Clinical and Biologic Characterization of a 3-Level Centiloid-Based Classification of Amyloid PET

Amyloid PET is commonly interpreted using binary visual classification. However, quantitative assessment on the Centiloid scale identifies an intermediate range of amyloid burden not always captured by dichotomous interpretation. The biologic and clinical relevance of this intermediate category remain unclear, despite its implications for biomarker interpretation, prognostic stratification, and eligibility for amyloid-targeting therapies. Our aim was to characterize individuals within this intermediate range using multimodal biomarker profiles and longitudinal cognitive outcomes, as well as deriving Centiloid thresholds for early tau PET positivity, advanced tau PET positivity, and cognitive decline. Methods: We retrospectively analyzed participants who underwent amyloid PET between 2016 and 2024 at the Geneva Memory Center, including cognitively unimpaired individuals and patients with mild cognitive impairment or dementia. Global amyloid burden was quantified in Centiloids, and participants were categorized into low (<12), intermediate (12-37), or high (>37) groups. Group comparisons, receiver-operating-characteristic curve analyses, and linear mixed-effects models were used to assess fluid biomarker differences, define Centiloid thresholds for tau positivity and cognitive decline, and evaluate longitudinal change in Mini-Mental State Examination scores. Results: Among the 512 participants, 202 (39%) had low, 63 (12%) had intermediate, and 247 (48%) had high Centiloid values. The intermediate-Centiloid group showed biomarker profiles, including cerebrospinal fluid and plasma markers of amyloid, tau pathology, and neurodegeneration, between those of the low and high groups. Both the intermediate- and high-Centiloid groups exhibited faster cognitive decline than did individuals in the low-Centiloid group, with a moderate rate in the intermediate-Centiloid group. Thresholds of 13 and 14 Centiloids best identified individuals with cognitive decline and tau accumulation in the mesial temporal lobe, respectively, whereas 51 Centiloids best discriminated advanced neocortical tau involvement. Conclusion: Centiloid-based classification delineates biologically and clinically distinct stages along the amyloid continuum. Individuals with intermediate Centiloid values already show tau involvement and increased risk of cognitive decline, whereas higher amyloid levels are associated with advanced tau pathology. These findings support a 3-level interpretation of amyloid PET beyond binary classification and highlight the clinical relevance of the intermediate zone.