Given the role of GABAergic inhibition in cortical information processing, and the recent cross-sectional observation of GABA levels decreasing with age, as measured using GABA-edited MRS [1], it is interesting to ask to what extent age-related functional decline is caused by loss of GABAergic control. One major difficulty in such studies is the appropriate handling of age-related atrophy which results in decreasing voxel tissue/CSF content. Additionally, the concentration of GABA differs between gray (GM) and white matter (WM) [2]. While Gao et al. [1] quantified GABA relative to Cr, making measures insensitive to the increasing CSF voxel content, this does not account for differences in GABA concentration between WM and GM. Recently Harris et al. [2] proposed a tissue correction that accounts for differences in GABA concentration between WM and GM, captured by the factor α (referred to as α-correction). In current study, we compare the impact of the typical CSF-correction and the α-correction in water-referenced GABA-edited MRS in a healthy, elderly cohort. Given the expectation that tissue fraction decreases with age, this cohort enables the comparison of tissue correction strategies for GABA-edited MRS with changing voxel-tissue content.

GABA-edited MRS data were collected in 94 participants (40 male) over an age range of 44 to 92, average age 73.1 ± 9.9 years. MEGA-PRESS data were collected on a 3T Philips Achieva scanner, using a 32-channel head coil and the following acquisition parameters: TR/TE = 2s/68ms, 14 ms-editing pulses at 1.9 ppm (‘On’) and 7.46 ppm (‘Off’), 320 averages, 2048 data points sampled at 2 kHz, VAPOR water suppression and 8 unsuppressed water averages for quantification. Data were acquired from an anterior voxel and a posterior voxel as shown in Figure 1. Data were processed using Gannet2.0 [3], with integrated voxel-to-image coregistration and segmentation using SPM [4]. Three different tissue correction strategies were compared: no tissue correction, CSF-correction (division by the voxel tissue fraction), and α-correction including the normalization to the group-average voxel-tissue fractions [2]. The α-correction (equation below) used an α = 0.5, thus assumes there is twice as much GABA in GM compared to WM. $$ \alpha-corrected GABA=\frac{I_{G}MM}{I_{w}\kappa}\frac{({\sum_i^{GM,WM,CSF}c_{w,i}}e^{-TE/T_{2w,i}}(1-e^{TR/T_{1w,i}})f_i)}{e^{-TE/T_{2G}}(1-e^{-TR/T_{1G}})}(\frac{\mu_{GM}+\alpha\mu_{WM}}{(f_{GM}+\alpha f_{WM})(\mu_{GM}+\alpha\mu_{WM})}) $$

where i designates the tissue compartments GM, WM and CSF, I_G and I_w are the signal integrals for GABA and water, respectively, c_w,i is the water visibility, MM is the macromolecular fraction and κ is the editing efficiency of GABA, f_i is the voxel fraction, and μGM and μWM is the group average voxel fraction for GM and WM.

Two anterior voxels were omitted due to poor segmentation. For the posterior voxel, all tissue correction approaches (no tissue correction, CSF-correction and α-correction) show a significant relationship between GABA and age (Figure 2A). In the anterior voxel the GABA values without tissue correction and the CSF-correction also show a significant relationship between age and GABA levels; however, after applying the α-correction, GABA and age are no longer correlated (Figure 2B). As shown in Figure 3, the anterior voxel shows a higher rate of age-related WM fraction decline and CSF fraction increase compared to the posterior voxel while GM fraction shows similar changes in both voxels.There is a need for consensus over whether to adjust MRS-measured concentrations for voxel content, and if so, how – here we demonstrate that the choice of tissue correction will impact conclusions. In this cohort, both regions show overall losses in tissue with age; however, the anterior voxel shows a steeper decline due to a faster rate of white-matter loss. In this voxel, the α-correction removes the relationship between GABA and age that is otherwise seen with no tissue correction or CSF-correction. This may indicate that WM-tissue fraction changes are driving the relationship between GABA and age in this voxel. In the posterior voxel, the same relationship between age and voxel WM is not apparent. Interestingly, the age-GABA relationship remains for the 3 different tissue correction strategies in the posterior voxel. These results provide compelling evidence for the need to consider tissue concentration differences in GABA concentration.