Among the three SIVD groups with lacunes, we found no association between the presence or severity of cognitive impairment and characteristics of radiologically defined lacunes. That is, similar numbers, volume, and anatomic location of lacunes were found in all three groups, irrespective of their cognitive status (CN+L, CI+L, or D+L). In contrast, we observed strong associations between the severity of cognitive impairment and hippocampal atrophy, ventricular dilation, increasing WMSH, and cortical gray mater atrophy. Hippocampal atrophy has been reported consistently among patients with prAD,10-12
but does not appear to be specific. In one study, hippocampal atrophy was also observed among nine patients with VaD.20
The current study confirms the occurrence of hippocampal atrophy in a much larger sample of patients with D+L (three with autopsied confirmed SIVD) (see ). Our data also confirm the association between dementia and ventricular enlargement, noted by other investigators studying heterogeneous samples of large- and small-artery type VaD.3,14,15
MRI and computerized segmentation methods used in the current study enable demonstration of a stronger association between dementia in SIVD and specifically neocortical as well as hippocampal atrophy.
According to the lacunar hypothesis, dementia is related to an increasing number and volume of lacunes, particularly when strategically located in networks subserving cognition.6,7,37
Despite our failure to marshal support for any part of this hypothesis, several caveats should be kept in mind. First, lacunes were defined, in this study, by their morphologic appearance in MRI. Criteria were established to distinguish lacunes from perivascular spaces, and all lesions in white matter and subcortical gray matter meeting these criteria were counted and sized. However, no attempt was made to separate symptomatic from silent lacunes and only the subset of cystic lacunes was volumed in the white matter. In the absence of adequate radiologic–pathologic correlation, one should not presume that all lesions meeting radiologic criteria for lacunes are histologically comparable. A “lacune” may represent a complete infarct, incomplete infarct, area of focal gliosis, or perivascular space. Thus, the absence of correlations between cognitive impairment and numbers or volume of lacunes could be explained by the differential distribution of “false-positive lacunes” among the three lacunar groups. A preliminary review of the clinical data does not support this explanation. In our sample, 44% of our subjects with lacunes had no symptomatic history of stroke or TIA. The proportion of symptomatic versus silent lacunes, however, was evenly distributed across all three cognitive groups. Because only cystic lacunes were volumed in the white matter, our methods do not adequately address the possibility that cognitive impairment correlates with the number and volume of lacunes in the white matter.
Second, the method used in this study to localize lacunes was based upon gross anatomic boundaries, rather than functional divisions within frontal–sub-cortical loops.37
For example, lacunes were localized to the thalamus, but not to specific nuclei such as the anterior and dorsomedial nuclei that project to the prefrontal lobe. Thus, our methods do not adequately test that component of the lacunar hypothesis related to the strategic importance of location.
Other investigators have questioned the validity of the lacunar hypothesis. No differences were found in the location or volume of lesions between demented and nondemented groups with lacunes.3,38
Other investigators reported an association between the number, but not the volume, of lacunes and cognitive function.17
These reports, taken together with our findings, suggest that radiologically defined lacunes, although an indicator of subcortical ischemic injury, may not be reliable markers for the overall severity of either ischemic brain injury or cognitive impairment.
The second question addressed in this study is whether dementia among patients with presumed SIVD simply indicates the presence of AD. Several pieces of evidence suggest not. The pattern of association between cognitive impairment and imaging variables differed in the lacune versus prAD groups. For the three groups with lacunes, cognitive impairment was associated with two partially independent atrophic processes. The primary process was reflected most strongly in hippocampal atrophy, whereas the secondary process was associated with greater volume of WMSH, increasing ventricular dilation, and cortical gray matter atrophy. For the prAD group, there was only evidence for one atrophic process, reflected most strongly in hippocampal atrophy. In this group, hippocampal and neocortical atrophy were strongly correlated with each other and did not contribute independently to cognitive impairment.
The regional pattern of neocortical gray matter atrophy also differed between the D+L and prAD groups. In the prAD group, primary motor and visual cortex were relatively spared. This finding is consistent with the lower densities of neurofibrillary tangles observed in these areas in histopathologically confirmed AD.39
By contrast, in the D+L group, all regions of neocortex evidenced comparable degrees of atrophy. Several factors may limit the accuracy of Talairach transformations for defining cortical regions of interest, including increasing distortion with increasing distance from the AC-PC line and disease-specific variations in brain morphology. However, identical operations were applied to all individual brains in the study. Therefore, limitations of the method do not explain the systematic differences observed in the pattern of cortical gray matter atrophy between the D+L and prAD groups. Although some of our lacune group may yet prove to have AD (or other neuropathology) as a contributing cause of atrophy, autopsies in eight of our lacune cases confirm that hippocampal and neocortical atrophy can and do occur in cases of relatively pure SIVD. This suggests that although cortical gray matter atrophy occurs in both AD and SIVD, the underlying cause is likely to differ.
The etiology of hippocampal and neocortical atrophy in D+L cases remains unknown. Several non-mutually exclusive possibilities may be considered: 1) concomitant AD (addressed above), 2) secondary degeneration, or 3) subclinical ischemia. Note that neurofibrillary tangles were found in the hippocampus in all three autopsied cases of D+L (Braak and Braak II–IV). In one case, hippocampal sclerosis was found as well. Thus, the pathogenesis of hippocampal atrophy in SIVD is variable and may reflect a combination of degenerative and ischemic pathologies.
Cortical atrophy in SIVD might result from secondary axonal and trans-synaptic degeneration following primary subcortical injury. This would represent the structural corollary to the traditional notion that cortical hypoperfusion in SIVD results from functional deafferentation of cerebral cortex.40
Based on anatomic studies of frontal–subcortical circuits in nonhuman primates,41
predominant atrophy of the prefrontal cortex might be expected. Regional analyses in our lacune groups did not reveal preferential atrophy of prefrontal cortex, as opposed to parietal, temporal, or occipital cortex (see ). However, despite this pattern of diffuse atrophy, secondary degeneration cannot be ruled out. Although prefrontal cortex is considered to be the primary area of connectivity, widespread areas of posterior multimodal association cortex and hippocampus are also interconnected within cortical–subcortical loops.41
Finally, tissue loss in neocortex might result from primary (albeit subclinical) ischemia. Occlusions of microvessels or chronic hypoperfusion might produce incomplete infarction (e.g., selective neuronal or axonal loss),42
undetected by clinical symptomatology or by focal neuroimaging changes. Patients with severe WMSH (i.e., Binswanger type vascular dementia) may be at particular risk for subclinical ischemia. Deficient autoregulatory reserve43
and increased oxygen extraction fraction44-46
have been demonstrated in this subgroup of SIVD. Thus, WMSH may be a marker for the severity of deep cerebral hypoperfusion.
Several investigators have noted an association between WMSH with VaD14
In our study, WMSH volume was twofold greater in the D+L versus CN+L groups, but not significantly different in the prAD versus NC groups. In all five groups studied, WMSH volume correlated with percent cortical gray matter, but not with hippocampal volume. Among the three lacunar groups, WMSH volume correlated negatively with MMSE score. These data suggest that in SIVD, the extent of WMSH is related to the pathogenesis of dementia, as well as the magnitude of cortical gray matter atrophy. Combined with pathophysiologic data in the literature, these findings suggest that cortical gray matter atrophy may (at least partially) reflect the magnitude of an ischemic process that is indexed by the extent of WMSH, whereas hippocampal atrophy reflects yet another pathogenetic processes.