Association between visceral adiposity index and cognitive dysfunction in US participants derived from NHANES data: A cross-sectional analysis

This study examines the cross-sectional association between the Visceral Adiposity Index (VAI) and domain-specific cognitive performance in US older adults. We analyzed data from 1323 participants aged ≥ 60 years in the National Health and Nutrition Examination Survey (NHANES) 2011 to 2014. Cognitive function was assessed with CERAD Word‑List Learning (memory/learning), Animal Fluency Test (semantic fluency/executive function), and the Digit Symbol Substitution Test (DSST; processing speed/attention). We fitted a sequence of multiple linear regression models: model 1 (crude), model 2 (adjusted for demographic covariates), model 3 (further adjusted for lifestyle factors), and model 4 (additionally adjusted for clinical comorbidities). Effect estimates are presented as β coefficients with 95% confidence intervals and P-values. VAI showed no significant link with composite cognition in age‑ and sex-unadjusted analyses (Model 1), while DSST approached significance (β = -0.16, 95% CI - 0.31 to - 0.02, P = .027). After adjusting for demographics (Model 2) DSST remained significant (β = -0.15, 95% CI - 0.29 to - 0.01, P = .031). Adding lifestyle factors (Model 3) gave a similar DSST effect (β = -0.15, P = .017) and also negative associations for immediate recall (β = -0.14, P = .039) and animal fluency (β = -0.18, P = .012). Full adjustment for sociodemographic and clinical comorbidities (Model 4) rendered these associations non‑significant (all P >.05), with Delayed Recall borderline (β = 0.15, P = .054). Sensitivity tests excluding clinical covariates partly restored the intermediate‑model effects, and removing the dyslipidemia covariate partly reversed attenuation, suggesting cardiometabolic comorbidities and overlap between VAI and lipid measures drive much of the attenuation. Age- and lifestyle-adjusted analyses showed inverse, domain‑specific links between higher VAI and cognition (most notably processing speed), but these weakened after full sociodemographic and clinical adjustment, suggesting measured sociodemographic and cardiometabolic factors largely explain the crude associations. Intermediate-model findings are hypothesis‑generating. Longitudinal studies with repeated measures, direct visceral-fat imaging, and mechanistic biomarkers are needed to separate confounding from mediation and to determine whether VAI or its metabolic components independently predict cognitive decline.