Towards a Universal Readout for Fluorine-18 Labelled Amyloid Tracers: The CAPTAINs Study.

Abstract

To date, three fluorine-18 labelled tracers have been approved for assessing cerebral amyloid plaques pathology to assist in the diagnosis of Alzheimer's Disease (AD). Although scanning protocols are relatively similar across tracers, FDA/EMA approved visual rating guidelines to render scans as positive or negative differ between the three tracers. The purpose of the present study was to assess the comparability of visual rating results when applying the three different FDA/EMA approved visual interpretation protocols to all three amyloid tracers both for experts and non-experts. Methods: In an international multicentre approach, both experts (N = 4) and non-experts (N = 3) rated scans acquired with fluorine-18 labelled florbetaben, florbetapir and flutemetamol. Scans obtained with each tracer were presented for reading according to all three approved visual rating protocols. In a randomized order, every single scan was rated by each reader according to all three protocols, resulting in a total of more than 700 image ratings. Raters were blinded for the amyloid tracer used and asked to rate each scan as positive or negative, giving a confidence judgement after each response. Percentage of visual reader agreement, inter-rater reliability and agreement of each visual read with binary quantitative measures of standard uptake value ratio based on a fixed threshold for positive and negative scans were computed. These metrics were analyzed separately for expert and non-expert groups. Results: No significant differences in visual ratings across the different metrics of agreement were observed in the group of experts. Nominal differences suggested that the Florbetaben visual rating protocol achieved the highest interrater reliability and accuracy especially under low confidence conditions. For the group of non-experts significant differences in visual ratings were observed with overall moderate-to-fair accuracy and with the highest reliability for the Florbetapir visual rating procedure. Conclusion: We observed high interrater agreement despite applying different visual rating protocols for all fluorine-18 labelled amyloid tracers, suggesting sufficient potiential for a standardization for visual assessment of cerebral amyloid plaques. Non-experts, however, may need extensive training on reading fluorine-18 labelled amyloid scans and may particularly benefit from a universal readout to ensure comparability across visual evaluation strategies.