Identification of Patients with a Histologically Unstable Carotid Plaque UsingUltrasonic Plaque Image Analysis

M.K. Salem a,*, M.J. Bown a,b, R.D. Sayers a, K. West c, D. Moore c, A. Nicolaides d,e, T.G. Robinson b, A.R. Naylor a

a Department of Cardiovascular Sciences, University of Leicester, Leicester, UKb NIHR Leicester Cardiovascular Biomedical Research Unit, Leicester, UKc Department of Histopathology, University Hospitals Leicester, Leicester, UKd Department of Surgery, St George’s University of London, London, UKe University of Nicosia Medical School, Nicosia, Cyprus

* CorCardiovLeicestE-ma1078

Surgeryhttp:

WHAT THIS PAPER ADDS

If validated, it may be possible to identify patients with a higher or lower likelihood of having a histologicallyunstable plaque using readily accessible ultrasound images and plaque software analysis.

Objectives: In patients with carotid stenosis the risk of stroke is highest in the first few days after onset ofsymptoms and it is low in asymptomatic patients. The ability to identify patients with a high (or low) probabilityof having a histologically unstable plaque might become a complimentary method that can refine the indicationsfor surgical intervention.Methods: Two histopathologists, using validated American Heart Association criteria, independently gradedplaques harvested during carotid endarterectomy. Preoperative Duplex images were independently assessed forjuxtaluminal black area, plaque type, plaque area, and grey-scale median (GSM) following image normalization.Logistic regression analysis was then performed to create a model for predicting predominantly histologicallyunstable or stable plaques.Results: A total of 126 patients were included in the study. Based on the presence and extent of histologicalfeatures including haemorrhage, thrombus, fibrous tissue, lipid core, inflammation, neovascularity, foam cells,and cap rupture, 39 plaques were graded as predominantly stable, while 87 were predominantly unstable.Unstable plaques were associated with a plaque area >95 mm2 (OR 4.15; 95% CI 1.34e12.8 p ¼ .009), ajuxtaluminal black area >6 mm2 (OR 2.77; 95% CI 1.24 to 6.17 p ¼ .01) and a GSM <25 (OR 3.76; 95% CI 1.14e12.39). Logistic regression indicated that patients with the first two features had a 90% probability of having ahistologically unstable plaque. The model was used to calculate the probability of having an unstable plaque ineach patient. The receiver operating characteristic curve using the p value was 0.68 (95% CI 0.59e0.78).Conclusions: Computerized plaque analysis has the potential to identify patients with histologically unstablecarotid plaques. This model requires validation, but offers the potential to influence patient selection foremergency interventions and the monitoring of medical therapy.� 2014 European Society for Vascular Surgery. Published by Elsevier Ltd. All rights reserved.Article history: Received 30 January 2014, Accepted 7 May 2014, Available online 16 June 2014Keywords: Carotid, Plaque instability, Ultrasound, Atherosclerosis

INTRODUCTION

Guidelines1,2 for treating patients with carotid artery dis-ease rely primarily on stenosis severity and clinical symp-toms. The majority of strokes in patients with moderate tosevere carotid stenoses are caused by thromboembolismfrom an unstable carotid plaque,3 defined as having severalof the following features: thinned/ruptured cap, large lipidcore, intraplaque haemorrhage, surface thrombus, cap/

responding author. M.K. Salem, Vascular Surgery Group, Division ofascular Sciences, Robert Kilpatrick Clinical Sciences Building,er Royal Infirmary, Leicester LE2 7LX, UK.il address: ms447@le.ac.uk (M.K. Salem).-5884/$ e see front matter � 2014 European Society for Vascular. Published by Elsevier Ltd. All rights reserved.//dx.doi.org/10.1016/j.ejvs.2014.05.015

plaque inflammation, and marked neovascularization.4,5

These features (especially dense macrophage inflamma-tion) were recently shown to be associated with a high riskof early recurrent stroke in patients with 50e99% stenoses,leading the authors to recommend that imaging strategiesbe developed for identifying patients with unstable plaquefeatures.6

Computed tomography (CT),7,8 magnetic resonance im-aging (MRI),9 positron emission tomography (PET), andinfrared spectroscopy10 have been used to identify the“unstable plaque”. Whilst showing reasonable concordancewith histological findings, most are not as accessible as ul-trasound. In earlier ultrasound studies, hypoechoic plaques(Types 1 and 2) were associated with symptoms.11,12 Morerecent studies using ultrasound and plaque characterization

European Journal of Vascular and Endovascular Surgery Volume 48 Issue 2 p. 118e125 August/2014 119

following image normalization have identified plaque fea-tures that were independent predictors of future cerebro-vascular events. It was hypothesized that these mightenable risk stratification in patients with asymptomatic ca-rotid stenosis.13,14 These features included; low grey scalemedian (GSM), larger plaque area, and the presence/size ofthe juxtaluminal plaque area. When the same scrutiny wasapplied to images of plaques from symptomatic patients, itidentified a subgroup with a higher rate of recurrent tran-sient ischaemic attacks (TIAs) whilst awaiting surgery.15

The aim of this study was to determine the associationbetween ultrasound-derived plaque features and histologi-cal plaque instability. We hypothesized that the ability toidentify a high (or low) probability of unstable plaque mightprovide an alternative (and accessible) method for planningthe nature and urgency of carotid interventions, as well asbeing a means of monitoring the effects of medical therapyon plaque morphology in other patient cohorts.

METHODS

Consecutive patients undergoing carotid endarterectomy(CEA) between August 2008 and March 2010 were includedafter giving informed consent. The study was authorized bythe Leicestershire, Northamptonshire, and Rutland EthicsCommittee. Patients were included if they were recentlysymptomatic (<6 months) with a 50e99% NASCET steno-sis,16 or asymptomatic with a 70e99% stenosis.

Symptomatic patients were admitted directly from thedaily TIA Clinic or from the Stroke Unit. Symptomatic pa-tients were started on 300 mg of aspirin and 40 mg of

Figure 1. (A) “Unstable” plaque detected on ultrasound analysis. Juxtalubright red areas within the black lumen. (B) “Stable” plaque features dwhite. DWAs are bright red areas within the black lumen.

simvastatin prior to transfer, and these medications werecontinued throughout the peri-operative period. In addi-tion, all patients received 75 mg of clopidogrel the nightbefore surgery.17 A fuller description of the reconfigured TIAclinic has been published elsewhere.18 CEA was performedas soon as possible after the index event. Asymptomaticpatients were started on medical therapy in the outpatientdepartment. Most had been taking aspirin, statin, and anti-hypertensive therapy for at least 4 weeks prior to surgery.

Duplex imaging

Accredited ultrasonographers performed the duplex exam-inations using an ATL HDI5000 ultrasound scanner and anL12e5 linear array probe (Philips Medical Systems, Andover,MA, USA). Stenosis severity was measured using NASCET-derived measurement criteria.16 In addition to gradingstenosis severity, B-mode longitudinal unenhanced imagesof the plaque were obtained with and without colourflow.14,19 Image analysis was then performed by a singleinvestigator (A.N.) who was completely blinded to all clinicaland histological data.

Image normalization and measurements of GSM and JBA(see Fig. 1a,b)

Image analysis was performed using the “Plaque TextureAnalysis software” (LifeQ Medical, 66 Metochiou, Engomi,2407, Cyprus; www.lifeqmedical.com), which is a dedicatedsoftware package.14 Image normalization was performed sothat the area representing blood had a grey scale of zero,whilst the brightest area of adventitia had a grey scale of

minal black area outlined in white. Discrete white areas (DWAs) areetected on ultrasound analysis. Juxtaluminal black area outlined in

120 M.K. Salem et al.

190. This was achieved by selecting a sample of “blood”from the vessel lumen avoiding areas of “noise” and thenthe middle two-fourths of the brightest part of the adven-titia adjacent to the plaque.

Ultrasound-derived plaque features

Grey-scale median.Median of the grey values of all pixels inthe plaque image.19

Plaque type (computer derived).

Figurethrom

Type 1. Uniformly echolucent (black): <15% of thepixels in the plaque area were occupied by pixels withgrey scale values >25.Type 2. Mainly echolucent: pixels with grey scalevalues >25 occupy 15e50% of the plaque area.Type 3. Mainly echogenic: pixels with grey scalevalues >25 occupy 50e85% of the plaque area.Type 4/5. Uniformly echogenic: pixels with grey scalevalues >25 occupy >85% of the plaque area.11,12

Plaque area (mm2). Calculated by the imaging softwareusing distance scale on the side of the image frame forcalibration and the plaque area outlined by the operator.

Juxtaluminal black areas (Fig. 1a,b). The juxtaluminal blackarea (JBA) (defined as the area adjacent to the lumen withpixels which after image normalization had a grey scale <25without a visible echogenic cap) was outlined and the areaautomatically calculated. If there were >2 discrete JBAs, thelarger value was used in the analyses. A large JBA has beenpreviously associated with a high prevalence of symptom-atic plaques in all grades of stenosis.20

Carotid endarterectomy

Carotid endarterectomy was performed under generalanaesthesia with routine patching, routine shunting,

2. Unstable plaque on haematoxylin and eosin staining showinbus (ST) (�10) and histological grading.

systemic heparinization (unfractionated) and distal intimaltacking sutures. Endarterectomy was performed via a lon-gitudinal arteriotomy and the plaque removed with theminimum of trauma. The lumen of the endarterectomizedsegment of artery was inspected using an angioscope priorto restoration of flow.

Histopathology

Plaques harvested during CEA were divided longitudinally.One half was immediately fixed in formalin for 24 hours andthen paraffin embedded. A 5-mm transverse section wastaken from the paraffin block and stained using haema-toxylin and eosin. Histological sections were analysed bytwo independent histopathologists (K.W., D.M.) who werecompletely blinded to the clinical and ultrasound findings.For each plaque (Fig. 2), a semi-quantitative 3- or 4-pointscore (Table 1) was assigned to assess the presence and/or amount of features of plaque instability, based on theAHA histological classification.4,21 Based on number of un-stable/stable features the plaque was subjectively scored byeach histopathologist independently as overall predomi-nantly stable or unstable. Where a difference was notedbetween scores, slides were placed on a large screen, andagreement reached after discussion.

Statistical analyses

The association between ultrasound-derived features (ste-nosis, GSM, plaque area, JBA, plaque type) and histologi-cally unstable/stable plaques was investigated by calculatingthe odds ratio for each group of the nominal variables orquartiles of the continuous variables. On the basis of thesignificant associations found between features, the cut-offpoints for the upper quartiles were then used to reclassifyeach of these features into two classes. These significantplaque features were then entered as dichotomous vari-ables into a multivariable logistic regression with the his-tological classification as the dependent variable using the

g cap rupture (arrow), intraplaque haemorrhage (IPH) and surface

Table 2. Demographics for patients with histologically stable andunstable plaques.

Patient demographics StableN ¼ 39(31%)

UnstableN ¼ 87(69%)

p Value

Male 23 (59%) 65 (75%) .09Age (median) 73 (48e90) 72 (66e96) .97Hypercholesterolaemiaa 29 (74%) 66 (76%) >.99Diabetes mellitus 10 (26%) 16 (18%) .35Ischaemic heart diseaseb 5 (13%) 24 (28%) .11Hypertensionc 30 (77%) 55 (87%) .15Ex/current smoker 30 (77%) 63 (72%) .67On anti-platelettherapy following admission

39 (100%) 87 (100%) >.99

On statin therapy followingadmission

39 (100%) 87 (100%) >.99

Symptomatic 31 (79%) 73 (83%) .80a On treatment by family GP/Hospital doctor for raised cholesterollevels.b Angiography/ECG/Biochemistry proven evidence of coronaryartery disease and on treatment.c On treatment for hypertension initiated by GP/Hospital doctor.

Table 1. Semi-quantitative scoring system for histological carotid plaque grading.19

Histological feature Grade 0 Grade 1 Grade 2 Grade 3Haemorrhage No haemorrhage Small haemorrhage Large haemorrhage eThrombus No thrombus Small thrombus Large thrombus eLipid core No lipid core Small lipid core Large lipid core eFibrous tissue Very little fibrous

tissueApprox. 50%fibrous tissue

Predominantlyfibrous tissue

e

Chronic plaque inflammation None Occasional cellsor one group �50

2e5 groups �50 �5 groups �50 or1 group �500

Chronic cap inflammation None <10 cells in cap 10e50 cells in cap �50 cells in capAcute plaque inflammation None Occasional cells

or one group �502e5 groups �50 >5 groups �50 or

1 group �500Acute cap inflammation None <10 cells in cap 10e50 cells in cap �50 cells in capFoam cells None <50 cells �50 cells eNeovascularity None <10/section �10/section eCap rupture Intact Probably intact Probably ruptured Definitely ruptured

European Journal of Vascular and Endovascular Surgery Volume 48 Issue 2 p. 118e125 August/2014 121

backward elimination method. Subsequently, the indepen-dent predictors were used as a final model to predict theprobability of having an unstable plaque for each patient. Areceiver operating characteristic (ROC) curve was thenproduced using calculated probability. Finally, a 2 � 2 tablewas constructed using the significant independent variablesto classify patients into four groups. The prevalence of un-stable plaques was determined for each of the four groupsof patients. The significance of the prevalence was thentested using the chi-square test and its magnitude using theodds ratio and 95% CI.

RESULTS

One hundred and sixty-nine patients were considered forinclusion. Twenty images from the initial part of the studywere of insufficient quality for computerized plaque analysisdue to problems associated with image acquisition (ultra-sound beam at incorrect angle, too much noise in lumen,inadequate visualization of the plaque and adjacentadventitia). An additional 13 images were lost during datatransfer. Ten patients were excluded as histology was un-obtainable. This left 126 patients for inclusion (TIA/monocular blindness, 75; stroke, 29; asymptomatic, 22).Forty-nine of 104 symptomatic patients (47%) underwentCEA <7 days after their most recent clinical event; 32 (31%)<8e14 days, 12 (12%) <15e28 days, while >29 days hadelapsed in 11 patients (11%). This, therefore, provided abroad spectrum of patients with regard to delays to surgery(<14 days, n ¼ 81), intermediate (15e28 days, n ¼ 12), late(>29 days, n ¼ 11) and asymptomatic (n ¼ 22). Patientdemographics (Table 2) were well matched among patientswho were found to have predominantly unstable plaques,compared to those with predominantly stable plaque fea-tures. The kappa value for interobserver variation betweenthe two histopathologists was 0.93 (95% CI 0.85e0.99).

Table 3 shows that GSM <25, plaque area >95 mm2, anda JBA >6 mm2 were associated with an increased preva-lence of unstable plaque features. Stenosis and Geroulakosplaque types were not associated with plaque instability. Onthe basis of the cut-off points provided by the upperquartiles of GSM and plaque area, and median JBA, a new

table was constructed (Table 4) which shows the magnitudeof the association. A multivariate logistic regression usingthe histological classification as the dependent variable andGSM, plaque area, and JBA as dichotomous explanatoryvariables using the backward elimination method showedthat only plaque area >95 mm2 and a JBA >6 mm2 werestatistically significant independent predictors of the pres-ence of a histologically unstable plaque. The model inTable 5 is a predictive model using only the above signifi-cant variables. This model was used to calculate the prob-ability of having an unstable plaque for each patient. TheROC curve using the value of p was 0.68 (95% CI 0.59e0.78).

Table 6 shows the prevalence of unstable plaques in thefour groups of patients defined by the dichotomous vari-ables: plaque area and JBA. When plaque area is <95 mm2,the presence of a JBA >6 mm2 increases the prevalence ofunstable plaques from 53% (when it is absent) to 76%. Italso shows that when the JBA is <6 mm2 the presence of a

Table 3. Odds ratios and 95% CI of unstable plaques in quartiles or groups of the ultrasonic features.

Ultrasonic measurement Range of measurement Patients studied N Unstable plaque N (%) p OR (95% CI)Stenosis (%)Three groups according to duplex grading

1 50e79 53 37 (70%)2 80e89 34 22 (65%) .619 0.79 (0.32 to 1.98)3 90e99 39 28 (72%) .83 1.10 (0.44 to 2.74)

p for trend ¼ .879GSMQuartiles

1 0e24 31 26 (84%) .030 3.76 (1.14 to 12.39)2 25e37 36 24 (67%) .469 1.44 (0.53 to 3.90)3 38e49 28 19 (68%) .43 1.52 (0.52 to 4.43)4 50e93 31 18 (58%)

p for trend [ .044Plaque area (mm2)Quartiles

1 0e27 31 17 (55%)2 28e50 32 22 (69%) .258 1.81 (0.65 to 5.07)3 51e95 33 22 (67%) 334 1.65 (0.60 to 4.53)4 96e178 30 26 (87%) .010 5.35 (1.50 to 19.03)

p for trend [ .015JBA (mm2)Quartiles

1 0e2.0 37 21 (57%)2 2.1e6.0 29 18 (62%) .663 1.25 (0.46 to 3.36)3 6.1e14.0 31 25 (81%) .040 3.17 (1.05 to 9.58)4 14.1e48 39 23 (80%) .058 2.92 (0.96 to 8.85)

p for trend [ .015Plaque typeHyperechoic 3 and 4 90 58 (64%) eHypoechoic 1 and 2 36 29 (80%) .082 2.29 (0.90 to 5.80)

122 M.K. Salem et al.

plaque area >95 mm2 increases the prevalence of unstableplaques from 53% (when it is absent) to 85%. In patientswith a plaque area >95 mm2 and a JBA >6 mm2, theprevalence of unstable plaque was 90%.

DISCUSSION

Guidelines1,2 for the management of symptomatic carotidartery disease are primarily based upon evidence frommeta-analyses of randomized trials showing that CEA con-fers significant benefit over medical therapy. Subgroup an-alyses have shown that the highest risk of stroke is in thevery early time period after onset of symptoms.22,23 Pa-tients suffering recurrent stroke (within hours/days of theindex event) have recently been found to have histologicalfeatures associated with plaque instability (especially dense

Table 4. Odds ratios and 95% CI of unstable plaques in the reclassifie

Ultrasonic measurement Range ofmeasurement

Patientsstudied N

GSM�25 0e25 95>25 26e93 31Plaque area (mm2)�95 0e95 94>95 96e178 32JBA (mm2)�6 0e6.0 66>6 6.1e48 60

macrophage inflammation).6 To date, easily accessible and/or reliable methods for predicting the likelihood of plaqueinstability have not been immediately available. In asymp-tomatic patients, randomized trials have shown that CEAdoes confer benefit (although much less than in symp-tomatic patients), though there is emerging evidence thatimprovements in medical therapy may have reduced theaverage annual rate of sroke.24

Notwithstanding the benefits conferred by CEA, therandomized trials have shown that the majority of patientswere never actually destined to suffer a stroke (88% ofasymptomatic patients with 60e99% stenoses were strokefree at 5 years; 84% of symptomatic patients with 70e99%stenoses were stroke free at 5 years). Accordingly, there hasbeen considerable interest in developing imaging criteria

d groups of the ultrasonic features.

Unstableplaque N (%)

p OR (95% CI)

61 (64%)26 (84%) .040 2.90 (1.02 to 8.24)

59 (63%)28 (87%) .009 4.15 (1.34 to 12.83)

39 (59%)48 (80%) .011 2.77 (1.24 to 6.17)

Table 5. Final linear logistic regression model using the significantfeatures plaque area and JBA with presence of unstable plaques asthe dependent variable.

Variable b OR (95% CI) p ValueJBA (�6 or >6 mm2) 0.931 2.54 (1.12 to 5.75) .026Plaque area(�95 or >95)

1.327 3.77 (1.20 to 11.84) .023

European Journal of Vascular and Endovascular Surgery Volume 48 Issue 2 p. 118e125 August/2014 123

that might identify “high risk for stroke” patients in whomto target CEA or stenting. Conversely, if it were possible toidentify “low risk for stroke” patients, they might be bettertreated medically and kept under surveillance.

Histological characterization can identify features associ-ated with plaque instability, but there has been inconsistentcorrelation between preoperative imaging and post-operative plaque analysis.25,26 Contrast-enhanced ultra-sound (CEUS) has been used to detect intraplaqueneovascularization.27,28 A recent review of six CEUS studieswith correlation against histological features demonstratedgood correlation between CEUS and histological diagnosesof neovascularization.29 An MRI diagnosis of intraplaquehaemorrhage (IPH) has also shown positive correlation be-tween recurrent clinical events30 and increased thrombo-embolic activity.31 There is also evidence that an MRIdiagnosis of IPH may identify asymptomatic patients with ahigher risk of late stroke.32 Preoperative MRI and histolog-ical correlation with “vulnerable” plaque features (includingintraplaque haemorrhage, large lipid core, and fibrous cap)have shown good sensitivity and specificity, but the tech-nique has had limited clinical applications outside theresearch environment.7,33 Not surprisingly, therefore,evidence-based guidelines do not include any role for so-phisticated imaging techniques for guiding patientselection.

By contrast, Duplex ultrasound is accessible, especially inthe outpatient setting. In this study, ultrasound imagingfeatures on univariate analysis that significantly predictedan unstable carotid plaque included: an echolucent plaque(GSM <25), a large juxtaluminal black area (>6 mm2), and alarge plaque area (>95 mm2). Geroulakos plaque Types 1and 2 showed a trend towards plaque instability, but did notreach statistical significance. Following multivariate analysis,

Table 6. The prevalence of histologically unstable plaques shown inbold in the 4 groups (cells a, b, c, and d) defined by Plaque Areaand JBA.

JBA <6 mm2 JBA >6 mm2

Plaque area >95 mm2 11 (85%)N ¼ 13 c

17 (90%)N ¼ 19 d

Plaque area <95 mm2 28 (53%)N ¼ 53 a

31 (76%)N ¼ 41 b

Comparison p (chi square) OR 95% CIa vs. b .023 2.77 1.32 to 6.77a vs. c .037 4.91 0.99 to 24.33a vs. d .005 7.59 1.59 to 36.16b vs. d .212 2.74 0.53 to 13.98c vs. d .683 1.54 0.19 to 12.64

only two ultrasound texture features remained significantlyassociated with an increased risk of having a histologicallyunstable plaque: (a) large plaque area (�95 mm2) and (b) alarge JBA (�6 mm2). These two features were also inde-pendent predictors for stroke in a recently published naturalhistory study (ACSRS) involving 1,121 patients withasymptomatic carotid stenosis followed up for a mean of 4years. When combined with degree of stenosis, these pla-que features were able to stratify the annual risk of strokein different asymptomatic patient cohorts from 0.5% to10%.13,14 This study has demonstrated that a plaque with alow GSM is dangerous but if the plaque area with low GSMis close to the lumen it is far more dangerous, so muchmore dangerous that GSM does not add more to the pre-diction of a histologically unstable plaque. In addition, fortwo plaques producing the same degree of stenosis, thelonger one which has a larger area on a longitudinal ultra-sound image is associated with a higher risk than the plaquewith a shorter length (smaller area) producing a localizedstenosis.

Clinical scoring methods for predicting early risk of strokefollowing TIA/minor stroke include the ABCD2 score(maximum score 7),34 the Essen Stroke Risk Score (ESRS) (9-point system)35 and the Stroke Prognosis Instrument II (SPI-II) score (15-point system).36 All have a predictive value inpatients who present with a TIA/minor stroke, though lim-itations have been found, including poor short-term pre-dictive value (ESRS and SPI-II) and not being specific topatients with a critical carotid artery stenosis (all).

The combination of carotid imaging and brain infarctionhas improved the prediction of early recurrent stroke. Gileset al.37 showed that even with a high ABCD2 score (6e7),the risk of stroke at 7 days was only 3.3% in patients whohad no evidence of infarction on neuroimaging, comparedwith 15% in patients who had a high ABCD2 score andrecent infarction. Merwick et al.38 demonstrated that thepresence of a 50e99% carotid stenosis in combination withabnormal diffusion-weighted imaging increased the likeli-hood of stroke following TIA. One limitation of Merwick’sstudy was that insufficient statistical power existed for lo-gistic regression of carotid imaging at 2 and 7 days, becauserelatively few patients underwent carotid imaging at thisvery early time period.

One of the limitations of the current study was that 20images were not of sufficient quality. These images weretaken during the earliest part of the study and reflect alearning curve (noise, angle of beam, and plaque area).After our technologists went through a 30-minute onlinestudy module, 100% of images were used in the study.

Ultrasonography, performed by well-trained, experiencedtechnologists, can provide an accurate and relatively inex-pensive assessment of the carotid arteries. The technique isnon-invasive and does not expose patients to ionizing ra-diation or potentially nephrotoxic contrast material. B-mode ultrasound has previously been used to identify high-risk plaques by identifying plaque echolucency which hasbeen linked to future ischaemic events.39 In a study byNicolaides et al.14 ultrasound plaque features, including

124 M.K. Salem et al.

GSM and plaque area, were independent predictors ofipsilateral stroke risk in asymptomatic individuals. However,no one (to date) has correlated computerized plaque find-ings with histology. The ACSRS study also found thatasymptomatic carotid stenosis and a history of contralateralTIAs were independent predictors of future stroke risk. Theadvantage of the current study is that it allows for predic-tion of plaque histology using ultrasound plaque features.

At present, this predictive scoring system requires vali-dation, but if it is shown to reliably identify unstable pla-ques, it has the potential to select high-risk patients foremergency surgery/stenting, whilst patients with a lowprobability of having an unstable plaque may benefit morefrom medical therapy. An added advantage is that the al-gorithm would also permit serial surveillance of plaquemorphology in patients treated medically.

In conclusion, this is one of the first studies to show thatit might be possible to identify patients with a higher orlower likelihood of having an unstable carotid plaque (onhistology) using accessible imaging technology.

FUNDING

We thank the consultants in the vascular surgery unit andthe staff in the vascular studies unit at Leicester Royal In-firmary for their help in this study, which was funded bygrants from the UK Circulation Foundation (the researchcharity affiliated to the Vascular Society of Great Britain andIreland) and a fellowship awarded from the ‘NationalInstitute for Health Research Collaborations for Leadershipin Applied Health Research and Care’ (NIHR CLAHRC).

CONFLICT OF INTEREST

None.

REFERENCES

1 Liapis CD, Bell PRF, Mikhailidis D, Sivenius J, Nicolaides A,Fernandes e Fernandes J, et al. ESVS guidelines. Invasivetreatment for carotid stenosis: indications, techniques. Eur JVasc Endovasc Surg 2009;37:S1e19.

2 Brott TG, Halperin JL, Abbara S, Bacharach JM, Barr JD, Bush RL,et al. 2011 ASA/ACCF/AHA/AANN/AANS/ACR/ASNR/CNS/SAIP/SCAI/SIR/SNIS/SVM/SVS guideline on the management of pa-tients with extracranial carotid and vertebral artery disease.Circulation 2011;124:e54e130.

3 Barnett HJM, Gunton R, Eliasziw M, Fleming L, Sharpe B,Gates P, et al. Causes and severity of ischemic stroke in patientswith internal carotid artery stenosis. JAMA 2000;283:1429e36.

4 Stary HC, Chandler AB, Dinsmore RE, Fuster V, Glagov S,Insull W, et al. A definition of advanced types of atheroscleroticlesions and a histological classification of atherosclerosis. Areport from the Committee on Vascular Lesions of the Councilon Arteriosclerosis, American Heart Association. Circulation1995;92(5):1355e74.

5 Golledge J, Greenhalgh RM, Davies AH. The symptomatic ca-rotid plaque. Stroke 2000;31(3):774e81.

6 Marnane M, Prendeville S, McDonnell C, Noone I, Barry M,Crowe M, et al. Plaque inflammation and unstable morphologyare associated with early stroke recurrence in symptomaticcarotid stenosis. Stroke 2014;45:801e6.

7 Millon A, Boussel L, Brevet M, Mathevet JL, Canet-Soulas E,Mory C, et al. Clinical and histological significance of gadolin-ium enhancement in carotid atherosclerotic plaque. Stroke2012;43(11):3023e8.

8 Adraktas DD, Tong E, Furtado AD, Cheng SC, Wintermark M.Evolution of CT imaging features of carotid atheroscleroticplaques in a 1-year prospective cohort study. J Neuroimaging2014;24:1e6.

9 Parmar JP, Rogers WJ, Mugler 3rd JP, Baskurt E, Altes TA,Nandalur KR, et al. Magnetic resonance imaging of carotidatherosclerotic plaque in clinically suspected acute transientischemic attack and acute ischemic stroke. Circulation 2010Nov 16;122(20):2031e8.

10 Wallis de Vries BM, van Dam GM, Tio RA, Hillebrands JL,Slart RH, Zeebregts CJ. Current imaging modalities to visualizevulnerability within the atherosclerotic carotid plaque. J VascSurg 2008;48(6):1620e9.

11 Geroulakos G, Ramaswami G, Nicolaides A, James K,Labropoulos N, Belcaro G, et al. Characterisation of symp-tomatic and asymptomatic carotid plaques using high-resolution real-time ultrasonography. Br J Surg 1993;80:1274e7.

12 Nicolaides AN, Kakkos SK, Griffin M, Sabetai M, Dhanjil S,Thomas DJ, et al. Effect of image normalization on carotidplaque classification and the risk of ipsilateral hemisphericischemic events: results from the asymptomatic carotid ste-nosis and risk of stroke study. Vascular 2005;13:211e21.

13 Kakkos SK, Griffin MB, Nicolaides AN, Kyriacou E, Sabetai MM,Tegos T, et al. The size of juxtaluminal hypoechoic area in ul-trasound images of asymptomatic carotid plaques predicts theoccurrence of stroke. J Vasc Surg 2013;57(3):609e18.

14 Nicolaides AN, Kakkos SK, Kyriacou E, Griffin M, Sabetai M,Thomas DJ, et al. Asymptomatic internal carotid artery stenosisand cerebrovascular risk stratification. J Vasc Surg 2010;52(6):1486e96.

15 Salem MK, Sayers RD, Bown MJ,West K, Moore D, Nicolaides A,et al. Patients with recurrent ischaemic events from carotidartery disease have a large lipid core and low GSM. Eur J VascEndovasc Surg 2012;43(2):147e53.

16 North American Symptomatic Carotid Endarterectomy TrialCollaborators. Beneficial effect of carotid endarterectomy insymptomatic patients with high grade stenosis. N Engl J Med1991;325:445e53.

17 Sharpe RY, Dennis MJ, Nasim A, McCarthy MJ, Sayers RD,London NJ, et al. Dual antiplatelet therapy prior to carotidendarterectomy reduces post-operative embolisation andthromboembolic events: post-operative transcranial Dopplermonitoring is now unnecessary. Eur J Vasc Endovasc Surg2010;40(2):162e7.

18 Salem MK, Sayers RD, Bown MJ, Eveson DJ, Robinson TG,Naylor AR. Rapid access carotid endarterectomy can be per-formed in the hyperacute periodwithout a significant increase inprocedural risks. Eur J Vasc Endovasc Surg 2011;41(2):222e8.

19 Griffin M, Nicolaides A, Kyriakou E. Normalisation of ultrasonicimages of atherosclerotic plaques and reproducibility of greyscale median using dedicated software. Int Angiol 2007;26:372e7.

20 Griffin MB, Kyriacou E, Pattichis C, Bond D, Kakkos SK,Sabetai M, et al. Juxtaluminal hypoechoic area in ultrasonicimages of carotid plaques and hemispheric symptoms. J VascSurg 2010;52(1):69e76.

21 Lovett JK, Gallagher PJ, Hands LJ, Walton J, Rothwell PM. His-tological correlates of carotid plaque surface morphology on

European Journal of Vascular and Endovascular Surgery Volume 48 Issue 2 p. 118e125 August/2014 125

lumen contrast imaging. Circulation 2004 Oct 12;110(15):2190e7.

22 Giles MF, Rothwell PM. Risk of stroke early after transientischaemic attack: a systematic review and meta-analysis. Lan-cet Neurol 2007;6(12):1063e72.

23 Wu CM, McLaughlin K, Lorenzetti DL, Hill MD, Manns BJ,Ghali WA. Early risk of stroke after transient ischemic attack: asystematic review and meta-analysis. Arch Intern Med2007;167(22):2417e22.

24 Naylor AR. Time to rethink management strategies in asymp-tomatic carotid artery disease. Nat Rev Cardiol 2012 Oct11;9(2):116e24.

25 Walker LJ, Ismail A, McMeekin W, Lambert D, Mendelow AD,Birchall D. Computed tomography angiography for the evalu-ation of carotid atherosclerotic plaque: correlation with his-topathology of endarterectomy specimens. Stroke 2002;33:977e81.

26 Lovett JK, Redgrave JN, Rothwell PM. A critical appraisal of theperformance, reporting, and interpretation of studiescomparing carotid plaque imaging with histology. Stroke2005;36:1091e7.

27 Vavuranakis M, Sigala F, Vrachatis DA, Papaioannou TG, Filis K,Kavantzas N, et al. Quantitative analysis of carotid plaque vasavasorum by CEUS and correlation with histology after endar-terectomy. Vasa 2013;42(3):184e95.

28 Varetto G, Gibello L, Bergamasco L, Sapino A, Castellano I,Garneri P, et al. Contrast enhanced ultrasound in atheroscle-rotic carotid artery disease. Int Angiol 2012;31(6):565e71.

29 Ten Kate GL, van den Oord SC, Sijbrands EJ, van der Lugt A, deJong N, Bosch JG, et al. Current status and future de-velopments of contrast enhanced ultrasound of carotidatherosclerosis. J Vasc Surg 2013;57(2):539e46.

30 Singh N, Moody AR, Gladstone DJ, Leung G, Ravikumar R,Zhan J, et al. Moderate carotid artery stenosis: MR imaging-depicted intraplaque hemorrhage predicts risk of cerebrovas-cular ischemic events in asymptomatic men. Radiology2009;252(2):502e8.

31 Altaf N, Goode SD, Beech A, Gladman JR, Morgan PS,MacSweeney ST, et al. Plaque hemorrhage is a marker ofthromboembolic activity in patients with symptomatic carotiddisease. Radiology 2011;258(2):538e45.

32 Esposito-Bauer L, Saam T, Ghodrati I, Pelisek J, Heider P,Bauer M, et al. MRI plaque imaging detects carotid plaqueswith a high risk for future cerebrovascular events in asymp-tomatic patients. PLoS One 2013 July 24;8(7).

33 den Hartog AG, Bovens SM, Koning W, Hendrikse J, Luijten PR,Moll FL, et al. Current status of clinical magnetic resonanceimaging for plaque characterisation in patients with carotidartery stenosis. Eur J Vasc Endovasc Surg 2013;45(1):7e21.

34 Johnston SC, Rothwell PM, Nguyen-Huynh MN, Giles MF,Elkins JS, Bernstein AL, et al. Validation and refinement ofscores to predict very early stroke risk after transient ischaemicattack. Lancet 2007 Jan 27;369(9558):283e92.

35 Diener HC, Ringleb PA, Savi P. Clopidogrel for the sec-ondary prevention of stroke. Expert Opin Pharmacother2005;6:755e64.

36 Kernan WN, Viscoli CM, Brass LM, Makuch RW, Sarrel PM,Roberts RS, et al. The stroke prognosis instrument II (SPI-II): aclinical prediction instrument for patients with transientischemia and non-disabling ischemic stroke. Stroke 2000;31(2):456e62.

37 Giles MF, Albers GW, Amarenco P, Arsava EM, Asimos AW,Ay H, et al. Early stroke risk and ABCD2 score performance intissue- vs time-defined TIA: a multicenter study. Neurology2011 Sep 27;77(13):1222e8.

38 Merwick A, Albers GW, Amarenco P, Arsava EM, Ay H, Calvet D,et al. Addition of brain and carotid imaging to the ABCD2 scoreto identify patients at early risk of stroke after transientischaemic attack: a multicentre observational study. LancetNeurol 2010;9(11):1060e9.

39 Reiter M, Effenberger I, Sabeti S, Mlekusch W, Schlager O,Dick P, et al. Increasing carotid plaque echolucency is predictiveof cardiovascular events in high-risk patients. Radiology2008;248(3):1050e5.