Autopsy studies in sporadic Alzheimer’s disease (AD) support the hypothesis that fibrillary amyloidosis (Aβ) occurs first in neocortex and later in striatum. We tested whether Aβ-PET could identify this sequence in-vivo, and potentially provide a staging measure for Aβ, in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Harvard Aging Brain Study (HABS).


      We included 1087 ADNI (367 cognitively normal (CN), 523 mild cognitive impairment (MCI), and 197AD) and 336 HABS (275CN, 46MCI, and 15AD) subjects with Aβ-PET (ADNI: [18F]-Florbetapir/AV45, ADNI composite reference; HABS: [C11]-PiB, cerebellar gray reference). Subjects were classified as high or low Aβ in neocortical and striatal (caudate and putamen) aggregates, using Gaussian mixture models. We evaluated changes in Aβ classification between baseline and follow-up Aβ-PET (0.9-5.1 years), and the association between Aβ classification at baseline and longitudinal memory performance, longitudinal hippocampal volume, and cross-sectional tau-PET ([18F]-T807/AV1451).


      Both cortical and striatal AV45/PiB have bimodal distribution (Lilliefors KS-test>0.1, p<0.001), allowing categorization. Striatal Aβ is elevated only if cortical Aβ is elevated, classifying subjects in three Aβ-stages: stage 0: low-Aβ in both regions, stage 1: high-cortical but low-striatal Aβ, stage 2: high-Aβ in both regions. Less than 1% of subjects have an indeterminate stage: high-striatal but low-cortical Aβ (Table 1, Figure 1). Subjects with Aβ-stage 0 at baseline are more likely to transition to high cortical than to high striatal Aβ after follow-up (ADNI: χ2=26.2, p<0.001; HABs: χ2=4.9, p=0.026). Aβ-stage 1 subjects are more likely to transition to high striatal Aβ than Aβ-stage 0 subjects (ADNI: χ2=19.6, p<0.001; HABs: χ2=33.5, p<0.001, Figure 2). Subjects with high-striatal Aβ at baseline are more at-risk for longitudinal memory decline, longitudinal hippocampal atrophy, and tau deposition than subjects with low-striatal Aβ, even among subjects with high-cortical Aβ (Table 2).


      Both datasets provide evidence that striatal Aβ accumulation occurs after neocortical Aβ, at later disease stages. Striatal Aβ is a risk factor for memory decline, hippocampal atrophy, and tau deposition, making it a potentially useful biomarker for tracking disease progression. Further work is ongoing to determine the potential added value of striatal Aβ for predicting cognitive decline in preclinical stages.
      Table 1Distribution of subjects by Aβ-stages, defined by neocortex (CX) and striatum (STR) baseline PET
      Aβ StagesStage 0Stage 1Stage 2Indeterminate
      Aβ Signal in CX/STRLow CX Low STRHigh CX Low STRHigh CX High STRLow CX High STR
      AD (ADNI, n=197)12%19%68%<1%
      MCI/AD (HABS, n=61)36%10%54%0%
      MCI (ADNI, n=523)40%29%31%<1%
      CN (ADNI, n=367)61%30%8%<1%
      CN (HABS, n=275)72%14%14%0%
      MCI and AD subjects in HABS have been grouped for power issues
      Table 2Association between Aβ-stages and: (1) longitudinal memory decline (2) longitudinal hippocampal volume and (3) cross-sectional tau (entorhinal and inferior temporal aggregate, cerebellar gray reference)
      ComparisonStage 1 vs 0Stage 2 vs 0Stage 2 vs 1
      Memory, ADNI (All: n=1087)T=-4.9, p=e-5T=-16.6, p=e-60T=-10.8, p=e-27
      Memory, ADNI (CN: n=367)T=-3.1, p=0.002T=-3.3, p=0.001T=-0.9, p=0.352
      Memory, HABS (All: n=332)T=-2.7, p=0.006T=-6.6, p=e-11T=-2.5, p=0.011
      Memory, HABS (CN: n=275)T=-2.9, p=0.004T=-4.9, p=e-5T=-1.3, p=0.193
      Hippocampus, HABS (CN: n=141)T=+0.4, p=0.707T=-2.0, p=0.046T=-1.8, p=0.077
      Tau, HABS (CN: n=121)T=+1.9, p=0.057T=+5.4, p=e-7T=+2.6, p=0.010
      Linear mixed models with random intercept and slopes, adjusted for age and sex (and education for memory)
      Figure thumbnail fx1
      Figure thumbnail fx2