Over the course of the last hundred years our understanding of dementia has undergone considerable change. For much of that time ‘senility’ was seen as an almost inevitable concomitant of ageing and was usually attributed to atherosclerosis. In the post-war period, a concept of multi-infarct dementia emerged, but at the same time Alzheimer’s disease (AD) came to be seen not as a rare young-onset disorder but as the cause of most cases of dementia. Indeed, the concept of dementia itself was built around the impairments seen in AD, notably the impairment of memory. With greater understanding of dementia, particularly its pathological basis, it came to be realized that many different kinds of pathology can produce a dementia syndrome and that multiple types of pathology are the rule rather than the exception. This chapter aims to review our current understanding of how cerebrovascular disease contributes to cognitive impairment in order to assist those involved in the assessment and management of people with cognitive impairments to understand the strengths and limitations of our current concepts, and to inform diagnosis and management.
Over the course of the last hundred years our understanding of dementia has undergone considerable change. For much of that time ‘senility’ was seen as an almost inevitable concomitant of ageing and was usually attributed to atherosclerosis. In the post-war period, a concept of multi-infarct dementia emerged,1 but at the same time Alzheimer’s disease (AD) came to be seen not as a rare young-onset disorder but as the cause of most cases of dementia. Indeed, the concept of dementia itself was built around the impairments seen in AD, notably the impairment of memory. With greater understanding of dementia, particularly its pathological basis, it came to be realized that many different kinds of pathology can produce a dementia syndrome and that multiple types of pathology are the rule rather than the exception. This chapter aims to review our current understanding of how cerebrovascular disease contributes to cognitive impairment in order to assist those involved in the assessment and management of people with cognitive impairments to understand the strengths and limitations of our current concepts, and to inform diagnosis and management. We try to reach pragmatic conclusions, not to provide a comprehensive account of a complex area of research.
Vascular cognitive impairment (VCI) is a term that includes the full range of cognitive impairments which occur in people with cerebrovascular disease, ranging from mild cognitive impairment (MCI) to vascular dementia (VaD).2 This is perhaps slightly confusing as the term ‘mild cognitive impairment’ refers to a condition which, unlike VCI, falls short of dementia in its severity.
The pathology of VaD and VCI is not well defined at the present time. There are several reasons why this is so: the clinical definitions of VaD and VCI are not well delineated and there are several different clinical schemes for assessing and diagnosing it; the clinico-pathological correlation is not very good and varies from one series of cases to another, in part based on the source of cases, e.g. whether in the context of post-stroke cognitive performance or memory clinics; and cerebrovascular pathology itself is highly heterogeneous with a lack of clear agreement about how to characterize it and how different components relate to cognitive problems. Some skewing of views has stemmed from the attempts, when arriving at definitions, to distinguish clearly between AD and VaD/VCI despite many cases of dementia, as is increasingly recognized, showing both AD pathology and vascular disease in the same brain. (This is shown schematically in Figure 2.1, which emphasizes how difficult it can be to find ‘pure’ cases of vascular pathology, especially when some pathology remains symptomatically silent.) Indeed, in a recent autopsy study of dementia nine different combinations of pathology were found and in more than 74% of cases there were two or more forms of neuropathology.3 An indication of the prevalence of vascular disease as a cause of cognitive impairment or dementia is provided by the extensive study of Power et al., who found, in a study of 1,362 autopsies on elderly subjects, 44% of whom had dementia, that vascular pathology accounted for 32% of the association between age and dementia while amyloid/tau pathology accounted for 24%, TDP-43 pathology and hippocampal sclerosis for 43% and Lewy body disease for 1%.4 However, differing prevalences of types of neuropathology are apparent, depending on the source of cases, as exemplified by another study of 132 autopsy cases of dementia, in which 84 cases had mixed pathology and Lewy body pathology figured prominently (with differing amounts of AD pathology), but vascular disease with AD was the commonest form of mixed pathology.5 Another level of complexity in linking vascular pathology to dementia arises from the very important added contribution to the debate that has come from neuroimaging studies over the last 20–30 years. The ease with which vascular lesions can be seen on imaging may lead to the overdiagnosis of vascular disease as the cause of dementia or cognitive impairment in some cases. At least this is the suggestion from a clinico-pathological study in which 42 autopsied subjects had the clinical diagnosis of the cause of their dementia compared with the autopsy findings. Thirty-eight per cent of these cases were given a clinical diagnosis of VaD, but at autopsy only 19% of the cases were so diagnosed.6
Figure 2.1 Coexistence of pathologies in dementia.Schematic diagram to show the extensive coexistence of different types of pathology in many individuals and the corresponding low frequency of pure forms of pathology. The degree of cognitive impairment increases from top to bottom of the diagram, and the horizontal line represents the point at which impairments are sufficiently great to be classified as dementia. CIND – cognitive impairment no dementia; AD – Alzheimer’s disease (presence of pathology, not necessarily with dementia syndrome); CVD – cerebrovascular disease (pathology, not necessarily with dementia syndrome). ‘Other’ includes many forms of pathology which can give rise to cognitive impairment, such as Lewy bodies
It could be said that in some respects our understanding of the pathological basis of VaD/VCI now is at the same stage as our understanding of the pathological basis of AD in the early 1980s. In both conditions there are complex, distinct components to the pathology (plaques and tangles in AD, different forms of vascular pathology in VaD/VCI), and we need to understand how these types of pathology in VaD/VCI are related and which are the key contributors. In AD we now know, following detailed clinico-pathological studies, that tangles correlate closely with the severity of dementia;7 but we are at an earlier stage of discerning which of the forms of vascular pathology are the critical ones. A recent step in the right direction was taken when a consensus study of the importance of different types of vascular pathology in VaD/VDI was published.8 We need this knowledge if we are to take steps to prevent VaD/VCI. In this review we shall try to tease apart the various strands of clinical, imaging and pathological evidence about VaD/VCI. Our aim is to achieve improved clarity about the nature of VaD.
Many authorities advocate neuroimaging as part of the routine clinical workup recommended for people presenting with memory disorders. Why? What exactly do such studies tell us? We suggest four types of answer.
First, it may pick up surgically treatable conditions such as tumours, subdural haemorrhage or normal pressure hydrocephalus (although it has to be said that in practice the pick-up rate of treatable causes is very low indeed, even though relevant findings may be made).
Second, it may help to improve the accuracy of diagnosis of the subtype of dementia, which may have implications for treatment.
Third, it can be helpful for those affected and their families to see for themselves that the problems they struggle with have a real physical cause and to feel that everything possible is being done to identify and treat the problem. The ability to use the picture archiving and communication system (PACS) to show images to patients in the clinic can be a source of information and explanation, and a valuable aid to discussing diagnosis. For some people with limited insight into their condition, it can help them to understand that there is a problem for which assistance might be beneficial.
Finally, neuroimaging has a growing role in research, and there is reason to hope that it can assist in early diagnosis and disease monitoring. As new treatments are developed which may be effective in the early stages of disease, it will become increasingly important to identify disease early. In the case of CVD, early identification of brain changes, perhaps even before clinical signs of dementia, may mean that there is stronger motivation to manage risk factors energetically at a relatively early stage. Neuroradiology has reached a level of development where some authors have argued that neuroradiology rather than pathology should be considered the gold standard against which new diagnostic criteria should be judged.9
Space does not allow a full treatment of the wide variety of new imaging techniques now available (but see Chapter 13).10 We will focus in this section on computerized tomography (CT) and magnetic resonance imaging (MRI) and on the pathology that these imaging techniques show.
Large vessel disease includes single and multiple infarcts that can be readily seen by CT or MRI. They can be in the territory of main arteries or in the watershed zones between the main arterial territories. Vascular dementia can be caused by multiple or single strategic infarcts (i.e. infarcts in locations critical for cognitive function such as thalamus, anterior thalamic radiation or angular gyrus), and the NINDS–AIREN criteria (devised by the National Institute of Neurological Disorders and Stroke and Association Internationale pour la Recherche et l’Enseignement en Neurosciences) specify in which brain regions infarcts should be for diagnosis of VaD. For large vessel infarcts to be sufficient for diagnosis of VaD they should be multiple, bilateral or in the dominant hemisphere. Studies of patients after they have suffered an ischaemic stroke show that progressive dementia is significantly more common than in an age- and sex-matched control population. Thus, in a study of 251 patients aged over 60 years, dementia was found in 26% of patients 3 months after they had suffered a stroke compared with 3.2% in a control sample.11 In a meta-analysis of dementia after stroke, dementia was present in 10% of patients before the stroke developed and in a further 10% there was evidence of dementia developing after one stroke, while this rose to more than 30% in those who had suffered more than one stroke.12 In a more recent study, post-stroke dementia was particularly likely after a severe stroke.13 This evidence points to the significant contribution that major ischaemic stroke can make to dementia. A deficiency in most of the stroke literature is that post-mortem studies have not been performed so that the relative contributions of AD pathology, perhaps ‘unmasked’ by vascular disease, and the vascular disease itself were not assessed.
Lacunes are small subcortical infarcts from 3 to 20 mm in size, usually round or oval. They may be clinically silent but may be accompanied by the history of a transient ischaemic attack (TIA) or stroke. Lacunar infarcts in the thalamus can cause prominent cognitive problems. At least two lacunes are required for a diagnosis of VaD. There is good evidence that lacunar infarcts can form a substrate for VaD. They are often multiple and bilateral and are associated with disease affecting the small arteries and arterioles subserving the deep grey matter and white matter (see Figure 2.2).
Figure 2.2 MRI appearance of vascular damage. Examples of two patients with dementia and diffuse white matter changes (a) and bilateral lacunar infarcts (b), and of a cognitively healthy control with mild white matter changes (c). MMSE – mini-mental state examination. Images were acquired with a FLAIR sequence in 3T MRI scanner
These are ischaemic lesions that are too small to be seen with the naked eye and are detected when microscopic histology is carried out. They can occur in the subcortical tissue but are most frequently found in the cortex, where it is estimated they can occur in their hundreds. They are difficult to detect in vivo. They can result from a number of different disease processes in vessel walls, e.g. vasculitis, but most are probably caused by cerebral amyloid angiopathy (CAA) or subcortical small vessel disease (see the following section) or by embolic occlusion of small arteries by microemboli.
Diffuse subcortical small vessel disease (SVD) is damage to the small subcortical vessels that supply the deep grey matter and white matter. The vessels affected are small arteries and arterioles that branch from proximal parts of the major cerebral arteries to supply the basal ganglia and thalamus, and penetrating arterioles from the pial surface that reach through the cerebral cortex to supply the white matter. In SVD the perivascular Virchow–Robin spaces around these vessels are frequently widened, and the vessel wall is thickened by fibrosis and the lumen narrowed. Smooth muscle cells in the walls of small arteries and arterioles are replaced by collagen. The tissue supplied by these small vessels may suffer from lacunar infarction or hypoperfusion of blood that renders it ischaemically damaged but not dead. The myelin in regions affected by diffuse SVD is reduced and microglia and macrophages infiltrate the tissue. In a study of brains from elderly subjects in whom pathology of AD and other neurodegenerative dementing conditions was absent, it was only those cases with the most severe extent of SVD that had dementia, and in those cases the severity of dementia within a short time of death was less than in AD.14 Neuroimaging changes of SVD are very commonly seen in scans of older people, and it is likely that this pathology accounts for the largest proportion of VaD.15, 16 Erkinjuntti et al. have proposed specific radiological criteria for subcortical ischaemic VaD.17
A number of types of small vessel disease are seen on imaging (see Figure 2.2). White matter hyperintensities (WMH) appear bright on T2-weighted and FLAIR MRI images (FLAIR images use a technique called ‘fluid attenuation inversion recovery’). When seen in CT, they appear dark and are referred to as hypodense areas or leukoaraiosis. They become more common with increasing age and are associated with cognitive impairment. As they are so commonly seen, the NINDS–AIREN criteria set an arbitrary cut-off that 25% of the total white matter should be affected for a diagnosis of VaD. White matter hyperintensities can be seen around the lateral ventricles, often starting at the poles, and in subcortical areas, although these two types of distribution are thought to have different aetiologies. They are often rated either by eye or using an automated process as mild (punctiform), moderate (beginning confluent) or severe (confluent).18 An unusual but informative form of subcortical small vessel disease occurs in the rare inherited condition of cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), in which a mutation in the Notch3 gene results in deposition of granular material in the walls of small arteries and arterioles, narrowing of vessel lumens and destruction of smooth muscle cells in their walls. Cerebral autosomal-dominant arteriopathy with subcortical infarcts and leukoencephalopathy is discussed further in Chapter 4.
Cerebral amyloid (congophilic) angiopathy refers to the deposition of abnormal protein in the form of twisted beta-pleated sheets in the walls of small arteries and arterioles where it eventually replaces the smooth muscle. It principally affects the leptomeningeal and cortical vessels. In most cases the protein is the amyloid beta protein that also forms the characteristic parenchymal amyloid plaques of AD. Cerebral amyloid angiopathy occurs to a variable extent in almost all cases of AD, but it also occurs in about a third of elderly subjects who do not meet criteria for pathological AD, and in these cases it can be shown to contribute to VaD or VCI. Criteria for clinical diagnosis of CAA have been developed.19, 20 With imaging, a key feature of CAA is microbleeds, which are small dots best seen in T2*-weighted images. They represent leakage of haemosiderin from damaged small blood vessels and, though present at low level in community populations, are much more common in stroke and VaD populations.
As indicated earlier, it is commonly the case that in an elderly brain there may be more than one of these forms of vascular pathology and it can be difficult to define the exact component of mixed pathology that contributes most to VaD or VCI. In the unselected community-based population of elderly subjects that contributed to the Medical Research Council (MRC) Cognitive Function and Ageing Study (CFAS) it was only in cases that had more than one form of vascular pathology that vascular disease contributed to dementia,21 but it is noteworthy that in almost all cases one form of vascular pathology was SVD. Thus, there is good reason to believe that SVD, including lacunes that it gives rise to, is a particularly important contributor to VaD/VCI.
Given the importance of SVD in VaD and VCI it is reasonable to suggest that cognitive dysfunction results at least in part from interruption of axonal connections between one part of the cerebral cortex and another and between the cerebral cortex and deep grey matter. Small vessel disease affects particularly the frontal lobe white matter and the closely related basal ganglia, so it is not surprising that the cognitive dysfunction commonly seen in VaD involves executive activity, which is known to be a function of the frontal lobe (see the following section). It is interesting that in CADASIL the cognitive dysfunction closely resembles that seen in elderly sporadic cases of VaD/VCI.22 Cerebral amyloid angiopathy may also affect white matter function because this is the final destination for the blood flowing in the cortical arterioles affected by this condition. In macroinfarction, lacunes and microinfarction it is loss of neurons that is thought to be important, and there is evidence that after a stroke dementia is more likely if the stroke was a large one, therefore destroying more neurons.
At least eight different sets of diagnostic criteria for VaD have been published, including the Hachinski Ischaemic Scale, the DSM-5 and its predecessors, the ICD 10 and its predecessors, and criteria from the State of California Alzheimer’s Disease Diagnostic and Treatment Centers (ADDTC) and from the NINDS–AIREN.23 In order to determine how effective these criteria are at identifying patients with VaD in life, Wiederkehr et al. conducted a literature search to identify studies which compared lifetime diagnoses using one or other of these criteria against post-mortem pathological diagnoses.23 Only six studies of this kind were identified. Sensitivities across the studies ranged from 20% to 100%, mostly in the range 20–60%, while specificities ranged from 13% to 100%, mostly in the range 60–90%. The NINDS–AIREN criteria tended to have the highest specificity, though lower sensitivity. Sensitivities were usually lower than specificities. One problem with the use of sensitivity and specificity is that they are affected by the prevalence of the disorder. In order to avoid this problem, likelihood ratios (LR) may be calculated. LR+ is the odds in favour of the disease being present given a positive test result, and LR− is the odds against this. Based on standard criteria, Widerkehr et al. found that, taking an LR+>10 and LR‒<0.1 as the threshold of acceptability for a diagnostic test, only the NINDS–AIREN criteria reached this threshold, with ADDTC coming close in one study.23
Two studies comparing the published criteria against clinical judgement found the DSM-IV criteria were most sensitive, while the NINDS–AIREN were the most specific.24, 25 Another comparative study found that all of seven sets of criteria agreed in only 8 out of 124 cases of dementia.26
Thus, even in research centres it is difficult to achieve accurate diagnoses in life. Among the factors that contribute to the difficulty are issues in the diagnosis of the dementia syndrome – that it may not be appropriate to base this on definitions of dementia derived largely from AD, the heterogeneity of pathology in cerebrovascular disease and the lack of agreement on pathological criteria for diagnosis. More research is needed to determine which of the many pathological cerebrovascular processes contribute to VCI and in what way.
Most diagnostic criteria specify that there must be a dementia syndrome and then set out criteria to classify it as cerebrovascular in origin. However, as AD is the predominant form of dementia, this can dominate the clinical concept of the dementia syndrome, e.g. by emphasizing the memory aspect. Studies comparing cognitive changes in AD with those in VaD have found that for similar levels of cognitive deficit VaD patients were more likely to have relatively better-preserved verbal episodic memory and poorer frontal-executive functioning.27, 28 That there should be differences is not surprising in view of the different pathologies and the tendency of Alzheimer’s pathology to start in hippocampal areas involved in processing memories. However, not all studies have confirmed that VCI does affect executive functioning in such a distinct way.29 DSM-5 draws attention to deficits prominent in complex attention (including processing speed) and frontal executive function.
There is widespread awareness that current criteria and definitions are an inadequate basis for the research that is necessary to resolve some of these complex issues, and so an expert group has proposed a set of harmonization standards for the collection of research data in this area.30 These do not constitute new diagnostic criteria but may inform their development.