Llorente Saguer, Isaac; Busche, Marc A; Oxtoby, Neil P; (2022) Composite SUVR: a new method for boosting Alzheimer's disease monitoring and diagnostic performance, applied to tau PET. Presented at: Alzheimer's Association International Conference (AAIC) 2022, San Diego, CA, USA.

Preview

Text AAIC21_ISTAART_Student_Poster_ills.pdf - Published Version Download (585kB) | Preview

Abstract

Background: Abnormal brain tau protein accumulation is strongly linked to multiple neurodegenerative disorders. Currently, brain tau pathology is quantified in vivo using tau PET by calculating the Standardized Uptake Value Ratio (SUVR) of target and reference regions of interest (ROIs). Recent work (Schwarz et al., 2021) in Alzheimer’s Disease (AD) explored various target and reference ROIs to report performance of SUVR as a biomarker for diagnosis, disease monitoring, and clinical trial efficacy/eligibility (sample size estimate, SSE). Here we introduce a new method and biomarker: Composite SUVR (CUVR). / Methods: We analyzed longitudinal SUV data from ADNI in the available 103 participants having three or more tau PET scans ([18F]AV-1451): 58 cognitively normal (CN); 21 mild cognitive impairment; 24 probable AD. In the spirit of SUVR and statistical ROIs (Chen, et al., NeuroImage 2010), we calculate CUVR as the SUV ratio of two composite regions. Our novel method is that the composite regions are determined by a genetic algorithm that searches the possible 3^96 combinations of regions from FreeSurfer’s default atlas. We compare performance of SUVR with CUVR. Performance metrics follow Schwarz et al.: a linear mixed-effects model quantifies longitudinal group separation by tau accumulation rate (t statistic between fixed effects for CN and AD) and longitudinal precision (model residuals’ standard deviation). CUVR and SUVR values were log-transformed before model fitting. We calculated SSE for a hypothetical clinical trial designed for 80% power to reduce tau PET accumulation by 20% (vs. placebo) in non-CN individuals. / Results: Our method identified a CUVR biomarker involving 60 regions. Figure-1 shows the vast performance improvement of CUVR versus the best-performing SUVR (inferior-temporal target; eroded subcortical white matter reference). Group separation improved by 2.9x (t = 9.57 vs 3.32); longitudinal precision by 6.5x (residual std = 0.331% vs 2.14%); and CUVR required a smaller sample size by 3.9x (83 vs 318). / Conclusions: Our simple data-driven approach discovered a new tau PET biomarker called CUVR. Experimental results show state-of-the-art longitudinal group separation, longitudinal precision, and clinical trial enrichment. The remarkable performance improvements provide compelling evidence for using CUVR for both eligibility and efficacy in Alzheimer’s disease clinical trials, particularly of anti-tau therapies.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item