Brain white matter hyperintensities-predicted age reflects neurovascular health in middle-to-old aged subjects

BACKGROUND: age-related neurovascular structural and functional impairment is a major aetiology of dementia and stroke in older people. There is no single marker representative of neurovascular biological age yet. OBJECTIVE: this study aims to develop and validate a white matter hyperintensities (WMH)-based model for characterising individuals' neurovascular biological age. METHODS: in this prospective single-site study, the WMH-based age-prediction model was constructed based on WMH volumes of 491 healthy participants (21-89 years). In the training dataset, the constructed linear-regression model with log-transformed WMH volumes showed well-balanced complexity and accuracy (root mean squared error, RMSE = 10.20 and mean absolute error, MAE = 7.76 years). This model of neurovascular age estimation was then applied to a middle-to-old aged testing dataset (n = 726, 50-92 years) as the testing dataset for external validation. RESULTS: the established age estimator also had comparable generalizability with the testing dataset (RMSE = 7.76 and MAE = 6.38 years). In the testing dataset, the WMH-predicted age difference was negatively associated with visual executive function. Individuals with older predicted-age for their chronological age had greater cardiovascular burden and cardiovascular disease risks than individuals with normal or delayed predicted age. These associations were independent of chronological age. CONCLUSIONS: our model is easy to use in clinical practice that helps to evaluate WMH severity objective to chronological age. Current findings support our WMH-based age measurement to reflect neurovascular health and have potential diagnostic and prognostic value for clinical or research purposes in age-related neurovascular disorders.