Visual versus quantitative tau-PET Braak staging in Alzheimer's disease using [¹⁸F]MK6240.

Abstract

The 2024 Alzheimer's Association Workgroup research framework designates tau proteinopathy (T₂) as a key element for Alzheimer's disease (AD) staging, but optimal staging approaches have yet to be determined. Here, we compared visual and quantitative tau-PET-based Braak staging as candidate strategies to implement T₂ biological staging in vivo. We included 140 participants from the TRIAD cohort who underwent [¹⁸F]MK6240 tau-PET. Quantitative Braak staging (qBraak) was derived from regional SUVR thresholds, whereas visual Braak staging (vBraak) was independently performed by three nuclear medicine physicians using an adapted interpretation algorithm. Inter-rater and inter-method agreement were assessed using Cohen's and Fleiss' κ statistics. Associations with clinical severity, cortical thickness, plasma pTau217, and cortical tau extent were examined. Diagnostic performance for identifying amyloid-positive cognitively impaired individuals was evaluated. vBraak demonstrated substantial to nearly perfect inter-rater agreement (κ = 0.65-0.93). Agreement between vBraak and qBraak was moderate when stages were treated categorically (κ = 0.51), but substantial when their ordinal nature was considered (weighted κ up to 0.73). Both strategies showed comparable associations with clinical severity and neurodegeneration. vBraak was more sensitive to amyloid-β-positive cognitive impairment and identified intermediate-stage involvement at lower global tau extent. Visual-quantitative discordant cases were primarily attributable to off-target binding or spatially heterogeneous tau patterns. Both vBraak and qBraak staging provide complementary and largely concordant approaches for operationalizing T₂ staging. Quantitative methods enable scalable, group-level analyses, whereas visual assessment remains essential for identifying atypical tau patterns and informing clinically relevant decision-making.