Sensitivity and specificity of color duplexultrasound measurement in the estimation ofinternal carotid artery stenosis: A systematic reviewand meta-analysisAfshin S. Jahromi, MD, MSc,a Claudio S. Cinà, MD, MSc,a Yan Liu, MD,c

and Catherine M. Clase, MB, BChir, MSc,b� Toronto and Hamilton, Ontario, Canada

Background: Duplex ultrasound is widely used for the diagnosis of internal carotid artery stenosis. Standard duplexultrasound criteria for the grading of internal carotid artery stenosis do not exist; thus, we conducted a systematic reviewand meta-analysis of the relation between the degree of internal carotid artery stenosis by duplex ultrasound criteria anddegree of stenosis by angiography.Methods: Data were gathered from Medline from January 1966 to January 2003, the Cochrane Central Register ofControlled Trials and Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, ACP Journal Club,UpToDate, reference lists, and authors’ files. Inclusion criteria were the comparison of color duplex ultrasound resultswith angiography by the North American Symptomatic Carotid Endarterectomy Trial method; peer-reviewed publica-tions, and >10 adults.Results: Variables extracted included internal carotid artery peak systolic velocity, internal carotid artery end diastolicvelocity, internal carotid artery/common carotid artery peak systolic velocity ratio, sensitivity and specificity of duplexultrasound scanning for internal carotid artery stenosis by angiography. The Standards for Reporting of DiagnosticAccuracy (STARD) criteria were used to assess study quality. Sensitivity and specificity for duplex ultrasound criteria werecombined as weighted means by using a random effects model. The threshold of peak systolic velocity >130 cm/s isassociated with sensitivity of 98% (95% confidence intervals [CI], 97% to 100%) and specificity of 88% (95% CI, 76% to100%) in the identification of angiographic stenosis of >50%. For the diagnosis of angiographic stenosis of >70%, a peaksystolic velocity >200 cm/s has a sensitivity of 90% (95% CI, 84% to 94%) and a specificity of 94% (95% CI, 88% to 97%).For each duplex ultrasound threshold, measurement properties vary widely between laboratories, and the magnitude ofthe variation is clinically important. The heterogeneity observed in the measurement properties of duplex ultrasound maybe caused by differences in patients, study design, equipment, techniques or training.Conclusions: Clinicians need to be aware of the limitations of duplex ultrasound scanning when making management

decisions. ( J Vasc Surg 2005;41:962-72.)

During the last 10 years, two large multicenter trials,the North American Symptomatic Carotid EndarterectomyTrial (NASCET)1 and the Asymptomatic Carotid ArteryStenosis trial (ACAS)2 have reported a reduction in the riskof stroke for patients with internal carotid artery (ICA)stenosis who underwent carotid endarterectomy comparedwith medical management. Dissemination of the results ofthe trials resulted in an estimated 74% increase between1994 and 1996 in patients undergoing carotid endarterec-tomy (data from Florida).3 In these trials, stenosis wasassessed before surgery by angiography, calculated by the

From the Division of Vascular Surgery,a Department of Medicine,b McMas-ter University, and Department of Medicine, Division of Rheumatology,University of Toronto.c

Dr Jahromi received financial support from a Fellowship from the Heart andStroke Foundation of Canada.

Competition of interest: none.Additional material for this article may be found online at www.mosby.com/jvs.Correspondence: CM Clase, MD, 708 25 Charlton Ave East, Hamilton ON

L8N 1Y2 Canada (e-mail: clase@mcmaster.ca)0741-5214/$30.00Copyright © 2005 by The Society for Vascular Surgery.

doi:10.1016/j.jvs.2005.02.044

962

NASCET method, and in North America this has becomethe gold standard for patient selection for endarterectomy.

Duplex ultrasonography (DUS) is the primary nonin-vasive screening procedure for evaluation of ICA stenosisand is widely used in clinical practice to select patients forangiography. Angiography, however, is resource intensiveand has an inherent risk of morbidity and mortality thatdecreases the potential benefit of carotid endarterectomy.Consequently, some clinicians have advocated endarterec-tomy on the basis of DUS findings alone, or in combinationwith magnetic resonance angiography (MRA) or computedtomographic angiography (CTA).4-8 The importance ofthe measurement properties of DUS in predicting angio-graphic stenosis has been recognized by the large numberof recent studies that have examined different DUS criteriaagainst the gold standard of angiography.

Despite the importance of the issue, no quantitativesummary of the measurement properties of different DUScriteria has previously, to our knowledge, been reported.We conducted a systematic review and meta-analysis ofstudies that compared DUS with the gold standard ofangiography. We also examined the methodologic quality

of the studies and the issue of generalizability of findings

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 963

between centers. We planned to summarize the literaturerelating degree of ICA stenosis by DUS criteria to degree ofstenosis by angiography calculated according to theNASCET method. We also assessed the quality of thedesign and reporting of these studies.

METHODS

Search strategy. Medline database from January 1966to 28 January 2003 was searched using OVID software forcarotid arteries, common carotid artery, carotid artery dis-eases, or carotid stenosis (MeSH heading) and ultrasonog-raphy terms Doppler, duplex, color, pulsed (MeSH head-ing) and sensitivity and specificity (MeSH heading, keyword) combined with a methodologic filter previouslyidentified as sensitive for the identification of diagnostic testreports in this database.9

A search of Cochrane Central Register of ControlledTrials and Cochrane Database of Systematic Reviews (Co-chrane library, 2002, fourth quarter), Database of Abstractsof Reviews of Effectiveness (DARE fourth quarter, 2002),ACP Journal Club (January 1991 to January 2003), andUpToDate (version 11.0), was performed using searchterms carotid stenosis and carotid ultrasound.

Reference lists of a textbook of vascular surgery,10 allrelevant articles, and review articles were also reviewed. Allsearches were performed in duplicate. Authors’ files werealso reviewed.

Eligibility criteria. Eligibility criteria were developeda priori. Studies were included if (1) color DUS was used toderive Doppler velocity criteria for classification of ICAstenosis; (2) the gold standard was angiography measuredby the NASCET method; (3) the study was a peer-reviewedpublication; and (4) the study reported �10 adults. If twoarticles presented duplicate data, the more recent was in-cluded. Studies that used DUS to distinguish between ICAocclusive and nonocclusive lesions with reference to an-giography were excluded if they did not develop criteria forless severe grades of ICA stenosis. Studies in all languageswere included.

DUS measurements. We extracted data on ICA peaksystolic velocity (PSV), ICA end diastolic velocity (EDV),and internal carotid artery/common carotid artery peaksystolic velocity (ICA/CCA) ratio.

Angiography. We extracted data on percentage ste-nosis by the NASCET method.

Relevance. Titles and abstracts were reviewed in du-plicate (A.S.J., Y.L.) and full text was retrieved for allpotentially-relevant studies. Final relevance was decidedindependently by each reviewer from the full text. Assess-ment of relevance and methodologic quality, and dataextraction were performed in duplicate (A.S.J., Y.L.). Dis-agreements were resolved by consensus.

Methodologic quality of studies and data extraction.To assess the quality of studies and for data extraction, astandardized form based upon the Standards for Reportingof Diagnostic Accuracy (STARD)11 criteria was developedand pretested. Reviewers were not masked to the titles and

authors of articles. Ten items from the STARD11 criteria

were selected as most relevant in quality assessment in thiscontext. If it could not be determined whether a method-ologic criterion was met from reading a study, it wasdecided that the criterion was not satisfied. Other studies ofmethodologic quality of diagnostic tests have used a similarapproach to determine if criteria for assessment of qualitywere satisfied.12,13

From the STARD criteria, studies were assessed fordescription of:

1. training of readers of ultrasound and angiographictests;

2. blinded assessment of index test (DUS) and gold stan-dard (angiography);

3. interobserver reliability for interpretation of angiogra-phy and DUS;

4. demographic (age, gender distribution) and clinical(symptomatic versus asymptomatic) characteristics—asatisfactory grade was assigned if both features weredescribed;

5. patient sampling (consecutive series versus a randomsample);

6. direction of data collection (prospective versus retro-spective);

7. distribution of severity of ICA disease according toangiography;

8. treatment of indeterminate results;9. patient selection - reason for performance of DUS;

10. details of methods of DUS and angiography.

For angiography, authors were expected to providedetails of measurement of carotid stenosis, technique (cath-eters, injection site), and type of angiography (arch angiog-raphy, selective angiography, digital subtraction angiogra-phy, or plain films). For DUS, authors were expected toprovide details of type of ultrasound machine, probe fre-quency, angle of insonation, and location of measurementof stenosis. A partial description of either diagnostic testreceived a partial grade of adequacy. If details were notprovided, studies were classified as inadequate in thisrespect.

Studies that developed DUS criteria and reported thesensitivity and specificity of these criteria in the same dataset were considered to be derivation studies. Studies thatreported sensitivity and specificity of previously derivedDUS criteria from data not used during the original deri-vation were considered to be validation studies. We distin-guished between validation of criteria within laboratoriesand between different laboratories.

The data extracted were: indication for DUS examina-tion (screening, carotid bruit, symptoms), number of pa-tients and arteries, description of personnel involved inexecution and reading of DUS and angiographic tests,sensitivity and specificity of DUS criteria for detection ofICA stenosis for all thresholds reported in the paper (re-gardless of the authors’ recommendations of optimalthresholds), methods and estimates of test reproducibility,

clinical (distribution of symptomatic versus asymptomatic

JOURNAL OF VASCULAR SURGERYJune 2005964 Jahromi et al

patients) and demographic (age, gender) characteristics,and distribution of the severity of ICA stenosis.

Statistical analysis. Microsoft Excel (MS Office 1997)was used for data collection. Box plots of sensitivity andspecificity ranges of different DUS criteria were createdusing Statistical Package for the Social Sciences (SPSS)software (version 11.5, 2002) (SPSS, Chicago, Ill).

When sensitivity, specificity, sample size, and preva-lence of disease for a particular DUS threshold could beextracted from at least four studies, we calculated for eachstudy the number of true positives, false positives, truenegatives, and false negatives. We performed a meta-analysis when at least four studies were available to identifymore widely used DUS thresholds. Sensitivity and specific-ity were combined across studies using weights that werethe inverse of the combined within-study and between-study variance (a random effects model).14,15 For a givenDUS threshold, heterogeneity of the sensitivity and speci-ficity results between studies was assessed by the Fisher-Freeman-Halton exact test (StatsDirect, version 2.2, 2003;Cheshire, UK), using P � .05 as the criterion for hetero-geneity.

RESULTS

Relevance. The search strategy yielded 798 articles,from which 72 articles were retrieved in full text. A reviewof reference lists identified an additional 48 articles, of

Table I. Methodologic quality of included studies

CriteriaNum

fulfill

Reason for DUS Participant recruitment based on patient having received:

DUS Angiography Angiography and DUS

Consecutive patients Prospective study Description of angiography:

Adequate Partial Not Reported

Description of DUS:Adequate Partial Not Reported

Description of readers of angiograms Description of DUS interpreters Blinding Reproducibility of results:

Angiography DUS

Description of clinical and demographic characteristics Distribution of carotid stenosis disease severity Description of uninterpretable results Derivation of laboratory specific criteria Intra-laboratory validation of criteria Inter-laboratory validation of criteria

DUS, Duplex ultrasound

which 38 were retrieved in full text after their abstracts were

reviewed. Review of the full text for the 110 articles selectedas possibly relevant on the basis of their titles or abstractsidentified 47 reports that met study eligibility criteria.These included six foreign language articles that were re-viewed by a single reviewer in conjunction with a translator;two were eligible. Interobserver agreement for decisions,which were made in duplicate, was very good for decisionsregarding potentially relevant articles (Cohen’s �, .85) andgood for final relevance (�, 0.77).

Methodologic quality. Overall, the two reviewers(A.S.J., Y.L.) agreed on quality ratings for 94% of items.Table I summarizes the methodologic quality of includedstudies. A detailed description of the studies is provided inTable II (available online). Thirty studies (68%) were ret-rospective and 39 (83%) enrolled consecutive patients.Forty-four studies (94%) recruited participants who hadreceived both DUS and angiography during their clinicalcare. Distribution of the severity of disease by angiographywas reported in 41 studies (87%). Clinical and demographiccharacteristics were described in 15 studies (32%). Ade-quate description of imaging technique, readers, and repro-ducibility for angiography appeared in 5 (10%), 24 (51%),and 16 (34%) studies, and for DUS in 21 (45%), 26 (55%),and 4 (9%) studies, respectively. Blinding was described in35 studies (75%). Fifteen studies (32%) described handlingof uninterpretable results. There were 23 descriptions (49%of studies) of the performance of criteria in a derivation

of studiesiteria (%) Study

43) 21,28-46

(2) 42

(4) 41,47

94) 18,21,28-40,43-46,48-72

83) 21,28,29,31-45,47-54,56,57,59-61,63-69,71

32) 21,31,32,35,37,41,44,47,50,52,56,57,59,60,67

10) 36,39,60,66,69

87) 18,21,28-35,37,38,40-53,55-59,61-65,67,68,70-72

(2) 54

45) 28,30,31,33,36,37,43-45,47,49,56-60,62,67-69,72

55) 18,21,29,32,34,35,38-42,46,48,50-55,61,63-66,70,71

(0)51) 21,28,29,35,37,39-43,46,47,49-52,55,57,58,62,63,66,67,71

55) 18,21,28,29,31-33,35,38,39,41-44,46-49,55,57,58,63,66,67,70,71

75) 21,28-31,33-37,39-45,47,50,51,53-57,59-61,63-68,70

34) 18,29,31,39-42,46,47,53,60,62,64,65,68,69

(9) 32,33,46,56

32) 18,28-31,34,35,38,41-44,49,52,60

87) 18,21,28-31,33-35,37-57,61-70,72

32) 18,42,43,46-49,51,54,56,60,66,67,70,72

49) 18,29-31,36,39,40,47-49,51-53,56,60,63-65,67-71

13) 21,28,31,34,43,59

53) 18,21,32-35,37,38,41-46,50,52,54,55,57,58,61-63,66,69,72

bering cr

20 (

12

44 (39 (15 (

5 (41 (

1

21 (26 (

024 (26 (35 (

16 (4

15 (41 (15 (23 (6 (

26 (

dataset, 6 descriptions (13% of studies) of the performance

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 965

of criteria derived at the same center in a new dataset, and26 descriptions (53% of studies) of the performance ofcriteria derived in other centers.

DUS thresholds. Figs 1, 2, and 3 are box plots ofsensitivity and specificity of DUS velocity parameters (PSV,ICA/CCA velocity ratio, EDV) for each degree of angio-graphic stenosis (�50%, �60%, �70%). Pooled weightedmeans of sensitivity and specificity for DUS criteria fordetection of stenosis �50% and �70%, for those studiesthat reported prevalence, are provided in Table III.

In 2003, the Society of Radiologists in Ultrasoundrecommended DUS velocity criteria for clinical applicationin the assessment of carotid stenosis (Table IV).4 Througha systematic review and meta-analysis, we have derivedpooled estimates of sensitivity and specificity for differentDUS criteria (Table III), including those recommended bythe society.

The threshold of PSV �130 cm/s is associated with asensitivity of 98% (95% confidence intervals [CI], 97% to100%) and a specificity of 88% (95% CI, 76% to 100%) inthe identification of angiographic stenosis �50%. PSV�230 cm/s has a sensitivity of 90% (95% CI, 83% to 96%)and a specificity of 85% (95% CI, 77% to 92%) in thediagnosis of stenosis �70%.

Broadly similar estimates for sensitivity and specificityin the diagnosis of stenosis �70% are also observed for PSVthresholds of 200 cm/s and 250 cm/s, all with widelyoverlapping confidence intervals.

An ICA/CCA ratio �4 is associated with an 80%sensitivity (95% CI, 70% to 90%) and 88% specificity (95%CI, 83% to 93%) in the diagnosis of stenosis �70%. Confi-dence intervals for the sensitivity and specificity of thecriterion ICA/CCA ratio �3 overlap substantially withthose for ICA/CCA ratio �4.

An EDV �100 cm/s has 82% sensitivity (95% CI, 70%to 93%) and 90% specificity (95% CI, 82% to 99%) in thediagnosis of stenosis �70%, with similar values obtained forthe threshold of �120 cm/s, and overlapping confidenceintervals for EDV �70 cm/s, EDV �100 cm/s, and EDV�120 cm/s. The results from the meta-analysis are consis-tent with data displayed graphically in Figs 1 to 3, whichshow all studies, not only those which were amenable tostatistical pooling.

DISCUSSION

In general, as expected, sensitivity decreased and spec-ificity rose with increasing thresholds for each parameter.However, many of the original studies that met eligibilitycriteria in this review are flawed. Poor documentation ofclinical features, variation in disease prevalence and severity,verification effects (the selection of patients to undergo thegold standard after the results of the index text are known),clinical review effects (the knowledge of clinical informa-tion in the interpretation of test results), and lack of blind-ing (the knowledge of one test result in the interpretationof the other) have each been empirically documented to beassociated with bias in studies of diagnostic test perfor-

mance.12,16 These problems were prevalent in the original

articles included in our review. The circumstances that leadto referral of patients for carotid DUS examination weredescribed in 49%, and patients’ clinical and demographiccharacteristics in 32%.

Severity of disease was reported in 87%, and mostpapers reported patients with a wide spectrum of severity ofcarotid stenosis. In 94% of studies, however, patients wererecruited if they had undergone both angiography andDUS, the decision for patients to receive the gold standardtest of angiography having been taken after a review ofDUS results, which might have led to verification bias.Blinding was not described in 25% of the studies.

Diagnostic estimates can also be influenced by exclu-sion of uninterpretable results. Treatment of uninterpret-able DUS results was described in 32% of the studies.Calcified lesions, tortuous or kinked ICA, or patient bodyhabitus can cause uninterpretable DUS findings. It is im-portant to obtain and report angiographic verification ofdisease status for indeterminate DUS cases. If carotid ste-nosis were associated with uninterpretable DUS results,then the reported sensitivity of DUS is overestimated whenpatients with uninterpretable results are excluded fromanalysis.

For each DUS threshold that has been studied on morethan one occasion, a wide range of point estimates forsensitivity and specificity have been reported (Figs 1 to 3).Our meta-analysis of those studies for which variance esti-mates could be calculated (Table III) shows pooled pointestimates with confidence intervals. The lack of precision inthese estimates is clinically important. We have previouslyused decision analysis to examine the effects of variation insensitivity and specificity on outcomes in carotid endarter-ectomy for asymptomatic disease.17 Though modelstended to be relatively insensitive to sensitivity within theclinically-plausible range, each of the models we examinedwas sensitive to changes in specificity within the range of75% to 95%.

In this decision analysis, for the strategies of performingcarotid endarterectomy on the basis of 70% stenosis byDUS, either confirmed by angiography or proceeding di-rectly to surgery, specificity as low as 75% was associatedwith overall harm, whereas if specificity were 85% orgreater, both strategies produced a small benefit. Table 3shows that for a number of thresholds in clinical use, thepooled confidence intervals include values below 85%, andFigures 1-3 shows that point estimates in this range havealso been reported in some studies.

Variability between studies,4 between laboratories inthe same study,18 and between observers using the sameequipment19,20 has been previously documented. Author-ities have interpreted the observed between-study variationas calibration differences and have recommended as a rem-edy the performance in each center of a derivation andvalidation study to determine appropriate, center-specificthresholds.4,18,21,22

However, no direct data support the hypothesis thatthe differences between laboratories and between studies

are due to systematic bias (ie, a calibration problem) rather

JOURNAL OF VASCULAR SURGERYJune 2005966 Jahromi et al

Fig 1. Reported diagnostic estimates of peak systolic velocity (cm/s) for measurement of internal carotid artery

stenosis: (A) �50%; (B) �50%; (C) �70%. N, Number of studies.

�60

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 967

than random error (ie, error of measurement). Randomerror is clearly prevalent. This has been documented in thelaboratory quality assurance performed for centers partici-

Fig 2. Reported diagnostic estimates of peak systolic vCCA) for measurement of ICA stenosis: (A) �50%; (B)

pating in ACAS.22

Each of the 37 ACAS centers provided data for eachdevice in use (a device was defined as the combination of aspecific machine in the hands of a particular group of

y internal carotid artery/common carotid artery (ICA/%; (C) �70%. N, Number of studies.

elocit

technicians). DUS results were compared with angio-

JOURNAL OF VASCULAR SURGERYJune 2005968 Jahromi et al

Fig 3. Reported diagnostic estimates of end diastolic velocity (cm/s) for measurement of internal carotid artery

stenosis: (A) �50%; (B) �60%; (C) �70%. N, Number of studies.

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 969

graphic stenosis for patients who underwent angiographywithin 6 weeks. The relationship between the two wasmodeled as a fourth order polynomial (which allows thespecification of a complex curvilinear relationship, ratherthan assuming that it is linear or takes some particularalgebraic form) specific for each device, and a device-specific threshold identified.

Of the 63 devices studied, sensitivity was poor in 9(14%) and no threshold could be established in a further 9(14%). No statistical relationship between DUS and an-giography could be established in approximately 20% ofdevices. However, excellent sensitivity was observed forothers: 13 (21%) had sensitivity �80%. No factors wereidentified that predicted poor performance.

Of the four manufacturers with at least eight devicesused in ACAS, each manufacturer was associated with atleast one excellent and one unusable device. This representsan important clinical problem for those who order andthose who interpret DUS results: substantial heterogeneity

Table III. Pooled weighted means of sensitivity and specifi

Angiogr

�50%

MeasureThreshold

cm/s N* Sensitivity 95% CI Specifici

PSV �120 3,001 96† 91-100 82†

�130 1,716 98† 97-100 88†

�150�200�230�250

Ratio‡ �3�4

EDV �70�100�120

CI, Confidence interval; EDV, end-diastolic velocity (cm/s); PSV, peak sys*Total number of subjects in combined studies.†Heterogeneity identified by Fischer-Freeman-Halton exact test, P � .05.‡Internal carotid artery PSV/common carotid artery PSV.

Table IV. Summary of recommendations of the Societyof Radiologists in Ultrasound Consensus Conference,with respect to key velocity thresholds, for the use ofduplex ultrasound criteria in the diagnosis of internalcarotid artery stenosis4

Degree ofStenosis

PrimaryParameters Additional Parameters

ICA PSV(cm/sec)

ICA/CCAPSV ratio

ICA EDV(cm/sec)

�50% �125 �2.0 �40�70 �230 �4.0 �100

ICA, Internal carotid artery; CCA, common carotid artery; PVS, peaksystolic velocity; EDV, end-diastolic velocity.

clearly exists in the measurement of stenosis by DUS, and

no predictors or markers of a reliable machine or reliablelaboratory have been identified.

Variation, which can lead to measurement error, occursat many different levels in the application of the test. Eachcenter usually has multiple observers whose training andqualifications may vary and frequently has more than onedevice-and-probe combination in use. Furthermore, thetechnology available has changed over time. Between cen-ters, each of these can be a source of variation, along withvariation in scanning protocols and quality assurance stan-dards. Of these sources of variation, to our knowledge, onlyinter-observer reliability has been systematically exam-ined.19,20 The generalizability of thresholds derived fromindividual reports and of our own summary estimates areseverely limited by the absence of information on theimportant sources of variation and how they may be mini-mized. However, generalization is a widespread clinicalpractice, and endorsed by the recent consensus statement.4

Our work represents a systematic summary of availableevidence from the era of widespread adoption of color DUStechnology. Rothwell13 previously assessed the design andmethodology of a random sample of published studies,found most to be of poor quality, and did not perform ameta-analysis. Two other systematic reviews reported over-all sensitivity for DUS of between 82% and 86% and aspecificity of 87% to 94%.8,23 In these reviews, DUS per-formance was summarized across all thresholds for detec-tion of any grade of ICA stenosis. The Society of Radiolo-gists in Ultrasound derived its summary and conclusionsfrom “30 representative articles” chosen by the consensusconference moderator, of which 10 are summarized in theconsensus statement.

We have used a comprehensive search strategy, used therigorous methodologic criteria developed by the STARDgroup for quality assessment,11 have described separately

for detection of stenosis �50% and �70%

stenosis

�70%

95% CI N* Sensitivity 95% CI Specificity 95% CI

72-9376-100

1,996 96† 93-98 80† 71-902,140 90† 84-94 94† 88-972,108 90† 83-96 85† 77-921,904 76† 63-89 93† 88-98

999 89† 81-96 84† 77-921,933 80† 70-90 88† 83-931,419 89† 84-94 80† 66-931,607 82 70-93 90† 82-991,478 79 71-87 92† 86-98

elocity (cm/s).

city

aphic

ty

tolic v

the measurement properties of every threshold reported in

JOURNAL OF VASCULAR SURGERYJune 2005970 Jahromi et al

the original article, regardless of the original authors’ con-clusions or choice of optimal threshold, and have providedstatistical summaries of those data which were amenable tometa-analysis.

Our study has a number of limitations:

● Our literature search was conducted in January 2003:it is likely that other studies have appeared since thistime. However, our principal findings that results areheterogeneous and differ between laboratories wouldnot be changed by the addition of further studies.

● We have relied upon authors’ reporting of methodol-ogy and results for quality assessment and data extrac-tion.

● Our meta-analysis of weighted means for DUS thresh-old was limited to studies from which we could recon-struct standard 2 � 2 tables for the calculation ofsensitivity and specificity.

● Publication bias may have affected our results, likelyleading to overestimation of DUS accuracy, thoughthe selection of all thresholds reported in every in-cluded study, rather than the authors’ recommended‘optimal’ threshold may have minimized this effect.

● Our analysis is a comparison of DUS properties withthe de facto gold standard of angiography rather thanwith pathologic findings. We recognize that interob-server variation is problematic for the gold standardtest itself, but felt that a detailed summary of data tothis effect was outside the scope of this meta-analysis.Comparison with angiography is useful and justifiedbecause of the clinical situation: influential trials doc-umenting the efficacy of carotid endarterectomy haveeither used angiography1,24,25 or DUS calibrated toangiography2,26 to define participants.

Implications for practice. For each DUS threshold,measurement properties vary widely between laboratories,and the magnitude of the variation is clinically important.Data from ACAS show that random error is prevalent.22

Whether calibration differences also exist is not known. Toprovide results that led to benefit rather than harm, webelieve individual laboratories must perform continuousquality improvement in the form of comparison of DUSwith angiographic data from patients who undergo the twoprocedures within a short time interval (perhaps 6 to 12weeks). (This differs from the conclusions reached by theSociety of Radiologists in Ultrasound4 which recommends:“For a particular laboratory setting, internal validation isencouraged when possible.”)

We suggest adopting the approach of Howard, inwhich the relation between the two is modeled with poly-nomial regression. This would permit a DUS report toinclude a predicted stenosis that is based on a complexrelationship between velocities and degree of stenosis, andthat is device specific.22 Whether the data are a reasonablefit to such a model (the model total r2) determines whethera device produces usable data. If the data are usable, theresults of this machine-specific regression should then be

used in the reporting of degree of stenosis measured by the

machine. This can be further assessed and described to theuser of the information as a � statistic comparing thepredicted percentage stenosis with measured percentagestenosis.

This differs from the current approach suggested by theIntersocietal Commission for the Accreditation of VascularLaboratories (ICAVL), which examines simple correlationsbetween DUS measurements and angiography as a check ofvalidity, and does not use the data to calibrate the instru-ment or to improve the validity of the measurement. If amachine-specific relationship between degree of stenosisand velocity can be elucidated, the issue of selection ofthresholds for the identification of stenosis no longer ap-plies, as the degree of stenosis is directly predicted by thevelocity measurement for that device, and actual estimateddegree of stenosis can be directly reported. Alternativeapproaches, such as the use of the thresholds suggested bythe Society of Radiologists in Ultrasound (Table IV) mayprovide adequate sensitivity and specificity, but no data areavailable to suggest the optimal combination of sensitivityand specificity in this setting. Adjacent thresholds providesimilar estimates of sensitivity and specificity with overlap-ping confidence intervals. Some machine-user combina-tions will (unpredictably) produce very poor data, whichwould be detected under current quality assurance stan-dards if the problem were random error, but which mightnot be obvious if the problem were calibration.

The ability to benefit from carotid endarterectomy isrelated to the degree of stenosis,24,25 so that patients whomake decisions about endarterectomy on the basis of mea-surements from devices that produce poor-quality or un-calibrated data do so on the basis of misinformation. Be-cause the risk-benefit ratio for carotid endarterectomy forasymptomatic patients is more marginal than for patientswith symptomatic disease, these considerations are partic-ularly relevant to asymptomatic patients.

The only method of ensuring that patients will experi-ence the same risk-benefit ratio as obtained in ACAS is toset thresholds by the same methodology, calculating andusing the device-specific threshold that produces a positivepredictive value of �90%.22 This is in keeping with re-sults of a decision analysis that suggest that specificity is acritical factor in the risk-benefit ratio of strategies for iden-tification of patients for endarterectomy.17 Failure to main-tain high specificity and high positive predictive value ex-poses patients without significant stenosis, who will notbenefit from endarterectomy, to the risk of stroke fromangiography, estimated at 1%.27

Implications for research. The quality of future re-search could be improved by application of the STARDcriteria by authors and editors. The effects of verificationbias may be studied by reporting results for contralateralarteries (ie, the artery with the lower degree of stenosis)separately. Variation due to disease severity may be exam-ined by reporting results separately for symptomatic andasymptomatic patients.28,29 Publication bias might beavoided by the creation of a study registry in which data are

recorded regardless of whether the test performed well or

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 971

poorly. Studies that examine the measurement propertiesof DUS compared with the criterion measure of angiogra-phy should include scatter plots showing the relationshipsbetween the two measures, and polynomial equations aswell as simple correlations should be used to describe theserelationships. Finally, there is a pressing need for reliabilitystudies that identify the important sources of random errorand determine whether calibration differences also exist.

Our thanks to Dr Keith Tsoi, and Robee Banerjee whoprovided invaluable help with translation of articles.

REFERENCES

1. North American Symptomatic Carotid Endarterectomy Trial Collabo-rators (NASCET). Beneficial effect of carotid endarterectomy in symp-tomatic patients with high-grade carotid stenosis. New Engl J Med1991;325:445.

2. Executive Committee for the Asymptomatic Carotid AtherosclerosisStudy. Endarterectomy for Asymptomatic Carotid Artery Stenosis.JAMA 1995;273:1421-28.

3. Morasch MD, Parker MA, Feinglass J, Manheim LM, Pearce WH.Carotid endarterectomy: Characteriation of recent increases in proce-dure rates. J Vasc Surg 2000;31:901-9.

4. Grant EG, Benson CB, Moneta GL, Alexandrov AV, Baker JD, BluthEI, et al. Society of Radiologists in Ultrasound Consensus Conferenceon Ultrasound and Doppler Diagnosis of Carotid Stenosis. Radiology2003;229:340-346.

5. Gertler JP, Cambria RP, Kistler JP, Geller SC, MacDonald NR, Brew-ster DC, et al. Carotid surgery without arteriography: noninvasiveselection of patients. Ann Vasc Surg 1991;5:253-56.

6. Nicholas GG, Osborne MA, Jaffe JW, Reed JF, III. Carotid arterystenosis: preoperative noninvasive evaluation in a community hospital. JVasc Surg 1995;22:9-16.

7. Mattle HP, Kent KC, Edelman RR, Atkinson DJ, Skillman JJ. Evalua-tion of the extracranial carotid arteries: correlation of magnetic reso-nance angiography, duplex ultrasonography, and conventional angiog-raphy. J Vasc Surg 1991;13:838-44.

8. Nederkoorn PJ, Mali WP, Eikelboom BC, Elgersma OE, Buskens E,Hunink MG, et al. Preoperative diagnosis of carotid artery stenosis:accuracy of noninvasive testing. Stroke 2002;33:2003-8.

9. Haynes RB, Wilczynski N, McKibbon KA, Walker CJ, Sinclair JC.Developing optimal search strategies for detecting clinically soundstudies in MEDLINE. J Am Med Inform Assoc 1994;1:447-58.

10. Rutherford RB. Rutherford’s Textbook on Vascular Surgery. 5th ed.Philadelphia: WB Saunders Co; 2000.

11. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, IrwigLM, et al. The STARD statement for reporting studies of diagnosticaccuracy: explanation and elaboration. Clin Chem 2003;49:7-18.

12. Lijmer JG, Mol BW, Heisterkamp S, Bonsel GJ, Prins MH, van derMeulen JHet al. Empirical evidence of design-related bias in studies ofdiagnostic tests. JAMA 1999;282:1061-66.

13. Rothwell PM, Pendlebury ST, Wardlaw J, Warlow CP. Critical appraisalof the design and reporting of studies of imaging and measurement ofcarotid stenosis. Stroke 2000;31:1444-50.

14. Laird NM, Mosteller F. Some statistical methods for combining exper-imental results. Int J Technol Assess Health Care 1990;6:5-30.

15. Kearon C, Julian JA, Newman TE, Ginsberg JS. Noninvasive diagnosisof deep venous thrombosis. McMaster Diagnostic Imaging PracticeGuidelines Initiative. Ann Intern Med 1998;128:663-77.

16. Whiting P, Rutjes AW, Reitsma JB, Glas AS, Bossuyt PM, Kleijnen J.Sources of variation and bias in studies of diagnostic accuracy: a system-atic review. Ann Intern Med. 2004;140:189-202.

17. Clase CM, Cina CS. Medical management versus investigate-and-operate strategy in asymptomatic carotid stenosis: a decision analysis. JVasc Surg 2002;36:541-48.

18. Kuntz KM, Polak JF, Whittemore AD, Skillman JJ, Kent KC. Duplexultrasound criteria for the identification of carotid stenosis should be

laboratory specific. Stroke 1997;28:597-602.

19. Mead GE, Lewis SC, Wardlaw JM. Variability in Doppler ultrasoundinfluences referral of patients for carotid surgery. Eur J Ultrasound2000;12:137-43.

20. Corriveau MM, Johnston KW. Interobserver variability of carotidDoppler peak velocity measurements among technologists in anICAVL-accredited vascular laboratory. J Vasc Surg 2004;39:735-41.

21. Hood DB, Mattos MA, Mansour A, Ramsey DE, Hodgson KJ, Bark-meier LD et al. Prospective evaluation of new duplex criteria to identify70% internal carotid artery stenosis. J Vasc Surg 1996;23:254-61.

22. Howard G, Baker WH, Chambless LE, Howard VJ, Jones AM, TooleJF. An approach for the use of Doppler ultrasound as a screening toolfor hemodynamically significant stenosis (despite heterogeneity ofDoppler performance). A multicenter experience. Asymptomatic Ca-rotid Atherosclerosis Study Investigators. Stroke 1996;27:1951-7.

23. Blakeley DD, Oddone EZ, Hasselblad V, Simel DL, Matchar DB.Noninvasive carotid artery testing. A meta-analytic review. Ann InternMed 1995;122:360-7.

24. MRC European Carotid Surgery Trial: interim results for symptomaticpatients with severe (70-99%) or with mild (0-29%) carotid stenosis.European Carotid Surgery Trialists’ Collaborative Group. Lancet 1991;337:1235-43.

25. Randomised trial of endarterectomy for recently symptomatic carotidstenosis: final results of the MRC European Carotid Surgery Trial(ECST). Lancet. 1998;351:1379-87.

26. Halliday A, Mansfield A, Marro J, Peto C, Peto R, Potter J, et al.Prevention of disabling and fatal strokes by successful carotid endarter-ectomy in patients without recent neurological symptoms: randomisedcontrolled trial. Lancet. 2004;363:1491-502.

27. Cina CS, Clase CM, Haynes BR. Refining the indications for carotidendarterectomy in patients with symptomatic carotid stenosis: a system-atic review. J Vasc Surg 1999;30:606-17.

28. Grant EG, Duerinckx AJ, El Saden SM, Melany ML, Hathout GM,Zimmerman PT, et al. Ability to use duplex US to quantify internalcarotid arterial stenoses: fact or fiction? Radiology 2000;214:247-52.

29. Grant EG, Duerinckx AJ, El Saden S, Melany ML, Hathout G, Zim-merman P, et al. Doppler sonographic parameters for detection ofcarotid stenosis: is there an optimum method for their selection? AJRAm J Roentgenol 1999;172:1123-29.

30. AbuRahma AF, Robinson PA, Strickler DL, Alberts S, Young L. Pro-posed new duplex classification for threshold stenoses used in varioussymptomatic and asymptomatic carotid endarterectomy trials. Ann VascSurg 1998;12:349-58.

31. Elgersma OE, van Leersum M, Buijs PC, van Leeuwen MS, van deSchouw YT, Eikelboom BC et al. Changes over time in optimal duplexthreshold for the identification of patients eligible for carotid endarter-ectomy. Stroke 1998;29:2352-56.

32. Elmore JR, Franklin DP, Thomas DD, Youkey JR. Carotid endarterec-tomy: the mandate for high quality duplex. Ann Vasc Surg 1998;12:156-62

33. Polak JF, Dobkin GR, O’Leary DH, Wang AM, Cutler SS. Internalcarotid artery stenosis: accuracy and reproducibility of color-Doppler-assisted duplex imaging. Radiology 1989;173:793-98.

34. AbuRahma AF, Richmond BK, Robinson PA, Khan S, Pollack JA,Alberts S. Effect of contralateral severe stenosis or carotid occlusion onduplex criteria of ipsilateral stenoses: comparative study of variousduplex parameters. J Vasc Surg 1995;22:751-61.

35. Huston J, III, Lewis BD, Wiebers DO, Meyer FB, Riederer SJ, WeaverAL. Carotid artery: prospective blinded comparison of two-dimensionaltime-of-flight MR angiography with conventional angiography andduplex US. Radiology 1993;186:339-44.

36. Hunink MG, Polak JF, Barlan MM, O’Leary DH. Detection andquantification of carotid artery stenosis: efficacy of various Dopplervelocity parameters. AJR Am J Roentgenol 1993;160:619-25.

37. Karam BG, Atallah N, Slaba S, Ashouh R, Jebara V, Tabet G et al. Theaccuracy of carotid duplex scanning in the diagnosis of 70% internalcarotid artery stenosis: evaluation of new duplex criteria. J Med Liban1996;44:223-26.

38. Mattos MA, Hodgson KJ, Faught WE, Mansour A, Barkmeier LD,Ramsey DE, et al. Carotid endarterectomy without angiography: is

color-flow duplex scanning sufficient? Surgery 1994;116:776-82.

JOURNAL OF VASCULAR SURGERYJune 2005972 Jahromi et al

39. MacKenzie KS, French-Sherry E, Burns K, Pooley T, Bassiouny HS.B-mode ultrasound measurement of carotid bifurcation stenoses: is itreliable? Vasc Endovascular Surg 2002;36:123-35.

40. Browman MW, Cooperberg PL, Harrison PB, Marsh JI, Mallek N.Duplex ultrasonography criteria for internal carotid stenosis of morethan 70% diameter: angiographic correlation and receiver operatingcharacteristic curve analysis. Can Assoc Radiol J 1995;46:291-95.

41. Bray JM, Galland F, Lhoste P, Nicolau S, Dubas F, Emile J, et al. ColourDoppler and duplex sonography and angiography of the carotid arterybifurcations. Prospective, double-blind study. Neuroradiology 1995;37:219-24.

42. Dippel DW, de Kinkelder A, Bakker SL, van Kooten F, van OverhagenH, Koudstaal PJ. The diagnostic value of colour duplex ultrasound forsymptomatic carotid stenosis in clinical practice. Neuroradiology 1999;41:1-8.

43. Filis KA, Arko FR, Johnson BL, Pipinos II, Harris EJ, Olcott C, et al.Duplex ultrasound criteria for defining the severity of carotid stenosis.Ann Vasc Surg 2002;16:413-21.

44. Link J, Brossmann J, Penselin V, Gluer CC, Heller M. Common carotidartery bifurcation: preliminary results of CT angiography and color-coded duplex sonography compared with digital subtraction angiogra-phy. AJR Am J Roentgenol 1997;168:361-5.

45. Jung EM, Kubale R, Clevert DA, Lutz R, Rupp N. [B-flow and contrastmedium-enhanced power Doppler (Optison(R))–preoperative diagno-sis of high-grade stenosis of the internal carotid artery]. Rofo FortschrGeb Rontgenstr Neuen Bildgeb Verfahr 2002;174:62-9.

46. Guo D, Wang Y, Fu W. Assessment of extracranial internal carotidartery stenosis by duplex scanning magnetic resonance angiography anddigital subtraction angiography: a comparative study. Chung Hua IHsueh Tsa Chih. 2000;80:98-100.

47. Soulez G, Therasse E, Robillard P, Fontaine A, Denbow N, BourgouinP, et al. The value of internal carotid systolic velocity ratio for assessingcarotid artery stenosis with Doppler sonography. AJR Am J Roentgenol1999;172:207-12.

48. Muller M, Ciccotti P, Reiche W, Hagen T. Comparison of color-flowDoppler scanning, power Doppler scanning, and frequency shift forassessment of carotid artery stenosis. J Vasc Surg 2001;34:1090-95.

49. Huston J, James EM, Brown RDJ, Lefsrud RD, Ilstrup DM, RobertsonEF, et al. Redefined duplex ultrasonographic criteria for diagnosis ofcarotid artery stenosis. Mayo Clin Proc 2000;75:1133-40.

50. Chen JC, Salvian AJ, Taylor DC, Teal PA, Marotta TR, Hsiang YN.Predictive ability of duplex ultrasonography for internal carotid arterystenosis of 70%-99%: a comparative study. Ann Vasc Surg 1998;12:244-47.

51. Winkelaar GB, Chen JC, Salvian AJ, Taylor DC, Teal PA, Hsiang YN.New duplex ultrasound scan criteria for managing symptomatic 50% orgreater carotid stenosis. J Vasc Surg 1999;29:986-94.

52. Jackson MR, Chang AS, Robles HA, Gillespie DL, Olsen SB, Kaiser WJ,et al. Determination of 60% or greater carotid stenosis: a prospectivecomparison of magnetic resonance angiography and duplex ultrasoundwith conventional angiography. Ann Vasc Surg 1998;12:236-43.

53. Neschis DG, Lexa FJ, Davis JT, Carpenter JP. Duplex criteria fordetermination of 50% or greater carotid stenosis. J Ultrasound Med2001;20:207-15.

54. Spadone DP, Barkmeier LD, Hodgson KJ, Ramsey DE, Sumner DS.Contralateral internal carotid artery stenosis or occlusion: pitfall ofcorrect ipsilateral classification–a study performed with color-flow im-aging. J Vasc Surg 1990;11:642-9.

55. Ballard JL, Fleig K, De Lange M, Killeen JD. The diagnostic accuracy ofduplex ultrasonography for evaluating carotid bifurcation. Am J Surg1994;168:123-6.

56. Ranke C, Creutzig A, Becker H, Trappe HJ. Standardization of carotidultrasound: a hemodynamic method to normalize for interindividual

and interequipment variability. Stroke 1999;30:402-6.

57. Lee VS, Hertzberg BS, Workman MJ, Smith TP, Kliewer MA,DeLong DM, et al. Variability of Doppler US measurements alongthe common carotid artery: effects on estimates of internal carotidarterial stenosis in patients with angiographically proved disease.Radiology 2000;214:387-92.

58. Thomas N, Taylor P, Padayachee S. The impact of theoretical errors onvelocity estimation and accuracy of duplex grading of carotid stenosis.Ultrasound Med Biol 2002;28:191-96.

59. Alexandrov AV, Vital D, Brodie DS, Hamilton P, Grotta JC. Gradingcarotid stenosis with ultrasound. An interlaboratory comparison. Stroke1997;28:1208-10.

60. Bonig L, Weder B, Schott D, Keel A, Nguyen T, Zaunbauer W.Prediction of angiographic carotid artery stenosis indexes by colourDoppler-assisted duplex imaging. A critical appraisal of the parametersused. Eur J Neurol 2000;7:183-90.

61. Londrey GL, Spadone DP, Hodgson KJ, Ramsey DE, Barkmeier LD,Sumner DS. Does color-flow imaging improve the accuracy of duplexcarotid evaluation? J Vasc Surg 1991;13:659-63.

62. Anderson CM, Saloner D, Lee RE, Griswold VJ, Shapeero LG, RappJH, et al. Assessment of carotid artery stenosis by MR angiography:comparison with x-ray angiography and color-coded Doppler ultra-sound. AJNR Am J Neuroradiol 1992;13:989-1003.

63. Faught WE, Mattos MA, van Bemmelen PS, Hodgson KJ, BarkmeierLD, Ramsey DE, et al. Color-flow duplex scanning of carotid arteries:new velocity criteria based on receiver operator characteristic analysis forthreshold stenoses used in the symptomatic and asymptomatic carotidtrials. J Vasc Surg 1994;19:818-27.

64. Carpenter JP, Lexa FJ, Davis JT. Determination of duplex Dopplerultrasound criteria appropriate to the North American SymptomaticCarotid Endarterectomy Trial. Stroke 1996;27:695-9.

65. Moneta GL, Edwards JM, Chitwood RW, Taylor LM Jr, Lee RW,Cummings CA, et al. Correlation of North American SymptomaticCarotid Endarterectomy Trial (NASCET) angiographic definition of70% to 99% internal carotid artery stenosis with duplex scanning. J VascSurg 1993;17:152-7.

66. Dinkel HP, Moll R, Debus S. Colour flow Doppler ultrasound of thecarotid bifurcation: can it replace routine angiography before carotidendarterectomy? Br J Radiol 2001;74:590-4.

67. Koga M, Kimura K, Minematsu K, Yamaguchi T. Diagnosis of internalcarotid artery stenosis greater than 70% with power Doppler duplexsonography. AJNR Am J Neuroradiol 2001;22:413-17.

68. Carpenter JP, Lexa FJ, Davis JT. Determination of sixty percent orgreater carotid artery stenosis by duplex Doppler ultrasonography. JVasc Surg 1995;22:697-703.

69. Neale ML, Chambers JL, Kelly AT, Connard S, Lawton MA, Roche J,et al. Reappraisal of duplex criteria to assess significant carotid stenosiswith special reference to reports from the North American SymptomaticCarotid Endarterectomy Trial and the European Carotid Surgery Trial.[see comments.]. J Vasc Surg 1994;20:642-9.

70. Moneta GL, Edwards JM, Papanicolaou G, Hatsukami T, Taylor LMJ,Strandness DEJ, et al. Screening for asymptomatic internal carotidartery stenosis: duplex criteria for discriminating 60% to 99% stenosis. JVasc Surg 1995;21:989-94.

71. Hwang CS, Shau WY, Tegeler CH. Doppler velocity criteria based onreceiver operating characteristic analysis for the detection of thresholdcarotid stenoses. J Neuroimaging 2002;12:124-30.

72. Erickson SJ, Mewissen MW, Foley WD, Lawson TL, Middleton WD,Quiroz FA, et al. Stenosis of the internal carotid artery: assessment usingcolor Doppler imaging compared with angiography. AJR Am J Roent-genol 1989;152:1299-305.

Submitted Nov 29, 2004; accepted Feb 13, 2005.

Additional material for this article may be found online at

www.mosby.com/jvs.

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 972.e1

Table II, online only. Description of included studies

Study

Participants Methods

N Patients;arteries

Setting/Reasonfor referral

Consecutive patients?Prospective/Retrospective

study?

DemographicsAverage age

Genderdistribution Severity

Reproducibility of testsAngiography – NASCET% Stenosis DUS – Doppler

velocity

AbuRahma34 178; 356 Symptoms,bruit

Consecutive patients.Retrospective study.

68 y, 58%male

49%: 50%-79%,24%: 80%-99% stenosis

Not reported

AbuRahma30 231; 404 Symptoms,bruit

Consecutive not reported.Direction of study not

reported.

67.9 y, 56%male

12%: 50%-59%,12%: 60%-69%, 9%:70-99%stenosis

Not reported

Alexandrov59 87; 174 Not reported Consecutive patients.Prospective study.

Not reported Not reported Not reported

Anderson62 31; 50 Not reported Consecutive not reported.Retrospective study.

95% male 10%: 40%-59%,22%: 60%-94% stenosis

Angiography Spearman’scorrelation 0.9-0.96,DUS not reported.

Ballard55 400; 774 Not reported Consecutive not reported.Retrospective study.

Not reported 22% 50%-79%,9.1% 80%-99% stenosis

Not reported

Bonig60 79; 158 TIA, AmF,stroke

Consecutive patients.Prospective Study.

66.3 y, 65%male

75% � 30%stenosis

Angiography Kappa0.89. DUS notreported

de Bray41 64; 128 Referral forangiography.

Consecutive patients.Prospective Study.

62 y, 84%male

20% �70%stenosis

Angiography Pearson’scorrelation 0.92, DUSnot reported.

Browman40 75; 145 Symptoms,bruit,screening

Consecutive patients.Direction of study not

reported.

66% male,age notreported

23% � 70%stenosis

Angiograms withinterobservermeasurementdifference �30% werere-examined, DUS notreported.

Carpenter68 110; 210 Not reported Consecutive patients.Retrospective study.

Not reported 46% � 60%stenosis

Angiography Kappa�0.86, DUS notreported.

Carpenter64 110; 210 Not reported Consecutive patients.Retrospective study.

Not reported 33% � 70%stenosis

Angiography Kappa �0.86, DUS notreported.

Chen50 99; 185 Not reported Consecutive patients.Prospective study.

Not reported 40% � 70%stenosis

Not reported

Dinkel66 116; 225 Not reported Consecutive patients.Retrospective study.

Not reported Unclear Not reported

Dippel42 152; 152 Symptoms Consecutive patients.Prospective study.

56 y, 65%male

22% � 70%stenosis

Angiography � 0.87-1.0,DUS not reported.

Elgersma31 121; 239 Symptoms Consecutive patients.Prospective study.

64 y, 66%male

1st period 26%,2nd period33% �70%stenosis

DUS thresholds derivedfrom differentinterpreters of sameangiograms werecompared, DUSreliability notreported.

Elmore32 24; notreported

Confirmationof priorDUS.

Consecutive patients.Prospective study.

Not reported Not reported Angiography notreported. DUSinterlaboratory � 0.62.

Erickson72 49; 95 Not reported Consecutive not reported.Direction of study not

reported.

Not reported 19;83 (23%)50%-75%,8;83 (10%)76%-99%stenosis

Not reported

Faught63 405; 770 Not reported Consecutive study.Retrospective study.

Not reported 13%: �50%-69%, 17%:�79%-99%stenosis

Not reported

JOURNAL OF VASCULAR SURGERYJune 2005972.e2 Jahromi et al

Table II, online only. Continued

Study

Participants Methods

N Patients;arteries

Setting/Reasonfor referral

Consecutive patients?Prospective/Retrospective

study?

DemographicsAverage age

Genderdistribution Severity

Reproducibility of testsAngiography – NASCET% Stenosis DUS – Doppler

velocity

Filis43 162; 326 Screening,symptoms

Consecutive patients.Retrospective study.

64 y, 67%male

158;296 (53%)�50%, 107;296 (36%)�70%

Not reported

Grant29 202; 333 Screening,symptoms,bruit

Consecutive patients.Retrospective study.

65 y, 99%male

51;201 (25%)�70%,39;132(30%)�60%

Angiography � 0.84.DUS not reported.

Grant28 220; 404 Screening,symptoms,bruit

Consecutive patients.Retrospective study.

64.8 y, 98%male

49% � 50%,21% � 70%stenosis

Not reported

Guo46 32; 64 Stroke Consecutive not reported.Retrospective study.

Not reported Not reported Not reported

Hood21 248; 457 Screening,symptoms,bruit

Consecutive patients.Direction of study not

reported.

Not reported 20% � 70%stenosis

Not reported

Hunink36 96; 166 Screening,symptoms,bruit

Consecutive patients.Retrospective study.

66.5 y, 59%male

Not reported Not reported

Huston35 50; 77 Symptoms Consecutive patients.Prospective study.

67 y, 70%male

31%� 50%stenosis

Not reported

Huston49 621; 1218 Not reported Consecutive patients.Retrospective study.

67.7 y, 61%male

62% � 70% Not reported

Hwang71 134; 268 Not reported Consecutive patients.Retrospective study.

69 y, 63%male

Not reported Not reported

Jackson52 50; 100 Not reported Consecutive patients.Prospective study.

69.5 y, 72%male

37% � 60%stenosis

Not reported

Jung45 88; 88 Symptoms,screening

Consecutive patients.Direction of study not

reported.

Not reported 49% � 70% Not reported

Karam37 49; 98 Symptoms,screening

Consecutive patients.Prospective study.

Not reported 26% � 50%,18% � 70%stenosis

Not reported

Koga67 75; 144 Not reported Consecutive patients.Prospective study.

63 y, 88%male

14% � 70%stenosis

Not reported

Kuntz49 64; 123 Not reported Consecutive not reported.Direction of study not

reported.

70 y, 66%male

laboratory A33% � 70%,laboratory B12.7% �70%

Angiography Kappa0.93. DUS notreported.

Lee57 59; 118 Not reported Consecutive patients.Prospective study.

69.6 y, 37;57 (65%)male

32; 57 (56%)� 60%stenosis

Not reported

Link44 28; 56 Symptoms,bruit

Consecutive patients.Prospective study.

63 y, 64%male

27% � 70%-99%

Not reported

Londrey61 Notreported;

307

Not reported Consecutive patients.Retrospective study.

Not reported 18% 50%-79%,14% 80%-99% stenosis

Not reported

Mackenzie39 192; 375 Screening,symptoms

Consecutive patients.Retrospective study.

47% male,age notreported

42.4% � 50%,20.9% �70% stenosis

Angiography Kappa0.82-0.88. DUS notreported

Mattos38 167; 167 Screening,symptoms,bruit

Consecutive patients.Retrospective study.

69 y, 58%male

23%: 50%-79%,65%: 80%-99% stenosis

Not reported

Moneta65 100; 184 Not reported Consecutive patients.Retrospective study.

Not reported 32% � 70% Angiography Kappa0.93. DUS notreported.

Moneta70 Notreported;

Not reported Consecutive not reported.Direction of study not

Not reported 46% � 60%stenosis

Not reported

352 reported.

JOURNAL OF VASCULAR SURGERYVolume 41, Number 6 Jahromi et al 972.e3

Table II, online only. Continued

Study

Participants Methods

N Patients;arteries

Setting/Reasonfor referral

Consecutive patients?Prospective/Retrospective

study?

DemographicsAverage age

Genderdistribution Severity

Reproducibility of testsAngiography – NASCET% Stenosis DUS – Doppler

velocity

Muller48 58; 101 Not reported Consecutive patients.Direction of study not

reported.

62 y, 84%male

21; 95 (22%)� 70% stenosis

Not reported

Neale69 60; 120 Not reported Consecutive patients.Retrospective study.

Not reported 24% � 70%stenosis

Not reported

Neschis53 110; 210 Not reported Consecutive not reported.Retrospective study.

Not reported 47% � 50%stenosis

Angiography Kappa0.86. DUS not reported.

Polak33 39; 72 Screening,symptoms,bruit

Consecutive patients.Direction of study not

reported.

45-85 y, 72%male

69% � 50%stenosis

Angiography notreported. DUSPearson’s correlation0.9-0.93.

Ranke56 44; 88 Not reported Consecutive patients.Prospective study.

69 y, 52%male

35% � 70%stenosis

Angiography notreported. DUSinterobserverPearson’s correlation0.94.

Soulez47 211; 365 Not reported Consecutive patients.Prospective study.

Not reported 41.2%: 30%-69%, 35.8%:60%-99%,24.4%: 70%-99%

Angiography Kappa0.78-0.91. DUS notreported.

Spadone54 183; 366 Not reported Consecutive patients.Retrospective study.

Not reported 11%: 50%-79%,16%: 80%-99% stenosis

Not reported

Thomas58 20; 38 Not reported Consecutive not reported.Direction of study not

reported.

69 y, 80%male

NR Not reported

Winkelaar51 99; 188 Not reported Consecutive patients.Direction of study

not reported.

Not reported 57% � 50%stenosis

Not reported

All ages are means. Denominators for percentages in severity is n arteries.

AmF, Amaurosis fugax; DUS, color duplex ultrasound; y � years (average age); N, sample size.