Our aim was to transform a conceptual model of CVL into an operational staging system for use in evaluating late-life cognitive disorders. We performed an extensive inventory and evaluate the frequency of CVL in a large series of postmortem brains from longitudinally followed patients with dementia and from 2 European brain banks. Since the definitions of CVL are often institution-specific and of variable quality,12
we chose to quantify those with most unequivocal description: modifications of small vessel wall (arteriolosclerosis and amyloid angiopathy), perivascular spacing, myelin loss in the WM, and infarcts. The latter were subdivided by size into microinfarcts (not visible on gross examination) and large infarcts (either lacunar or territorial). These lesions were also selected according to their clinicopathologic relevance in VCI.6
Infarcts and WM changes have long been associated with cognitive impairment.13
The clinical relevance of perivascular spaces dilatations is still controversial, but these modifications are associated with age, blood pressure, and MRI of SVD.14
Perivascular hemosiderin leakage has recently been shown to represent a neuropathologic marker of vascular frailty in dementia, linked to both amyloid angiopathy and vascular risk factors.15
Our study confirms the high frequency and the widespread heterogeneity of CVL among brains from patients with dementia, as published in previous postmortem cohorts.2,16,17
Our systematic semiquantitative scoring provides a precise overview of the relative frequency of each CVL across 4 strategic regions in various conditions defined by routine neuropathologic examination. The most frequent CVL were arteriolosclerosis and enlarged perivascular spaces. Arteriolosclerosis was largely predominant in the basal ganglia. CVL were more frequent in the temporal lobe from mixed cases (AD or DLB + VaD) than the VaD cases, suggesting that the neurodegenerative process by itself may contribute to WM damage in this location.18
Conversely, CVL were more frequent in the frontal lobe in VaD cases than mixed cases. Hippocampal microinfarcts were more frequent in this (17.8% of all cases) than those in a recently reported cohort.19
The analysis did not address the contribution to cognitive impairment by the burden of cerebrovascular pathology. However, the development of a grading scheme for CVL that reflects sequential worsening of pathologic changes is necessary antecedent to understanding clinical–pathologic correlations. Our demonstration of the spectrum of pathologic changes ranging from very common to less common shows that there may be a quantitative way of representing CVL. Initial lesions are vessel wall modifications: arteriolosclerosis, amyloid angiopathy, or a combination of both. These 2 types of SVD, even if they are completely different in nature, are both linked to parenchyma damage, either WM changes or cortical microinfarcts,5,20–22
although their precise pathogenetic mechanisms are unclear.16,23
Modifications of perivascular spaces could represent early consequences of altered vessel wall permeability in arteriolosclerosis, leading to hemosiderin leakage and alteration in the perivascular parenchyma.24
Hypoperfusion may then explain modifications and dilatation of the perivascular spaces, myelin loss in the WM. Cortical or subcortical microinfarcts and large infarcts seem to be the latest events in the time course of cerebral SVD. However, as illustrated by our rare cases with cortical microinfarcts but without vessel wall modification or WM damage, systemic microemboli represent an alternative etiology of small ischemic cortical lesions. Based on this distribution, we devised a scoring system for cerebral SVD with a special focus on microscopic lesions. Our vascular score, defined by the addition of the SVD stages in the 4 regions of interest, was in good agreement with the previous standard neuropathologic assessment of cerebrovascular pathology in these cases.
Our study has limitations. The fact that our assessment of CVL was only performed on a limited but practicable number of sections, i.e., 8 per brain (4 regions stained twice), could represent a critical constraint. However, rather than aiming at an extensive numbering of lesions, which may be neither practicable nor more informative, we focused on the hierarchy of the pathologic processes in each major arterial territory of the brain. Alternate sampling was not available in this retrospective postmortem cohort which thus does not entirely comply with the harmonization standards.6
However, in SVD, microscopic CVL are usually symmetrically distributed25,26
as even judged by the examination of whole-brain sections to reveal all the lesions and their extent.27
In our more recent study on poststroke stable and dementia cases there was also no major impact on laterality of medial temporal atrophy or hippocampal microscopic changes.28
In agreement with the harmonization standards,6
we recommend that alternate sampling should be now widely used in prospective cohorts. We recognize, however, that even the limited number of brain regions we studied may exceed available resources on a clinical neuropathologic service. Unfortunately, cerebrovascular pathology requires more intensive histologic examination. Second, the staging mostly relies on the natural history of arteriolosclerosis and atherosclerosis and may not reflect other causes of vascular injuries of the brain. Cardioembolic or hemodynamic infarcts may not perfectly fit with the hierarchical extent we describe. Nevertheless, the vast majority of cases with VCI are, at least in part, in causal relationship with arteriolosclerotic SVD. Pure cardioembolic or hemodynamic VaD cases are rare and should be considered separately. Finally, we did not include brains from aged controls without dementia and were not able to compare the distribution of CVL between cognitively impaired and nonimpaired groups. The extent of overlap in vascular scores between cognitively normal persons and those with dementia could be considerable.29,30
The vascular scores proposed here as an index for the vascular burden are presented to the clinicians, pathologists, and dementia researchers who called for a long time for a relevant way to assess cerebrovascular pathology despite several factors preventing the creation of neuropathologic criteria for VaD.31
Replication of our studies worldwide is needed to assess if the staging system is useful, since the lack of adherence was one of the reasons explaining why none of the previous published criteria for VaD were widely accepted. However, our procedure relies on routine examination of the brain and does not necessarily require extensive sampling or supplemental or time-consuming techniques. Similar assessment in a cognitively unimpaired, age-matched sample of subjects would represent an important addition to the literature.
Our proposed scheme of cerebral SVD reflects that there is probably not a clear cutoff threshold between cases with significant CVL burden and cases with pure neurodegenerative diseases. On the contrary, cerebral SVD more reliably represents a spectrum from initial vessel wall modifications to the later occurrence of WM damage and cortical infarcts. Further clinical pathologic studies are required to assess the clinical relevance of evaluating the CVL burden for diagnostic practice or research purposes.