Brain structural pathways linking dietary patterns and cognitive performance in adolescents

Dietary patterns covary with cognitive performance during adolescence, but the structural brain features associated with this covariance remain incompletely characterized. To identify structural brain correlates of cognition-aligned dietary patterns in adolescents while addressing adiposity, ethnicity adjustment, family relatedness, and multiple testing. We analyzed 3,308 adolescents from ABCD Release 5.1. Ten food-group measures were summarized by five dietary principal components (PCs). We constructed sample-derived, cognition-aligned dietary scores (PolyIQ) and used repeated five-fold family-level cross-fitting so that each participant's score was calculated using cognition-derived weights estimated without that participant or members of the same family. The original PCs provided an outcome-independent benchmark. Primary models adjusted for age, sex, pubertal stage, socioeconomic status, income-to-needs ratio, sleep, and age- and sex-standardized BMI; ethnicity was excluded from primary models and added in sensitivity analyses. Brain-cognition models and family-cluster bootstrap indirect-effect models examined fractional anisotropy (FA), cortical thickness, surface area, and regional volume. All pre-specified indirect-effect models, including nonsignificant results, were retained. Cross-fitted dietary scores were associated with crystallized cognition (β = 0.099, SE = 0.018, P < 0.001) and fluid cognition (β = 0.064, SE = 0.015, P < 0.001); the five original dietary PCs were jointly associated with both domains. Brain-cognition associations in the bilateral fornix and hippocampal cingulum remained after BMI adjustment. In the all-ROI analysis, indirect associations through right and left fornix FA survived false-discovery-rate correction for both crystallized cognition (right: 0.0067, 95% CI [0.0016, 0.0134]; left: 0.0063, [0.0013, 0.0126]) and fluid cognition (right: 0.0061, [0.0019, 0.0127]; left: 0.0052, [0.0011, 0.0106]). No gray-matter indirect effect survived correction across all tested regions. A sample-derived dietary pattern aligned with cognition covaried with FA in bilateral fornix pathways. The findings were robust to adiposity adjustment and were similar when ethnicity was added, but the cross-sectional design supports only a decomposition of shared variance, not a causal or temporally ordered mechanism.