HAI Book 2025 - Flipbook - Page 492
Otero Svaldi, Diana
Regional tau quantification methodologies in early symptomatic
participants with presence of tau pathology
Diana Otero Svaldi1, Leonardo Iaccarino1, Ixavier Alonzo Higgins1, Nicolai Franzmeier2, Michael
Ewers2, Sergey Shcherbinin1
1
Eli Lilly and Company, Indianapolis, IN, US
Institute for Stroke and Dementia Research, University Hospital, Ludwig Maximilian University of Munich, Munich, DE
2
Background: This study compares approaches for defining target regions in their ability to consistently capture
tau progression relative to anatomical composites used in Alzheimer9s disease trials.
Methods: Paired, baseline and 18-month regional flortaucipir data were included from TRAILBLAZER-ALZ2
(NCT04437511, Low/Medium/High Tau) and TRAILBLAZER-ALZ (NCT03367403, Low/Medium Tau) placebo
participants. Regional SUVRs were combined into four composite regions according to individual- versus grouplevel and anatomical- versus data-driven approaches. Composites were defined and ranked according to
hypothetical longitudinal spatial gradients based on descending mean regional SUVR (Q1→Q4). Anatomical
composites were defined by lobes, whereas data-driven composites (quartiles) were defined by either baseline
SUVR or functional connectivity (FC) to the highest baseline SUVR quartile. Group-level methodologies used mean
regional SUVRs across participants, whereas individual-level used single participant regional SUVRs. Finally, the
minimum positive composite (lowest SUVR>2 standard deviations (SD) above amyloid-negative cohort mean) was
considered a target region. Longitudinal signal-to-noise (SNR) for each composite (mean/SD of change) and mean
SNR across composites were compared to group lobar composites (reference method, Figure1A). Homogeneity
between individual-level spatial gradients and across trial-level spatial gradients were evaluated for methods
showing similar/superior SNR to reference.
Results: Group-level, data driven approaches exhibited lower mean SNR relative to the reference but provided
increases in Q1 SNR (Figure1). Individualized methods exhibited similar/higher mean SNR, increased Q3-Q4 SNR
(Figure1), and increased spatial homogeneity between participants (Figure2) and across trials (Figure3) versus the
reference. Minimum positive composites exhibited superior SNR across individualized and anatomical
methodologies (Figure1B). Spatial gradients were similar between SUVR-based and FC-based methodologies
(Figures1-3).
Conclusions: Participant specific target regions may improve consistency in longitudinal tau PET analyses. For
each participant, focusing on regions that have recently reached positivity may improve SNR. The similarity
between SUVR-based and FC-based methods continues to suggest connectivity between regions is a significant
mechanism contributing to tau spread.
HAI2025 - 492
