Population studies of sporadic cerebral amyloid angiopathy and dementia: a systematic review

A systematic review of population-based neuropathological studies of ageing and dementia in older people was published in 2006 [39]. Six studies were identified as being fully population-based – the Hisayama Study (Japan), Vantaa 85+ (Finland), the Cambridge City over 75 Cohort (CC75C; England), the Honolulu-Asia Aging Study (HAAS; USA), the Cache Country Study (USA) and the MRC Cognitive Function and Ageing Study (CFAS; England and Wales). To identify articles for the current review, PsychInfo (1806–April Week 3 2008), OVID MEDLINE (1950–April Week 2 2008) and Pubmed (searched 21 April 2008) databases were used to search for papers employing any of these six study populations to assess CAA. Study titles were searched for as keywords, along with 'amyloid angiopathy'. When no study could be found using the key term 'amyloid angiopathy', it was replaced by 'neuropathology' and the relevant articles were read and searched to determine if CAA was assessed. To ensure that no further prospective population-based clinicopathological studies had been published since Zaccai et al. [39], a search was conducted using the same databases detailed above using the terms 'population' and 'amyloid angiopathy' or 'pathology'. No new studies were identified.

To compare CAA prevalence rates in population-based studies with those in studies of selected samples, we conducted a systematic review of studies in which CAA prevalence was determined relative to dementia status.

To identify studies using selected samples the PsychInfo (1806–April Week 3 2008), OVID MEDLINE (1950–April Week 2 2008) and Pubmed (searched 21 April 2008) databases were searched. One thousand three-hundred and seventy-three studies were identified using the search term 'amyloid angiopathy'. Titles and abstracts from these studies were read and studies were excluded if they did not assess human cases, the prevalence of CAA in the demented and/or non-demented specifically related to hereditary/familial CAA, reported a sample with a n less than 10, reported CAA prevalence relative to a condition other than dementia or only a feature of dementia (e.g., Parkinson's disease or APOE genotype), selected cases based on the presence of CAA, if they clearly reported the same cases as a previously selected study [e.g., [30, 40, 41]], if analyses were not neuropathological (e.g., MRI), or had been included as one of the population-based studies described above. The search was restricted to the English language. When a decision could not be made as to whether the article should be excluded, full articles were read. When articles were read, any cited articles that were not identified during the search were included.

Five of the six population-based studies reported CAA prevalence rates [13, 14, 35, 42, 43], of which four calculated these rates relative to dementia status (demented and/or non-demented) prior to death [all but [43]]. The findings from the four studies identified are reviewed below and summarised in Additional file 1. None of the studies assessed sex differences and only one study provided prevalence relative to age-groups [14] and thus prevalence relative to age and sex factors are not shown.

Thirty-eight studies were identified for review and are summarised in Additional file 2. It can be seen in Additional file 2 that the prevalence of CAA, regardless of severity, is higher in the demented (32–100% regardless of subtype or 47–100% for AD-only) than the non-demented (0–77%) in selected samples. The studies that assessed severe CAA reported prevalence rates of 0–50% in the demented (6–50% for AD-only) and 0–11% in the non-demented. Only five of the thirty-eight selected studies tested for an association between CAA and dementia [34, 37, 40, 44, 45]. All of those that did reported that cases with dementia had a significantly higher prevalence than those without, or that CAA prevalence correlated significantly with dementia. The only exception was the study by Jellinger and Attems [40] in which this was true for capillary CAA but not other forms of CAA.

There was greater variability in CAA prevalence rates in selected-sample studies than population-based studies. This may be due to differential bias between selected samples and the paucity of clinical information prior to death in most studies.

Estimates of CAA prevalence in the population studies were lower than in those from selected samples. This may be due to diagnostic differences (primarily neuropathological methods in studies of selected samples as compared to clinical methods in population-based studies) and younger populations being assessed, perhaps including some cases of hereditary rather sporadic CAA. Only eight studies of selected samples employed a clinical diagnosis of dementia rather than a neuropathological confirmation of AD or VaD. This is an important difference from population-based studies, all of which employed a clinical diagnosis, as neuropathological and clinical AD classification methods correspond imprecisely [46, 47].

The mean age of the cohorts in selected-sample studies was between 69–91, with participants in their 40s and 50s commonly included [e.g., [44, 48‐50]]. These studies on selected samples therefore included much younger cases than population-based studies. Given that population-based studies most closely reflect the population at risk of disease, these findings highlight their important contribution to the understanding of pathological correlates of clinically determined dementia relevant to populations and patients. Evidence relating to the pathology of dementia needs to be obtained from both population-based as well as from selected samples.

Both the population-based studies and those on selected samples employed various stains to assess CAA, including Congo red, thioflavin-S, anti-Aβ immunohistochemistry and Weigert's haematoxylin. Some studies used multiple stains – either different stains for different cases [e.g., [51]] or anti-Aβ immunostaining to validate Congo red-positive cases [e.g., [34]]. To the authors' knowledge, no study has directly compared the sensitivity and specificity of these staining methods in relation to the measurement of CAA. Haglund and Englund [49] assessed a small sample (n = 10) with both Congo red and anti-Aβ immunohistochemistry. They reported that for 8 of the 10 cases, CAA severity appeared similar with both methods (Aβ immunohistochemistry perhaps giving slightly higher severity grades), while the other 2 cases showed little Congo red positivity but strong labelling of vascular Aβ.

It is not only the staining method that may produce variability between studies, the pre-treatment methods and sampling strategies also contribute [52, 53]. The selection of parenchymal and/or leptomeningeal samples is also important, as CAA severity is usually higher in the later [54]. The choice of cortical region(s) and the grading system employed in the scoring of CAA in terms of its scale (e.g., 0–3 or 0–4) and definition of severity, would have also contributed to inter-study variability.

At the population level, it appears as though CAA prevalence rates are consistently higher in the demented than the non-demented, suggesting that CAA is an important dementia-related abnormality in a population context, particularly severe CAA. This finding is pertinent in light of recent reports of overlap in the prevalence rates between demented and non-demented cohorts of other dementia-related abnormalities such as plaques and tangles [14, 47].

Determination of the presence and severity of CAA may aid in neuropathological assessment that more closely resembles the clinical assessment of dementia. The MRC-CFAS study reported the OR for risk of a clinical dementia diagnosis to be higher for severe CAA (OR = 9.3, 95%CI 2.7–41.0) than either neocortical neuritic plaques (OR = 5.0, 95%CI 1.2–29.8) or neocortical tangles (OR = 4.6, 95%CI 1.5–15.8) [14]. Future work should elucidate whether the inclusion of CAA in the neuropathological assessments of dementias such as AD improves the specificity and sensitivity of classification relative to clinical dementia during life. Further, given that CAA is a potential discriminator of dementia from normal ageing, imaging methods that detect CAA in living individuals may aid in the differential diagnosis of clinical dementia syndromes [5, 12, 55]. CAA is a potential predictor of the development of dementia during life, for which treatment and management strategies might be developed. The presence of CAA in the living non-demented may also indicate a specific syndrome such as Vascular Cognitive Impairment No-Dementia (VCI-ND) [56] – this also needs to be investigated in the future.

Some studies [6, 34, 43, 57‐60] but not all [14, 48, 61] have found that increasing age in the elderly is a risk factor for CAA. In a community sample of 100 individuals 50–91 years, Mastaglia et al. [62] found CAA prevalence to increase with age to a maximum in those over 90. As yet there is no evidence that the prevalence of CAA varies with sex [7, 43]. Cerebrovascular disease (e.g., atherosclerosis) is another possible risk factor for CAA. It has been suggested to affect the efficiency of Aβ removal by perivascular drainage, leading to CAA [63]. However, hypertension, diabetes mellitus and hyperlipidemia do not appear to be risk factors for CAA [60].

The ε4 allele of the APOE gene has been reported to be a risk factor for CAA as it is for AD [64‐68] and this was seen in two of the population-based studies reviewed here [13, 35]. It has been suggested that the association between the APOE ε4 allele and CAA is due to the ε4 allele being related to vascular rather than parenchymal accumulation of Aβ [64]. The APOE ε4/ε4 genotype has been associated with CAA-related inflammation [69], while the ε2 allele has been reported to be a risk factor for CAA-related haemorrhage in both those with and those without AD [65, 70].

Studies of clinic and/or necropsy cohorts are valuable for determining the implications of CAA at a population level. For instance, these studies have found that many neurological events induced by CAA can themselves compromise neurological integrity and cognitive function. For example, CAA increases vessel wall fragility and increases the risk of cortical and subcortical intracerebral haemorrhage [5, 71], carrying a high risk of poor neurological and cognitive outcome [72]. CAA may impair blood flow, leading to ischemic damage to the cerebral cortex and white matter, which is also associated with neurological and cognitive impairment [73‐75]. CAA is associated with failure of drainage of interstitial fluid from cerebral white matter in AD and the pathogenesis of leukoaraiosis [76]. Further, the Aβ peptide that accumulates in CAA can provoke inflammation/vasculitis in a small subset of individuals [69, 77] with both microglia and T-cells involved [78]. CAA-related inflammation has also been proposed to be related to severe cognitive decline due to circulatory dysfunction [69, 78]. Thus, patients who develop haemorrhage, ischaemia, white matter damage or inflammation as complications of CAA are at risk of further cognitive impairment.

The type of vessel involved and distribution of CAA may differentially affect cognition. It was reported that capillary CAA is more closely associated with impaired cognition than is CAA of larger blood vessels [30, 79]. Further, although CAA is most prevalent in the posterior regions of the cerebral hemispheres, the association between CAA and clinical dementia has been reported to be strongest for frontal CAA [37].

The four population-based studies reviewed here employed relatively crude measures of CAA. There were varying attempts to take into account the interactions with a wide range of other dementia-related pathologies and the severity of CAA. No study took into account the distribution of CAA, the presence of inflammation and the type of vessel affected. All these factors have been reported to influence the association with cognition [e.g., [30, 37, 79, 80]].

Population-based studies, as with studies on selected samples, face some important stumbling blocks in the investigation of CAA. The distribution of CAA is variable [4] and its detection critically dependent of the extent and distribution of histological sampling. There is likely to be an under-diagnosis of CAA even in severe cases [18]. There is currently no consensus as to how to sample for or detect CAA or grade its severity. Full evaluation of CAA is most satisfactorily assessed by isolating cerebral and leptomeningeal vessels from the brain and staining for amyloid with Thioflavin [22, 76]. A standard consensus method for detecting and classifying CAA would greatly facilitate future population-based multi-centre studies of CAA and such plans are underway.