Revising the amyloid buildup curve in Alzheimer’s disease? - Partial volume effect correction of longitudinal amyloid PET data

Objectives: Many researchers believe that, in Alzheimer’s disease (AD), brain β-amyloid load accumulation over time reaches a plateau at an early symptomatic disease stage. In hot-spot imaging techniques like amyloid PET, brain atrophy-related partial volume effects (PVEs) negatively affect the image signal. We aimed at testing the shape of the amyloid buildup curve in AD after PVE correction in longitudinal ADNI data.

Methods: We included ADNI data of 216 subjects (17 with AD dementia: 76±6 years, 8 females, 22.4±2 MMSE; 122 with early MCI: 71±8 years, 48 females, 28.3±2 MMSE; 77 with late MCI: 71±8 years, 36 females, 27.5±2 MMSE) who had at least two longitudinal 18F-florbetapir PET scans available, as well as corresponding MRIs (within ±50 days), and corresponding MMSE scores. PVEs were corrected using the modified Müller-Gärtner method implemented in PVELab. SUVRs were obtained for all brain regions in PVELab’s anatomical atlas and a cortical composite region using the cerebellar cortex as reference region. To include amyloid-positive subjects only, we applied a composite SUVR threshold of 1.1.

Results: PVE correction revealed significantly higher regional and composite SUVR changes over time when compared to non-corrected PET data (composite SUVRs: P = 0.0002). With decreasing MMSE scores, the slopes of the composite SUVR increases were significantly higher with vs. without PVE correction (F = 7.1, P = 0.008). Composite SUVR changes over time were significantly correlated to the baseline MMSE scores only with PVE correction (r = -0.11, P = 0.049), but not without PVEC (r = -0.07, P = 0.16).

Conclusions: These PVE correction results of longitudinal amyloid PET data indicate that the amyloid buildup curve in AD may not have reached a plateau at the early symptomatic disease stage. A further prospective evaluation of the impact of PVE correction on the in vivo characterization of time-dependent AD pathology is warranted.