O3‐01‐08: Hippocampal glucose metabolism predicts cognitive decline and correlates to disease progression in the ADNI population

Declines in regional cerebral metabolism (rCMglc) measured with Positron Emission Tomography (PET) imaging have been demonstrated to predict and correlate to the progression of Alzheimer's Disease (AD), and to differentiate between dementias. Recent studies have shown that the diagnostic accuracy is improved by including measurement of the hippocampus (HIP) (Mosconi et al, 2005, 2008). However, HIP rCMglc cannot be accurately and practically sampled in broad populations using conventional techniques. An automated method has been developed and demonstrated that achieves accurate, rapid sampling, and optimize sensitivity and specificity without compromise from spatial normalization, smoothing, spillover, and atrophy (Mosconi, 2005). By including HIP rCMglc, highly sensitive and specific differential diagnostic accuracies have been achieved (Mosconi, 2008). We applied our automated sampling method to 250 subjects from the ADNI database (78 female, 172 male, age 59 - 88, average 76 years; 79 NL, 111 MCI, 60 AD at initial diagnosis). Using the automated sampling approach, rCMglc was measured in 32 regions of interest in the baseline FDG-PET scans of 250 ADNI subjects. Regions included left and right hippocampi (HIP), amygdala, posterior cingulate cortex, inferior parietal lobes, medial, lateral, and superior temporal lobes, normalized to the cerebellum and age-corrected. Subjects were stratified into 7 subgroups across normal (NL), MCI, and AD categories, based upon initial diagnosis and progressive CDR, GDS scale, and MMSE scores. We observed a significant correlation between rCMglc in several regions and the clinical progression from stable NL (NL-nonDecl, n = 14) to NL with subsequent clinical decline (NL-Decl, n = 74, subdivided by extent of decline), to subcategories of MCI (MCI-nonDecl, n = 77 and MCI-Decl and mild-AD, n = 37) and AD. HIP rCMglc was found to be the most sensitive predictor of decline and discriminator between disease stages. As compared to the stable NL, HIP CMRglc was reduced by 5% in NL-Decl, 12% in stable MCI, 14% in MCI-Decl (P≤0.05), and 24% in AD (P≤0.001). These results demonstrate in a broad population the feasibility of achieving highly specific diagnostic accuracies by incorporating HIP rCMGlc, and provide insight to the continuum of decline through which normalcy advances to late stage dementia.