Structural MRI and amyloid PET scans have been used as short-term, but not long-term predictors of incident dementia in elderly individuals. The purpose of this study is to examine the long-term association between MRI brain structural integrity and PET measures of amyloid-β (Aβ) deposition, and incident dementia
183 non-demented elderly subjects (mean age 85.5) were followed from 2009 to 2015 with annual clinical evaluation (mean follow up 5.2 years). Pittsburgh compound-B (PiB) PET was analyzed as a dichotomous and continuous variable. Neurodegeneration was based on hippocampal (HIP) volumes (<25th percentile), and small vessel disease on white matter hyperintensities (WMH) (>75thpercentile).
In 2009, 148 participants were categorized as cognitively normal (CN) and 35 as mild cognitive impairment (MCI). By 2015, 25% of the CN participants and 66% of MCI developed dementia during follow-up. Risk of dementia was linearly related to amyloid deposition. Dementia incidence was 32/1000 age adjusted person-years (PYs) in participants with a PiB SUVR <1.34, and 97/1000 in subjects with PiB SUVR >2.16. HIP and WMH were also associated with increased risk of dementia. In combined models, the CN who had normal PiB, HIP and WMH had very low risk of dementia (35/1000 PYs). By contrast, when all three biomarkers were abnormal, the risk was 192.3/1000 PYs. PiB-positivity in isolation (SUVR >1.57) was not associated with an increased risk of dementia among CN (37/1000 PYs), nor was HIP atrophy (40.6/1000 PYs) or WMH volume alone (20/1000 PYs). However, using continuous variables, greater amyloid deposition (HR: 2.62), smaller HIP (HR: 0.69), and more WMH (HR: 3.37) independently increased the risk of incident dementia.
We found that when the dichotomous measures of the common biomarkers for Alzheimer’s disease (AD) are considered in isolation, they are unrelated to clinical dementia risk. It is only when all three abnormalities are present, and the AD pathology is established, and the clinical symptoms of dementia are imminent.
© 2017 Published by Elsevier Inc.