Concordance between visual interpretation of [¹⁸F]flortaucipir pet images and the Universal centaur scale

Abstract Background Positron emission tomography (PET) imaging with [ 18 F]flortaucipir allows for in‐vivo visualization of aggregated tau in Alzheimer’s disease (AD). The FDA‐approved label for [ 18 F]flortaucipir PET provides a standardized, clinically applicable definition of tau‐PET positivity by visual interpretation. Here, we studied the concordance between this clinically approved definition of tau PET positivity and a recently proposed universal scale—CenTauR—for the standardized quantification of abnormal tau aggregates. Method We included 1849 participants (cognitively unimpaired [CU] and impaired [CI, MCI or AD dementia]) from four cohorts (A4 study, ADNI, OASIS‐3, and HABS) with available [ 18 F]flortaucipir PET (mean age: 72.2 y, 55% females). Three trained readers scored each [ 18 F]flortaucipir PET scan as positive/negative using an FDA‐approved visual interpretation method. Visual reads were compared to [ 18 F]flortaucipir PET quantification using the Universal CenTauR z scale. Specifically, we analysed concordance by 1) performing a receiver operating characteristic (ROC) curve analysis of continuous CenTauR z (CTR z ) values discriminating between positive/negative visual reads; and 2) computing sensitivity and specificity when using a previously proposed cut‐point of 2 CTR z . Result Concordance between visual interpretation and continuous CTR z measures was limited (AUC = 0.82 for CU and AUC = 0.87 for CI, Figure 1). When using a previously proposed cut‐point of 2 CTR z , this quantitative approach yielded a sensitivity of 49% [95% CI, 42% to 57%] and a specificity of 95% [93% to 96%] in CU individuals and a sensitivity of 73% [95% CI, 66% to 79%] and a specificity of 94% [95% CI, 90% to 96%] in CI individuals. Visual‐positive, CenTauR z ‐negative individuals were more frequently Aβ‐positive than visual‐negative, CenTauR z ‐positive individuals (87% vs 48%, p< 0.001). Conclusion Visual interpretation of [ 18 F]flortaucipir PET images and quantification using the universal CenTauR z scale exhibited limited agreement, particularly among CU individuals. Tau‐PET positivity based on visual interpretation aligns better with the presence of Aβ‐pathology. Visual interpretation seems to be more sensitive to early tau aggregation compared to universal CenTauR z quantification. Future analyses will compare clinical outcomes associated to each definition of tau‐PET positivity.