In this well characterized autopsy sample of 190 elderly individuals only presence of leukoencephalopathy, among the six types of pathological vascular lesions examined, was associated with an antemortem diagnosis of dementia and AD after accounting for other vascular and neurodegenerative pathologies. As the severity of leukoencephalopathy increased so did the risk of all causes of dementia, AD, and VaD. This dose-response relationship strengthens our findings. Previous studies that have examined the contributions of leukoencephalopathy to antemortem cognitive status have shown mixed results with some (Englund et al., 1988
; Garcia and Brown, 1992
) but not all studies (Tatemichi, 1990
) showing a significant association.
Most elderly individuals coming to autopsy have vascular lesions (Fernando and Ince, 2004
; Kalaria et al., 2004
; Knopman et al., 2003
; Riekse et al., 2004
; Schneider et al., 2004
; White et al., 2002
). One or more of the six types of vascular lesions were seen in 68% of our sample, comparable to other community-based samples (Neuropathology_group, 2001
; Schneider et al., 2007
). Previous studies have focused mostly on ischemic infarctions in autopsied brains. Macroscopic vascular lesions have been linked to clinical dementia (Chui et al., 2006
; Petrovitch et al., 2005
; Schneider et al., 2004
; Snowdon et al., 1997
) as well as with severity of cognitive impairment (Heyman et al., 1998
; Schneider et al., 2004
). However, like one previous study (Neuropathology_group, 2001
), we did not find an association of large infarcts and lacunar infarcts with a clinical diagnosis of dementia. Large infarcts were associated with antemortem clinical diagnosis of VaD. In a prior analysis of this autopsy sample, lacunar infarcts, but not leukoencephalopathy were associated with dementia (Crystal et al., 1993
). However, we now have a four-fold larger sample, and controlled for other previously unaccounted covariates in our multivariate analyses. In our effort to understand the independent effect of specific vascular lesions, we controlled analysis of each vascular lesion for the presence of others. If vascular disease has a common pathophysiological substrate this may represent over-adjustment. To address this issue we developed a vascular score which summarized three types of macroscopic pathology rather than adjusting for it. The success of the vascular risk score supports this approach.
Our results suggest that accumulating macroscopic vascular pathology, regardless of type, contributes to the diagnosis of dementia, and does so even in presence of other neurodegenerative pathology. The association between increasing vascular score and antemortem AD, supports vascular contributions to this clinical dementia subtype (Beach et al., 2007
; Honig et al., 2005
; Roher et al., 2004
). Another community based clinicopathological study with comparable age range and with analyses adjusted for other co-existing pathological processes did not find an association between dementia and individual types of vascular pathology, whereas presence of multiple vascular pathologies was associated with dementia (Neuropathology_group, 2001
). While our observations may be driven by the presence of large infarcts in the vascular score, our subgroup analyses excluding infarcts from the vascular score did not materially change findings. Our results, which look at lesions in aggregate in our sample, do not exclude the possibility that a single strategically located vascular lesion may result in dementia in individual patients. Silent thalamic infarcts detected on MRI were associated with increased rate of memory decline in the Rotterdam Study (Vermeer et al., 2003
In an autopsy series restricted to older Japanese-American men, multiple large infarcts were associated with antemortem dementia, and this association was nearly equal in magnitude to that seen with Alzheimer lesions (Petrovitch et al., 2005
). The Religious Order Study reported that large and lacunar cerebral infarctions were associated with a twofold increased risk for dementia (Schneider et al., 2003
). Another study reported a borderline association of subcortical vascular burden with antemortem cognitive status when adjusted for Alzheimer pathology. However, the investigators used a composite vascular score restricted to cystic infarcts, lacunar infarcts, and microinfarcts (Chui et al., 2006
). Unlike some studies, we did not find a significant association of microinfarctions with dementia subtypes (Kovari et al., 2007
). Strict comparisons with previous studies may be limited by differing study populations, smaller samples in previous studies, limited set of confounders considered in analyses, various definitions of vascular disease burden, differences in clinical diagnostic procedures and pathological techniques.
The antemortem correlates of our vascular lesions should be further explored using techniques such as neuroimaging. Epidemiological studies have reported an association between periventricular white matter lesions on imaging studies and cognitive decline (Prins et al., 2005
; Vermeer et al., 2003
), mirroring our findings with respect to leukoencephalopathy. Radiographic white matter lesions have been correlated with neuropathological findings such as axonal loss, myelin rarefaction, small lacunes, cerebral amyloid angiopathy, and perivascular demyelination and gliosis (Merino and Hachinski, 2000
). Although cerebral amyloid angiopathy was strongly associated with dementia in the initial model, we did not incorporate it in the vascular score for two reasons. Firstly, cerebral amyloid angiopathy represents vessel pathology while infarcts and leukoencephalopathy are parenchymal, and we did not want to mix those two pathologies in the score. Secondly, is not yet possible to verify or quantify cerebral amyloid angiopathy pathology during lifetime. As resolution of imaging techniques improves, correlations between MRI and microscopic vascular lesions should be re-examined.
Some possible limitations of our study merit discussion. The median interval between the last clinical examination and death was 13 months. Hence, it is possible that some individuals may have experienced cognitive decline after their last assessment. We controlled all analyses for clinicopathological interval. The significant associations were unchanged in our subgroup analysis restricted to subjects evaluated within a year of death. In common with other autopsy studies selection bias is the norm rather than the exception. We accounted for possible cohort effects in our analyses since our source populations included community volunteers, institutionalized participants, and population-based randomly recruited individuals. Presence or absence of clinical strokes or infarcts on imaging studies is used to assign dementia subtypes, raising issues of diagnostic circularity. This limitation does not apply to our overall clinical dementia diagnosis, which is assigned prior to subtyping cases as AD or VaD. We conducted a number of sensitivity analyses to address this issue. The results were not materially different when we excluded pathological large infarcts that may have been recognized during life or when cognitive impairment was defined independent of the dementia subtyping process by using cut scores on the Blessed test. The vascular lesions were known to the pathologist at the time of assigning neuropathological diagnoses. Hence, correlation of vascular scores with neuropathological diagnoses is limited by diagnostic circularity. While our dementia diagnostic procedures have been reported to correlate highly with pathological findings, clinical diagnostic misclassification may have reduced the strength of some of the observations. Our results should be confirmed in other independent autopsy samples. Lastly, neuropathologic evaluations were performed on one hemisphere in many cases. Moreover, identification of vascular pathology was based on gross brain sections. This is likely to have resulted in under estimation of the association between vascular lesions and clinical diagnosis of dementia. Utilization of techniques such as myelin staining of brain sections will improve quantification of white matter pathology. However, these techniques may not necessarily have clinical or imaging correlates.
The strengths of our study include the detailed assessment of clinical and neuropathologic data by experienced examiners, validated diagnostic procedures, and the inclusion of a wide range of pathological lesions associated with dementia.
In summary, our study shows that individual types of vascular lesions as well as increasing vascular burden are associated with antemortem diagnosis of clinical dementia. The individual types of lesions as well as the score described can be used to improve definitions of vascular dementia as well as dementia assessment procedures.