Advertisement

18F-FLORBETABEN SUVR CUTOFF QUANTIFICATION ACROSS REFERENCE REGIONS AND STANDARDS OF TRUTH AND COMPARISON TO VISUAL ASSESSMENT

      Background

      The aims of this study were 1) to generate 18F-Florbetaben (FBB) SUVR cutoff values for different reference regions (RRs) using visual assessment (VA) as standard of truth (SoT); 2) to validate them across different clinical populations and SoTs and 3) to compare VA and SUVR cutoffs.

      Methods

      For aim 1, SUVR cutoff was generated using FBB scans from a sample of 143 subjects (69.5±7.5 yrs; n=75 Alzheimer’s disease (AD), n=68 healthy volunteers (HVs)) who were assessed by 3 blinded independent readers SUVR was calculated for different RRs: cerebellar gray matter (GCER), whole cerebellum (WCER), pons (PONS), and subcortical white matter (SWM). A SUVR cutoff value for each RR was generated using Receiver Operating Characteristic analysis, VA as the SoT, and highest accuracy as the selection criteria. For aim 2, SUVR cutoffs were validated in two different samples: Sample a) n=78 end-of life subjects (80.1±10.4 yrs) using VA from 5 readers and histopathological confirmation of the presence of neuritic beta-amyloid plaques as SoT and Sample b) n=233 subjects (72.6±7.3 yrs; n=118 HVs, n=115 AD) using VA from 8 readers and clinical diagnosis as SoT. For aim 3, FBB scans were visually assessed by 8 blinded readers. The number of correctly or incorrectly classified scans according to pathology results using VA and SUVR cutoffs for each RR was compared.

      Results

      Aim 1: Composite SUVR cutoffs were 1.43 (GCER); 0.96 (WCER); 0.78 (PONS); 0.71 (SWM). Aim2: Accuracy values were high and consistent across SUVR cutoffs and VA methods and SoTs (range 83-94% for histopathology, 80-87% for clinical diagnosis and 86-97% for VA). Aim 3: A range of 63-71 cases were correctly classified by both SUVR cutoff (across RRs) and VA, while 1-3 cases (across RRs) were incorrectly classified by both SUVR cutoff and VA. SUVR cutoff method misclassified more cases (range:3-10 cases) than VA (range:1-3 cases) for any RR.

      Conclusions

      18F-FBB SUVR cutoff quantification is very robust. SUVR cutoff performance was high for all RRs, independent of the SoTs, and consistent with VA. The lack of differences across RR supports using the cerebellum as the most biologically appropriate one for 18F-FBB SUVR.