Premium
O3‐12‐02: When an amyloid PET threshold of 1.5 becomes 1.4 and longitudinal accumulation is not what it appears: Interpreting and reconciling values amidst scanner variability
Author(s) -
Matthews Dawn,
Andrews Randolph,
Mosconi Lisa,
Schmidt Mark
Publication year - 2012
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2012.05.1207
Subject(s) - precuneus , white matter , nuclear medicine , scanner , posterior cingulate , psychology , grey matter , neuroscience , cortex (anatomy) , medicine , magnetic resonance imaging , radiology , artificial intelligence , computer science , cognition
Background: Measurement of amyloid burden using 11C-PiB has been incorporated as an endpoint in Alzheimer’s Disease clinical research and therapeutic trials. Studies of Normal, MCI, and AD populations have been used to establish thresholds for amyloid positivity, impacting subject inclusion in trials and contributing to diagnosis. Yet, these values are highly dependent upon several factors including the scanner. Methods: One hundred eightyseven PiB scans from 94 ADNI subjects were evaluated from five scanner models: HR+ (76 scans), HRRT (42), GE Advance (32), GE Discovery (15), and Biograph HiRez (22). Using 50-70 minute summed images, values were sampled at 27 individual slices of gray matter cerebellum, 27 slices of subcortical white matter, 5 regions of interest (anterior cingulate, posterior cingulate/precuneus, frontal cortex, lateral temporal cortex, parietal cortex), and 8 additional reference regions including combinations of gray and white matter cerebellum and pons. Standardized Uptake Value Ratio (SUVR) values were compared across scanner models using PIBand PIB+ scans together and separated into subgroups, and across sites within scanner model. A subgroup of subjects each having scans from both a GEAdvance and HR+ scanner was evaluated. Differences between scanners were mathematically modeled and tested to predict cross-scanner equivalent values. Results: Significant SUVR differences were found across scanner models. Using PiBand PiBscans combined, averagewhite matter referenced to graymatter cerebellum SUVRs ranked: HR+ >GE Advance 1⁄4GE Discovery >BiographHiRez >HRRT(site-specific). Differences between HR+ and all other scanner models were significant (p<0.03 vs. Discovery to p<0.00001 vs. HRRT-site 1) and 20% in some cases. Significant differences were found between HRRT sites and between HR+ sites. Differences between 3 scanner types were reduced when pons was used as reference. Modeling showed potential to compensate for inter-scanner differences. Conclusions: Amyloid values measured using 11C-PiB must be evaluated in the context of the scanner used to collect data, particularly when the SUVR value is close to a positivity threshold. In longitudinal studies, within-subject data must be reconciled to account for variability in SUVR values that arise from change in the scanner, changes in acquisition on the same camera, and impact of different corrections applied to the emission data.