Longitudinal associations between amyloid and tau‐PET: Impact for prevention trials

Abstract Background In Alzheimer’s disease, cognitive decline is associated with a rapid increase in tau pathology, making tau‐PET a potentially useful outcome in prevention trials. To identify individuals most at‐risk of tau accumulation, we investigated the longitudinal associations between amyloid and tau in clinically normal (CN) and impaired (MCI/AD) participants from the Harvard Aging Brain Study. Method Serial Flortaucipir tau‐PET measures (2‐4 observations over a median 2.2years [0.8 to 6.0]) were obtained from 128CN and 11MCI/AD participants (Table 1). PiB (amyloid‐PET) data were expressed as Centiloid (CL) values. Flortaucipir data were expressed as PVC‐SUVr in inferior temporal neocortex scaled to subcortical white matter. We obtained tau slopes per subject from linear mixed‐effect models with random intercept and slopes. We tested in the entire sample, and in CN only, the association between Tau slope and baseline PiB and Tau‐PET. Linear, quadratic, and cubic relationships were tested. Age and sex were used as covariates. We computed power curves simulating therapeutic trials. Result In the entire sample (n=139), both baseline Tau (R 2 =0.34, Fig. 1 red‐line) and PiB (R 2 =0.30, Fig. 2 red‐line) were linearly associated with subsequent tau slope. In CN (n=128), these associations were lower than in MCI/AD, but significant (Tau: R 2 =0.16, PiB: R 2 =0.15, p≤0.00001). Importantly, a cubic fit best matched the PiB data (Fig. 2 blue‐line): A positive association between baseline PiB and Tau changes was only observed between PiB CL=‐2 and CL=66, with a maximal association at CL=32. In CN, the greatest tau changes were observed at CL=66. There was evidence of slowing in tau accumulation rate in the small number of CN with higher PiB CL. As a result, trials simulations demonstrated that increasing the Centiloid threshold to enroll CN participants did not increase power (Fig. 3 blue‐lines); in contrast, power was increased when excluding CN with CL>66 (Fig. 3 red‐line) or when enrolling CN with high baseline tau (Fig. 4 red‐lines). Enrolling MCI patients in therapeutic trials had a similar effect than enrolling high‐amyloid CN participants with baseline Tau SUVr>1.30 (Fig. 4 blue‐line). Conclusion A combination of amyloid and tau thresholds proved more effective to increase the power of Tau‐PET outcomes in prevention trials than using only amyloid.