Comparison of three diagnostic methods for determining amyloid positivity in cognitively intact elderly controls : Flutemetamol in Flander's Aging population (FLUFLAG study)

Background: Previous work has shown clinical and neuroimaging-based measures that display differences between MCI individuals who will later show progressive decline from those who will remain stable. The goal of this project was to identify changes in normal subjects which are associated with subsequent cognitive decline and conversion to MCI or AD.Methods: Utilizing the ADNI database, we identified 41 individuals who remained stable for 48-months (NC) and 16 who converted to MCI (CNV). Of these 57 subjects, all had available baseline clinical andMRI data, but only 16 NC and 11 CNV had available FDG-PET data. Wilcoxon Rank Sum tests assessed baseline demographic and clinical imbalances between CNV and NC. The effect of conversion status on neuroimaging measures at baseline was tested using linear regression modeling. Finally, linear discriminant analysis (LDA) models were created using features from an a-priori subset of clinical metrics, MRI measures, and FDG-PET measures obtained at baseline to predict which individuals would later convert to MCI and which would remain stable. We used a leave-one-out cross validation strategy and permutation analysis to confirm significance. Results: Significant differences between CNV and NC were found in ADAS-cog (P 1⁄4 0.006), FAQ (P 1⁄4 0.009), and AVLT (P 1⁄4 0.047) with no accompanying differences in age, gender, or education level. There was a marginal difference between groups in APOE status, but it did not reach significance (likely due to sample size). Conversion status was found to have a significant effect on several apriori selected neuroimaging measures including hippocampal (P 1⁄4 0.014) and entorhinal (P1⁄4 0.017) cortical volumes, entorhinal cortical thickness (P 1⁄4 0.021), and regional FDG-PET average of posterior cingulate (P 1⁄4 0.003). LDA models created using a combination of baseline MRI and FDG-PET measures were able to predict MCI conversion with up to 81% accuracy (P 1⁄4 0.005). Conclusions: The present study indicates that individuals destined for future cognitive decline may begin to express subtle but informative, and even predictive, differences in clinical assessments as well as brain structure and metabolism up to 4 years prior to displaying overt neuropsychological symptoms of decline.