We found that the PSD group has more evidence of vascular pathology indicated by WMH volumes and lower cortical perfusion. In the AD group, there was reduced perfusion specifically in the parietal and prefrontal areas, consistent with previous studies,6
but no increase in WMH volume.
Differences between the PSD and AD groups are that, relative to control subjects, the stroke group had lower perfusion throughout the whole GM and specifically reduced perfusion relative to that in the AD group in the central gyrus region, which is typically spared in AD. WMH volume, typical of small vessel disease, was increased in stroke but not in AD, and the PSND group had a total brain volume of the same size as that of the AD group (and almost significantly different from that of the control group) (), whereas their hippocampus size, although smaller than that for the control group, was larger than that in the AD group.
The PSND group showed subtle changes in cognitive function, with low CBF rather than hippocampus volume predicting a low CAMCOG-R score in the stroke group. The data in the stroke group, i.e., the presence of WMH, lower brain volume, and decreased cognitive ability, are in accord with the suggestion2
that stroke increases the risk of dementia because vascular damage increases the susceptibility of the brain to further insult. Hence, either a second stroke, which markedly increases dementia incidence,1
or development of Alzheimer pathology will lead to earlier dementia or cognitive decline than would otherwise occur. None of the PSD group and only one of the PSND group reported a subsequent stroke, and hence in this cohort, development of AD seems a more likely cause of dementia or cognitive decline.
Hippocampus atrophy has been found to be a good surrogate of AD pathology in postmortem studies.22
In the PSND group, there was evidence of mild hippocampus atrophy and hypoperfusion in the parietal cortex, both typical of early AD.6,22
MTA just after the stroke was correlated specifically with parietal hypoperfusion at 6 years, again supporting the idea that the atrophy is related to AD. Furthermore, the addition of baseline cognition to the regression changed the imaging predictor of CAMCOG-R score from CBF to hippocampus, perhaps suggesting that stroke caused a simultaneous lowering in perfusion and cognitive ability and that subsequent cognitive decline is related to hippocampus atrophy. Some of the cognitive decline in the stroke group is likely to be due to AD-type changes, with the probability that the ischemic changes (increased WMH volume and tendency to reduced perfusion) has reduced the cognitive reserve, making the brain more sensitive to the additional pathology. The best predictor of low global CBF in the stroke group was hippocampus volume, again suggesting that presence of AD-type pathology may play a causal role in CBF reduction. Incidence of AD shows a stronger increase with age than does incidence of VaD,23
and this may explain the relatively strong presence of Alzheimer-type neuroimaging features in the stroke group of this old age group. The strong correlation of baseline CAMCOG-R score with 6-year CAMCOG-R score in the stroke group suggests that baseline cognitive testing shortly after stroke may be a good predictor of relative cognitive performance a number of years later.
Research on the relationship of CBF to dementia in vascular disease has mostly been done in subcortical ischemic vascular disease (SIVD). Methodology has varied, with voxel-based analysis being used to find a reduction in the thalamus7,9
and parietal and temporal lobes.9
An ROI study found reductions in parietal, temporal, and frontal lobes in SIVD similar to those in AD compared with those in control subjects8
with a slightly greater magnitude of reduction in parietal CBF in SIVD. A study of a larger cohort of stroke subjects on average 1 year after stroke looked at whole-hemisphere CBF and showed that bilateral CBF reduction was associated with cognitive decline.10
Our study is one of the few to examine CBF in elderly stroke survivors long term after stroke, and its findings are broadly consistent with these previous studies, with parietal and global deficits in CBF related to dementia, and in the cognitively normal stroke group, there are CBF deficits 6 years poststroke.
A number of studies have linked perfusion or metabolic changes to subcortical GM structures, including the thalamus in VaD.7,9,24,25
We did not find significant differences in thalamic perfusion between groups; however, the thalamic perfusion was lowest in the PSD group. There may be some variation in perfusion deficits according to different stroke type (the 2 poststroke studies that found thalamic perfusion reductions were of small vessel disease); however, not all studies examined the thalamus. We did not find any differences in thalamic perfusion between the different stroke classifications albeit with relatively low numbers in each group. It may be that the heterogeneity of our stroke group reduced the significance of any thalamic perfusion changes.
Limitations of the study include the few stroke subjects with dementia. Given the heterogeneity of stroke, this means that the PSD group is not necessarily representative of patients with PSD as a whole. Because this study investigated poststroke delayed dementia, imaging was, of necessity, performed 6 years after the initial stroke. There is thus a survivor effect, with those subjects experiencing ill health being more likely to have dropped out. This may have biased the profile of the remaining subjects with dementia to a more Alzheimer type, because those with severe vascular disease dropped out. Unfortunately, we did not have any postmortem or Pittsburgh compound B amyloid imaging confirmation of Alzheimer-type pathology. The MRI group had a relatively large proportion of subjects with posterior syndrome stroke compared with the original cohort, which may also have biased the results toward a more posterior deficit in CBF. Although we tried to recruit all the surviving stroke subjects, a number of them were too ill to tolerate imaging, and incidence of dementia was higher in this group. It is likely that this group with poorer health and greater dementia incidence may have had a lower CBF than those with imaging, and hence our findings probably underestimate the degree of hypoperfusion in poststroke survivors.
Strengths of the study include a well-characterized group of older poststroke survivors who did not have dementia immediately after stroke and well-matched comparison groups of subjects with AD and healthy control subjects. We used information from the anatomic image to identify GM pixels in the perfusion image, and thus changes in perfusion values should represent changes in CBF rather than brain atrophy.
We found evidence of both vascular and AD-type changes in the PSD group, suggesting that both the direct impact of the stroke and, more importantly, the subsequent development of AD in a compromised brain play a role in the etiology of PSD.