Healthy and unhealthy plant-based diet quality and whole-brain imaging-derived phenotypes: an imaging-wide association study

OBJECTIVES: Plant-based diets are widely promoted for cardiometabolic and brain health, yet their associations with health may depend on the quality of the plant foods consumed. Whether healthy and unhealthy plant-based diet quality differ in their associations with brain structure across the whole brain remains unclear. METHODS: In a cross-sectional analysis of 23,460 participants with 24-h dietary recalls and brain MRI, we performed an imaging-wide association study of standardized healthy (hPDI) and unhealthy (uPDI) plant-based diet indices against 910 imaging-derived phenotypes (IDPs) spanning structural, diffusion, and functional MRI. Multivariable linear regression (adjusting for age, sex, ethnicity, household income, education, smoking, alcohol, physical activity, and total energy intake) yielded standardized beta coefficients; category-level and global false-discovery-rate (FDR) correction controlled multiple testing. The overall plant-based diet index (PDI), energy-adjusted Dietary Inflammatory Index (E-DII), DII, DASH, and AHEI served as comparison scores. RESULTS: uPDI showed widespread associations, with 187 white-matter diffusion and 110 structural-volume IDPs reaching category-FDR < 0.05, versus 1 and 19, respectively, for hPDI. Higher uPDI was associated with greater normalized grey-matter volume (beta=0.038, 95% CI 0.028 to 0.047) and higher fornix fractional anisotropy (beta=0.029, 95% CI 0.018 to 0.040). hPDI and uPDI were weakly correlated (Pearson r=0.230), and variance inflation factors were 1.16 and 1.13, respectively. In models including both indices, uPDI retained category-FDR-significant associations with 89 structural and 46 diffusion IDPs. The overall PDI and E-DII also showed widespread imaging associations; after mutual adjustment for overall PDI, uPDI retained associations with only 20 structural and 1 diffusion IDP. CONCLUSIONS: Unhealthy plant-based diet quality was associated with broad structural and white-matter imaging signatures that were substantially more numerous than those observed for healthy plant-based diet quality and remained evident after mutual adjustment for hPDI. Higher uPDI was associated with greater normalized grey-matter volume and higher fornix fractional anisotropy, a direction that was contrary to expectations based on prior literature. The findings from the overall PDI further indicate that much of this uPDI signal overlapped with overall plant-versus-animal dietary composition, leaving only a limited independent signal after mutual adjustment; these associations therefore cannot be attributed specifically to the quality of plant foods and warrant longitudinal confirmation.