Post-mortem assessment in vascular dementia: advances and aspirations

In clinical population-based series, the prevalence of VaD/VCI averages 8–15.8 % (in Japan, 23.6–35 %) with standardised incidence rates between 0.42 and 2.68 per 1000/year, increasing with age [23]. The range is broader in clinical studies using convenience series from western memory clinics, varying from 4.5 to 39 % [23]. However, the prevalence rates of VaD/VCI are unlikely to be accurate in any of these series because even the best clinical diagnostic criteria show only moderate sensitivity (approximately 50 %) and variable specificity (range 64–98 %) [23, 24]. VaD in autopsy series also varies tremendously, ranging from 0.03 to 58 % [23], and this variation is partly due to the lack of internationally accepted consensus criteria for the neuropathological diagnosis of VaD. In elderly patients, the prevalence of ‘pure’ VaD ranges from 5 to 78 %. In the oldest-old, that is, ≥90 years, the prevalence of pure VaD drops (to 4.5–46.8 %) but that of mixed AD/VaD increases, reflecting a constant age-related increase of neurodegenerative changes. Rigorous population-based clinico-pathological correlative studies addressing the prevalence of VaD are few, but they are arguably more informative about the actual prevalence of VaD/VCI. In population-based clinico-pathological series, the prevalence of pure VaD ranges from 2.4 to 23.7 %, and that of mixed AD/VaD from 4.1 to 21.6 % [25, 26]. The range is still wide and this may reflect regional differences in managing cardiovascular risk factors and ethnic-related genetic variances. In general terms, these studies show that the prevalence of VaD/VCI is higher in developing countries and Japan. For instance, in a clinico-pathological study from Brazil, where cardiovascular risks are poorly managed, the prevalence of pure VaD was 21.2 %, one of the highest detected in population-based studies [26]. On the other hand, in a retrospective hospital-based study in 1700 consecutive autopsy cases of elderly patients with dementia in Vienna, Austria (mean age 84.3 ± 5.4 years; 90 % over 70 years), pure VaD was observed in 10.7 %, decreasing between age 60 and 90+ from 15.0 to 8.7 % [27]. VaD and VCI are potentially preventable diseases; therefore, studies focusing on its prevalence, incidence and risks factors in the different populations are essential to guide public policies.

At present there are two fundamental issues regarding the assessment and diagnosis of VaD. First, there are no currently accepted neuropathological consensus criteria regarding the assessment of VaD, VCI, cerebrovascular pathology or related lesions [28]. Neuropathological assessment of the post-mortem brain is required to reach a definitive diagnosis and must be carried out in a standardised manner, applying reproducible methods and following generally accepted consensus criteria [29]. Widely used consensus criteria for the pathological diagnosis of common neurodegenerative disease, such as AD and Lewy body disease, have been available for some time [19‐21, 30‐33]. However, despite several attempts being made without major success [16, 34‐36], generally accepted neuropathological criteria for diagnosing VaD are still unavailable. Second, general assumptions regarding the underlying pathology of frequently observed in vivo magnetic resonance imaging (MRI) findings might not always be accurate. Neuroimaging is indeed an important tool in the clinical diagnosis of CVLs and imaging-pathological correlative studies are aiming to bridge the gap between in vivo imaging and post-mortem neuropathology. However, general assumptions regarding the underlying pathogenesis of common in vivo MRI findings are not unequivocally corroborated by neuropathological findings and this may result in inadequate clinical diagnosis and treatment.

Various forms of cerebrovascular disorders may lead to cognitive impairment and dementia in the elderly [17]. While pure VaD – most frequently caused by infarctions – is rare, it is generally assumed that cerebrovascular pathology contributes to the development of cognitive impairment in other neurodegenerative diseases, in particular in mixed AD/VaD. Such mixed disorders are frequently observed in the brains of elderly individuals and their prevalence and severity increase with advancing age [37]. In aged individuals, lacunes, microbleeds, WMLs and microinfarcts have been associated with cognitive decline, including reduced mental speed and impaired executive functions [38]. Cerebral SVD may interact with pathophysiological processes in AD either independently of each other or through additive or synergistic effects on cognitive decline [39, 40]. There are several clinical classification criteria for VaD/VCI, such as the NINDS-AIREN criteria, the State of California Disease Diagnostic and Treatment Centers (ADDTC) criteria, the International Classification of Diseases, Tenth Edition ICD-10 criteria and the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) criteria. They distinguish between the following: possible VaD – clinical criteria of dementia with focal clinical or imaging signs of one or more infarcts, gait disorder, pseudobulbar palsy, personality and mood changes; probable VaD – all signs of dementia, two or more infarcts followed by dementia and imaging signs of at least one extracerebellar infarct; and proven VaD – clinically proven dementia and pathological demonstration of multiple CVLs and mixed dementia. The diagnosis of VaD/VCI is reflected by recent clinical criteria [41] that are based on evidence of infarcts, white matter hyperintensities (WMH) and microbleeds, using structural MRI. Several autopsy studies have demonstrated that microinfarcts are major risks for VCI; however, microinfarcts can not be detected by 1.5 and 3.0 T MRI or naked eye examination, whereas they may be seen on novel high-resolution 7.0 T MRI [42‐45]. However, no accepted and pathologically validated criteria for the diagnosis of VaD/VCI are currently available [46]; therefore, the diagnostic accuracy of possible VaD is still relatively poor, with an average sensitivity of 0.49 (range 0.20–0.89) and an average specificity of 0.88 (range 0.64–0.98) [47, 48]. Cognitive decline has been shown to be weighted on specific pathological lesions in the following ranked order: NFT > Lewy bodies > Aβ plaques > macroscopic infarcts [49]. In neuropathologically defined mixed AD/VaD and SVD, the cognitive impairment profile mirrors that seen in AD cases, that is, all cognitive domains are equally impaired but memory scores are lower than executive scores [50]. This indicates that, regarding the combination of AD and SVD, it is the AD pathology that has the greatest impact on the severity and profile of cognitive impairment. Longitudinal, clinical and neuropathological studies have previously illustrated the impact of AD pathology in mixed AD/VaD, and demonstrate the usefulness of multivariate approaches to understand clinico-pathological profiles, as well as highlighting the current limitations to modelling and predicting cognitive decline and clinical profiles [49]. Nevertheless, the detection of the preclinical stages of cognitive impairment and early AD changes became a reality with the inception of amyloid PET tracers and various Aβ ligands, for example, Pittsburgh Imaging Compound B (PiB), fluorbetapir and flutemetamol [51]. Several studies have illustrated how amyloid PET imaging will improve differentiation between AD and mixed AD/VaD cases of dementia.

Converging evidence suggests that cerebrovascular and AD pathology exert an additive (and/or synergistic) effect on cognitive impairment. Does CVD merely reduce the cognitive threshold needed for overt clinical dementia in AD, or do both factors potentiate AD-specific pathophysiological pathways? Recent neuroimaging studies in cognitively normal elderly people aged 70–90 years suggested that vascular and amyloid pathologies are at least partly independent predictors of cognitive decline in the elderly, and that cognitive reserve seems to offset the deterioration effect of both pathologies on the cognitive trajectories [52].

Concomitant CVLs increase the risk and severity of clinical dementia in elderly individuals meeting the neuropathological criteria for AD [53‐55]. However, many studies emphasise additional pathogenesis in older people without dementia, in particular CVLs, with, for example, small or large cerebral infarcts, lacunar infarcts and WMLs reported in 22 to almost 100 % of cases [48, 55‐61]. Cerebral infarcts were seen in 21–48 % of seniors without dementia, with a higher frequency of large infarcts [48, 55, 58, 60, 62‐64] and CAA [55, 58]. Among 418 participants without dementia in the Religious Order Study (mean age 88.5 ± 5.3 years), 35 % showed macroscopic infarcts; those without macroscopic infarcts had microinfarcts (7.9 %), arteriosclerosis (14.8 %) or both (5.7 %), with only 37.5 % being free of CVLs [63]. In a study of 336 cognitively normal elderly adults, cerebral microinfarcts were seen in 33 % and high-level microinfarcts in 10 % [65]. In another study of 100 elderly participants without dementia (mean age 81.2 ± 5.4 years), CVLs including basal ganglia/deep white matter lacunes were seen in 73 % and CAA in 39 %; only 9 % of these participants were free of CVLs [66]. There were no correlations between CVLs and AD-related pathology in this latter cohort, whereas others reported an inverse relationship between Braak NFT stage and CVLs in autopsy-proven AD [67, 68]. The profile of AD and vascular changes becomes more complex with increased cognitive impairment in older people without dementia and these changes are likely to constitute a major substrate for age-associated cognitive impairment, suggesting a need for rigorous investigation of both neurodegenerative and vascular risk factors in old age [61]. However, the interactions in the pathophysiology between vascular risk factors, CVD and AD pathology, while plausible, are still unresolved.

In contrast to AD, less is known about the impact of CVD in other common neurodegenerative diseases, such as dementia with Lewy bodies (DLB) and frontotemporal lobar degeneration (FTLD). Prevalence reports of CVD in DLB are scarce, but autopsy studies reported a frequency of 20.2–34.4 % [69, 70], which does not differ significantly from controls [70]. In addition, an autopsy study indicated that more advanced Lewy body pathology is less likely to show severe CVD, and therefore suggested that cognitive impairment in DLB appears to be independent of CVD [71]. With regards to the heterogeneous group of FTLD, data in relation to the prevalence and patho-mechanistic role of CVD are very limited and contradictory. One autopsy study reported a frequency of 5.2 % for FTLD-tau and 17.3 % for FTLD-TDP-43 [69]. Some data support a role for SVD in FTLD disease progression [72], while others could not confirm this [69]. Therefore, further studies are necessary to clarify the role of CVD in non-AD neurodegenerative diseases.

In conclusion, the co-occurrence of CVD and AD in the elderly is very frequent [73]. There is evidence suggesting that both lead, in an additive as well as an independent fashion, to cognitive dysfunction. The characteristic pattern of HPτ-related neurodegeneration (i.e. Braak NFT stages) in AD corresponds to a pattern of memory loss that spreads to other cognitive domains. By contrast, the neuropsychological profile associated with VaD shows considerable variation; for example, executive dysfunction often equals or may exceed memory impairment in the SVD-subtype of VaD, but depending on location and severity of CVL all possible types of cognitive impairment may ensue. We anticipate that the availability of comparable measures of AD and VaD pathology from in vivo neuroimaging studies in the future will replace dichotomous classifications of diseases with more sophisticated modelling. However, as of today, the best available models predict less than half of the variance in cognitive performance [49].

WMLs histologically encompass structural damage of the cerebral white matter as a result of white matter rarefaction [3]. They are visualised as WMHs on pre- and post-mortem T2-weighted MRI, and they have been associated with a wide range of cognitive deficits [74]. Interestingly, WMHs are frequently seen in individuals both with and without dementia, although WMHs seen in AD are significantly more severe than the ones seen in so-called normal ageing [75‐77]. The pathogenesis of WMHs is generally thought to be associated with SVD because vessel wall alterations may lead to chronic hypoperfusion of the surrounding white matter [35]. Although WMHs are currently assumed to reflect SVD, WMHs on T2-weighted MRI are a visualisation of white matter abnormalities and cannot determine the underlying pathogenesis. Previous studies have suggested a multifactorial aetiology of WMHs [78‐82] inclusive of SVD-related ischaemia, but also degenerative axonal loss secondary to cortical AD pathology, that is, deposits of HPτ and Aβ. The exact pathological mechanism of degenerative axonal loss is still unclear, but it has been suggested that axonal death occurs simultaneous to grey matter atrophy or via calpain-mediated degradation, activated by AD pathology-related axonal transport dysfunction [83, 84]. Evidence from neuroimaging has shown region-specific white matter changes in patients with AD, most frequently in the posterior deep white matter [75, 85, 86] and corpus callosum [75], which have been directly associated with AD-related cortical atrophy [85, 86].

HPτ has been implicated as a principle instigator of degenerative axonal loss in AD. An extensive quantitative neuropathological study revealed that the burden of cortical HPτ in the temporal and parietal lobes was a predictor of WMH severity in AD [87], corroborating previous studies reporting an association between higher Braak NFT stage and increased WMH severity [77, 78, 88], and degenerative axonal loss in temporal [89] and parietal [84] white matter when in proximity to high cortical HPτ pathology burden. Furthermore, the combination of high cerebrospinal fluid (CSF) total-tau and higher parietal WMH volume was shown to predict the clinical conversion from mild cognitive impairment to AD [89],further supporting an association between the two pathologies. Although SVD-related ischaemic damage has long been assumed to be the main factor for the development of WMHs (for review see [90]), neuropathological investigations of patients with AD with severe WMH usually revealed only minimal SVD pathology [84, 89, 91]. However, in cases with minimal neocortical HPτ pathology (Braak NFT stage 0–II), SVD was found to be associated with WMH (Fig. 2) [92].

While theoretically both cortical HPτ pathology and SVD may lead to the development of WMH, it appears that in neurodegenerative diseases such as AD, WMHs are likely to be primarily associated with cortical HPτ pathology. On the other hand, in cases without dementia and in VaD cases, SVD seems to play a role in the development of WMH, which may relate to gliovascular abnormalities and BBB damage [93]. The clarification of the underlying pathogenesis of WMH and respective MRI characteristics is warranted to allow for clear interpretation of white matter neuroimaging and subsequent adequate management of patients.

The term cerebral microbleeds describes the radiological phenomenon of small, well-demarcated, hypointense, round or ovoid lesions detected on T2*-weighted gradient-recalled echo (T2*-GRE) and susceptibility-weighted imaging (SWI) MRI sequences [10]. Microbleeds create a ‘blooming’ effect on T2*-GRE/SWI, but are generally difficult to see on T1-weighted or T2-weighted sequences [10, 92]. Microbleeds have generated interest as a marker of the haemorrhagic consequences of SVD. Microbleeds are common in many different patient populations (healthy elderly, ischaemic stroke, intracerebral haemorrhage [94, 95], AD [96, 97] and VCI [98]). Of note, microbleeds are more prevalent in patients with recurrent stroke than in those with first-ever stroke, and they tend to accumulate over time, indicating a relationship with the progression and severity of cerebrovascular pathology [94]. Microbleeds generate increasingly common clinical dilemmas due to the concern that they may be a marker of future intracerebral bleeding risk [99‐104]. In a meta-analysis of 10 prospective studies including 3067 patients with ischaemic stroke or transient ischaemic attack, the presence of microbleeds was associated with a high risk of intracerebral haemorrhage (pooled odds ratio 8.53), raising questions regarding the safety of antithrombotic drugs [105, 106]. Moreover, most available studies suggest that microbleeds are associated with impairment of cognitive function [107, 108], although whether they are directly and independently implicated – or simply reflect more severe SVD – remains uncertain.

Similar to other SVD markers, microbleeds appear to represent a potential link between stroke, brain ageing, dementia and AD [97, 109], but they have not yet resulted in high-quality evidence-based recommendations for stroke and dementia clinical practice nor emerged as a valid surrogate marker for clinical trials in SVD, for example, in intracerebral haemorrhage and VCI. This might be due to the significant gap between the clearly defined markers seen on MRI and their as-yet uncertain pathological basis and pathophysiological mechanisms [109‐112]. It is consistently emphasised in the literature that microbleeds are the MRI correlate of extravasation of red blood cells from arterioles and capillaries damaged by a primary haemorrhagic SVD process and, therefore, are potentially strongly associated with haemorrhagic stroke risk. However, microbleeds are also associated with increased subsequent ischaemic stroke risk [113‐116], highlighting that they are a marker of a CVD that is simultaneously ischaemic and haemorrhagic, a phenomenon sometimes termed mixed CVD [109, 117]. Nonetheless, histopathological correlation studies suggest that radiologically defined microbleeds generally correlate with focal deposits of blood-breakdown products, predominantly haemosiderin-iron [110, 118]. MRI-histopathological correlation has been underutilised [119, 120], with a total of <70 microbleeds analysed in just a small sample of patients [110‐112], often detected using relatively insensitive T2*-GRE sequences at 1.5 T [118]. Technical challenges involved in correlating MRI with histopathology for such small lesions with a widespread distribution in the brain probably account for the small number of brains with microbleeds that have been analysed. Notwithstanding these limitations, when systematic neuropathological examination of SWI-visualised microbleeds is undertaken, the underlying pathologic substrates are actually rather variable, including not only focal accumulations of blood-breakdown products, but also (albeit much less commonly) microaneurysms, small lacunes, vessel wall dissections or (pseudo-) microaneurysms [112, 118, 121, 122].

Although most microbleed pathological correlation studies emphasise blood leakage from nearby damaged small vessels into the brain parenchyma as a mechanism, it must not be assumed that a primary haemorrhagic process fundamentally produces all microbleeds or that the most severely affected vessels are the culprits. Alternative non-haemorrhagic mechanisms for microbleeds, particularly if no tissue damage surrounds the vessel and haemosiderin is limited to the perivascular space, include ischaemia-mediated iron store release by oligodendrocytes [123], phagocytosis of red blood cell microemboli into the perivascular space (termed angiophagy) [121, 124], or even haemorrhagic transformation of small microinfarcts (Fig. 3) [125].

It is widely accepted that, by analogy with spontaneous intracerebral haemorrhage, the pathological processes underlying microbleeds differ according to their location in the brain, with CAA being the most notable correlate of exclusively lobar microbleeds (most often in the occipital and posterior temporo-parietal regions), while ‘hypertensive arteriopathy’ (including a spectrum of neuropathological processes affecting deep perforating vessels such as AS and lipohyalinosis) is strongly associated with predominantly deep microbleeds. The majority of data to date support this hypothesis, but much of the evidence is indirect and largely based on clinical and imaging studies [10, 112, 126‐130], rather than extensive direct morphological-pathological analyses [131]. A recent neuropathological study found no direct topographical association between CAA presence or severity and microbleeds (defined only pathologically as haemosiderin-laden macrophages in any brain region) [132]. Whether these microscopic lesions have the same biological significance and underlying mechanisms as radiologically defined microbleeds is not clear [120]. Further exploration of the neuropathological basis of microbleeds will be a key step in clarifying their mechanisms and nature. Along with well-designed observational clinical studies, this greater understanding should allow microbleeds to become useful in clinical management decisions [133]. Until then, the main question of whether a radiologically defined microbleed is always a true microbleed or whether it may also represent haemosiderin deposits, which in turn may or may not stem from a microbleeding event, remains unanswered.

With regards to CVL, novel applications of neuroimaging and biochemical methods, as well as additional investigation of neuroinflammation, have been suggested for the assessment of human post-mortem brains. Although these methods are beyond the scope of basic routine diagnostic procedures, the addition of such novel techniques may help to further elucidate the impact of CVD on cognitive performance.

Aside from the hallmark pathological lesions, there is evidence to suggest a role for immunological and inflammatory mechanisms in the pathophysiology of VaD/VCI. Neuroinflammation encompasses local endothelial activation, leading to the extravasation of fluid (and, sometimes, cells) via a dysfunctional BBB, resulting in oedema and tissue damage in the surrounding parenchyma and eventually leading to the activation of perivascular macrophages, microglia and other glial subtypes (Fig. 5a, b) [156‐158].

Clinical studies in patients with symptomatic SVD [159, 160] or WMH [161‐163] found elevated levels of circulating biomarkers of endothelial activation, that is, ICAM1, soluble thrombomodulin, interleukin-6 (IL-6) and PAI-1. This suggests that endothelial activation, and a possible inflammatory process, might contribute to SVD and to cognitive decline. A neuropathological study by Giwa and colleagues assessed endothelial activation in small perforating arteries in cases with moderate-severe SVD, and with minimal AD pathology (Braak NFT stage 0–II, and insufficient neuritic plaque pathology to meet CERAD criteria for AD). They found that endothelia were rarely immunoreactive for ICAM1 or IL-6; however, levels of luminal thrombomodulin (depletion of which is a hallmark of activated endothelium) were more pronounced, especially in individual vessels with severe high sclerotic index (Fig. 5c) [164]. The study concluded that local endothelial activation is not a feature of the arteriolosclerosis form of SVD, which is in agreement with evidence from a previous study of brain lysates demonstrating attenuation of inflammatory mediators (MCP-1 and IL-6) in individuals with VaD and mixed dementia, relative to aged control subjects [165]. While BBB dysfunction is often claimed to be part of SVD pathology, neuropathology studies show no conclusive association of BBB markers (fibrinogen, IgG, albumin; Fig. 5d) with SVD. Some neuropathology reports found a positive association between SVD severity and extravascular plasma proteins [166, 167] while others did not [139, 168, 169]. In subcortical white matter, fibrinogen labelling was associated with clinical dementia diagnosis in an AD-free cohort where dementia was likely to be primarily VaD [169]. Observationally, little evidence of leukocyte infiltration has been associated with SVD. Microglia have been shown to be significantly higher in number in the brains of persons with VaD and widespread WMH [79, 170, 171]. Activated microglia (CD68+) are strongly associated with WMLs (Fig. 5e, f) [79, 142].

Elucidation of the role of neuroinflammation in the pathogenesis and pathophysiology of SVD will enable the evaluation of immunotherapies as potential therapeutic options for prevention or treatment of VCI/VaD.