Comparison of SUVR Methods and Reference Regions in Amyloid PET

Objectives We compare results of standard uptake value ratio (SUVR) analyses of Alzheimer’s Disease Neuroimaging Initiative (ADNI) florbetapir PET scans using a native space compared to SPM template methods and a variety of possible SUVR reference regions. The objective is to find a method with highest longitudinal effect size to allow a sufficiently powered clinical trial efficacy measure using the smallest number of subjects.

Methods Freesurfer (FS) was used to obtain an ROI parcellation on T1 MRI data from 476 ADNI subjects (179 Normal, 160 EMCI, 93 LMCI, 44 AD). Florbetapir data from two time points approximately 24 months apart were registered to the MRI data in T1 native space, and SUVR’s were computed for 110 subregions. A composite SUVR was computed using a grouping of 4 larger cortical regions, equally weighted as described by Landau¹. Eight reference regions including whole cerebellum, cerebellar grey matter, brainstem and subcortical white matter were evaluated. SUVR’s were also computed in template space using a PET-only (PO) method and SPM on regions described by Clark² as well as on grey matter-masked AAL regions^3,4. Effect sizes between AD and Normal groups were evaluated using Cohen’s d. Effect size of longitudinal change in SUVR in the AD group was also compared.

Results In both the cross-sectional and longitudinal analyses, and for both native space and templates space SUVR methods, effect size was greatest when using a reference region that included white matter. Methodological differences in effect size for the cross-sectional analysis were modest, however in the longitudinal analysis, large differences were seen. Consistently, the native space Freesurfer method outperformed the template space methods over all tested reference regions.

Conclusions Use of reference regions including white matter appears essential for increased effect size and therefore possibly reduced sample size in a clinical trial using longitudinal amyloid SUVR as an efficacy measure. Additionally a native space approach using Freesurfer-defined cortical and reference regions appears to be superior to template space methods.