Role of cerebral amyloid angiopathy in intracerebral hemorrhage in hypertensive patients

Subject selection.

We investigated brains of patients who had died between 1997 and 2000 in the Neurology Department of the University Hospital of Debrecen, Hungary. One hundred twenty-one patients with the admission diagnosis of ICH had died in the study period. The brains of the patients were stored in 10% formalin according to federal law in the brain archives of the Department of Neuropathology. Brains of patients in the archives were entered in chronologic order in the logbook with name, age, admission diagnosis, and accompanying diseases. We only investigated brains of hypertensive patients. Fifty-seven patients were excluded for the following reasons: 25 patients were not hypertensive, 10 patients had had subarachnoid hemorrhage, two had a vascular malformation, two had metastatic malignancies, and 15 patients had incomplete medical records. In three additional subjects, it was not possible to clarify whether a primary ICH or a hemorrhagic transformation of a primarily ischemic infarction was present. Therefore, 65 patients with primary ICH and confirmed HT were selected from the archives log. Forty-nine patients in this group had died as a direct consequence of the hemorrhage. Other reasons for death were myocardial infarction (1), pneumonia (6), multiorgan failure (3), and pulmonary embolism (5).

For each case with ICH and HT, the next consecutive patient who had died after a case patient was selected as a control if the following conditions were present: 1) patient had never had ICH, 2) patient was hypertensive, 3) patient was of a similar age (±5 years), and 4) complete medical records were available. The controls were not matched for gender. Most of the controls had had ischemic stroke on admission (53 of 65 patients, 82%). The most frequent causes of death were herniation (20), pneumonia (17), pulmonary embolism (9), and myocardial infarction (6). One patient with cerebellar ICH was excluded after the study because cerebellar ICH is difficult to assign to either the lobar or the deep ICH group. In total, there remained 64 cases and 65 controls.

Risk factor analysis.

Complete medical records including a neuropathologic examination and a full postmortem autopsy report of all subjects were available. Charts were reviewed for age, gender, control of HT, bleeding disorders, intake of anticoagulants, and chronic alcoholism. The risk factors were considered present or absent according to the algorithms below.

Location of the bleeding.

The ICHs were classified as being predominantly lobar or deep. Subcortical and cortical ICHs of the cerebral hemispheres and subinsular ICH (hematoma of the insular subcortex with close relation to the basal ganglia) were classified as lobar ICHs. ICHs of the basal ganglia, the capsula interna, and the brainstem were classified as deep ICHs. Cranial CT performed on admission of the patients was available for 109 of the 129 subjects (56 of 64 with ICH, 53 of 65 controls). To classify the location of the bleeding, CT scans were assessed if available. If no CT was available, the appearance of the ICH on gross neuropathology was assessed.

Tissue sampling and preparation.

All brains had been stored in coronal slices of ∼1 cm thickness in 10% formalin until preparation. From each brain, four to five tissue samples were taken from one side of the following regions: 1) frontoparietal lobe (junction of medial frontal gyrus and precentral gyrus), 2) occipital lobe (around calcarine fissure), 3) basal ganglia (at the coronal level of the mamillary bodies), 4) cerebellar hemisphere, and 5) tissue from the presumed area of the ICH origin, if not already covered by one of the first four samples.

In addition to the sample from the bleeding origin, the other four samples per brain were randomly taken from the left or right hemispheres. The side was determined before tissue extraction using a binary, four-digit random number generator with 1 determining the right side and 0 the left side. Tissue blocks were 1 × 1 × 1.5 cm. Care was taken to make sure that cortical samples also contained leptomeningeal tissue. The samples were embedded in paraffin and thin sections of 6 μm were cut and stained with Congo red and hematoxylin-eosin (HE).

Microscopic evaluation.

Stained thin sections were evaluated using light microscopy. The presence and severity of CAA were analyzed based on Congo red–stained samples. Fibrinoid necrosis, wall thickness, and preservation or derangement of vessel architecture was noted in HE-stained samples.

CAA was scored as being definitely present (Congo positive) if a characteristic pink or salmon-like staining was found in the vessel wall and not in the tissue itself.

We graded the severity of CAA using the method proposed by Vonsattel et al.¹⁶ scoring the severity of CAA in single arteries. In short, severity was judged using scores of 0 to 3. A score of 0 was given if there was no CAA in the entire sample, i.e., Congo negative. A score of 1 was given if the most severely affected vessels showed a thin Congo-positive rim restricted to the vessel wall (figure, A). A score of 2 was applied if the entire wall was replaced by amyloid, but the architecture of the vessel was preserved (see figure, B). A score of 3 was given if secondary changes could also be observed in amyloid-positive vessels. These changes included the presence of microbleedings, microaneurysms, fibrinoid necrosis, or a double-barrel appearance^16,19,25,27 (see figure, C).

• Download figure
• Open in new tab
• Download powerpoint

Figure. Severity grading of cerebral amyloid angiopathy. (A) Only a fine rim of salmon-like staining is found in the vessel wall. There is no change in the vessel architecture or thickening of the wall. Score = 1. Also note Alzheimer plaques in the vicinity of the vessels (arrows) (Congo red, ×100). (B) The entire wall of the vessel is dominated by Congo-positive material. There are no vessel fractures or fibrinoid necrosis or paravascular blood. Score = 2 (Congo red, ×100). (C) The vessel architecture (arrow) is destroyed. The wall is Congo positive. Fibrinoid necrosis can be seen (*). Score = 3 (Congo red, ×100; small section in upper left corner, hematoxylin-eosin. ×100).

Statistics.

The primary outcome variable was the presence of CAA in the entire ICH group compared with the non-ICH group using the χ² test. For the secondary evaluation, the entire ICH group was divided into lobar and deep ICH. The presence of CAA was compared between subgroups and the control group using the χ² test.

Frequency of risk factors was also assessed for the entire ICH group and for the subgroups using χ² tests for the dichotomous variable HT management status, bleeding diathesis, anticoagulant therapy, and alcoholism. The Mann–Whitney U test was used to test for significant differences in age.

All the above factors were included in multivariate logistic regression models separately applied to the entire ICH group, the group of lobar ICH, and the group of deep ICH. For these models, age was dichotomized into age ≤ or >70 years. The decision as to which factors to include in the model was made before sample analysis including all variables shown to be important risk factors for ICH.^1,7,8,13

For the comparison of CAA severity, the total scores of all investigated brain regions were calculated for each Congo-positive subject. These total scores were tested for differences between the subgroups and controls using the Mann–Whitney U test. p Values of <0.05 were considered significant.

Risk factor analysis.

Table 1 gives the detailed risk factors for all 129 subjects including p values of the statistical tests.

Single-factor analysis of the entire ICH group vs controls revealed a higher prevalence of CAA in subjects with ICH (15 of 64, 23%) than in controls (5 of 65, 8%, p = 0.026). Uncontrolled HT was also more prevalent in subjects with ICH (34 of 64, 53%) than in controls (19 of 65, 29%, p = 0.01).

Single-factor analysis of the subgroups vs controls showed more subjects with CAA in the lobar ICH group compared with controls (p = 0.007). CAA was not more prevalent in deep ICH compared with controls in single-factor analysis (p = 0.224). Uncontrolled HT was more prevalent in the deep ICH group than in controls (p = 0.016) but not in the lobar ICH group compared with controls (p = 0.11). There was a trend toward a higher prevalence of anticoagulant intake in the deep ICH group compared with controls (p = 0.055).

All recorded risk factors were included in multivariate regression analyses. The results are presented in table 2.

The presence of CAA (odds ratio [OR]: 5.6), uncontrolled HT (OR: 3.1), and anticoagulant intake was significantly more frequent in the entire ICH group than in controls. The statistical association of CAA and lobar hemorrhage was strong (OR: 9.5;, p = 0.001) after logistic regression analysis. The association of CAA and deep ICH was weaker (OR: 4.2, 95% CI: 1.1 to 17.7; p = 0.047) but still significant. In subjects with deep ICH, uncontrolled hypertension (OR: 3.6; p = 0.008) and anticoagulant intake before hospital admission (OR: 26.9; p = 0.007) was also significantly more frequent than in controls but not in subjects with lobar ICH.

Distribution and severity of CAA.

No obvious differences in the severity total scores (Σ) of CAA in the ICH subgroups compared with the control group were detected (entire ICH: Σ = 3.2 ± 2.8, lobar ICH: Σ = 3.9 ± 3, deep ICH: Σ = 2.2 ± 2.4, controls: Σ = 3.6 ± 1.5; p ≫ 0.05).

CAA was predominantly present in the occipital and frontoparietal sections of the brains investigated. In addition, CAA tended to be most severe here whenever it was present in more than one region.

We did not find CAA in any of the basal ganglia samples but found CAA in the cerebellum of five of 20 subjects (four of 15 with ICH, one of five without ICH).