Abstract
485
Objectives: Three Fluorine 18-labelled tracers are currently commercially available to measure cerebral amyloid pathology to assist in the diagnosis of possible Alzheimer’s Disease (AD); Amyvid, Vizamyl and Neuraceq. Although scanning protocols are relatively similar across the tracers, FDA-/EMA approved visual rating guidelines to render a scan as positive or negative considerably differ. These differences include color scale used, intensity-scaling, definitions of a region as well as spatial and signal thresholds to determine positivity. Standardization of visual reading would however be highly desirable in order to ensure that diagnostic classification, patient selection and therapeutic decisions are taken on a comparable basis, independently of the tracer used. To this end, we examined the comparability of different FDA-/EMA visual interpretation protocols on the three different commercially available amyloid tracers. Methods: In a multicentric approach, a total of 18 PET-Scans were compiled for this study (6 scans for each of the three tracers, matched for diagnostic categories). Each scan was pre-categorized with tracer-specific thresholds into amyloid positive or negative and submitted to an in-house-created online rating tool. On this platform, every single scan was presented for reading according to each of the three visual ratings schemes to a multi-center team of certified expert readers from Australia, Europe and the US. The certified raters were blind to clinical category and type of tracer used and were instructed to give a binary classification if a scan was positive or not. We analyzed intra- and interrater reliability and further assessed which visual rating schemes produces the highest reliability across tracers. Finally, we compared the correspondence of each response with the binary pre-classification based on a quantitative threshold. Results: Intrarater-reliability showed good to very good agreement (kapparange= .74-.80; kappaSE = .05-.05) across all reads. Irrespective of the tracer all three visual rating approaches also reached good interrater-reliability, with Vizamyl and Neuraceq criteria leading to the highest values (kapparange=.70-.80; kappaSE = .08-.07). When comparing the agreement of each expert rater’s response with the quantitatively pre-determined category of amyloid-positive or -negative, agreement rates although very high, varied among expert readers (Neuraceq approach: 80-95 %; Vizamyl approach: 90-93 %; Amyvid approach: 85-90 %). Conclusion: Our preliminary data of this ongoing study suggests that for visual assessment with regard to non-quantitative evaluation, established FDA-/EMA-approved interpretation guidelines may be sufficient to ensure relatively good levels of standardization with regard to classification into amyloid-positive and -negative subjects. However, a certain degree of variance remained. To potentially further improve standardization of visual rating, we are currently developing and evaluating a unified visual rating scheme that may enhance interrater-reliability and simplify interpretation of amyloid-scans also for the less-experienced user. We further investigate how interrater-reliability may vary as a function of diagnostic group. Together, the CAPATINs Study aims to further improve standardization of visual rating of amyloid-PET studies.