An updated roadmap for brain-mental health associations in the UK Biobank

The UK Biobank offers an unprecedented opportunity to examine associations between brain imaging and mental health at scale, but their magnitude and replicability remain unclear. We correlated 17,580 imaging-derived phenotypes across seven MRI modalities with 309 mental health phenotypes in 8,415 participants, split into discovery and validation subsets. Associations were generally small (median |r| = 0.018; top 1% of 0.108), and they replicated poorly: only 9.0% of significant discovery associations were significant in validation. Eating disorders showed the largest effects of any mental disorders but replicated in only 6.0% of cases, whereas substance use showed the highest replication (14.5%). Susceptibility-weighted imaging produced the most replicable imaging associations. Longitudinal changes in brain and mental health measures were weakly related. We provide an interactive dashboard that returns the effect-size and replicability benchmarks for any combination of mental health category and imaging modality, supporting variable selection and study design in UK Biobank research.