Background: Results of previous studies have shown associations between PET imaging of amyloid plaques and amyloid-β pathology measured at autopsy. However, these studies were small and not designed to prospectively measure sensitivity or specificity of amyloid PET imaging against a reference standard. We therefore prospectively compared the sensitivity and specificity of amyloid PET imaging with neuropathology at autopsy.
Methods: This study was an extension of our previous imaging-to-autopsy study of participants recruited at 22 centres in the USA who had a life expectancy of less than 6 months at enrolment. Participants had autopsy within 2 years of PET imaging with florbetapir ((18)F). For one of the primary analyses, the interpretation of the florbetapir scans (majority interpretation of five nuclear medicine physicians, who classified each scan as amyloid positive or amyloid negative) was compared with amyloid pathology (assessed according to the Consortium to Establish a Registry for Alzheimer's Disease standards, and classed as amyloid positive for moderate or frequent plaques or amyloid negative for no or sparse plaques); correlation of the image analysis results with amyloid burden was tested as a coprimary endpoint. Correlation, sensitivity, and specificity analyses were also done in the subset of participants who had autopsy within 1 year of imaging as secondary endpoints. The study is registered with ClinicalTrials.gov, number NCT 01447719 (original study NCT 00857415).
Findings: We included 59 participants (aged 47-103 years; cognitive status ranging from normal to advanced dementia). The sensitivity and specificity of florbetapir PET imaging for detection of moderate to frequent plaques were 92% (36 of 39; 95% CI 78-98) and 100% (20 of 20; 80-100%), respectively, in people who had autopsy within 2 years of PET imaging, and 96% (27 of 28; 80-100%) and 100% (18 of 18; 78-100%), respectively, for those who had autopsy within 1 year. Amyloid assessed semiquantitatively with florbetapir PET was correlated with the post-mortem amyloid burden in the participants who had an autopsy within 2 years (Spearman ρ=0·76; p<0·0001) and within 12 months between imaging and autopsy (0·79; p<0·0001).
Interpretation: The results of this study validate the binary visual reading method approved in the USA for clinical use with florbetapir and suggest that florbetapir could be used to distinguish individuals with no or sparse amyloid plaques from those with moderate to frequent plaques. Additional research is needed to understand the prognostic implications of moderate to frequent plaque density.
Funding: Avid Radiopharmaceuticals.
Copyright © 2012 Elsevier Ltd. All rights reserved.