Reserve and Maintenance in the Aging Brain: A Longitudinal Study of Healthy Older Adults

Abstract

The aging brain undergoes structural changes even in very healthy individuals. Quantifying these changes could help disentangle pathologic changes from those associated with the normal human aging process. Using longitudinal magnetic resonance imaging (MRI) data from 227 carefully selected healthy human cohort with age ranging from 50 to 80 years old at baseline scan, we quantified age-related volumetric changes in the brain of healthy human older adults. Longitudinally, the rates of tissue loss in total gray matter (GM) and white matter (WM) were 2497.5 and 2579.8 mm³ per year, respectively. Across the whole brain, the rates of GM decline varied with regions in the frontal and parietal lobes having faster rates of decline, whereas some regions in the occipital and temporal lobes appeared relatively preserved. In contrast, cross-sectional changes were mainly observed in the temporal-occipital regions. Similar longitudinal atrophic changes were also observed in subcortical regions including thalamus, hippocampus, putamen, and caudate, whereas the pallidum showed an increasing volume with age. Overall, regions maturing late in development (frontal, parietal) are more vulnerable to longitudinal decline, whereas those that fully mature in the early stage (temporal, occipital) are mainly affected by cross-sectional changes in healthy older cohort. This may suggest that, for a successful healthy aging, the former needs to be maximally developed at an earlier age to compensate for the longitudinal decline later in life and the latter to remain relatively preserved even in old age, consistent with both concepts of reserve and brain maintenance.

Keywords: MRI; aging; brain; healthy adults; maintenance; reserve.

Figure 1.

Age-related changes in total GM volume (first row), total WM volume (second row), and total CSF volume (third row). The first column showed LME model estimates using all participants’ data, the second column using only that of male participants, and the third column for female participants. Dots correspond to raw volumes adjusted for sex and eTIV. Colored thin lines are conditional fit of the longitudinal data using the estimated LME model. Longer green thick lines represent association with baseline age, whereas shorter black thick lines represent association with time for the longitudinal component. Data from male participants are shown in blue, whereas that of female participants are shown in red. Estimated model parameters, t values, and p-values are given in Extended Data Figure 1-1.

Figure 2.

Linear relationship between (a) regional GM volumes and baseline age as well as (b) regional GM volumes and time from baseline scan. The cortex is divided into 34 ROIs for each hemisphere based on the Desikan–Killany atlas. Estimated model parameters, t values, and p-values are given in Extended Data Figures 2-1, 2-2, 2-3 using data from all participants, male only, and female only, respectively.

Figure 3.

Estimated LME models for selected subcortical ROIs using all participants’ data. Volumes are adjusted for sex and eTIV. Dots correspond to adjusted raw volumes. Colored thin lines are conditional fit of the longitudinal data using the estimated LME model. Longer green thick lines represent association with baseline age, whereas shorter black thick lines represent association with time for the longitudinal component. Data from male participants are shown in blue, whereas data from female participants are shown in red. Estimated model parameters, t values, and p-values are given in Extended Data Figure 3-1.

Figure 4.

Estimates of LME model for cerebellar volumes: (a) cortex and (b) WM. Sex-specific LME estimates are also shown in the insets. Dots correspond to raw volumes adjusted for sex and eTIV. Colored thin lines are conditional fit of the longitudinal data using the estimated LME model. Longer green thick lines represent association with baseline age, whereas shorter black thick lines represent association with time for the longitudinal component. Data from male participants are shown in blue, whereas that of female participants are shown in red. Estimated model parameters, t values, and p-values are also given in Extended Data Figure 3-1.