In this cross-sectional study of more than 575 elderly individuals from a biracial population sample, we found that when considered singly, WMHV, TBV, and having multiple cerebral infarcts were associated with global cognitive function. Overall and among those without dementia, cognition was inversely associated with WMHV and number of infarcts but was positively associated with TBV. When all 3 measures were simultaneously added to the model, the association of global cognition with WMHV and TBV remained significant and unchanged but was no longer significant with infarcts. There was a non-significant trend for having multiple infarcts in a higher proportion of white subjects, but WMHV was similar in both groups. While African Americans performed more poorly on all cognitive tests, the associations of all 3 MRI measures with cognitive function were not modified by race. Finally, the inverse associations of both WMHV and multiple infarcts with global cognition were much stronger among subjects without dementia, and the strength of the associations with cognition varied across specific cognitive domains.
Few studies have examined the associations of multiple MRI measures with cognitive function in general population samples of older persons. Results of the 3 studies published in the last decade suggest that WMHV, TBV, and infarcts each are associated with cognitive function, but when considered simultaneously, such associations are mixed.2,17
One study found that the association of cognition with WMHV was independent of its association with infarcts but not with TBV17
; another found that the association of global cognition with infarcts was independent of both its associations with TBV and WMHV.2
Neither of these studies used quantitative measures of WMHV or TBV; thus, the extent to which the differences in findings are attributable to methodological differences is uncertain. The results presented from this study, that the strength of the association of WMHV with cognitive function was unchanged in the presence of cerebral infarcts, are in keeping with those of others5
and suggest that WMHV may better reflect overall vascular brain injury and therefore have a stronger association with cognitive function.
The association of WMHV with global cognition and with performance in specific cognitive domains (episodic memory, perceptual speed, and visuospatial ability) was much stronger among subjects without dementia. A possible partial explanation is the disparity in these groups' sizes (81 subjects with dementia and 494 without). Another possibility, however, is that the association of WMHV with cognition may be more important at earlier stages (ie, before subjects meet criteria for clinical diagnosis of dementia). Recent data suggest that higher WMHV is associated with increased risk of progression from no cognitive impairment to mild cognitive impairment but not from mild cognitive impairment to dementia.13,57
The underlying mechanism may relate to vascular brain injury, in view of the association of WMHV with vascular disease10,23,48,51
and the reported association of vascular disease and vascular risk factors with impaired cognition.25
White matter hyperintensities are thought to reflect diffuse ischemic changes in the brain, which disrupt both the frontal subcortical circuits that subserve predominantly executive cognitive abilities58–60
and the connections between the frontal lobes and other cortical regions. White matter hyperintensity distribution over many anatomical areas of the brain could explain the correspondingly wide effects on the cognitive systems subserved by these areas, with maximal effects occurring before the pathological processes associated with dementia (reflected by brain volume loss) begin. Also possible is that many subjects with dementia had mixed pathological features61
including AD. This could explain the continued significant association between brain atrophy and cognition, but not WMHs or infarcts and cognition, in this subgroup, since brain atrophy is affected by both pathological processes.62
Some studies suggest that WMHs and TBV are associated with function in multiple cognitive domains,18,19,23,24
in contrast to cerebral infarcts, which affect predominantly measures of executive function.2,12,25
Our results generally agree: the measure of executive function performance (perceptual speed) was the domain most consistently associated with each MRI measure. We found that having multiple infarcts was associated with performance in episodic memory (typically prominently affected in AD) in the entire sample, and preliminary data suggest a similar, albeit weaker, association among subjects without dementia. These results are similar to those reported from 2 recent studies in older persons without dementia.63,64
The underlying mechanisms by which cerebral infarcts could affect cognition in elderly individuals, specifically in the domain of memory, are unclear; one possibility is that infarcts have an additive effect on subclinical preexisting AD pathology that varies with the anatomical locations of the infarcts. Although a high proportion of our subjects without dementia (130 of 494) had infarcts seen on MRI, information on the location of these infarcts will be needed to address this issue further.
Few studies have examined differences in brain MRI findings between African Americans and whites.7,62,65-67
In the present study, we examined 2 potential differences: first, differences in the amount or degree of a brain measure between these racial groups, and second, differences according to race in the association of brain MRI measures with cognition. Unlike other studies suggesting that both WMHs and cerebral infarcts66,67
may be more prevalent among African Americans, in this sample we found a trend toward a higher proportion of whites with multiple infarcts. One possible explanation is that the overall occurrence of vascular risk factors in African Americans in our sample may have been lower than in previous studies. Further studies in this sample are planned to closely examine the association of vascular disease and risk factors and their associations with brain MRI measures.
Consistent with a previous study,5
the association of each MRI measure with cognitive function was similar between African Americans and whites. Despite differences in baseline cognitive test results that persisted after controlling for educational achievement, both groups showed similar associations between cognitive performance and MRI pathological measures, suggesting that race itself does not modify the association of MRI measures with cognitive function. Longitudinal assessment of cognitive function in relation to brain measures is needed to further examine these preliminary observations.
This study has important strengths and limitations. Strengths include the inclusion of a large, well-characterized, biracial cohort of older persons; the use of previously established and validated measures of cognitive function; and the use of identical neuroimaging protocols at the same MRI facility. Limitations include the cross-sectional design, losses to MRI participation, and use of rough measures of infarct status (number and size vs volume measurements). Finally, though this study examined the association of multiple MRI measures with cognitive function in persons with dementia (the majority of which were AD), the paucity of cases of mixed dementia and vascular dementia limited our ability to perform specific analyses to examine the association between multiple MRI measures and cognitive function in these dementia subtypes. It will be important to examine these associations in future studies with larger sample sizes.