The future of clinical trials for Alzheimer's disease. A blood-based biomarker perspective

Objectives: The primary objective was to investigate the utility of blood-based
biomarkers of amyloid, tau, and neurodegeneration for (i) screening, (ii)
enrichment, and (iii) tracking response to treatment in clinical trials of Alzheimer’s
disease.
Methods: Longitudinal, participant-level data used in these studies was drawn from
the Swedish BioFINDER study and the ADNI study. Participants were classified as
cognitively unimpaired, mild cognitive impairment, or Alzheimer’s disease
dementia. For screening, logistic regression was used to predict amyloid PET status
in CU individuals from plasma Aβ42/Aβ40, APOE status, and age. For enrichment,
Linear mixed effects models were used to predict longitudinal cognitive decline and
future risk of AD dementia in CU individuals or in MCI individuals from a basic
model (age, sex, education, APOE status) and varying combinations of blood-based
biomarkers (plasma Aβ42/Aβ40, plasma pTau181, plasma pTau217, plasma NfL).
For treatment response, plasma NfL was measured longitudinally in MCI or AD
patients and properties such as slope, inter-subject variability, and intra-subject
variability were calculated. Plasma NfL was then compared with MRI and
cognition.
Results: The amyloid PET screening model had an AUC of 0.87, with a significant
independent effect for plasma Aβ42/Aβ40 and APOE status, but not age. This model
was estimated to reduce total cost of recruiting 500 amyloid-positive CU
participants by 31 – 42%, depending on the relative cost of amyloid scanning to
plasma measurement. For enrichment, plasma pTau181 and pTau217 had the largest
effect on predicting cognitive decline in CU and MCI participants, with Aβ42/Aβ40
and NfL having significant effects in some scenarios. Using these biomarkers in a
clinical trial could reduce the required sample size of a clinical trial in CU
participants by up to 70%. Finally, plasma NfL was shown to have worse theoretical
performance as a trial progression marker compared to MRI-based measures,
primarily due to its high within-subject variability. NfL compared better to cognitive
measures as endpoints.
Discussion: The future of AD clinical trials will likely leverage plasma biomarkers
for initial screening. Their utility for enrichment and tracking treatment response
still needs to be evaluated in the context of other biomarkers measured in CSF, MRI,
or PET. The plasma ATN biomarkers evaluated here all appear to be independently
useful, but there is strong potential for more plasma biomarkers to be added to such
a panel.