CI: -0.115 to -0.068, p_BH <0.001). For TFM, significant consistent adverse associations were identified in tertiles 2 and 3, with the inverse relationship observed for memory performance in T3 (β = -0.075, 95% CI: -0.094 to -0.056, p_BH <0.001). Critically, these associations between BMI, WHR and FIS remained statistically significant following Benjamini-Hochberg correction for multiple testing (p_MR_BH <0.05) and were further corroborated by the Mendelian randomisation robust adjusted profile score approach, with consistent findings observed across all sensitivity analyses. CONCLUSIONS: This study demonstrates a negative association between BMI, WHR, TFM and cognitive function in middle-aged populations. Moreover, BMI, WHR and TFM appear to have a causal relationship with specific domain cognitive decline. Further multicentre, multiregional and large-scale studies are required to validate these findings.