Modeling longitudinal BP and impact on brain aging: findings from the Framingham Heart Study
Kim, Hyun (Monica)
MetadataShow full item record
While the association between blood pressure (BP) and brain health is increasingly strengthened by various clinical and epidemiologic research findings, less is known about the relationship between longitudinal patterns of BP across midlife and their impact on cognitive aging. Therefore, the current project used a large-scale, prospective longitudinal dataset of the Framingham Heart Study to model various long-term BP patterns using traditional methods and a novel machine-learning approach and investigated their impact on the development of dementia, as well as cognitive performance and brain volumes in late life. Study 1 examined intra-individual BP variability (BPV) across midlife and assessed its association with neuropsychological test performance, brain volumetric measures, and the development of dementia in late life. Contrary to previous findings in the elderly population, increased BP variability across midlife was not significantly associated with any brain aging measures. However, greater mean BP across midlife significantly predicted a greater risk of dementia. This finding led to the hypothesis that elevated BP in midlife, rather than BPV, may predict poorer brain and cognitive outcomes in late life. Study 2 investigated a long-term pattern of elevated BP using a cumulative exposure model, which has been well-recognized as a summary measure of longitudinal variation and cumulative burden associated with elevated cardiovascular risk. Consistent with the hypothesis driven from Study 1, elevated values of cumulative BP were associated with increased risk of dementia, along with poorer performance in most cognitive domains and reduced brain volume in areas including frontal, occipital, and temporal regions. Finally, Study 3 capitalized on a machine-learning approach, and namely, the dynamic time warping algorithm, to analyze BP data over the course of midlife using various pattern clusters. Although preliminary in nature, analyses using this novel approach detected various shapes of BP patterns across midlife. Clinical utility of these shapes and advantages of the machine-learning tool in BP research are discussed. Together, the results from these three studies suggest that BP pattern over the course of midlife, especially regarding long-term elevation of BP, is significantly associated with brain aging outcomes in late life.