Determining Amyloid-β Positivity Using¹⁸F-AZD4694 PET Imaging

Abstract

Amyloid-β deposition into plaques is a pathologic hallmark of Alzheimer disease appearing years before the onset of symptoms. Although cerebral amyloid-β deposition occurs on a continuum, dichotomization into positive and negative groups has advantages for diagnosis, clinical management, and population enrichment for clinical trials. ¹⁸F-AZD4694 (also known as ¹⁸F-NAV4694) is an amyloid-β imaging ligand with high affinity for amyloid-β plaques. Despite being used in multiple academic centers, no studies have assessed a quantitative cutoff for amyloid-β positivity using ¹⁸F-AZD4694 PET. <b>Methods:</b> We assessed 176 individuals [young adults (<i>n</i> = 22), cognitively unimpaired elderly (<i>n</i> = 89), and cognitively impaired (<i>n</i> = 65)] who underwent amyloid-β PET with ¹⁸F-AZD4694, lumbar puncture, structural MRI, and genotyping for <i>APOEε4</i>¹⁸F-AZD4694 values were normalized using the cerebellar gray matter as a reference region. We compared 5 methods for deriving a quantitative threshold for ¹⁸F-AZD4694 PET positivity: comparison with young-control SUV ratios (SUVRs), receiver-operating-characteristic (ROC) curves based on clinical classification of cognitively unimpaired elderly versus Alzheimer disease dementia, ROC curves based on visual Aβ-positive/Aβ-negative classification, gaussian mixture modeling, and comparison with cerebrospinal fluid measures of amyloid-β, specifically the Aβ₄₂/Aβ₄₀ ratio. <b>Results:</b> We observed good convergence among the 4 methods: ROC curves based on visual classification (optimal cut point, 1.55 SUVR), ROC curves based on clinical classification (optimal cut point, 1.56 SUVR) gaussian mixture modeling (optimal cut point, 1.55 SUVR), and comparison with cerebrospinal fluid measures of amyloid-β (optimal cut point, 1.51 SUVR). Means and 2 SDs from young controls resulted in a lower threshold (1.33 SUVR) that did not agree with the other methods and labeled most elderly individuals as Aβ-positive. <b>Conclusion:</b> Good convergence was obtained among several methods for determining an optimal cutoff for ¹⁸F-AZD4694 PET positivity. Despite conceptual and analytic idiosyncrasies linked with dichotomization of continuous variables, an ¹⁸F-AZD4694 threshold of 1.55 SUVR had reliable discriminative accuracy. Although clinical use of amyloid PET is currently by visual inspection of scans, quantitative thresholds may be helpful to arbitrate disagreement among raters or in borderline cases.