Background and Purpose Examination of cortical tissue obtained surgically is an important tool for diagnosis of cerebral amyloid angiopathy (CAA) during life. Analysis of a single sample of cortical tissue, however, might lead to conclusions that are either falsely positive (because of the high frequency of CAA in the healthy elderly) or falsely negative (because of the patchy distribution of CAA pathology). We therefore attempted to estimate the sensitivity and specificity of cortical biopsy for diagnosis of CAA as the cause of intracerebral hemorrhage.
Methods To simulate biopsy in CAA, we took biopsy-sized cortical samples from postmortem brains with known extents of CAA: either CAA-related hemorrhage or mild to severe CAA without hemorrhage. Samples were stained with the use of methods routinely available in surgical pathology laboratories and blindly examined for vascular amyloid and amyloid-related vasculopathic changes.
Results The presence of vascular amyloid was a sensitive marker for CAA-related hemorrhage, occurring in all 28 specimens from brains with hemorrhage. Conversely, the appearance of fibrinoid necrosis in amyloid-laden vessels was relatively specific for CAA-related hemorrhage. This finding occurred in 13 of the 28 specimens (46%) from brains with hemorrhage but in none of 27 sections from brains with mild CAA and in only 4 of 42 specimens with moderate to severe CAA without hemorrhage.
Conclusions These data help to define criteria for the diagnosis of CAA-related hemorrhage from surgical specimens.
Cerebral amyloid angiopathy refers to deposition of amyloid in the walls of vessels in the cerebral cortex and leptomeninges. Severe CAA is associated with vasculopathic changes, vessel rupture, and cerebral hemorrhage.1 2
Clinical diagnosis of CAA as the cause of intracerebral hemorrhage presents a challenge. CAA-related hemorrhage can be difficult to distinguish clinically from other potential causes of lobar hemorrhage such as trauma, hemorrhagic infarction, lobar extension of a hypertensive hemorrhage, hemorrhagic tumor, vascular malformation, or cerebral vasculitis.3 CAA can be suggested by radiographic demonstration of multiple hemorrhages restricted to lobar regions, but this pattern is probably not present in all cases4 5 and may be mimicked by other disorders. Among molecular markers for AD, apolipoprotein E genotype has emerged as an important risk factor for CAA-related hemorrhage6 7 8 but lacks sufficient specificity and sensitivity for the diagnosis of this disorder. Other potential markers for AD such as amyloid peptide and tau protein9 have not been studied in CAA but also seem unlikely to provide definitive diagnostic information.
Pathological demonstration of vascular amyloid thus remains a cornerstone to the diagnosis of CAA and figures prominently in the proposed Boston diagnostic criteria.3 A sample of cortical tissue can be obtained safely in CAA by biopsy or hematoma evacuation.10 11 12 13 Its interpretation is complicated, however, by the possibility of sampling error. On the one hand, the distribution of CAA is characteristically patchy and segmental,1 such that even in severe cases it might be possible for a section of brain not to contain vascular amyloid. On the other hand, some degree of amyloid in blood vessels is a common incidental finding in the elderly,14 15 so that the presence of vascular amyloid does not prove it to be the cause of hemorrhage. Potentially the most specific pathological findings for CAA-related hemorrhage are those CAA-related vasculopathic changes, such as cracking and fibrinoid necrosis of the amyloid-laden vessel wall, that appear to identify cases at risk for hemorrhage.15 16 17
The present study represents an attempt to estimate the sensitivity and specificity of cortical biopsy for diagnosis of CAA-related hemorrhage. Because full postmortem examination is the “gold standard” for diagnosis of CAA, the conditions of biopsy were simulated by the analysis of biopsy-sized samples taken from postmortem brains with known severities of CAA.
The overall scheme for the present study was (1) to select postmortem brains with known severities of CAA ranging from mild to severe, (2) cut new blocks from multiple cortical regions of the selected brains and prepare biopsy-sized sections from each block, and (3) analyze each new section (by routinely available techniques and without knowledge of its brain of origin) for the presence of CAA and CAA-related vasculopathy.
Seven brains were chosen to reflect the full range of CAA severities from cases referred to the BTRC at McLean Hospital (Belmont, Mass) between May 1995 and June 1995. Each brain had previously been systematically sectioned and graded for severity of CAA as described.15 18 Two of the selected brains carried the grade of mild CAA, three moderate to severe CAA without evident hemorrhage, and two CAA-related hemorrhage (defined as severe CAA and hemorrhage in corticosubcortical regions without other evident cause3 ). The brains also demonstrated various degrees of atherosclerotic vascular changes but no evidence of hypertensive vasculopathy (lipohyalinosis, fibrinoid necrosis, or microaneurysms) in penetrating vessels from sections of the lenticular nuclei, pons, thalamus, or cerebellar dentate nucleus.
New blocks for the present analysis were cut from formalin-fixed half-brains of the above cases. The block size used was 1.3×1.0×0.3 cm, corresponding to the approximate mean size of 18 consecutive open biopsy specimens obtained and reviewed at the Massachusetts General Hospital in June 1995. The samples taken included cortex, subcortical white matter, and whenever possible, leptomeninges. A total of 14 new cortical blocks were cut from each brain, one from each of the following areas: 2 anterior frontal cortex (corresponding to Brodmann’s areas 9 and 10/46), 2 posterior frontal cortex (areas 6 and 6/4), 2 anterior temporal cortex (areas 22 and 21), 2 posterior temporal cortex (areas 41/42/22 and 21), anterior and posterior superior parietal lobules (areas 3/1/2/5), anterior and posterior inferior parietal lobules (areas 40/22), lateral occipital cortex (areas 37/19), and occipital pole (area 17). In addition, a single block was taken from the lateral cerebellum, including the horizontal fissure.
Two sections were prepared from each paraffin-embedded block and stained with LHE (7-μm-thick sections) and Congo red (10-μm-sections). Sections from the various brains were examined in random order with and without polarized light. Grading was performed by a single pathologist (J-P.G.V.) without knowledge of the brain from which the section derived. The identity of material staining positively with Congo red was subsequently confirmed to be β-amyloid by immunocytochemistry (monoclonal antibody 6F3D, DAKO Corp, 1:150). To confine the analysis to routinely available techniques, however, findings from the immunocytochemical slides were not used to classify CAA severity in the present study.
The following grading scale was modified from previously described methods15 for use on single tissue specimens (Figure⇓). This method scores the most advanced degree of CAA present within the specimen, an approach previously demonstrated to correlate with risk of hemorrhage.15 Separate scores were given to the leptomeningeal and the parenchymal vessels, and a total score was given for the most affected vessel overall. Grade 0 CAA denotes the absence of congophilic staining in vessels, grade 1 (Figure⇓, panel A) the presence of some congophilic staining in an otherwise normal-appearing vessel, and grade 2 (Figure⇓, panel B) complete replacement of the media by congophilic material. Grade 3 (Figure⇓, panel C) refers to cracking of the amyloid-laden vessel wall (creating a vessel-within-vessel appearance) affecting at least 50% of the circumference of the vessel. Grade 4 (Figure⇓, panel D) denotes the presence in an amyloid-laden vessel of fibrinoid necrosis, recognized as homogeneous discrete foci or segments of the vascular wall that contain smudgy eosinophilic material obscuring the cytoarchitecture. This material has previously been observed to stain positively with phosphotungstic acid hematoxylin.15
Sensitivity was calculated as the likelihood of finding a given degree of CAA in specimens taken from brain with CAA-related hemorrhage, and specificity was calculated as 100% minus the percent likelihood of encountering that degree of CAA in a brain sample from the general elderly population. Estimates of the distribution of CAA in the general elderly population (required for the calculation of specificity) were obtained from cases examined at the BTRC. Severity of CAA was analyzed in 784 consecutive postmortem brains from patients aged 65 years or older referred to the BTRC between September 1989 and July 1996 for diagnoses other than cerebral hemorrhage. Since the BTRC receives a disproportionate number of brains with AD, this sample was reweighted to reflect the age-specific prevalence of AD, taken as the average of two community studies in the Massachusetts area.19 20
Comparisons of degree of amyloid severity between parenchymal and leptomeningeal vessels and between anterior and posterior brain sections were performed with the use of the Wilcoxon signed-rank test for matched samples. For comparison of anterior and posterior sections, the following eight tissue sections from each brain were matched: anterior frontal (area 9) with lateral occipital (area 37/19), anterior frontal (area 10/46) with occipital pole (area 17), posterior frontal (area 6) with posterior superior parietal lobule (area 3/1/2/5), and posterior frontal (area 6/4) with posterior inferior parietal lobule (area 40/22). All analyses were performed with Stata software (Stata Corp), and all significance tests were two-tailed.
We systematically examined biopsy-sized sections of cortex derived from brains with previously identified degrees of CAA, grading each section as shown in the Figure⇑. Comparison between vessels in the brain parenchyma and those in the overlying leptomeninges of the same specimen revealed significantly more severe CAA in leptomeningeal than parenchymal vessels (P<.005). A comparison was also performed between matched sections from anterior (frontal lobe) and posterior (occipital or posterior parietal lobe) cortical regions within each brain (see “Methods”). This analysis revealed posterior greater than anterior CAA in leptomeningeal (P<.02) but not parenchymal vessels (P>.7).
The results in Table 1⇓ depict the frequency of each grade of CAA in biopsy-sized cortical sections according to the extent of CAA in its brain of origin. There was substantial variability among sections within each level of CAA. Thus, rare sections from brains with mild CAA demonstrated grade 2 or 3 amyloid angiopathy, while sections from brains with CAA-related hemorrhage occasionally demonstrated grade 1 amyloid only. (Because the blocks in this study were entirely independent of the sections originally used to rate extent of CAA, it was possible for the present study to reveal vessels of greater severity than seen in the initial grading.) Overall, 26 of 28 sections (including all 24 sections that included leptomeningeal tissue as well as 2 of 2 sections of cerebellar cortex) from brains with CAA-related hemorrhage demonstrated at least grade 2 vascular amyloid. Conversely, 0 of 27 sections from brains with mild CAA and only 4 of 42 (9.5%) with moderate to severe CAA without hemorrhage contained the vascular deposits suggestive of fibrinoid necrosis (grade 4; Figure⇑, panel D), a finding in 13 of the 28 sections (46.4%) from brains with CAA-related hemorrhage.
To estimate sensitivity and specificity for diagnosis of CAA-related hemorrhage, we first assessed the distribution of CAA in the general elderly population. Table 2⇓ shows estimates of age-specific prevalence for the various severities of CAA based on a series of 784 consecutive postmortem cases. Sensitivity was then calculated from the data in Table 1⇑ as the likelihood of detecting a given degree of CAA in specimens from patients with CAA-related hemorrhage, and specificity was calculated from data in Tables 1⇑ and 2⇓ according to the probability of finding that degree of CAA in samples from patients in the general elderly population.
The estimates of sensitivity and specificity with various degrees of CAA used as cutoff are listed in Table 3⇓. Requiring at least grade 2 CAA (complete replacement of a vessel wall with amyloid) in the examined specimen to make the diagnosis of CAA-related hemorrhage resulted in a sensitivity of 93% and a specificity of 88%, even in patients aged 85 years or older.
Pathological data are considered the firmest evidence during life for CAA.1 Our results, demonstrating considerable section-to-section variability within each class of CAA (Table 1⇑), emphasize the complexities in interpreting surgical specimens. Overall, however, those cases with CAA-related hemorrhage had both more severe and more widespread CAA pathology than brains without hemorrhage. The data thus suggest several tentative rules for interpretation of surgical specimens in patients with suspected CAA.
First, the presence of mild to moderate vascular amyloid (Figure⇑, panels A and B) in a section of cortical tissue is a sensitive marker for the presence of CAA-related hemorrhage. All samples from brains with CAA-related hemorrhage contained at least grade 1 CAA, and all samples that included leptomeningeal tissue demonstrated at least grade 2. The complete absence of vascular amyloid in an adequate tissue specimen would thus weigh strongly against a diagnosis of CAA-related hemorrhage.
Second, the appearance of fibrinoid necrosis (Figure⇑, panel D) in an amyloid-laden vessel is reasonably specific for CAA-related hemorrhage, appearing only rarely in sections with moderate to severe CAA without hemorrhage. The finding is uncommon enough, however, that its absence does not preclude CAA as the cause of hemorrhage.
Fibrinoid necrosis of blood vessels has been noted as a marker of severe CAA in several postmortem studies.15 16 21 Ultrastructural analysis suggests that plasma components such as fibrinogen enter the vessel wall within a dilated region of the amyloid-laden vessel, producing an area devoid of amyloid.17 Our data indicate that these deposits are relatively specific correlates of CAA-related hemorrhage in biopsy specimens, as they appear to be in the postmortem brain.15 16 17 While fibrinoid necrosis can characterize other types of hemorrhage such as those that are hypertension related,22 its appearance within vessels that also stain positively for amyloid (Figure⇑, panel D) and within sections demonstrating other areas of CAA-related vasculopathy would seem unlikely in hemorrhagic conditions other than CAA. It should be noted that none of the brains used in this study had evidence for hypertensive vasculopathy (see “Methods”).
The appearance of fibrinoid necrosis and cracking of the vessel wall (Figure⇑, panel C) were the most helpful markers for CAA-related vasculopathy in our analysis. Several other described vasculopathic changes, such as the presence of ectatic regions, microaneurysms, or inflammatory infiltrates,15 16 were less useful. Inflammation involving the vessel wall, while potentially an important step in vessel rupture,23 is a relatively rare pathological finding and could not be identified by routine methods in any of our specimens. Ectatic or aneurysmal changes are more common but are difficult to identify in vessels (usually seen in cross section) in the single section typically available for microscopic review.
Several issues arise in applying these data to clinical situations. Characteristics of the biopsy tissue such as size and location are likely to affect the chances of detecting vascular amyloid. In particular, the predilection of CAA for leptomeningeal vessels makes it useful to include leptomeninges in the resected specimen when possible. As demonstrated in Table 3⇑, the specificity of biopsy also depends on patient age because of the effect of age on frequency of CAA in the general population. Thus, the detection of vascular amyloid in a 65-year-old patient carries greater specificity than its detection in an 85-year-old, since CAA is less likely to be encountered by chance in an unaffected 65-year-old. A final issue in applying our data is that they derive from only a small number of brains, although relatively little brain-to-brain variation was seen within each class of CAA severity.
The behavior of leptomeningeal vessels in CAA presents a bit of a paradox. Vessels of the leptomeninges are extensively involved with amyloid; indeed, we found CAA to be significantly more severe in leptomeninges than in parenchymal vessels of the same section. Leptomeningeal vessels nonetheless appear rarely, if ever, to rupture in CAA.1 24 Our data indicate that while fibrinoid necrosis may occur somewhat less frequently in leptomeningeal vessels relative to their extent of amyloid deposition, these vessels are not immune to this vasculopathic change (see Table 1⇑ and also Reference 1616 ). The explanation for this apparent inconsistency awaits future study.
It is interesting in this context to note that while we, like others,14 observed a tendency for more severe CAA to occur in posterior relative to anterior cortical regions, this pattern applied only to leptomeningeal vessels. The essentially uniform distribution of amyloid in parenchymal vessels in our material is consistent with the relatively even distribution of CAA-related hemorrhages.1 If confirmed, this observation might help to resolve the apparent discrepancy between the locations of amyloid deposition and hemorrhage in CAA.
Selected Abbreviations and Acronyms
|BTRC||=||Brain Tissue Resource Center|
|CAA||=||cerebral amyloid angiopathy|
|LHE||=||Luxol fast blue–hematoxylin-eosin|
This study was supported in part by an American Academy of Neurology research fellowship (Dr Greenberg) and National Institute of Neurological Disorders and Stroke grant 31862 to the BTRC (Dr Vonsattel). We are grateful to Larry Cherkas for assistance in preparing figures.
- Received February 10, 1997.
- Revision received April 10, 1997.
- Accepted April 10, 1997.
- Copyright © 1997 by American Heart Association
Vinters HV. Cerebral amyloid angiopathy: a critical review. Stroke. 1987;18:311-324.
Kase CS. Cerebral amyloid angiopathy. In: CS Kase, LR Caplan, eds. Intracerebral Hemorrhage. Boston, Mass: Butterworth-Heinemann; 1994:179-200
Greenberg SM, Finklestein SP, Schaefer PW. Petechial hemorrhages accompanying lobar hemorrhages: detection by gradient-echo MRI. Neurology. 1996;46:1751-1754.
Offenbacher H, Fazekas F, Schmidt R, Koch M, Fazekas G, Kapeller P. MR of cerebral abnormalities concomitant with primary intracerebral hematomas. AJNR Am J Neuroradiol.. 1996;17:573-578.
Greenberg SM, Briggs ME, Hyman BT, Kokoris GJ, Takis C, Kanter DS, Kase CS, Pessin MS. Apolipoprotein E ε4 is associated with the presence and earlier onset of hemorrhage in cerebral amyloid angiopathy. Stroke. 1996;27:1333-1337.
Motter R, Vigo-Pelfrey C, Kholodenko D, Barbour R, Johnson-Wood K, Galasko D, Chang L, Miller B, Clark C, Green R, Olson D, Southwick P, Wolfert R, Munroe B, Lieberburg I, Seubert P, Schenk D. Reduction of beta-amyloid peptide42 in the cerebrospinal fluid of patients with Alzheimer’s disease. Ann Neurol. 1995;38:643-648.
Greene GM, Godersky JC, Biller J, Hart MN, Adams HPJ. Surgical experience with cerebral amyloid angiopathy. Stroke. 1990;21:1545-1549.
Matkovic Z, Davis S, Gonzales M, Kalnins R, Masters CL. Surgical risk of hemorrhage in cerebral amyloid angiopathy. Stroke. 1991;22:456-461.
Vinters HV, Gilbert JJ. Cerebral amyloid angiopathy: incidence and complications in the aging brain, II: the distribution of amyloid vascular changes. Stroke. 1983;14:924-928.
Maeda A, Yamada M, Itoh Y, Otomo E, Hayakawa M, Miyatake T. Computer-assisted three-dimensional image analysis of cerebral amyloid angiopathy. Stroke. 1993;24:1857-1864.
Bachman DL, Wolf PA, Linn R, Knoefel JE, Cobb J, Belanger A, D’Agostino RB, White LR. Prevalence of dementia and probable senile dementia of the Alzheimer type in the Framingham Study. Neurology. 1992;42:115-119.
Takebayashi S. Ultrastructural morphometry of hypertensive medial damage in lenticulostriate and other arteries. Stroke. 1985;16:449-53.
Yamada M, Itoh Y, Shintaku M, Kawamura J, Jensson O, Thorsteinsson L, Suematsu N, Matsushita M, Otomo E. Immune reactions associated with cerebral amyloid angiopathy. Stroke. 1996;27:1155-1162.
Yamada M, Itoh Y, Otomo E, Hayakawa M, Miyatake T. Subarachnoid haemorrhage in the elderly: a necropsy study of the association with cerebral amyloid angiopathy. J Neurol Neurosurg Psychiatry. 1993;56:543-547.