Large-Scale Brain Network Connectivity Mediates the Association Between Metabolic Risk Factors and Cognition: An fMRI Study of the Human Connectome Project.
Subclinical metabolic disturbances have been associated with functional brain changes; however, the neural pathways mediating their cognitive effects remain unclear.
This study investigated whether large-scale brain network connectivity mediates associations between metabolic risk factors and cognition in healthy young adults.
Data were obtained from 676 participants (22 - 37 years) in the Human Connectome Project. Metabolic indices (body mass index (BMI), glycated hemoglobin, thyroid-stimulating hormone (TSH), systolic/diastolic blood pressure, and hematocrit) were assessed alongside the National Institutes of Health (NIH) Toolbox Cognition Battery scores, the Mini-Mental State Examination, and fluid intelligence measures. Resting-state fMRI data underwent group independent component analysis to identify 14 intrinsic connectivity networks. Partial correlations were computed between network time series, and multiple regression models were adjusted for age, sex, race, and education. Mediation analyses with bootstrapped confidence intervals were conducted to test whether network connectivity explained the metabolic-cognition relationships.
We found that large-scale resting-state networks significantly mediated the associations between specific metabolic risk factors and cognitive performance. Specifically, BMI, hematocrit, and TSH showed significant associations with both cognition and network connectivity (all P < 0.05). BMI-related reductions in vocabulary performance were fully mediated by altered connectivity in the frontoparietal-language, sensorimotor-frontoparietal, primary visual-auditory, and default mode-executive attention networks and partially mediated (compensatory effect) by salience-sensorimotor connectivity (indirect effects: -0.0238, -0.0115, -0.0110, -0.0096, and 0.0068, respectively; 95% CIs: [-0.0421, -0.0085], [-0.0239, -0.0030], [-0.0237, -0.00016], [-0.0206, -0.0014], and [0.0001, 0.0165], respectively). Hematocrit's positive association with vocabulary performance was partially mediated by the primary visual-auditory, salience-auditory, and language-dorsal attention networks, with primary visual-ventral attention connectivity exerting a suppressive effect (indirect effects: 0.0094, 0.0149, 0.0074, and -0.0113, respectively; 95% CIs: [0.0007, 0.0215], [0.0032, 0.0319], [0.0001, 0.0197], and [-0.0249, -0.0012], respectively). TSH was positively associated with fluid intelligence via a direct pathway (P = 0.010), without mediation by resting-state connectivity.
These findings highlight large-scale brain networks as potential intermediate phenotypes that link metabolic health to cognition, suggesting that targeted neuromodulation of vulnerable circuits, combined with metabolic interventions, may offer novel strategies for preserving cognition in at-risk populations.
This study investigated whether large-scale brain network connectivity mediates associations between metabolic risk factors and cognition in healthy young adults.
Data were obtained from 676 participants (22 - 37 years) in the Human Connectome Project. Metabolic indices (body mass index (BMI), glycated hemoglobin, thyroid-stimulating hormone (TSH), systolic/diastolic blood pressure, and hematocrit) were assessed alongside the National Institutes of Health (NIH) Toolbox Cognition Battery scores, the Mini-Mental State Examination, and fluid intelligence measures. Resting-state fMRI data underwent group independent component analysis to identify 14 intrinsic connectivity networks. Partial correlations were computed between network time series, and multiple regression models were adjusted for age, sex, race, and education. Mediation analyses with bootstrapped confidence intervals were conducted to test whether network connectivity explained the metabolic-cognition relationships.
We found that large-scale resting-state networks significantly mediated the associations between specific metabolic risk factors and cognitive performance. Specifically, BMI, hematocrit, and TSH showed significant associations with both cognition and network connectivity (all P < 0.05). BMI-related reductions in vocabulary performance were fully mediated by altered connectivity in the frontoparietal-language, sensorimotor-frontoparietal, primary visual-auditory, and default mode-executive attention networks and partially mediated (compensatory effect) by salience-sensorimotor connectivity (indirect effects: -0.0238, -0.0115, -0.0110, -0.0096, and 0.0068, respectively; 95% CIs: [-0.0421, -0.0085], [-0.0239, -0.0030], [-0.0237, -0.00016], [-0.0206, -0.0014], and [0.0001, 0.0165], respectively). Hematocrit's positive association with vocabulary performance was partially mediated by the primary visual-auditory, salience-auditory, and language-dorsal attention networks, with primary visual-ventral attention connectivity exerting a suppressive effect (indirect effects: 0.0094, 0.0149, 0.0074, and -0.0113, respectively; 95% CIs: [0.0007, 0.0215], [0.0032, 0.0319], [0.0001, 0.0197], and [-0.0249, -0.0012], respectively). TSH was positively associated with fluid intelligence via a direct pathway (P = 0.010), without mediation by resting-state connectivity.
These findings highlight large-scale brain networks as potential intermediate phenotypes that link metabolic health to cognition, suggesting that targeted neuromodulation of vulnerable circuits, combined with metabolic interventions, may offer novel strategies for preserving cognition in at-risk populations.