What's the cut‐point? A systematic review of tau pet thresholding methods: Development of new models and analysis methods/tau. (7th December 2020)

• Record Type: Journal Article
• Authors: Weigand, Alexandra J.; Eglit, Graham M.L.; Maass, Anne; Bondi, Mark W.
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 3
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.046270
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: With increasing use of Alzheimer's disease (AD) biomarker staging schemes, derivation of thresholds to determine biomarker positivity is becoming a critical methodological step. Recent advances in tau PET have made this modality a primary method for in vivo assessment of tau pathology. Given the heterogeneity in methodological decisions around tau PET positivity thresholding across studies, we conducted a systematic review of articles deriving tau PET thresholds to chronicle and compare these methods. Method: Studies were excluded if they used animal models, included non‐AD and related dementias patients (e.g., Parkinson's disease), used cerebrospinal fluid data, used a tracer other than flortaucipir, or included tau PET as a continuous variable only. Search terms included "tau PET positivity, " "tau PET threshold, " "tau PET cutoff, " and "tau PET cutpoint, " with these searches then repeated after replacing "tau PET" with "flortaucipir." Of the 405 studies reviewed, a total of 19 studies were identified that matched inclusion criteria for the review. Result: Of the 19 studies identified, 7 thresholded tau positivity in a manner conditional on amyloid positivity (e.g., maximally discriminating amyloid‐negative and amyloid‐positive), leading to SUVR cut‐points ranging from 1.19 to 1.33. The other most used method involved thresholding at 2‐2.5 SDs above the SUVR for young (n = 6) or old (n = 1) controls (cut‐points from 1.17 to 1.36). Other approaches included using conditional inference regression discriminating clinical diagnoses or global cognition (n = 4; cut‐points from 1.13 to 1.40) and clustering into high and low tau (n = 1; cut‐point = 1.25). Regions used differed greatly between studies, ranging from individual cortical regions (n = 5) to a temporal meta‐ROI (n = 10) to PET‐recapitulated Braak stages (n = 4). Conclusion: There exists significant variability in tau PET cut‐points derived even within a particular thresholding method, which may be attributable to differences in sample composition, preprocessing methods (e.g., normalization, partial‐volume correction), or regions selected. Notably, although a large proportion of studies thresholded tau conditionally upon amyloid status, this may lead to bias in the estimated proportion of AT(N) groups. Consensus in thresholding methods, and particularly those that threshold tau and amyloid independently of the other, are needed given the increasing use of biomarker cut‐points within AD diagnostic frameworks.