Cognitive processing speed in older adults: relationship with white matter integrity

Abstract

Cognitive processing slows with age. We sought to determine the importance of white matter integrity, assessed by diffusion tensor imaging (DTI), at influencing cognitive processing speed among normal older adults, assessed using a novel battery of computerized, non-verbal, choice reaction time tasks. We studied 131 cognitively normal adults aged 55-87 using a cross-sectional design. Each participant underwent our test battery, as well as MRI with DTI. We carried out cross-subject comparisons using tract-based spatial statistics. As expected, reaction time slowed significantly with age. In diffuse areas of frontal and parietal white matter, especially the anterior corpus callosum, fractional anisotropy values correlated negatively with reaction time. The genu and body of the corpus callosum, superior longitudinal fasciculus, and inferior fronto-occipital fasciculus were among the areas most involved. This relationship was not explained by gray or white matter atrophy or by white matter lesion volume. In a statistical mediation analysis, loss of white matter integrity mediated the relationship between age and cognitive processing speed.

Figure 1. Computerized processing speed tasks.

Screenshots are illustrated for the seven visuospatial choice reaction time tests described in detail in the text (Subjects and Methods).

Figure 2. Processing speed correlates with age.

A composite response latency score, calculated as a z-score relative to young normal controls, is plotted against age for the 131 subjects in this study. The line represents the linear regression, bounded by a 95% confidence interval.

Figure 3. Processing speed correlates with white matter integrity.

Voxel-wise regressions compared the composite scaled reaction time with various parameters of white matter integrity: fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (DR). In red are voxels that correlated with scaled reaction time (p<0.01 after family-wise error correction); correlations were negative for FA and positive for MD and DR. These significant areas are thickened for ease of illustration. The TBSS white matter skeleton used for voxel-wise comparisons is illustrated in blue on axial images. Regression models included age, gender, education, and TIV as nuisance variables. Axial diffusivity was also tested, but was not illustrated because there was no area of significance (p>0.05). Axial slices are illustrated in anatomical (left-is-left) orientation.

Figure 4. Individual processing speed tasks correlate with white matter integrity.

Voxel-wise regressions compared scaled reaction time from the indicated tasks (see Subjects and Methods) with FA. In red are voxels that correlated with scaled reaction time (p<0.05 after family-wise error correction); correlations were negative. These significant areas are thickened for ease of illustration. The TBSS white matter skeleton used for voxel-wise comparisons is illustrated in blue. Regression models included age, gender, education, and TIV as nuisance variables. Results from the other four tasks are not illustrated because there was no area of significance (p>0.05). Axial slices are illustrated in anatomical (left-is-left) orientation.

Figure 5. Relationship of processing speed with the integrity of individual white matter regions of interest.

In yellow are voxels from the TBSS white matter skeleton contained within each indicated white matter ROI, according to a probabilistic atlas (see Subjects and Methods). Superimposed in red are the voxels within each ROI in which FA correlated inversely with scaled reaction time (p<0.01 after family-wise error correction). Forceps major and minor overlap with and contain fibers from the splenium and genu of the corpus callosum.

Figure 6. Processing speed is independent of white or gray matter atrophy.

Axial image slices are illustrated in anatomical (left-is-left) orientation. (A) In red are areas of white matter (WM) or gray matter (GM) that exhibited a significant degree of volume loss in relation to age or scaled reaction time by voxel-based morphometric analysis (p<0.05 after family-wise error correction). White matter maps are superimposed on the white matter template derived from the subjects in this study, and the gray matter maps are superimposed on a standard brain template. (B) Red voxels indicate where scaled reaction time correlated inversely with FA (p<0.01 after family-wise error correction), after taking into account any contribution from white matter atrophy by entering white matter partial volume maps into the regression as a voxel-wise nuisance variable. As in Figure 3, other standard nuisance variables (age, gender, education, and TIV) were also included in the model. (C) Axial FLAIR images are illustrated for the individual in this study with the greatest extent of white matter signal change. Not all subjects in this study had a FLAIR sequence, and the T1 scan was used for determination of white matter lesion volume, as detailed in Subjects and Methods.