Intra- and inter-site reproducibility of human brain single-voxel proton MRS at 3T.

Abstract

Clinical trials that involve participants from multiple sites necessitate standardized and reliable quantitative MRI outcomes to detect significant group differences over time. Metabolite concentrations measured by proton MRS (1 H-MRS) provide valuable information about in vivo metabolism of the central nervous system, but can vary based on the acquisition and quantitation methods used by different MR sites. Therefore, we investigated the intra- and inter-site reproducibility of metabolite concentrations measured by 1 H-MRS on MRI scanners from a single manufacturer across six sites. Five healthy controls were scanned twice within 24h at six participating 3T MR sites with large single-voxel PRESS (TE/TR/NSA=36ms/4000ms/56) and anatomical images for voxel positioning and correction of partial volume relaxation. Absolute metabolite concentrations were calculated relative to the T1 and T2 relaxation corrected signal from water. Intra- and inter-site reproducibility was assessed using Bland-Altman plots and intra- and inter-site coefficient of variation (CoV) as well as intra- and inter-site intra-class correlation coefficient. The median intra-site CoVs for the five major metabolite concentrations ([NAA], [tCr], [Glu], [tCho] and [Ins]) were between 2.5 and 5.3%. Inter-site CoVs were also low, with the median CoVs for all metabolites between 3.7 and 6.4%. Metabolite concentrations were robust to small inconsistencies in voxel placement and site was not the driving factor in the variance of the measurement of any metabolite concentration. Between-subject differences accounted for the majority of the concentration variability for creatine, choline and myo-inositol (42-65% of the variance). A large single-voxel 1 H-MRS acquisition from a single manufacturer's MRI scanner is highly reproducible and reliable for multi-site clinical trials.