Background: The amyloid cascade hypothesis of Alzheimer's Disease was initially proposed over 20 years ago. Recent data from studies such as the Alzheimer's Disease Neuroimaging Initiative (ADNI) suggest that the relationship between brain amyloid burden, biomarkers of neurodegeneration, genetics, age and disease stage may be complex.
      Methods: Data used in this study, including brain measures, was downloaded from the ADNI public database. Subjects were 154 with late mild cognitive impairment (lMCI), 238 with early MCI (eMCI), and 183 who were cognitively normal (CN). Neuropsychological evaluations included tests of memory (Rey Auditory Verbal Learning Test) and executive function (Trails B score). Subjects also had an F-18 AV-45 amyloid PET scan, an FDG PET scan and a structural MRI scan. Imaging measures included hippocampal volume, the average normalized CMRglu of 6 AD-sensitive regions, and average SUVR for the amyloid scans of the following regions: anterior and posterior cingulate, prefrontal, lateral temporal and parietal (AV-45 SUVR). Path analysis was carried out to explore the dependencies of cognitive scores (RAVLT and Trails B),hippocampal volume and cerebral glucose metabolism (CMRglu) with respect to APOE genotype (epsilon4+ or epsilon4-), age and AV-45 SUVR. The following path model was considered: APOE genotype and age as predictor variables; AV-45 SUVR as a mediator variable; and Trails B Score, RAVLT, hippocampal volume and mean FDG SUVR as outcomes.
      Results: Standardized regression weights for significant correlation are reported in Figure 1. In the path model, APOE genotype was strongly associated with amyloid SUVR, weakly associated with CMRglu and cognitive scores (and not at all to hippocampal volume). In contrast there was a stronger relationship between AV-45 SUVR, on the one hand, and CMRglu, hippocampal volume and cognitive function, on the other hand, independent of APOE genotype. Age was more strongly associated with hippocampal volume thanwith CMRglu and the cognitive measures.
      Conclusions: The effect of APOE genotype on cognition, hippocampal volume and CMRglu is mediated primarily by its effect on increasing amyloid deposition, with relatively little independent contribution by APOE genotype.
      Figure thumbnail fx1
      Figure 1Path Analytic Models addressing dependencies of Cognitive Variables [Trails B (A) and RAVLT (B)], Volumetric variable [hippocampal volume] (C) and Metabolic variables [rCMRglu](D) to APOE e4 genotype, age and mean [18F]-florbetapir AV45 SUVRs, showing the Standardized Regression Weights for each correlation. The associations of Age (0.179, p<0.001) and APOE genotype (0.440 p<0.001) with AV-45 SUVR were found to be significant across all path models.