Concordance Between [18F]Flortaucipir PET Visual Reads and CenTauR-Based Quantification.
TL;DR
Generalized additive models examined clinical progression on the basis of visual and CenTauRz-based assessments of tau PET positivity and highlighted the need for new quantification approaches to better capture the complex patterns of tau deposition, important for early AD detection and monitoring.
Abstract
Tau PET imaging with [18F]flortaucipir allows for the visualization and mapping of aggregated tau deposits, a key neuropathologic feature of Alzheimer disease (AD). A visual interpretation method for [18F]flortaucipir was approved by the Food and Drug Administration and European Medicines Agency and has been implemented for a standardized, clinically usable definition of tau PET positivity. The CenTauRz method, on the other hand, offers the possibility to harmonize the definition of tau PET positivity across different radiotracers using quantitative metrics, but its concordance with clinically relevant [18F]flortaucipir visual reads remains unclear. Methods: A convenience sample of 3991 participants, including cognitively unimpaired (CU) and cognitively impaired (CI) individuals (i.e., those with mild cognitive impairment or AD dementia), underwent [18F]flortaucipir PET imaging. Each [18F]flortaucipir scan was assessed by 3 trained readers using the approved visual interpretation method and quantified using the CenTauRz quantification pipeline in different regions of interest (ROIs). Concordance between positive visual reads and CenTauRz-defined positivity (i.e., >2 on the CenTauRz scale) was assessed using Cohen κ. Receiver-operating-characteristic (ROC) analysis evaluated the discriminative power of continuous CenTauRz values in distinguishing between negative and positive visual reads. Generalized additive models examined clinical progression on the basis of visual and CenTauRz-based assessments of tau PET positivity. Results: Concordance between visual reads and CenTauRz-based assessments of tau PET positivity was moderate, particularly in CU individuals (κ = 0.23-0.55 for CU participants, κ = 0.57-0.82 for CI participants, depending on the ROI). ROC analysis revealed that the agreement remained moderate, independent of the CenTauRz cut point used (area under the ROC curve, 0.72-0.87 vs. 0.87-0.97 for CU and CI individuals, respectively). Among discordant cases, participants with visually positive/CenTauRz-negative tau PET scans were more frequently amyloid-positive and exhibited faster clinical progression compared with visually negative/CenTauRz-positive individuals. Conclusion: Our findings highlight the relatively limited agreement between the visual assessment of [18F]flortaucipir PET images and CenTauRz-based quantification, particularly in CU individuals. Participants with visually positive/CenTauRz-negative tau PET scans showed a high frequency of amyloid positivity and faster clinical progression, suggesting that visual reads are more sensitive to heterogeneous, clinically relevant tau accumulation patterns not captured by ROI-based methods. These findings underscore the need for new quantification approaches to better capture the complex patterns of tau deposition, important for early AD detection and monitoring.