Article Text

Research paper
APOE associations with severe CAA-associated vasculopathic changes: collaborative meta-analysis
  1. Kristiina Rannikmäe1,
  2. Rajesh N Kalaria2,
  3. Steven M Greenberg3,
  4. Helena C Chui4,
  5. Frederick A Schmitt5,
  6. Neshika Samarasekera1,
  7. Rustam Al-Shahi Salman1,
  8. Cathie L M Sudlow1,6
  1. 1Division of Clinical Neurosciences, Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK
  2. 2Institute for Ageing and Health, Newcastle University, Newcastle upon Tyne, UK
  3. 3Department of Neurology, Stroke Research Centre, Massachusetts General Hospital, Boston, Massachusetts, USA
  4. 4Department of Neurology, University of Southern California, Los Angeles, California, USA
  5. 5Department of Neurology, Sanders-Brown Center on Aging, University of Kentucky, Lexington, Kentucky, USA
  6. 6Institute of Genetics and Molecular Medicine, University of Edinburgh, Edinburgh, UK
  1. Correspondence to Professor Cathie Sudlow, Division of Clinical Neurosciences, Centre for Clinical Brain Sciences, University of Edinburgh, Western General Hospital, Crewe Road, Edinburgh EH4 2XU, UK; cathie.sudlow{at}


Objectives Cerebral amyloid angiopathy (CAA) is associated with lobar intracerebral haemorrhage (ICH). While only the ε4 allele of the apolipoprotein E (APOE) gene is associated with the presence of CAA, both APOE-ε4 and ε2 are associated with lobar ICH. The generally accepted explanation is that APOE-ε4 promotes vascular amyloid deposition, while APOE-ε2 promotes progression to severe CAA with associated vasculopathic changes that cause vessel rupture and ICH. We assessed the evidence for these allele-specific effects.

Methods We systematically identified published studies with data on APOE genotype and histopathological assessment of postmortem brains for CAA severity. We obtained unpublished data from these for meta-analyses of the effects of ε4-containing (ε4+) and ε2-containing (ε2+) genotypes on progression to severe CAA.

Results Of six eligible studies (543 eligible participants), data were available from 5 (497 participants, 353 with CAA). Meta-analyses showed a possible association of ε4+ genotypes with severe CAA (ε4+ vs ε4−: severe vs mild/moderate CAA, OR 2.5, 95% CI 1.4 to 4.5, p=0.002; severe vs moderate CAA, OR 1.7, 95% CI 0.9 to 3.1, p=0.11). For ε2+ versus ε2− genotypes, there was no significant association, but the very small number of participants with ε2+ genotypes (22) precluded reliable estimates.

Conclusions We found a possible association of severe CAA with APOE-ε4 but not APOE-ε2. However, our findings do not exclude a biologically meaningful association between APOE-ε2 and severe CAA. Further work is needed to elucidate fully the allele-specific associations of APOE with CAA and their mechanisms.

  • Amyloid
  • Cerebrovascular Disease
  • Apolipoproteins
  • Meta-Analysis
  • Systematic Reviews

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Sporadic cerebral amyloid angiopathy (CAA) is characterised by deposition of amyloid β-protein in leptomeningeal and cortical blood vessels, with a prevalence in population-based autopsy studies of 20%–40% in non-demented and 50%–60% in demented elderly people.1 Neuropathological case control and cross-sectional studies, as well as the increased incidence of intracerebral haemorrhage (ICH) in patients with Alzheimer's disease, suggest that CAA causes lobar ICH.2 ,3

CAA is thought to be responsible for up to a third of all ICH in elderly people.4 The prevalence of ICH in cases with CAA in a recent systematic review was about 11%.2 It is unknown why only a few people with CAA pathology develop ICH, but it seems likely to involve biological pathways additional to and distinct from those involved in vascular amyloid deposition. Cases of CAA with ICH have a greater proportion of amyloid-laden blood vessels,5 and more often demonstrate severe CAA with associated vasculopathy (figure 1).5–8

Figure 1

Proposed theory about associations between APOE genotype and CAA phenotype. APOE, apolipoprotein E; CAA, cerebral amyloid angiopathy; ε4+: ε4 allele-containing genotypes; ε2+: ε2 allele-containing genotypes; The currently generally accepted explanation is that while APOE-ε4 promotes vascular amyloid deposition, ε2 promotes progression to severe CAA with associated vasculopathy leading to vessel rupture and intracerebral haemorrhage (ICH).

Apolipoprotein E (APOE) genotype is associated with both histopathologically-confirmed CAA and CAA-related clinical phenotypes, including lobar ICH. Our recent systematic review demonstrated a dose-dependent association between histopathologically-confirmed CAA and ɛ4- but not ɛ2-containing genotypes.9 However, a recent large-scale genetic association study found that both ε4- and ε2-containing genotypes were associated with lobar ICH, particularly when attributed to CAA.10 Furthermore, APOE-ɛ2 but not ɛ4 predicted initial haematoma volume, haematoma expansion, increased mortality and poor functional outcome of lobar ICH.11 ,12 The generally accepted explanation for these findings is that while APOE-ε4 promotes vascular amyloid deposition, ε2 promotes progression to severe CAA with associated vasculopathy leading to vessel rupture and ICH (figure 1).5 ,13 We aimed to assess the evidence for this hypothesis through a systematic review and meta-analyses of all available, relevant, published, histopathological studies.


Study identification and inclusion/exclusion criteria

We sought all studies of adult humans published in any language which had conducted both APOE genotyping and histopathological assessment for CAA, including assessment for severe CAA with associated vasculopathic changes (blood vessel dilatation, microaneurysm formation, fibrinoid degeneration, cracking and double-barrelling of the vessel wall, and paravascular leakage of blood) (figure 1). The assessment for severe CAA could have occurred either as part of the Vonsattel grading scale, which includes such changes in its ‘severe’ category (see online supplementary appendix e-1), or through specifically reporting on some or all of the relevant histopathological characteristics. From now on when using the term ‘severe CAA’, we are referring to severe CAA with associated vasculopathic changes.

To avoid the effects of possible reporting bias (whereby positive results are more likely to be included within publications than negative ones) and to increase the size of the relevant dataset, we considered studies to be eligible for inclusion whether or not they actually reported on any association of APOE with severe CAA.

We searched OVID Medline (1950 to March 2012) and Embase (1980 to March 2012), using a combination of search terms for APOE, genes and CAA (see online supplementary appendix e-2). We also checked the bibliographies of all relevant studies and reviews identified, and searched Google Scholar for other studies citing relevant studies.

Our focus was on assessing the potential influence of APOE genotypes (in particular APOE-ε2) on severe CAA preceding rupture. Because APOE-ε2 is already known to be associated with CAA-related ICH (perhaps—but not definitely—causally), and severe CAA is more commonly found in such cases, we reasoned that including brains with ICH in our analyses would inevitably yield an association of APOE-ε2 with severe CAA, without necessarily meaning that APOE-ε2 influences risk through promoting the most severe stages of CAA pathology that precede rupture. Hence, to avoid selection bias, we excluded studies that had selected participants on the basis of having had a CAA-related ICH. For the same reason, we excluded participants selected on the basis of having had a CAA-related ICH from the included studies. We also excluded studies of hereditary CAA cases and those with less than 10 eligible participants (thereby excluding <3% of overall eligible participants). Two authors independently selected studies for inclusion, resolving disagreements by discussion.

Data extraction

For each study included, we extracted information on: first author; publication year; country in which the study was conducted; source of participants; and methodological characteristics. We assessed the quality of reporting of genotyping based on the STREGA (Strengthening the Reporting of Genetic Association Studies) and MOOSE (Meta-analysis of Observational Studies in Epidemiology) recommendations14 ,15 and the quality and characteristics of CAA pathology assessment using our own previously published criteria.9

For our planned meta-analyses, we required summary data on the numbers of participants with each APOE genotype (ɛ3ɛ3, ɛ2ɛ3, ɛ3ɛ4, ɛ4ɛ4, ɛ2ɛ2, ɛ2ɛ4) and whether or not CAA was present or absent. For those with CAA, we required, for each genotype, their CAA severity on the Vonsattel scale and/or data on the presence or absence of specific vasculopathic changes associated with severe CAA. These data were not all provided in the publications we identified. To facilitate the sharing and analysis of the unpublished data required, we therefore formed a collaborative group including the principal investigators of the relevant studies. We created a structured data extraction form and completed it as far as possible by entering data from the relevant publication(s). The principal investigators then checked the information entered, made any necessary corrections, and entered additional unpublished data if required and available.

Statistical analyses: APOE genotype and CAA severity on the Vonsattel scale

We used Cochrane RevMan (V.5) software. In our primary analyses, we included only individuals with CAA present on histopathological assessment, calculating study-specific and random effects pooled, unadjusted ORs for severe versus mild or moderate CAA among ɛ4 carriers (ε4+) versus those with other genotypes (ε4−) and among ɛ2 carriers (ε2+) versus others (ε2−). In secondary analyses, we repeated the primary analyses: comparing severe versus moderate CAA (but excluding mild CAA and so focusing on the severe end of the CAA spectrum) and comparing severe versus mild/moderate CAA, and severe versus moderate CAA among ε4+ versus the wild-type ɛ3ɛ3 and among ɛ2+ versus ɛ3ɛ3 to avoid potential confounding by mixed effects of ɛ2 and ɛ4 in the comparison group. In a further analysis, we included individuals both with and without CAA, comparing the presence versus absence of CAA among ɛ4+ versus ε4− and among ɛ2+ versus ε2− genotypes. To assess robustness of the methods, we repeated all the above analyses using the fixed effects method. We assessed heterogeneity with I2 and χ2 statistics. We considered p<0.05 to imply statistical significance.


From 1754 publications identified by our search, we identified eight relevant studies. We excluded two with <10 eligible participants,16 ,17 leaving six eligible studies. These six studies included 645 unique participants, of whom 102 had been selected on the basis of having had an ICH, leaving 543 who were eligible for our study (figure 2).5 ,13 ,18–21 Only one of the six studies had previously reported on the association between the APOE genotype and severe CAA (assessed using Vonsattel scale), finding a significantly greater frequency of APOE-ε2 in severe versus moderate CAA cases.13 This study and four others that had rated CAA on the Vonsattel scale, between them including 497 eligible participants (92% of all 543 potentially eligible participants),18–21 were able to share their unpublished data in collaborative meta-analyses. Data were unavailable from one additional study (46 eligible participants) that had assessed various different CAA-associated vasculopathic changes (double-barrelling of the vessel wall, fibrinoid necrosis and microaneurysm formation), including those that are not part of the Vonsattel scale.5

Figure 2

Selection of included studies.

Characteristics and quality of studies included in meta-analyses

All of the five studies included in our meta-analyses used autopsy brains from brain tissue banks or a population-based prospective study with an autopsy component. There were between 57 and 227 eligible participants per study. Mean age at death was between 78 and 84 years and about half of all participants were men. Three studies (357 participants) were conducted in predominantly white populations in the USA while information on ethnicity was unavailable for two studies (140 participants). About 50% of participants had clinical dementia (mainly neuropathologically-confirmed Alzheimer's disease), about 20% were known not to be demented and in the remaining 30% dementia status was unknown (table 1). The quality of genotyping and of pathology assessment was generally very good when assessed against current reporting standards.14 ,15 Methods for pathological assessment were variable, reflecting a lack of agreed standards for CAA pathology assessment at the time these studies were conducted (table 2).

Table 1

Study characteristics

Table 2

Study genotyping and pathology characteristics and quality

Meta-analyses: APOE genotype and CAA severity on the Vonsattel scale

353 out of 497 participants in five studies had CAA present and meta-analysis of data from these 353 participants showed for ε4+ versus ε4− genotypes, a significant association with severe versus mild/moderate CAA (OR 2.5, 95% CI 1.4 to 4.5, p=0.002) but no significant association with severe versus moderate CAA (OR 1.7, 95% CI 0.9 to 3.1, p=0.11) (figure 3A, see online supplementary figure e-1). There was no significant heterogeneity between individual studies’ results (I2=0%; χ23df=1.2; p=0.75 and I2=0%; χ23df=1.2; p=0.76, respectively). For ε2+ versus ε2− genotypes, the associations were non-significant (severe CAA vs mild/moderate CAA: OR 2.3, 95% CI 0.5 to 11.3, p=0.3; severe CAA vs moderate CAA: OR 2.7, 95% CI 0.6 to 11.4, p=0.19), with wide CIs due to small numbers of participants, particularly in the ε2+ group, which included 22 individuals, only seven of whom had severe CAA (figure 3B, see online supplementary figure e-2). There was moderate heterogeneity between individual studies’ results for severe CAA versus mild/moderate CAA (I2=52%; χ23df=6.2; p=0.1) and minimal heterogeneity for severe versus moderate CAA (I2=11%; χ23df=3.4; p=0.3). Results were similar and conclusions unchanged for the ε4+ and ε2+ genotypes when we used ε3ε3 genotypes as the comparison group (rather than ε4− or ε2−) (see online supplementary table e-1), and when we performed the analyses using the fixed effects method.

Figure 3

(A) Meta-analysis of APOE-ε4 associations with severe CAA. (B) Meta-analysis of APOE-ε2 associations with severe CAA.

Associations with the presence versus absence of CAA were consistent with results from our previous published systematic review,9 showing a clearly significant association with ε4+ (ε4+ vs ε4−: OR 4.8, 95% CI 3.0 to 7.6, p<0.00001) but not with ε2+ genotypes (ε2+ vs ε2−: OR 0.38, 95% CI 0.1 to 1.0, p=0.05).


After collecting and analysing data from five out of six existing relevant studies identified through a systematic search (including >90% of all eligible participants), our meta-analyses suggested a possible association of APOE-ε4 with progression to severe CAA as well as the expected association of ε4 (and not of ε2) with the presence of CAA.9 However, we were unable to confirm a statistically significant association between APOE-ε2+ genotypes and severe CAA. It should be emphasised that, although we included all available data from relevant publications, the relative rarity of ε2+ genotypes (only 22 of 353 individuals with CAA in our collaborative dataset) made the ε2-based analyses very imprecise.

The strengths of our study are the thorough search methods along with critical appraisal of the quality of included studies (which was generally good), inclusion of unpublished data through establishing a collaborative group to share data and inclusion of these data in meta-analyses. In addition, we were careful to avoid selection bias by excluding cases selected on the basis of having had an ICH, since this phenotype has known associations both with severe CAA and with APOE-ε2 and ε4.

Our study has a number of limitations. First, although our analysis includes data from >90% of eligible cases identified, we were unable to include cases from one small study.5 In that study, APOE-ε2 allele carriers had a statistically significant excess of fibrinoid necrosis compared with non-ε2 carriers, but CAA-related ICH cases were included in these analyses. Second, despite including data from almost all relevant cases from the published literature, the total numbers were relatively small and CIs wide, especially for analyses of the effects of APOE-ε2. Third, methods for histopathological assessment varied between studies, potentially introducing heterogeneity and reducing the likelihood of detecting a consistent effect across studies. Fourth, APOE allele-specific effects on severe CAA may differ according to the presence or absence of Alzheimer's disease, particularly for APOE-ε2, which has been associated with a decreased risk of Alzheimer's dementia.22 We could not perform informative subgroup analysis, however, because of the small overall numbers of participants and because dementia status was unknown for a large number of participants. Fifth, while the studies included here assessed those severe CAA-associated vasculopathic changes that are specifically alluded to in the Vonsattel scale, other vasculopathic changes may also be relevant. Sixth, both APOE allele-specific and other genetic associations may differ by CAA subtype. For example, there is preliminary evidence that APOE-ɛ4 may be associated with CAA type 1 (where CAA is found in cortical capillaries) and ɛ2 with CAA type 2 (where amyloid is deposited in leptomeningeal and cortical vessels with the exception of cortical capillaries).23 If CAA types 1 and 2 represent different pathological entities, the mechanisms and genetic risk factors for severe CAA and ICH could also differ. We did not have the necessary data to explore this in our study. Seventh, the issue of potential confounding due to population stratification was not addressed in the included studies. However, such confounding is unlikely because these were not case control studies but autopsy series where comparison groups within each study came from the same population. Finally, there may be other genetic influences that interact with APOE-ε2 to increase risk of or protect against severe CAA and ICH.

Our study therefore confirms existing strong evidence that APOE-ε4 promotes CAA, and further suggests that ε4 may increase the risk of developing severe CAA among those with CAA. However, while our findings do not exclude a biologically meaningful association between the APOE-ε2 allele and progression to severe CAA, we did not find convincing evidence to support this. Much larger numbers of individuals will need to be included in CAA histopathology studies before reliable conclusions can be drawn about the specific effects of APOE-ε2 on CAA. In particular, the hypothesis that APOE-ε2 influences risk of ICH through promoting progression of CAA to its severe form is not supported by the existing, relevant, unbiased data. Future research efforts in this area will also be helped substantially by the development and use of an internationally-agreed, standardised histopathological grading system for CAA (including assessment of CAA types 1 and 2), and by the consistent reporting of dementia—and specifically Alzheimer's disease—status among individuals included in histopathology studies.24


The authors would like to acknowledge Professor Jonathan Rosand and his team at the Center for Human Genetic Research, Massachusetts General Hospital, for their helpful comments on the manuscript. Professor Schmitt gratefully acknowledges the assistance of Dr Peter Nelson, Ms Erin Abner, Dr Daron Davis and Dr Richard Kryscio from the Sanders-Brown Center on Aging, University of Kentucky, for their help with data collection, funding acquisition and supervision of the personnel. Professor Chui gratefully acknowledges the assistance of Professors Harry Vinters (University of California Los Angeles), William Ellis (University of California Davis) and Chris Zarow (University of Southern California) for the original report of the cases and Professors Ling Zheng and Wendy Mack (University of Southern California) for data collection. Professor Greenberg gratefully acknowledges the assistance of Professor Jean Paul Vonsattel (Professor of pathology at Columbia University Medical Center and Director of the New York Brain Bank) for his contributions towards collecting the data and original reports of the cases.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors KR: design and conceptualisation of the study, acquisition, analysis and interpretation of the data, drafting and revising the manuscript; RNK, SMG, HCC and FAS: acquisition, analysis and interpretation of the data, revising the manuscript; NS and RA-SS: design and conceptualisation of the study, analysis and interpretation of the data, revising the manuscript; CLMS: senior supervision and coordination of the study, design and conceptualisation of the study, acquisition, analysis and interpretation of the data, drafting and revising the manuscript.

  • Funding This work was supported by: European Neurological Society Scientific Fellowship (KR); UK Medical Research Council research training fellowship (G0900428, NS); UK Medical Research Council senior clinical fellowship (G1002605, RA-SS); Scottish Funding Council (CLMS); and National Institutes of Health (2R01 AG26484, SMG). National Institute on Aging (P3OAGO28383; RO1AGO19241; RO1AGO38651; FAS).

  • Competing interests None.

  • Ethics approval In accordance with NRES and MRC guidance, this study does not require ethics approval as it does not directly involve human participants. The data used in the systematic review and meta-analysis have come from studies which have previously satisfied regulatory requirements and been peer reviewed.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing The authors are open to discussion regarding data sharing.