|Home | About | Journals | Submit | Contact Us | Français|
Cerebrovascular disease plays an important role in cognitive disorders in the elderly. Cerebrovascular disease and Alzheimer’s disease interact on several levels, one important level being the overlap in risk factors. The major vascular risk factors such as diabetes and impaired glycemic control, hypertension, obesity and hyper- or dyslipidemia have been associated both with Alzheimer’s disease and vascular dementia. The purpose of this review is to consider the context in which vascular dementia is diagnosed, place the pathophysiological consequences of cerebrovascular disease on cognition in the context of clinical and pathological Alzheimer’s disease, and then to consider the evidence for the role of major vascular risk factors in late-life cognitive impairment, changes in brain imaging and neuropathological changes. Midlife diabetes mellitus, hypertension and obesity are established risk factors for clinically defined Alzheimer’s disease as well as vascular dementia. The basis for these relationships could either be that the risk factors lead to microvascular brain disease, promote Alzheimer pathology or both. The associations of late-life onset diabetes mellitus, hypertension and obesity with cognitive impairment are either attenuated or reversed. The role of vascular risk factors in midlife should be the focus of public health efforts to reduce the burden of late-life cognitive impairment.
Stroke and cerebrovascular disease play important roles in late-life cognitive decline, but their relationship to Alzheimer’s Disease (AD), the most commonly diagnosed etiology of late-life cognitive impairment and dementia, has been enigmatic. Up until recently, the pathophysiologies of the two conditions have been regarded as distinct. However, “vascular” risk factors have also been shown to be associated with AD, further blurring the distinction between cerebrovascular and Alzheimer-type clinical diagnoses. The purpose of this review is to first consider the context in which cognitive impairment of vascular origin is diagnosed, place it in the context of clinical and pathological AD, and then to consider the evidence for the role of major vascular risk factors in late-life cognitive (after age 65 years) impairment. This is an area of growing interest because treatment of vascular risk factors, already important in their own right, would gain increased urgency for prevention of late-life cognitive impairment.
In order to avoid confusion, some conventions for terminology should be clarified. Clinical diagnoses and pathological diagnoses must have unique designations because of the lack of perfect correspondence between the two. Hereafter, whenever we refer to Alzheimer’s disease, we will specify whether we mean clinically diagnosed dementia due to AD (AD*c) or pathologically-defined AD (AD*p) based on the presence of neuritic plaques and neurofibrillary tangles in isocortex. The term “vascular dementia” (VaD) will refer to a clinical diagnosis of dementia in which CVD is presumed to be the etiology, whereas cerebrovascular pathology (CVP) refers to the pathologically observed processes of infarction and ischemia. The combination and possible interactions of CVP and AD*p are ultimately of the most interest. The label “Alzheimer’s disease with cerebrovascular disease”  captures the overlap between the two, but doesn’t do it in a very artful way. This review will focus on these two etiologies of cognitive impairment and their combination.
Stroke is one of the most common neurological diseases of advancing age. Strokes cause cognitive and motor impairments both acutely and persistently. Survivors of stroke have a far higher rate of incident dementia than age-matched peers [2–5]. A history of stroke at any point in the past is a risk factor for the development of cognitive impairment [6, 7], MCI  and dementia [9, 10]. MR imaging in population-based cohorts of elderly individuals has shown that 20% or more have clinically silent infarcts . The burden of imaging-detectable cerebral infarction is even larger than clinically overt stroke . Prospective observational studies show that numbers of lacunar infarcts are associated with risk for future cognitive impairment and dementia [12, 13]. There is evidence that the burden of macrovascular  as well as microvascular pathology independently accounts for cognitive impairment [15–17], even after taking AD*p into account.
Despite this wealth of observation, CVP as a cause of cognitive impairment may still be under-appreciated for several reasons. It has been very difficult – both in life with imaging and at autopsy neuropathologically – to quantitate the burden of CVP. There is no validated scheme for assigning different grades to CVP burden though progress is being made [16, 18, 19] in developing pathological ratings that reflect increasing burden of CVP pathology. Not only is there the problem of properly accounting for the role of small but strategically placed infarcts versus the total burden of ischemic pathology, but quantitating microscopic evidence of infarction is prohibitively labor-intensive and therefore, most neuropathological examinations use only qualitative estimates of CVP. There is a nagging concern that current neuropathological methods may tend to undercount ischemic lesions. Analyses that attempt clinical-pathological correlations with CVP, AD*p and their overlap are inherently more complex and require larger numbers of subjects than ones that look at AD*p alone.
To talk about CVP and dementia, it is necessary to put AD*p in perspective. While the focus of this review is to highlight why AD*p cannot be considered in isolation from CVP, AD*p is nonetheless a major neuropathologic process in late-life dementia [18, 20]. The loss of neurons, synapses, dendrites, axons in heteromodal association isocortices is the proximate cause of cognitive impairment from AD*p . Neurodegeneration in AD*p is closely linked to the formation of neurofibrillary tangles (NFT) in neurons in heteromodal isocortex . Modulators of the AD*p neurodegenerative cascade that could have either genetic or environmental bases affect the rate of accumulation of brain pathology or the time when the pathological changes express themselves clinically.
CVP could modulate the appearance of cognitive impairment in conjunction with AD*p by one of several mechanisms. If a certain volume of cortical or subcortical grey matter undergoes infarction, there will be a reduction in intact brain volume, loss of synapses and reduced connectivity between remaining neurons, and hence brain reserve. An increasing burden of CVP would either accelerate the appearance of cognitive impairment or increase its severity at a particular level of AD*p. There is substantial evidence that as the burden of CVP increases, the abundance of AD*p will be less, for a given severity of dementia [14, 17, 22–24]. The interaction of CVP and AD*p might account for the generally lower burden of AD*p for a given level of cognitive impairment after age 75–80 yrs . There appears to be no interactive effect on dementia severity between the burden of cerebral infarcts and the amount of AD*p [19, 25]. Conceivably, microscopic CVP could accelerate the neurodegenerative process in the AD*p cascade, but there is no direct evidence that such an interactive mechanistic pathway exists.
Postmortem studies of patients dying with dementia that have compared clinical diagnoses with neuropathological findings have shown that the clinical diagnosis of VaD is very insensitive for predicting CVP [26–28], and furthermore, AD*c is often associated with some degree of CVP [17, 29]. Problems with the diagnosis of VaD raise questions about the validity of the diagnosis, particularly when neuroimaging is not available. The clinical mis-recognition of CVP pathology in dementia patients has major implications for understanding the role vascular risk factors in the epidemiology of dementia. The presence of CVP in AD*c means that epidemiological studies that claim to study observations on risk factors for AD*c are actually targeting a dementing illness that includes some element of CVP. In other words, when an epidemiological study asserts that a risk factor is “associated with AD,” that really means “associations with a dementing illness that mostly includes persons with AD*p and some amount of CVP.” Conversely, factors reported to be associated with VaD may not be as specific for a dementing illness due only to CVP. An “association with VaD” means that the associations are with persons who have a higher burden of CVP but probably have coexistent AD*p. Therefore, in the absence of information about underlying pathology, investigations of risk factors using epidemiologic methods are limited in their inferences about the underlying dementia etiology.
A number of mid-life risk factors that have been traditionally viewed as causing vascular disease including the components of the metabolic syndrome –diabetes and impaired glycemic control, hypertension, hyper- or dyslipidemia and obesity – are associated with an increased incidence of cognitive impairment and dementia. This review focusses on individual vascular risk factors and their impact on structural changes on imaging and neuropathology.
Diabetes mellitus (DM) is a common condition from middle age and beyond. By age 65 about 20% of the population carries a diagnosis of DM, mainly type 2 diabetes [30, 31]. DM affects the kidney, heart, peripheral nerves and retina, in addition to the brain.
Alterations in midlife glycemic control and DM almost certainly have greater consequences for cognitive function than late-onset DM presumably because of the longer duration of exposure, although a more metabolically aggressive disease might also occur at younger ages. A metaanalysis  observed that the magnitude of risks for dementia were generally higher when DM was diagnosed at midlife versus later life.
The evidence linking DM to late-life cognitive impairment includes its associations with cognitive decline in middle age [6, 33–37], cognitive decline in later life, mild cognitive impairment [38, 39], AD*c [40–48] and VaD [40, 49–51]. In general, associations have been stronger for VaD than for AD*c. Not all studies show an association between dementia and DM , however, perhaps because of the age when DM was ascertained.
The nature of the cognitive deficits associated with DM provides some clues as to DM’s brain targets. DM is associated with impairment of both amnestic and non-amnestic cognitive functions, but the association with non-amnestic dysfunction has been shown to be of greater magnitude [38, 39]. Less amnestic and greater non-amnestic cognitive dysfunction would be more consistent with involvement of white matter pathways, basal ganglia or thalamus, which in turn are loci expected to undergo infarction as a result of CVP.
DM is associated with structural brain changes that can be detected on imaging [53–57]. In a metaanalysis, the major changes on imaging associated with DM were brain atrophy, lacunar infarcts and to a less extent, white matter hyperintensities . In the Atherosclerosis Risk in Communities (ARIC) study that included persons in late middle age (50–73 yr), there was a 4 to 7% increased risk of ventricular enlargement for each 10mg/dL of fasting blood sugar elevation . In the Cardiovascular Health Study (CHS), DM was associated with brain atrophy only in women. The CHS population was 65 and older . Perhaps the less consistent associations in the CHS are a result of the older age of the subjects.
Neuropathological studies of diabetics also show an increased number of brain microinfarcts and lacunar infarcts compared to non-diabetics [60–63]. Several studies have shown that diabetics with dementia have a lower burden of β-amyloid pathology [60–62].
DM could cause also brain injury through non-CVP mechanisms. Hyperglycemia and impaired control of insulin homeostasis might have a direct effect on brain β-amyloidosis [64–67]. Insulin degrading enzyme (IDE), an enzyme that is involved in insulin trafficking, is also a key enzyme that degrades β-amyloid. With increases in peripheral insulin levels in midlife in type 2 diabetics, brain insulin levels could be raised . Insulin might compete with β-amyloid for binding to IDE; thus, elevated insulin levels would lead to decreased β-amyloid clearance, and raised β-amyloid levels in the brain.
Because most patients with DM are treated with either oral hypoglycemic agents or insulin, it may be very difficult to determine whether treatments alter clinical-pathological correlations in DM. Severity of DM is more likely to be associated with the use of insulin, whereas hyperglycemia that doesn’t exceed 125 mg/dL might not be treated with medication at all. Since more severe DM will almost always be treated with insulin, the unique contributions of DM severity and insulin use to structural and neuropathological changes may not be resolvable. With insulin use, the impact of DM severity versus episodes of hypoglycemia must also be considered in evaluating dementia risk.
Based on imaging and neuropathological evidence, microvascular disease is a very strong candidate mechanism to account for the impact of DM on late-life cognitive impairment. DM as a microvascular disease could influence the course of AD*p by decreasing the threshold at which AD*p produces clinical effects, even if DM itself had no direct impact on AD*p-mediated neurodegeneration. If this were the case, then for a given level of cognitive impairment, persons with diabetes should have less AD*p, which in fact has been observed [60, 61]. An alternative hypothesis is that the metabolic or ischemic injury caused by DM facilitates β-amyloid pathology, tau pathology or both. In this alternative model, DM would be associated with more AD*p. The evidence so far does not support this latter model.
Hypertension (HT) affects a majority of middle-aged and elderly populations . While there are various ways to represent abnormal BP such as history of hypertension, systolic BP, diastolic BP, pulse pressure or mean arterial pressure, most of the data we will cite simply used either history of hypertension, systolic BP or diastolic BP. HT is a well-known major risk factor for ischemic heart disease, peripheral vascular disease, chronic kidney disease and CVP, via both ischemic and hemorrhagic mechanisms. HT in midlife is associated with later life cognitive impairment and dementia [69–71]. Associations are generally stronger for VaD than AD*c [72–77].
Interpretation of the effects of HT in later life is confounded by the rising prevalence of hypotension in late-life. In late-life both HT and hypotension may cause brain injury. Orthostatic hypotension becomes common with advancing age, and hypotension is also deleterious to the brain [78–80]. Even among normotensives, there may be an admixture of borderline hypertensives and hypotensives, making it quite difficult to identify the deleterious effects of HT in the very elderly. Several studies have failed to show an association of HT with poor cognition when assessed in late-life [78, 79, 81–84].
Different classes of antihypertensive medications have shown associations with lower rates of dementia in observational studies [85–87]. Similarly, clinical trials have demonstrated that treatment of persons with HT with several classes of antihypertensive medications reduces the incidence of dementia [88–91]. It is not clear which class or classes of antihypertensive provides the greatest protective effects for the brain. Duration of antihypertensive therapy also has an impact on associations of hypertension with dementia; especially in younger persons, those on antihypertensive therapies for longer periods of time had reduced rates of incident dementia . Because therapy has an impact on blood pressure measurements themselves and also appears to impact cognitive function, naturalistic observations with either imaging or neuropathology are confounded by treatment.
In imaging studies involving both middle-aged and elderly persons, HT is more closely associated with white matter hyperintensities (WMH) than atrophy [53, 54, 93–99]. HT is the most consistent correlate of excess burden of WMH [97, 99, 100]. In the cross-sectional findings of the ARIC study, HT was not associated with increased ventricular volume, but there were strong associations with increasing burden of WMH . In the longitudinal assessments of the ARIC cohort, systolic BP measured cumulatively over 14 years was strongly associated with progression of WMH. BP measurements from early years of follow-up showed stronger associations with WMH burden than later ones .
The effects of systolic versus diastolic BP on brain structure may differ. In elderly subjects (age 72 ± 7 yrs) from the Rotterdam Scan study, systolic blood pressure was not associated with brain atrophy. However, diastolic BP exhibited a J-shaped relationship to brain atrophy. Subjects with high and low diastolic BP had greater brain atrophy than subjects with diastolic BPs in the 65–74 mm Hg range. In addition, in a subset of subjects with BPs measured 20 years earlier, those with elevated diastolic BPs in midlife but who were not treated had greater brain atrophy than non-hypertensives. Subjects with elevated diastolic BP in midlife who were treated did not show any brain volume loss compared to non-hypertensives. Both elevated systolic and elevated diastolic BP were associated with WMH burden. The association between 20-year change in diastolic blood pressure and subcortical WMH was also J-shaped, indicating that both declines and elevations in BP over time were related to increasing burdens of WMH .
Several studies have found that midlife hypertension was associated with greater burdens of NFT and neuritic plaques in late life [103, 104], but medication effects make interpretation complex. Neuropathological studies that have compared subjects with and without antihypertensive medications have shown that medicated patients had either less cognitive impairment or less AD pathology than the un-medicated hypertensive peers . Persons receiving antihypertensive therapy had less AD*p than normotensives.
Despite the likelihood that the effects of HT on the brain are microvascular and are similar to that seen in other microvascular beds such as in the heart or kidney, the neuropathological evidence suggests that HT has direct effects on AD*p that are independent of infarction. These findings suggest that HT could alter β-amyloid or abnormal tau production or clearance. A final speculation is that of reverse causality: AD*p in brainstem structures  might promote HT.
Rising levels of obesity across all age groups is a major public health issue. Obesity is usually defined by body mass index (BMI), but abdominal girth has been shown to have greater predictive ability for disease outcomes such as heart disease . Obesity in midlife has been associated with later life cognitive impairment [106–108]. This association persists even when DM, HT and hyperlipidemia are included in the analytic models. The observation suggests that midlife obesity is associated with cognitive impairment in some ways that are unique from the other related risk factors. Similarly, obesity could promote microvascular disease by mechanisms that are distinct from DM and HT. Various circulating factors released by adipose tissue could play a role, perhaps by promoting inflammation. Increased fat mass, particularly visceral fat, elevates blood levels of inflammatory cytokines such as TNFα and IL6. Elevations of circulating levels of these factors could alter endothelial function, which in turn could lead to insulin resistance . Leptin, a hormone released by adipose tissue, regulates lean body mass, complements insulin action in the peripheral circulation, decreases brain β secretase levels, and modulates β-amyloid turnover. In obesity, chronically elevated leptin levels result in leptin resistance and an inability to regulate weight. In the late-life, consistent with the decreased BMI that occurs prior to dementia, leptin levels decrease and are inversely associated with dementia . Similarly, adiponectin, a complement related protein produced by adipose tissue, is a risk factor for coronary heart disease. The associations with dementia are yet to be determined.
Studies that have examined weight, body-mass index or other measures of adiposity in late-life have found either no association with dementia, or a protective effect [111–114]. Obesity may appear to be a protective factor for impaired cognitive function in late-life because of the competing effects of the associations of weight loss with illnesses such as cancer. Furthermore, some studies have shown that persons destined to develop dementia lose weight in the few years preceding the diagnosis. Persons with dementia are highly likely to be thinner than non-demented peers [111–114]. It is uncertain whether reverse causality could also be acting here, whereby subtle cognitive or behavioral changes that occurred prior to dementia altered dietary habits.
There were no significant associations between elevated BMI and structural brain changes in either ARIC  or an earlier analysis of the CHS  cohort. A more recent study of a smaller number of CHS participants (mean age = 77 yrs) observed reductions in grey matter volume in persons with high BMI (>30 kg/m2), even after controlling for DM . In younger persons there also is an association of obesity with reduced brain volume [116, 117]. To our knowledge, there are no neuropathological studies that have assessed the role of midlife obesity on neuropathology in late-life. Because late-life obesity appears to be protective for dementia, understanding the competing forces of midlife and late-life obesity may be very difficult.
Whether obesity plays an independent role in the genesis of dementia, AD*p or CVP, separate from DM and HT is not clear. Midlife obesity, to the extent that it promotes DM and HT, has adverse effects on the brain, regardless of whether these effects are direct or indirect.
The relationship between elevated lipid levels and dementia is uncertain, as many studies fail to show associations between elevated LDL cholesterol or triglycerides and dementia, cognitive impairment or abnormalities on brain imaging. A few studies have shown that midlife hypercholesterolemia was associated with later life MCI or dementia [77, 118, 119]. In contrast, the Framingham study failed to observe an association between midlife cholesterol levels and AD . The ARIC study involving, mainly middle aged subjects, found that elevated LDL-C was not associated with cognitive impairment . When lipid levels are measured in later life, the association is even more uncertain . Elevated levels of total cholesterol were protective of dementia in late-life . Low levels of cholesterol were associated with dementia [123, 124], as well as a number of serious systemic diseases, in particular cancer and malnutrition, so that an admixture of persons with high and low cholesterol in epidemiological studies may obscure true associations between high cholesterol levels and dementia in late-life. Once again, processes that lower lipid levels and potential survival biases compete with hyperlipidemia and confound analyses.
Although hypercholesterolemia is a potent risk factor for cardiac disease, its impact on cerebrovascular disease has been inconsistent. In neither Northern Manhattan Study , ARIC  nor CHS  was there an association between HDL-C or LDL-C and either clinical strokes or infarctions on imaging. Why elevated cholesterol is not as strongly linked to cerebrovascular disease as it is to cardiovascular disease is a mystery that is beyond the scope of this essay.
Cholesterol or triglyceride level have not been associated with larger ventricular size or increased white matter hyperintensities in the ARIC cohort . In the CHS cohort, high HDL and low LDL levels were associated with progression of WMH . A few studies that investigated the relationship between lipid levels in late life and neuropathology have claimed associations between neuritic plaques and elevated HDL-C  or total cholesterol and LDL-C levels . There are no studies of midlife lipid levels and neuropathology. Given the biphasic nature of the relationship between lipids and dementia, it may be midlife levels of HDL-C or LDL-C that are most relevant.
It is not clear at this point what specific role hyperlipidemia plays in the development of dementia, AD*p and CVP. Because of the multiple demonstrations of either neutral or protective effects of hyperlipidemia on cognition and brain structure, this risk factor does not offer many insights into the mechanisms of cognitive impairment.
Late-life dementia due to AD*p is strongly associated with educational attainment, occupation and socioeconomic status (SES) [129, 130]. CVP and VRF are also moderately strongly associated with SES [131, 132]. DM, HT, obesity, and heart disease itself are all more likely in persons of low educational attainment, the usual proxy for SES. Persons of lower SES are more likely to have a poor diet, to be overweight, to smoke, and to be sedentary, all of which contribute to higher burdens of VRF. Perhaps, VRFs are associated with cognitive impairment because they stand for an array of dysfunctional health behaviors that are more important than the VRFs themselves. It seems likely that dysfunctional health behaviors in persons with lower education or lower SES promote DM, HT, and obesity. The description of the associations of low intelligence test scores in childhood with late-life VaD rather than AD*c  in a Scottish cohort suggests that the impact of low education and low SES on brain integrity might actually begin in childhood. It is possible that SES shares variance with brain reserve, and as a consequence different levels of brain reserve have an impact on the threshold for the appearance of dementia for a given level of AD*p or CVP.
VRFs, when present in midlife, exert a consistent deleterious effect on late-life cognition and dementia. For DM, HT and obesity parallel effects on subclinical infarction, cognition and brain structure support their distinct roles in the pathogenesis of late-life cognitive impairment. These disorders might either promote microvascular disease, AD*p or both. In contrast, while midlife hyperlipidemia may be associated with late-life cognitive impairment, hyperlipidemia does not seem to alter brain structure and is not clearly associated with an increased risk of stroke. The dissociation raises some doubts on how relevant hyperlipidemia is for cognitive impairment. In later life, the role of the VRFs is much more complex. For all four conditions when measured in late-life, their associations with cognitive impairment are either attenuated, neutral or reversed. The inverse of three of them – hypotension, low cholesterol and triglyceride levels and cachexia – are associated with serious systemic illnesses that worsen survival. The relationship of SES to VRF implies that the genesis of VRFs has its roots in childhood social class, education and culture.
The evidence from imaging studies suggests that treatment of VRFs in persons who already have late-life cognitive impairment may be too late to be curative, but there still could be lesser clinically important benefits. Treatment of VRFs in persons with both AD*c and VaD should be considered because of the etiological overlap of AD*p and CVP in both.
However, VRF can be treated in midlife, and that should be grounds for optimism. Public health initiatives to treat the VRFs in midlife are already well-justified on the grounds of preventing cardiovascular disease. Despite the cardiovascular imperatives, rates of effective treatment of DM, HT, obesity and hyperlipidemia are rather low [30, 31]. Perhaps fear of cognitive impairment would spur more aggressive attempts to treat VRF in midlife.
Support: U01 AG 06786 (Mayo Alzheimer’s Disease Patient Registry), P50 AG 16574 (Mayo Alzheimer’s Disease Research Center) and K01 AG028573 from the National Institute on Aging
Disclosures: DSK serves on a Data Safety Monitoring Board for Lilly Pharmaceuticals, and is an investigator for clinical trials sponsored by Elan Pharmaceuticals, Forest Pharmaceuticals and Baxter Healthcare. He is an associate editor of Neurology, and receives compensation for editorial activities. The other authors have no relationships to disclose.