Longitudinal magnetic resonance imaging studies of older adults: a shrinking brain

Abstract

Age-related loss of brain tissue has been inferred from cross-sectional neuroimaging studies, but direct measurements of gray and white matter changes from longitudinal studies are lacking. We quantified longitudinal magnetic resonance imaging (MRI) scans of 92 nondemented older adults (age 59-85 years at baseline) in the Baltimore Longitudinal Study of Aging to determine the rates and regional distribution of gray and white matter tissue loss in older adults. Using images from baseline, 2 year, and 4 year follow-up, we found significant age changes in gray (p < 0.001) and white (p < 0.001) volumes even in a subgroup of 24 very healthy elderly. Annual rates of tissue loss were 5.4 +/- 0.3, 2.4 +/- 0.4, and 3.1 +/- 0.4 cm3 per year for total brain, gray, and white volumes, respectively, and ventricles increased by 1.4 +/- 0.1 cm3 per year (3.7, 1.3, 2.4, and 1.2 cm3, respectively, in very healthy). Frontal and parietal, compared with temporal and occipital, lobar regions showed greater decline. Gray matter loss was most pronounced for orbital and inferior frontal, cingulate, insular, inferior parietal, and to a lesser extent mesial temporal regions, whereas white matter changes were widespread. In this first study of gray and white matter volume changes, we demonstrate significant longitudinal tissue loss for both gray and white matter even in very healthy older adults. These data provide essential information on the rate and regional pattern of age-associated changes against which pathology can be evaluated and suggest slower rates of brain atrophy in individuals who remain medically and cognitively healthy.

Fig. 1.

Scatterplots showing stability of brain (top) and ventricular (bottom) volumes over 4 years. Against a background of highly stable measurement, brain volumes decrease and ventricular volumes increase over the 4 year interval.

Fig. 2.

Magnitudes of longitudinal change as a function of brain (gray + white) region. To illustrate the significant region-by-time interaction, mean longitudinal change for each lobar brain region is plotted as a z-score, using the mean and SD from lobar brain volumes at baseline to control for differences in absolute volumes. Frontal and parietal brain volumes show greater relative tissue loss compared with temporal and occipital regions.

Fig. 3.

Annual rates of change in brain, gray, white, and ventricular volumes (cubic centimeters) for the entire sample, the subgroup with some medical problems, and the subgroup of very healthy individuals. Mean values are displayed at the axis ends. All values are significant at p < 0.001 except gray (NS) and white (p < 0.05) in the healthy group. Note that the nonsignificant trend for gray matter tissue loss in the healthy subgroup is significant in the more sensitive mixed-effects regression analysis (see Results).

Fig. 4.

Local changes in gray matter volume. Longitudinal declines in tissue volumes over the 4 year interval are shown by the color-coded t score values, calculated from a voxel-based comparison between baseline and year 5 gray matter images. Images are thresholded at p < 0.001 orz = 3.18. To facilitate anatomic localization, significant voxels are superimposed on transaxial slices of a map of the average gray matter distribution of the baseline images after segmentation and stereotaxic normalization. Brighter regions of the gray matter image are more likely to contain gray matter.

Fig. 5.

Local changes in white matter volumes. Longitudinal declines in white matter tissue volume from baseline to year 5 are shown as t values for each voxel. See Figure4 for a more detailed description.

Fig. 6.

Three-dimensional views of significant longitudinal tissue loss in specific gray matter regions. Three-dimensional views of the t statistic are shown projected on the outer cortical surface of a representative subject in our sample. The projection on the surface was performed by averaging the value of the t statistic along the normal of each surface vertex; only voxels with t statistic >3.18 were included in this averaging procedure. The color bar shows the colors corresponding to these calculated average t statistics.Bottom, Views of the right and left hemispheres highlight tissue loss in inferior frontal, insular, and posterior temporal regions (right > left) and the inferior parietal (left > right) region. Top, Gray matter volume loss in the insula (right > left; inferior and coronal views), orbital frontal cortex (inferior and sagittal views), and cingulate cortex (sagittal and coronal views) are highlighted. In the inferior view, the anterior temporal lobe is cut away to expose the surface of the insular cortex.