Brain-age predicts subsequent dementia in memory clinic patients
INTRODUCTION: Research into quantitative neuroimaging biomarkers of dementia risk rarely uses data representative of everyday clinic practice. METHODS: We analysed T1-weighted MRI scans from memory clinic patients (n=1140; 60.2% female and mean [SD] age of 70.0 [10.8] years) to derive "brain-age", an index of age-related brain health. We determined which patients went on to develop dementia (n=476) via linkage to electronic health records. RESULTS: Cox regression indicated a 3% increased risk of dementia per brain-PAD year (brain-PAD = brain-age minus chronological age), HR(95% CI)=1.03(1.02, 1.04), p<0.001, adjusted for age, age^2, sex, MMSE and normalised brain volume. Brain-PAD remained significant even with a minimum time-to-diagnosis of 3 years (HR=1.06) and with MMSE score above 26 (HR=1.03). DISCUSSION: Memory clinic patients with older appearing brains are more likely to receive a subsequent dementia diagnosis. These results from a "real-world" dataset suggest quantitative neuroimaging biomarkers like brain-age could be readily used in the clinic.