Association between Carotid Artery Plaque Type and Cerebral Microbleeds

Study Design and Patient Population

Institutional review board approval for this study was obtained. Based on a power calculation (type I error, α = 0.05; type II error, β = 0.1; AUC null hypothesis value = 0.5; AUC significant value = 0.75, pooled group), we estimated that a sample size of at least 70 carotid arteries would be sufficient for investigating the relationship between carotid artery plaque type and CMBs. Therefore, 70 consecutive patients (47 men, 23 women; mean age, 65 years; range, 42–86 years) were prospectively enrolled in this study from May 2009 to June 2010. All patients underwent MDCTA of supra-aortic vessels and brain MR imaging.

At our institute (Azienda Ospedaliera Universitaria di Cagliari), inclusion criteria for performing MDCTA of carotid arteries were previously published^22,23 and are as follows: presence of a sonography examination that showed a pathologic stenosis (50% of stenosis according to the NASCET criteria^22⇓–24) and/or a plaque alteration that was considered as the presence of a heterogeneous plaque, an irregular surface, intraplaque hemorrhage, and/or the presence of ulceration in the plaque. Another criterion for performing MDCTA was identified when the ultrasound could not provide adequate information about the degree of stenosis and plaque type. Known allergies to iodinated contrast material or elevated renal function tests were considered as exclusion criteria for the MDCTA examination.

We did not include those patients with a potentially confounding condition (ie, suspected embolism from a cardiac source, follow-up after carotid endarterectomy, intracerebral aneurysms, brain tumors) or with posterior cerebrovascular symptoms in this study.

At our institution, carotid arteries of asymptomatic patients were also studied in those patients who underwent cardiac interventions for coronary artery disease, aortic interventions, and lower-leg artery surgery, and in diabetics who were >50 years old.

Classification of Cerebrovascular Symptoms

Vascular risk factors and coexisting comorbidities and treatment known before stroke or transient ischemic attack are systematically recorded in our institution. Causes of stroke were classified according to Trial of Org 10172 in Acute Stroke Treatment criteria.²⁵ We classified patients as symptomatic or asymptomatic according to the neurologic assessment documented in the clinical chart review. In the case of a TIA or stroke, a patient was considered symptomatic. TIA was regarded as an episode of neurologic dysfunction (hemiparesis, hemiparesthesia, dysarthria, dysphasia, or monocular blindness) not exceeding 24 hours. If the episode of neurologic dysfunction exceeded 24 hours, it was classified as a stroke. The time window to be included in the symptomatic group was 3 months; when a patient underwent MDCTA examination >3 months after the TIA or stroke event, he or she was excluded from the symptomatic group and included in the asymptomatic group. Diabetes was indicated by an abnormal fasting plasma glucose level (>7.9 mmol/l) or the current use of insulin or an oral hypoglycemic agent. Dyslipidemia was defined as abnormal fasting plasma cholesterol (low-attenuation lipoprotein cholesterol) levels (fasting cholesterol >5.0 mmol/l) or the current use of lipid-lowering agents (only if these agents were prescribed for a history of hyperlipidemia or dyslipidemia). Essential hypertension included those patients with systolic blood pressure >140 mm Hg and/or diastolic blood pressure >90 mm Hg, or who were being treated with blood-pressure-lowering drugs. Cigarette smoking status was categorized as never/former (24 months) or current.

MDCTA Technique

All patients underwent MDCTA of the supra-aortic vessels with a 16 multidetector row CT system (Brilliance; Philips, Best, the Netherlands) using a technique previously described.^24,26 Angiographic phase was obtained by injecting 80 mL of contrast medium (Ultravist 370; Bayer, Leverkusen, Germany) at a flow rate of 5 mL/s. A bolus tracking technique was used to calculate the correct timing of the scan. CT technical parameters included matrix 512 × 512; FOV 14–19 cm; mAs 180–220; kV 120–140; an intermediate reconstruction algorithm (C-filter) was used. Angiographic acquisition included the carotid siphon. None of the patients included in the study had a medical history of cardiac output failure or any contraindications to iodinated contrast media.

Brain MR Technique

Imaging examinations were performed with a Gyroscan 1.5T superconducting magnet (Philips) with a head coil, according to a standardized protocol. In all participants, conventional 2D T2*-weighted GRE MR imaging was performed (TR shortest; TE 23 ms; flip angle 15°; in-plane resolution 0.9 × 0.9 mm; FOV 230 mm; matrix 512 × 512; section thickness 5 mm; 0.5 mm intersection gap). Two-dimensional T2*-weighted GRE was performed with flow compensation, with an echo time of 20 msec and a section thickness of 5 mm, in line with conventional settings.^27⇓–29 DWI was performed with single-shot spin-echo with 2 diffusion sensitivity values of 0 and 1000 seconds/mm² along the transverse axis. In addition to T2*-weighted GRE and DWI sequences, axial spin-echo T1-weighted images (500–600/15/2 for TR/TE/excitations) and fast spin-echo T2-weighted images (2200–3200/80–120/1,2 for TR/TE/excitations; turbo factor, 24) were also obtained with the same section thickness.

CMBs Definition and Rating

Acquisition date and participant identification were removed from all images. Two experienced radiologists (L.S. and E.R., with 10 and 8 years of experience, respectively), unaware of the clinical data, reviewed, blinded to each other, all images for presence and location of CMBs. This analysis was performed to evaluate the interobserver variability. After a 3-week gap, datasets were reanalyzed in consensus to obtain the reference parameter for the statistical analysis. At this time, the T1-weighted images were additionally used to confirm the location of microbleeds, as the inherent properties of microbleeds will cause these to appear as a focus of low signal intensity (without marked blooming) on the T1-weighted image as well. Thus, the T1-weighted image further facilitated differentiation of microbleeds from calcification in the ventricle or from sulcal vessels.³⁰ The presence, location, and number of all cerebral microbleeds was recorded. In our study, CMBs were considered as focal, rounded, or circular areas of very low signal intensity that were smaller than 10 mm in size, as suggested by Jeerakathil et al³¹ (Fig 1). In accordance with consensus guidelines on standards for neuroimaging in vascular cognitive impairment, we considered 2 mm as the lower size limit.³² We also categorized the locations of the CMBs^27,31 as lobar (cortical gray matter and subcortical white matter), deep (deep gray matter [basal ganglia and thalamus] and deep white matter [corpus callosum]), or infratentorial (brain stem and cerebellum). Symmetric hypointensities in the globi pallidi, likely to represent calcification, and sulcal flow voids from cortical vessels were disregarded.³³.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

MR axial images that demonstrate the presence of CMBs. The white arrows indicate the CMBs.

Microbleeds were classified in each cerebral hemisphere (and for infratentorial lesions, these were considered in the right/left cerebellum of right/left side of the brain stem) as absent (grade 1), mild (grade 2; total number of microbleeds, 1–2), moderate (grade 3; total number of microbleeds, 3–10), and severe (grade 4; total number of microbleeds, >10), according to Lee and colleagues.³⁴

Carotid Artery Plaque Classification

A single experienced radiologist (L.S., 10 years of experience), blinded to the clinical information and MR brain analysis, performed all MDCTA measurements. MDCTA axial images were analyzed, with a varying magnification from 120%–400%, in comparison with the acquisition. Window parameters (width and level) were not preset.^18,35 We categorized different types of plaques, dividing them into fatty plaques, with attenuation <60 HU; mixed plaques, with attenuation between 60–120 HU; and calcified plaques, with attenuation >130 HU (Fig 2 ). To measure HU value, a circular or elliptical ROI cursor in the predominant area of plaque was used by the method outlined according to Saba et al.¹² Regions of beam hardening in calcified areas were also excluded. The radiologist took the measurements at the point of major stenosis, which was identified by elaborating oblique axial images normal to the ICA lumen centerline.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

CTA axial images that demonstrate (A) fatty plaque, (B) mixed plaque, and (C) calcified plaque.

Statistical Analysis

For the analysis of the data, we considered the carotid and cerebral hemisphere of each side as an independent unit. Therefore, the total number of studied carotid arteries and brain hemispheres was 140. In symptomatic patients, we considered the ipsilateral carotid artery and brain hemisphere as “symptomatic,” and the contralateral carotid artery and brain hemisphere as “asymptomatic,” whereas in asymptomatic patients, both carotids and brain hemispheres were considered “asymptomatic.” In symptomatic patients, we considered a symptomatic side and asymptomatic side according to the clinical symptoms, whereas in asymptomatic patients, both sides were considered asymptomatic. We tested the interobserver reliability for CMB rating using the Cohen κ test.³⁶ Continuous data were described as the mean value ± SD and were compared using the Wilcoxon signed rank test. ROC curve analysis was performed to identify thresholds in number of CMBs associated with presence of symptoms. The Student t test for paired samples was used to test the difference in the number of CMBs in the same subjects between the side with the fatty plaques and the side without the fatty plaques. Presence or absence of cerebrovascular symptoms was compared between patients with and without CMBs, according to the different plaque types, using a Cochran-Mantel-Haenszel test. Logistic regression and stepwise logistic regression analysis was used to identify the most accurate linear combinations of distinguishing variables. R software (http://www.r-project.org) was used for statistical analyses.

Relationship of Cerebrovascular Symptoms and CMBs

We performed an analysis to evaluate the relationship between the number of CMBs (and their grade, from 1 to 4) and the presence of cerebrovascular symptoms. We observed a statistically significant correlation between the presence of symptoms and the number of CMB (correlation coefficient = 0.319; P value = .0001) and the CMBs grade (correlation coefficient = 0.314; P value = .0001). We also analyzed the association between the different subtypes of cerebrovascular symptoms (stroke, TIA, and amaurosis) with the presence of CMBs, and found a strong association between stroke (P = .0001; OR = 10.71), but no association with TIA (P = .347; OR = 1.62) and amaurosis (P = .895; OR = 1.12).

Relationship of Plaque Type and CMBs

In the plaque type analysis of the 140 carotid arteries, we excluded 14 because there was no evidence of plaque. In the remaining 126 carotid arteries, we found 37 fatty plaques (29.4%), 49 mixed plaques (38.9%), and 40 calcified plaques (31.8%). In 2 cases CMBs were found in the 14 excluded hemispheres, supplied by an regular ICA, and therefore the plaque type analysis was performed by considering 35 hemispheres with CMBs. All 39 symptomatic hemispheres had a plaque in the carotid artery and therefore were included in the analysis. A summary of the plaque type analysis, according to the presence of CMBs and symptoms, is given in Table 2. The Cochran-Mantel-Haenszel test showed that there was a statistically significant increase (P = .012; OR = 3.465) in CMB prevalence in symptomatic patients versus that in asymptomatic patients.

In analyzing the relationship between CMBs and the type of plaque, we performed a correlation analysis and found a statistically significant correlation with the fatty plaques (correlation coefficient = 0.271; P value = .0019), whereas an inverse statistical trend was observed between CMBs and calcified plaques (correlation coefficient = −0.152; P value = .0851). No statistically significant association was detected between CMBs and mixed plaque type (correlation coefficient = −0.117; P value = .188).

In analyzing whether CMBs are an expression of diffuse disease or the consequence of an unstable plaque, we performed a Student t test for paired samples and observed a statistical trend (P value of .0699).

ROC Curve Analysis

The ROC curve analysis for symptoms versus number of CMBs is shown in Fig 3A. The ROC area under the curve (Az) for the direct-mm method was 0.646 (95% CI 0.561–0.724; SD = 0.054; P = .0069). Using the ROC analysis, we calculated the sensitivity, specificity, positive likelihood ratio and negative likelihood ratio values for the direct-mm method; a summary of these values is provided in Table 3. By analyzing threshold values, we observed that the presence of 5 CMBs are associated with a specificity of 90.2% for the presence of cerebrovascular symptoms. We also performed the ROC curve analysis for stroke, TIA, and amaurosis (Fig 3B-D) and we obtained the following AUCs, respectively: 0.802 (P = .0001), 0.464 (P = .61), and 0.526 (P = .819), by confirming that there is an association between CMBs and stroke but not between CMBs and TIA or amaurosis

• Download figure
• Open in new tab
• Download powerpoint

Fig 3.

ROC curves between presence of CMBs and (A) cerebrovascular symptoms, (B) stroke, (C) TIA, and (D) amaurosis.

Logistic Regression Analysis

To exclude any confounding effect between CMBs and other potentially related variables, logistic regression analysis was performed with CMB as the dependent variable, plaque type as the independent variable, and other factors such as hypertension, diabetes, tobacco use, coronary artery disease, and dyslipidemia as cofactors (Table 4). This analysis demonstrated that the only statististically significant association was present between CMBs and fatty plaques. Next, the stepwise logistic regression analysis was performed; the results are provided in Table 5. In the stepwise logistic regression model, the dependent variable was the presence of cerebrovascular symptoms, and we included the following parameters: sex of the patient, coronary artery disease, diabetes, hypertension, dyslipidemia, tobacco use, type of plaque (fatty, mixed, calcified), and presence of CMBs. The stepwise logistic regression model indicated that CMBs and fatty plaques were both independent variables associated with cerebrovascular symptoms (P values, respectively, .0026 and .0018). Other variables did not demonstrate a significant statistical association with cerebrovascular symptoms.

Results Based on CMB Location

In 13 cases (9.3%), we had only lobar CMBs; in 14 cases (10%), we found only deep CMBs; whereas in 9 cases (8.2%), we found combined (both deep and lobar) CMBs. We observed CMBs in the cerebellum (right side) in only 1 case (0.7%).