Background: Anorexia nervosa (AN) is characterised by disturbances in cognition and behaviour surrounding eating and weight, which may relate to the structural connectivity of the brain that supports effective information processing and transfer.
Methods: Diffusion-weighted MRI data acquired from female patients with AN (n = 148) and female healthy controls (HC; n = 119), aged 12-40 years, were combined across five cross-sectional studies. Probabilistic tractography was completed, and full cortex connectomes describing streamline counts between 84 brain regions generated and harmonised. The network-based statistic tested between-group differences in connectivity strength of brain subnetworks. Whole-brain connectivity of brain regions was indexed using graph theory tools, and compared between groups using multiple linear regression. Associations between structural connectivity variables that differed between groups, and illness severity markers, were explored amongst AN patients using multiple linear regression. Statistical models included age, motion, and study as covariates.
Outcomes: The network-based statistic indicated AN patients, relative to HC, had reduced connectivity in a network comprising subcortical regions and greater connectivity between frontal cortical regions (p < 0.05, FWE corrected). Graph theory analyses supported reduced connectivity of subcortical regions, and greater connectivity of left occipital cortex, in patients relative to HC (p < 0.05, permutation corrected). Reduced subcortical network connectivity was associated with lower BMI among the AN group.
Interpretation: Structural differences in subcortical and cortical networks are present in AN, and may reflect illness mechanisms.