Alzheimer's Imaging Consortium.

Area of Science:

• Neuroimaging
• Machine Learning
• Cardiovascular Health

Background:

• Hypertension (HTN) is a known risk factor for neurodegeneration and dementia.
• Individual outcomes vary significantly despite established links.
• Machine learning (ML) models can quantify HTN-related neurodegeneration from structural MRI (sMRI).

Purpose of the Study:

• To investigate the predictive capacity of the SPARE-HTN ML model for incident hypertension.
• To assess the mediating role of SPARE-HTN in the relationship between hypertension and cognition.

Main Methods:

• Developed SPARE-HTN model using sMRI from 37,098 individuals.
• Evaluated SPARE-HTN in 968 individuals with longitudinal clinical data.
• Used Cox regression and mediation models to analyze HTN incidence and cognitive effects.

Main Results:

• SPARE-HTN was elevated in individuals who developed hypertension within 3-7 years.
• Elevated baseline SPARE-HTN predicted higher risk of incident hypertension.
• SPARE-HTN mediated up to 26% of the effect of HTN on cognition, outperforming WMH volume.

Conclusions:

• The ML-based SPARE-HTN marker predicts incident hypertension before clinical diagnosis.
• Subclinical cerebrovascular changes associated with blood pressure variations are detectable.
• SPARE-HTN offers potential for individualized dementia risk stratification and therapeutic efficacy measurement in midlife.