Amyloid Imaging in Therapeutic Trials: the Quest for the Optimal Reference Region

Victor L. Villemagne,Pierrick Bourgeat,Vincent Doré,Lance Macaulay,Robert Williams,David Ames,Ralph N. Martins,Olivier Salvado,Kewei Chen,Eric M. Reiman,Colin L. Masters,Christopher C. Rowe

DOI: https://doi.org/10.1016/j.jalz.2015.06.036

2015-01-01

Abstract:The reference region (RR) approach was initially proposed for the kinetic analysis of neuroimaging radiotracers to remove the need of a metabolite-corrected plasma input function. The RR was described as a brain region with similar cellular and blood flow characteristics as the target region but lacking specific (saturable) binding sites. It relied on two basic assumptions: that the degree of nonspecific binding and the volume of distribution of the free compartment were the same in the RR and target regions. The use of a RR was later applied to semiquantitative tissue ratio approaches, as the ones widely used for Aβ imaging studies. With the advent of therapeutic anti-Aβ trials there has been a renewed interest in optimizing outcomes by reducing the variance of Aβ burden measurements. The purpose of this study was to first assess the stability of different RR and then look at the stable RR that yielded the lowest variance for Aβ burden estimates obtained with three FDA-approved Aβ imaging tracers. 653 participants were evaluated (258 w/flutemetamol-FLUTE-; 184 w/florbetapir-FBP- and 211 w/florbetaben-FBB-) where 237 had longitudinal scans (81 w/FLUTE; 87 w/FBP and 69 w/FBB). We assessed the SUV of 11 either pure grey (GM) or white matter (WM) RR, and their combinations (Table 1) across clinical conditions, across Aβ status, and across time. Stable RR were then used to assess the variance of the Aβ burden estimates. For FLUTE, GM and most WM RR were stable under all conditions, with the composite SWM+pons yielding the lowest Aβ burden variance both cross-sectionally and longitudinally. While a subcortical WM RR (SWMKCER, extending from the centrum semiovale to the corpus callosum as proposed by Kewei Chen and Eric Reiman) was stable under all conditions for FBP, Cerebellar GM (CbGM) was the only stable RR for FBB. The Aβ burden variances obtained with the aforementioned RR were similar for all tracers. SWM+pons for FLUTE, CbGM for FBB, and SWMKCER for FBP remained stable across the examined conditions, yielding the lowest variance of the Aβ burden estimates. To optimize outcomes in ongoing therapeutic trials, tracer-specific RR should be applied.