Abstract
284
Background: Understanding the time course of amyloid-beta (Aβ) aggregation in Alzheimer’s disease (AD) is important for understanding the chronological sequence of events in the AD continuum. This work uses serial [11C]Pittsburgh Compound B (PiB) PET imaging data to 1) characterize the relationship between rates of Aβ accumulation and Aβ burden, and 2) estimate the time course of PiB-measure Aβ accumulation by applying a novel algorithm that uses a combination of Euler’s method with discrete sampling.
Methods: 181 initially cognitively unimpaired participants from the Wisconsin Registry for Alzheimer’s Prevention underwent serial [11C]PiB PET imaging between 2010 and 2019 (baseline age mean±SD = 61.0±6.2, 66% female, median 3 serial scans, follow-up duration mean±SD = 6.2±2.1 years). Aβ burden was quantified for each scan by averaging distribution volume ratios (DVR; reference Logan, cerebellum gray reference) across 8 bi-lateral regions of interest in subject space. Discrete sampling was used for a-priori DVR values to characterize the relationship between ΔDVR/yr and DVR and the proportion of participants with ΔDVR/yr >0 for each DVR value. The average time-course of Aβ accumulation for the sample was estimated using an iterative algorithm. The algorithm calculates the mean ΔDVR/yr for longitudinal scan pairs that cross a DVR query value and then applies Euler’s method with a fixed time interval (0.5 yr) to generate the next query value. The algorithm terminates once too few scan pairs exist for an average or once the maximum number of iterations has been reached and outputs DVR vs time data. The goodness of model fit was assessed by mapping individual observed trajectories onto the modeled DVR vs. time function, and estimating the percentage of observed DVR variance explained by the model for all participants with data in the modeled range. Results: PiB ΔDVR/yr plotted as a function of DVR indicated minimal variability in ΔDVR for DVR <1.6, with a marked increase in variability for DVR>1.6, potentially due to fewer participants in this range (Fig 1). ΔDVR/yr was near 0.0 for DVR<1.1, increased gradually for 1.1<DVR<1.18, and were nearly constant at mean±SD = 0.03±0.003 ΔDVR/yr for 1.18<DVR<1.6. The proportion of participants with positive longitudinal ΔDVR/yr was lowest for 1.0<DVR<1.1 and increased to all participants having positive ΔDVR/yr for DVR>1.2. The iterative algorithm produced an Aβ accumulation curve (PiB DVR vs. time; Fig 2) that was nearly constant for DVR<1.11, began to increase for 1.1<DVR<1.15, and increased linearly for DVR>1.15. The proportion of DVR variance explained by the model was estimated to be R2=0.94 using participants with data represented in the range of the model (1.09<DVR<1.80; n=65). Conclusions: This work demonstrates a novel algorithm for modeling the time-course of Aβ accumulation measured by serial PET imaging. These results suggest that Aβ accumulation rates are largely homogeneous with respect to the level of Aβ burden, and that the level of Aβ may be informative for estimating the cumulative duration of amyloidosis. Future work is ongoing cross-validating this approach with other methods and testing this methodology in other cohorts.