Prognostic value of Amyloid PET in Real-World Amnestic MCI: Influence of reader experience and interpretation method.

Abstract

This study assessed the prognostic value and inter-/intra-observer reliability of three amyloid PET interpretation methods-conventional visual reading, assisted visual interpretation and standardized uptake value ratio (SUVR)-based quantification-for predicting conversion to Alzheimer's disease (AD). This retrospective longitudinal study included 145 patients with amnestic mild cognitive impairment who underwent amyloid PET and were followed for up to 8 years at a single memory clinic. Two certified nuclear medicine physicians performed structured visual reads and assisted visual interpretations. Semiquantitative SUVR thresholds validated in prior studies were applied as a third method. Inter- and intra-observer agreement was assessed using Cohen's kappa (κ). Cox regression models estimated the association of conversion to AD in PET-positive versus PET-negative patients for each method, adjusting for age, sex and education. Inter-observer agreement was excellent for structured visual reading (κ = 0.877) and assisted visual interpretation (κ = 0.904). Intra-observer reliability was also high (κ = 0.891 for Reader 1; κ = 0.859 for Reader 2). Consensus agreement for the assisted method reached κ = 0.936. Amyloid PET positivity was strongly associated with progression to AD across all methods. Hazard ratios were highest for assisted visual interpretation (HR = 3.23, 95% CI 1.94-5.38, p < 0.0001), followed by SUVR quantification (HR = 2.19, 95% CI 1.55-3.10, p < 0.0001) and structured visual reading (HR = 1.83, 95% CI 1.10-3.06, p = 0.02). Amyloid PET positivity consistently predicted conversion to AD regardless of interpretation method. Assisted visual and semiquantitative approaches showed numerically higher prognostic discrimination, while expert structured visual reading remained clinically reliable.