Relationship Between PET-Assessed Amyloid Burden and Visual and Verbal Episodic Memory Performance in Elderly Subjects

Background: Studies of elderly subjects using biomarkers that are proxies for Alzheimer's disease (AD) pathology have the potential to document meaningful relationships between cognitive performance and biomarker changes along the AD continuum. Objective: To document cognitive performance differences across distinct AD stages using a categorization based on the presence of PET-assessed amyloid-beta (A beta) burden and neurodegeneration. Methods: Patients with mild dementia compatible with AD (n = 38) or amnestic mild cognitive impairment (aMCI; n = 43) and a cognitively unimpaired group (n = 27) underwent PET with Pittsburgh compound-B (PiB) assessing A beta aggregation (A+) and [F-18]FDG-PET assessing neurodegeneration ((N)+). Cognitive performance was assessed with verbal and visual episodic memory tests and the Mini-Mental State Examination. Results: The A+(N)+ subgroup (n = 32) showed decreased (p < 0.001) cognitive test scores compared to both A+(N)-(n = 18) and A-(N)- (n = 49) subjects, who presented highly similar mean cognitive scores. Despite its modest size (n = 9), the A-(N)+ subgroup showed lower (p < 0.043) verbal memory scores relative to A-(N)- subjects, and trend lower (p = 0.096) scores relative to A+(N)- subjects. Continuous A beta measures (standard uptake value ratios of PiB uptake) were correlated most significantly with visual memory scores both in the overall sample and when analyses were restricted to dementia or (N)+ subjects, but not in non-dementia or (N)- groups. Conclusion: These results demonstrate that significant A beta-cognition relationships are highly salient at disease stages involving neurodegeneration. The fact that findings relating A beta burden to memory performance were detected only at (N)+ stages, together with the similarity of test scores between A+(N)- and A-(N)- subjects, reinforce the view that A beta-cognition relationships during early AD stages may remain undetectable unless substantially large samples are evaluated.