Predicting Cognitive Decline Using Brain Age
Post by Andrew Vo
The takeaway
Brain age is a measure of an individual brain’s deviation from a normative aging trajectory. This measure may serve as a biomarker for personalized prediction, diagnosis, and intervention of age-related cognitive decline and dementia.
What's the science?
Brain age refers to the degree to which an individual’s brain deviates from the average aging process. Greater brain age is associated with cognitive impairment and neurodegenerative diseases such as Alzheimer’s. Previous studies have introduced machine-learning-based brain age models and demonstrated their sensitivity to cognitive functioning as well as amyloid positivity – an indicator of Alzheimer’s disease. However, such models have yet to be independently validated for their generalizability in different ethnic and clinical samples. This week in Molecular Psychiatry, Karim et al. employed their existing brain age model in an independent clinical sample and tested its performance in discriminating patient diagnoses and predicting future cognitive decline.
How did they do it?
The authors applied their brain age model to a sample of 650 patients from a South Korean memory clinic. Diagnoses ranged from subjective cognitive decline to mild cognitive impairment to overt dementia. All patients underwent magnetic resonance imaging (MRI) to measure grey matter volumes, positron emission tomography (PET) imaging to detect amyloid status, genotyping for APOE4 (the APOE4 genotype places someone at high risk for Alzheimer’s disease), and clinical assessments. A subset of these patients was followed up for cognitive testing one year after baseline. The brain age model used was previously trained on data from a largely Caucasian sample with PET-confirmed amyloid negativity. Each patient’s brain age was estimated as the residual error after regressing out expected effects of age and sex from the normative brain age model. Thus, a higher brain age reflected a brain that appeared “older” than expected at that chronological age. Each patient’s brain age was then related to their clinical measure of cognitive function and longitudinal decline.
What did they find?
Brain ages estimated for the dementia group showed a more pronounced deviation from their chronological age compared to the non-dementia group, particularly for younger patients. Similarly, brain age residuals were greater in patients with dementia compared to those with either subjective cognitive decline or mild cognitive impairment. Greater brain age residuals were associated with more severe cognitive impairment as well as higher amyloid deposition. Examining those patients with longitudinal follow-up data, brain age residual at baseline predicted future cognitive decline even after adjusting for APOE4 or amyloid status.
What's the impact?
This study validated a previous machine-learning-based brain age model in an independent ethnic and clinical sample, demonstrating its generalizability. Brain age differences could be discerned across the dementia continuum and predicted future cognitive decline. Brain age modeling shows promise as a useful tool for predicting and tracking age-related cognitive impairment and neurogenerative disease.