Total Magnetic Resonance Imaging Burden of Small Vessel Disease in Cerebral Amyloid Angiopathy: An Imaging-Pathologic Study of Concept Validation

Importance  Cerebral amyloid angiopathy (CAA) is characteristically associated with magnetic resonance imaging (MRI) biomarkers of small vessel brain injury, including strictly lobar cerebral microbleeds, cortical superficial siderosis, centrum semiovale perivascular spaces, and white matter hyperintensities. Although these neuroimaging markers reflect distinct pathophysiologic aspects in CAA, no studies to date have combined these structural imaging features to gauge total brain small vessel disease burden in CAA.

Objectives  To investigate whether a composite score can be developed to capture the total brain MRI burden of small vessel disease in CAA and to explore whether this score contributes independent and complementary information about CAA severity, defined as intracerebral hemorrhage during life or bleeding-related neuropathologic changes.

Design, Setting, and Participants  This retrospective, cross-sectional study examined a single-center neuropathologic CAA cohort of eligible patients from the Massachusetts General Hospital from January 1, 1997, through December 31, 2012. Data analysis was performed from January 2, 2015, to January 9, 2016. Patients with pathologic evidence of CAA (ie, any presence of CAA from routinely collected brain biopsy specimen, biopsy specimen at hematoma evacuation, or autopsy) and available brain MRI sequences of adequate quality, including T2-weighted, T2*-weighted gradient-recalled echo, and/or susceptibility-weighted imaging and fluid-attenuated inversion recovery sequences, were considered for the study.

Main Outcomes and Measures  Brain MRIs were rated for lobar cerebral microbleeds, cortical superficial siderosis, centrum semiovale perivascular spaces, and white matter hyperintensities. All 4 MRI lesions were incorporated into a prespecified ordinal total small vessel disease score, ranging from 0 to 6 points. Associations with severity of CAA-associated vasculopathic changes (fibrinoid necrosis and concentric splitting of the wall), clinical presentation, number of intracerebral hemorrhages, and other imaging markers not included in the score were explored using logistic and ordinal regression.

Results  In total, 105 patients with pathologically defined CAA were included: 52 with autopsies, 22 with brain biopsy specimens, and 31 with pathologic samples from hematoma evacuations. The mean (range) age of the patients was 73 (71-74) years, and 55 (52.4%) were women. In multivariable ordinal regression analysis, severity of CAA-associated vasculopathic changes (odds ratio, 2.40; 95% CI, 1.06-5.45; P = .04) and CAA presentation with symptomatic intracerebral hemorrhage (odds ratio, 2.23; 95% CI, 1.07-4.64; P = .03) were independently associated with the total MRI small vessel disease score. The score was associated with small, acute, diffusion-weighted imaging lesions and posterior white matter hyperintensities in adjusted analyses.

Conclusions and Relevance  This study provides evidence of concept validity of a total MRI small vessel disease score in CAA. After further validation, this approach can be potentially used in prospective clinical studies.

Sporadic cerebral amyloid angiopathy (CAA) is the most common cause of symptomatic lobar intracerebral hemorrhage (ICH) in elderly people and results from cerebrovascular amyloid-β deposition, preferentially involving cortical and leptomeningeal small vessels.^1-4 Frequently, CAA is present in the brain of patients with Alzheimer disease, but it can also have an independent contribution in age-related cognitive decline.^5,6

Cerebral amyloid angiopathy is characteristically associated with magnetic resonance imaging (MRI) biomarkers of small vessel brain injury, including multiple, strictly lobar cerebral microbleeds (CMBs),⁷ cortical superficial siderosis (cSS),⁸ centrum semiovale perivascular spaces (CSO-PVSs),^9,10 and white matter hyperintensities (WMHs).^1,11 Although these neuroimaging markers probably reflect distinct pathophysiologic aspects or steps in CAA,^4,12 they are often closely related. No studies to date have combined these structural imaging features to gauge total brain small vessel disease burden in CAA. A comprehensive approach might have potential advantages over individual markers, providing a practical framework to better assess the effect of CAA-related damage on clinical outcomes.¹³

Recently, the approach of assessing total MRI small vessel disease burden has been developed and applied in patients at high risk for ischemic small vessel damage, including lacunar or nondisabling cortical stroke.^14-16 By summing different MRI features of ischemic small vessel disease in one measure (with 1 point assigned for each of the following: lacunar infarct, CMBs, basal ganglia PVSs, and WMHs), Staals et al¹⁴ demonstrated that a higher score is independently related to lacunar stroke subtype and certain vascular risk factors (eg, age, hypertension, or smoking). In patients with lacunar stroke, this score was associated with blood pressure¹⁵ and cognition.¹⁶ Furthermore, the score was associated with lower general cognitive ability in 680 older participants.¹⁷

In this study, we prespecified and developed a total MRI small vessel disease score specifically tailored for patients with CAA based on the 4 major imaging signatures of the disease as described in a consensus report on CAA neuroimaging biomarkers. In a cohort with pathologic evidence of CAA, we investigated whether this is a valid construct by examining (1) whether the total MRI CAA score is associated with symptomatic CAA-related intraparenchymal bleeding events during life and severe CAA on neuropathologic examination and (2) the association between the total MRI CAA score and other characteristic imaging findings in CAA, including small positive lesions on diffusion-weighted imaging (DWI)¹⁸ and occipital-predominant WMHs.^19,20

Box Section Ref ID

In this retrospective, cross-sectional study, we used data from a neuropathologically defined CAA cohort based on eligible patients from the Massachusetts General Hospital from January 1, 1997, through December 31, 2012, who were systematically identified using overlapping methods as described.^12,21 Data analysis was performed from January 2, 2015, to January 9, 2016. Patients with pathologic evidence of CAA (ie, any presence of CAA from routinely collected brain biopsy specimen, biopsy specimen at hematoma evacuation, or autopsy) and available in vivo brain MRI sequences of adequate quality, including T2-weighted, T2*-weighted gradient-recalled echo, and/or susceptibility-weighted imaging and fluid-attenuated inversion recovery sequences, were considered for the study, as previously described.^12,21 The clinical presentation of included patients was ascertained based on all available data and was classified as (1) symptomatic lobar ICH (confirmed on neuroimaging) or (2) presentation without ICH (including cognitive impairment, transient focal neurologic episodes, or other neurologic symptoms) at the time of the MRI.¹²

The study received ethical approval by the institutional review board of Massachusetts General Hospital. All patients provided written informed consent, and all data were deidentified for statistical analysis but not for imaging analysis.

Morphologic assessment was performed in routine hematoxylin-eosin staining, and the presence or absence and severity of vascular amyloid-β deposition were confirmed by immunohistochemical detection (anti–antibody 6E10; 1:200; Signet Laboratories) and Congo red staining. Cases were considered positive for CAA when they had at least 1 leptomeningeal or cortical vessel with amyloid-β reported by an experienced neuropathologist, providing enough information to reliably classify CAA severity according to the modified Vonsattel grading system.^22,23 We systematically extracted information on the presence of CAA-related vasculopathic changes, defined as vessel-within-vessel appearance (sometimes called double-barreling) and fibrinoid necrosis, corresponding to modified Vonsattel grade 3 and grade 4, respectively, and hence severe CAA.²³

Imaging for all patients included T2-weighted, fluid-attenuated inversion recovery, T2*-weighted gradient-recalled echo (field strength, 1.5 T; section thickness, 5 mm; section gap, 1 mm; and echo time, 24 milliseconds), and/or susceptibility-weighted imaging (field strength, 3 T; section thickness, 1.2 mm; section gap, 0 mm; and echo time, 21 milliseconds).²¹ Two different MRI scanners were used during the study period: a Siemens Trio at 3 T (used in 25% of the patients) and a GE Signa at 1.5 T. Review of the MRIs were masked to all clinical and histopathologic findings by trained observers, according to Standards for Reporting Vascular Changes on Neuroimaging (STRIVE).²⁴ For patients with multiple MRIs available, the MRI closest to the neuropathologic assessment was examined.

The presence and number of CMBs were evaluated according to current consensus criteria.¹¹ The presence and number of macro-ICHs (>5 mm in diameter)²⁵ were also noted. Cortical superficial siderosis was defined as linear residues of blood products in the superficial layers of the cerebral cortex showing a characteristic gyriform pattern of low signal on blood-sensitive sequences as previously described.²⁶ The distribution and severity of cSS were classified as focal (restricted to ≤3 sulci) or disseminated (≥4 sulci).⁸ Contiguous or potentially anatomically connected cSS with any lobar ICH was not included in the aforementioned categories.

Perivascular spaces were assessed in line with STRIVE definitions²⁴ and rated on axial T2-weighted MRIs using a previously described, validated 4-point visual rating scale (0 indicating no PVS; 1, ≤10 PVSs; 2, 11-20 PVSs; 3, 21-40 PVSs; and 4, >40 PVSs) in the basal ganglia and CSO.^9,27 Deep and periventricular WMHs were assessed according to the 4-point Fazekas rating scale.²⁸ The WMHs in the frontal and occipital lobe were further evaluated separately on axial fluid-attenuated inversion recovery as described by Zhu et al.¹⁹ The frontal-occipital gradient (the WMH score in the frontal lobe minus that in the occipital lobe) was calculated (range, −6 to 6; >0 implies frontal dominance and <0 implies occipital dominance).¹⁹ All available DWIs were reviewed for the presence of small (generally <4 mm) hyperintensities by trained raters.²⁹

We prespecified an ordinal scale that represented the total burden of small vessel disease in CAA by incorporating the 4 most characteristic MRI markers of the disease (lobar CMBs, cSS, CSO-PVSs, and WMHs) (Figure 1). For lobar CMBs, a point was awarded if 2 to 4 CMBs were present and 2 points for 5 or more CMBs. The presence of cSS was awarded with 1 point if focal and 2 points if disseminated. The presence of CSO-PVSs was counted if there were moderate to severe (grade 3–4, ie, >20) PVSs (1 point if present). The presence of WMHs was defined as (early) confluent deep (ie, the region between juxtacortical and ventricular areas) WMHs (Fazekas score ≥2) or irregular periventricular WMHs extending into the deep white matter (Fazekas score of 3) (1 point if either present), as previously described.¹⁴ Hence, the score ranged from a minimum of 0 to a maximum of 6 points, creating an ordinal scale.

In defining these cutoffs, we took into account current evidence from cross-sectional and longitudinal studies in CAA. We aimed to distinguish the weights of hemorrhagic imaging markers of CAA (lobar CMBs, cSS) from other nonhemorrhagic markers (CSO-PVSs, WMHs), which might also occur with high frequency in other diseases. For example, multiple (≥2) strictly lobar CMBs are the most common and characteristic imaging feature of CAA and are part of the Boston criteria^8,31; the presence of 5 or more CMBs is associated with recurrent lobar ICH.³⁴ Cortical superficial siderosis, particularly disseminated, is a common hemorrhagic marker of CAA²⁶ associated with future lobar ICH risk.³³ The prespecified definition of CSO-PVS (ie, >20, grade 3) was found to be more strongly related to CAA.^9,10,34 The WMH cutoffs chosen are also in line with the scores used in the suggested total MRI scale for ischemic small vessel disease^14-17 based on those Fazekas scores related to small vessel disease in a histopathologic study.³⁵

As an exploratory analysis, we tested the effect of using 2 alternative scores: (1) an expanded score that also included the presence of occipital predominant WMHs¹⁹ as another element (1 extra point; hence, a total score ranging from 0-7) and (2) a simplified version of the CAA score in line with the original MRI score suggested for ischemic small vessel disease.^14-16 For this simplified score, we rated the presence (yes vs no) of each of the 4 MRI features above and assigned 1 point for each (ie, presence of any lobar CMBs, any cSS, moderate to severe CSO-PVSs, and extensive WMHs), resulting in an ordinal scale of 0 to 4.

To investigate the construct validity of this score, our statistical approach was to use the total MRI score as the dependent variable against clinical and neuropathologic markers of CAA severity. We performed univariable ordinal logistic regression to investigate the association between the MRI small vessel disease score with clinical characteristics, presence of CAA-associated vasculopathic changes, and CAA clinical presentation with ICH. We performed multivariable ordinal logistic regression that explored the independent association between the presence of vasculopathic changes and CAA clinical presentation with the small vessel disease score (as dependent variable), controlling for age, sex, and hypertension. To further account for potential biases based on the size of the pathologic specimen (biopsy vs autopsy specimen), with full autopsy being more likely to reveal some CAA and severe CAA pathologic findings than biopsy, we adjusted for this factor in a sensitivity analysis.

Separate univariable and multivariable ordinal regression analyses were used to assess the association between the small vessel disease score and DWI lesions, occipital predominant WMH presence, and symptomatic ICH number, adjusting for age and sex and additionally for CAA-related vasculopathic changes and clinical presentation. We tested the effect of the occipital predominant WMH as an additional score element and the alternative simplified total MRI small vessel disease score using a similar approach.

Significance level was set at P < .05. We used STATA software, version 11.2 (StataCorp). In the ordinal regression models, the proportional odds assumption was checked using relevant tests (omodel command). Colinearity was tested using the variance inflation factor (<10 for all variables). The article was prepared with reference to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.³⁵

In total, 105 patients with pathologically defined CAA were included: 52 with autopsies, 22 with brain biopsy specimens, and 31 with pathologic samples from hematoma evacuations. Fifty-four patients presented with symptomatic, spontaneous lobar ICH, whereas 51 patients presented without any symptomatic ICH at baseline. Patients without ICH presented with cognitive impairment (n = 42; median Clinical Dementia Rating score, 1; interquartile range [IQR], 0.5–2), transient focal neurologic episodes (n = 3), or a combination of other symptoms (n = 6; including altered mental status or seizures consistent with inflammatory CAA). Clinical and imaging characteristics are given in Table 1. As previously reported, mild CAA (Vonsattel grade 1) and moderate to severe CAA (Vonsattel grades 2–4) were equally represented in the groups (P = .43).¹² Among the 52 patients with autopsy data, 39 (75%) had evidence of arteriolosclerosis (ie, concentric hyaline thickening of small arteries) in deep (basal ganglia or deep white matter) perforating arterioles, which was moderate to severe in 31 cases.

Eighteen patients (17.3%) had none of the MRI markers of small vessel disease included in the score. Most patients (15 of 18 [83.3%]) had only mild CAA. The distribution of total small vessel disease score severity is shown in Figure 2. Patients with CAA and ICH had higher ratings of small vessel disease burden compared with those presenting without ICH (median score, 3 [IQR, 1-5] vs 2 [IQR, 1-4], respectively; P = .02). In univariable analysis, total small vessel disease score was associated with the presence of CAA-related vasculopathic changes on pathologic analysis (odds ratio [OR], 2.21; 95% CI, 1.02-4.78; P = .04) and CAA presentation with ICH (OR, 2.37; 95% CI, 1.19-4.72; P = .02) but not with age, hypertension, sex, or antithrombotic drug use. None of the different MRI markers comprising the score were individually associated with vasculopathic changes in univariable or multivariable logistic regression analyses, including all 4 markers.

In multivariable ordinal regression, the total score was associated with the presence of vasculopathic changes and CAA presentation with ICH (Table 2). The results remained consistent and of similar effect size when hypertension was not included in the model and when the size of the pathologic specimen (biopsy vs autopsy), time from MRI to pathologic analysis and whether different blood-sensitive MRI sequences were included in sensitivity analyses. There was an increasing MRI score with increasing macrobleed count on MRI (median, 1; range, 1-6) in the patient group with symptomatic ICH (coefficient, 1.77; 95% CI, 0.08-3.45; P = .04). Small acute DWI lesions and occipital predominant WMHs were associated with small vessel disease score in unadjusted ordinal regression and remained so in fully adjusted models (Table 3).

In the same multivariable regression model as above, the alternative score, including occipital predominant WMHs, was associated with CAA-related vasculopathic changes (OR, 2.64; 95% CI, 1.13-6.14; P = .02) and symptomatic ICH presentation (OR, 2.28; 95% CI, 1.09-4.79; P = .03). However, although the simplified total MRI small vessel disease score was associated with CAA presentation with ICH (OR, 2.51; 95% CI, 1.19-5.30; P = .02), there was only a trend for an association with the presence of vasculopathic changes (OR, 1.99; 95% CI, 0.88-4.48; P = .10).

In this study, we developed an MRI small vessel disease score based on the 4 most characteristic neuroimaging signatures of sporadic CAA to compile the overall brain burden of the disease. Using a neuropathologically defined CAA cohort of patients presenting with and without ICH, we provide evidence of the construct validity of our approach. The total small vessel disease score was found to be independently associated with CAA-related vasculopathic changes on pathologic analysis (a marker of CAA-related microangiopathy severity) and clinical presentation with symptomatic lobar ICH.

Studies^37,38 have suggested staging schemes based on pathologic analysis for the extent of cerebral small vessel disease in aging brains. In the Oxford Project to Investigate Memory and Ageing cohort, small vessel disease severity assessed using 1 of these pathology schemes was correlated with cognitive impairment.³⁸ Approaches to integrate a range of imaging manifestations of small vessel disease into 1 score have been recently advocated by an international working group (STRIVE).²⁴ Some studies^14-17 have presented a first attempt to capture the total MRI burden in ischemic small vessel disease, validating this approach mainly in populations with lacunar stroke. An assessment of total small vessel disease load on MRI has several potential advantages because it avoids overreliance on any 1 individual marker of small vessel disease. Hence, a more complete evaluation of the total burden may be important to understand the effect of small vessel disease on clinical outcomes, such as disability and cognition.

In CAA, studies^8,9 have found that the combination of 2 different MRI markers of the disease (lobar CMBs and cSS, lobar CMBs and high-grade CSO-PVSs) increase diagnostic sensitivity. Our score provides a practical and easily applied structural MRI visual tool to comprehensively evaluate small vessel disease in CAA using validated rating systems for each feature.²⁴ We acknowledge that the assessment of the total MRI small vessel disease burden is complex, and different imaging features probably reflect distinct pathophysiologic mechanisms in CAA.^4,12 For example, cSS likely represents blood-leaking episodes into the subarachnoid space from CAA-affected leptomeningeal vessels³⁰ (as opposed to lobar CMBs, which may arise from CAA-laden parenchymal vessels), whereas CSO-PVSs might be related to cerebrovascular amyloid burden and drainage impairment^10,39,40 and WMHs to chronic ischemia.⁴¹ The pathologic substrates of small vessel disease imaging markers are, however, heterogenenous,^42,43 and direct histopathologic imaging correlation studies are notably underused in the field.⁴⁴ Despite different mechanisms, these features are often related and co-occur in patients with CAA.⁶ We also note that the chosen cutoffs and weightings in our CAA score might not be optimal and to a certain extent are arbitrary, although we did take into account the totality of current clinical, neuroimaging, and neuropathologic evidence from cross-sectional and longitudinal studies^1,9,14,31,32 in defining different weights for these markers. In support of our approach, none of the individual MRI lesions included in the score seem to be significantly driving the associations with CAA severity in isolation.

Clinically silent small areas of restricted diffusion on DWI MRI, thought to represent acute microinfarcts, are sometimes considered an additional marker of CAA.¹⁸ Their transient nature, combined with the lack of neuropathologic confirmation and their occurrence in patients with hypertensive arteriopathy and healthy individuals,⁴⁵ may render it difficult to reliably use them as a specific biomarker of the disease. On the basis of these considerations, we did not include DWI lesions in the MRI small vessel disease score. We did, however, find an association between DWI lesions and total MRI score, indicating that they may be an additional marker of disease severity. Similarly, a more posterior distribution of WMHs may be an additional promising marker for CAA,^19,20 potentially associated with more severe disease.⁴⁶

Our study has limitations. The fact that our score was tested on a clinical pathologic series of patients with CAA, including patients with the most salient clinical manifestations of the disease, makes the current analysis powerful and informative for construct validity. However, this approach could also represent a constraint because of selection bias because only symptomatic patients with pathologic evaluation and brain MRI were included, likely representing patients at the more severe end of the disease spectrum. The differences in pathologic sampling between autopsied brains and biopsy specimens may have contributed to bias in that some patients with CAA who had negative brain biopsy results may have been excluded. Another limitation of the current study is the retrospective (which means that not all MRI sequence parameters were standardized) and cross-sectional design, which precludes any analysis on the prognostic significance of the small vessel disease scores. T1 sequences were not systematically obtained as part of the MRI protocol, precluding any assessment of lacunes and atrophy. Finally, the tacit implication of the current and previous scores is that there are only 2 major types of cerebral microvascular disease in the brain, CAA and everything else (so-called hypertensive arteriopathy, including arteriolosclerosis, lipohyalinosis, fibrinoid necrosis, and microaneurysm), which is an oversimplification.⁴⁴

This study provides evidence of the concept validity of a total MRI small vessel disease score in CAA. Despite limitations, the suggested scale for CAA performed reasonably well in the current study and raises several questions and opportunities for further testing and development in large-scale studies. Longitudinal cohorts should investigate whether this system, with or without modifications, can be used to assess the effect of CAA on clinical outcomes in different settings, including cognitive impairment, ICH risk, and disability.^13,47

Accepted for Publication: March 1, 2016.

Corresponding Author: Andreas Charidimou, MD, PhD, Hemorrhagic Stroke Research Program, Department of Neurology, Massachusetts General Hospital Stroke Research Center, Harvard Medical School, 175 Cambridge St, Ste 300, Boston, MA 02114 (andreas.charidimou.09@ucl.ac.uk).

Published Online: June 27, 2016. doi:10.1001/jamaneurol.2016.0832.

Author Contributions: Drs Charidimou and Viswanathan had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Charidimou, Greenberg, Viswanathan.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscripts: Charidimou, Viswanathan.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Charidimou.

Obtained funding: Greenberg, Viswanathan.

Administrative, technical, or material support: Charidimou, Oliveira-Filho, Lauer.

Study supervision: Rosand, Gurol, Greenberg, Viswanathan.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grants 5R01AG047975, 5P50AG005134, 5K23AG028726 (Dr Viswanathan), and R01AG26484 (Dr Greenberg) from the National Institutes of Health.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.