Carotid Stenosis Index
A New Method of Measuring Internal Carotid Artery Stenosis
Background and Purpose Current methods of measuring carotid stenosis such as those used in the North American Symptomatic Carotid Endarterectomy Trial (NASCET) and European Carotid Surgery Trial (ECST) have limitations caused by difficulties in measuring the normal width of the distal internal carotid artery (ICA) or the carotid bulb.
Methods We developed a new technique, the Carotid Stenosis Index (CSI), based on the known anatomic relationship between the common carotid artery (CCA) and ICA (1.2×CCA diameter=proximal ICA diameter). The normal ICA diameter can therefore be calculated from direct measurement of the CCA. Three blinded observers evaluated the angiograms of 57 patients (114 carotid arteries), previously screened with duplex ultrasonography, using the NASCET, ECST, and CSI methods. In a subset of 30 patients undergoing carotid endarterectomy, comparison was also made to computerized carotid plaque planimetry.
Results The NASCET method could only be applied correctly in 89% and the ECST method in 95% of cases because of overlying vessels or inadequate views of the distal ICA or carotid bulb. An additional 9% of NASCET cases had a “negative” stenosis, in which the stenosis is wider than the distal ICA. The CSI method was applicable in 99% of cases. Interobserver comparison using ANOVA revealed significant differences using NASCET (P<.0001) and ECST (P<.001) but not CSI (P=NS). NASCET had a sevenfold variation (P<.01) and ECST a twofold variation (P<.01) in results compared with CSI . The intraobserver reliability was 0.87 for NASCET, 0.86 for ECST, and 0.90 for CSI. However, the 95% confidence intervals for an individual measurement by an observer were ±30% for NASCET, ±19% for ECST, and ±15% for CSI. With linear methods of measurement there were significant differences between NASCET and CSI (P<.0001) and ECST (P<.0001) but not between CSI and ECST. A comparison of area derivations of these methods to carotid plaque planimetry revealed significant differences from NASCET (P<.0001) but not ECST, CSI, or duplex methods. A CSI nomogram was created, allowing measurement of both linear and area percent stenosis.
Conclusions CSI is the most reliable validated method of measuring carotid stenosis, and it correlates with duplex and carotid pathology.
Recent randomized trials of carotid endarterectomy have focused interest on the accuracy of measuring carotid stenosis by angiography1 2 3 and the validity of carotid ultrasound as a screening test.4 5 6 There are significant discrepancies between the North American Symptomatic Carotid Endarterectomy Trial (NASCET) and the European Carotid Surgery Trial (ECST) angiographic methods, such that a 50% stenosis by NASCET is equivalent to approximately 70% by ECST.6 7 This may produce a management dilemma for the clinician considering carotid surgery for an individual symptomatic patient because, in the absence of data for the moderate stenosis group, 70% is the present limit for surgical treatment.
Both the NASCET and ECST angiographic methods have limitations.6 7 8 With the NASCET method, one problem is deciding the appropriate segment of distal internal carotid artery (ICA) on which to base a measurement. Inadequate contrast enhancement or the presence of overlying vessels may obscure the distal ICA. Furthermore, since the carotid bulb is larger than the distal ICA, minor degrees of stenosis (<50% bulb diameter reduction) may result in a “negative stenosis.”6 For ECST measurements, the problem lies in predicting the normal position of the carotid bulb; variations in anatomy or irregular stenoses can make this difficult even for the experienced observer.
The principal difficulty, therefore, with both methods is establishing the diameter of the normal ICA. An alternative approach is to use the well-established physiological and anatomic relationship between the common carotid artery (CCA) and proximal ICA, so that the normal ICA diameter can be derived from measurement of the CCA. For example, duplex ultrasound studies have consistently shown normal ICA/CCA peak blood flow velocity ratios of 0.7,9 10 and studies of normal angiograms11 have shown the ICA/CCA diameter ratio is 1:1.19 (±0.09).
The aim of this study was to develop a technique based on this principle, termed the Carotid Stenosis Index (CSI), and compare this with current methods (NASCET, ECST, duplex ultrasound) and with the “gold standard,” carotid plaque pathology.
Subjects and Methods
We prospectively compared the angiographic findings of consecutive patients undergoing carotid angiography. All patients had transient ischemic attack or minor stroke. Before angiography all patients had color-coded carotid duplex ultrasound (Diasonics Inc). Angiography was performed by experienced radiologists using the intra-arterial digital subtraction technique through the femoral route with selective catheterization of the extracranial arteries. Biplanar images were obtained of each carotid bifurcation.
Three blinded observers made measurements from the printed x-ray film at the following sites (Fig 1⇓): ICA stenosis (at the narrowest point), distal ICA (as recommended by NASCET1 ), predicted outline of the ICA bulb (as recommended by ECST2 ), and proximal CCA (3 to 5 cm proximal to the carotid bifurcation). Percent stenosis was then calculated using the formula (1−D/N)×100%, where D is the diameter of the stenosis lumen and N is the normal ICA diameter. This formula was applied according to the methods of NASCET and ECST.
To use the CSI, the proximal CCA diameter is first multiplied by 1.2 (approximated from 1.19, following published data11 ) so as to derive the expected width of the normal proximal ICA. Percent stenosis is then calculated as in the other methods by comparing this with the diameter of the ICA stenosis (D) using the formula (1−D/N)×100%, with N=1.2×CCA diameter.
Each observer made note of any methodological difficulties they encountered in trying to use each of these angiographic methods. To assess intraobserver reliability, two of the observers remeasured 58 vessels 5 months after the initial measurements were made.
A subgroup of patients undergoing carotid endarterectomy had the carotid plaque removed intact, which was sent for processing and sectioning in the pathology laboratory as previously described.6 Using a computerized technique, “Sigma scan” (Jandel Scientific), the cross-sectional area (mm2) or planimetry of the stenosed lumen and the proximal ICA were measured. Comparison was then made between plaque pathology and the area derivations of CSI, NASCET, ECST, and duplex methods.6
Alterations in plaque morphology with processing produced uniform shrinkage of the specimen by 8±9.5%, consistent with previous studies,12 such that the degree of stenosis was not affected.
Statistics were performed using a two-tailed t test for continuous variables and ANOVA for comparison between observers using each method, using the sas statistical software.13 Comparison of the variation of each method was made using the Fmax statistic.14 Interobserver and intraobserver reliability was measured using the intraclass correlation coefficient.15
A total of 57 patients (114 carotid arteries) were evaluated. Five percent (6/114) of vessels were occluded. Carotid plaque planimetry was available for comparison for 30 patients having carotid endarterectomy.
Using ANOVA, there were significant observer effects for NASCET (F2,112=14, P<.0001) and ECST (F2,112=6.85, P<.001) but not CSI (F2,112=2.53, P=NS). For NASCET the mean±SD of each observer was R1=31±40%, R2=36±40%, R3=41±35%; for ECST, R1=61±25%, R2=63±26%, R3=65±24%; and for CSI, R1=64±24%, R2=66±23%, R3=66±24%.
Another test of the variability of each method can be assessed by comparing the mean square errors (MSE), using the Fmax statistic.14 Compared with CSI (MSE, 378.7), there is a sevenfold variation in NASCET (MSE, 2787.3; Fmax P<.01) and a twofold variation in ECST (MSE, 915.6; Fmax P<.01).
Analysis of interobserver and intraobserver reliability revealed intraclass correlation coefficients for each method: NASCET, 0.81 and 0.87; ECST, 0.86 and 0.86; and CSI, 0.88 and 0.90. However, the range of possible error by an observer for an individual measurement (the 95% confidence intervals derived from the standard error of the mean) was greatest for NASCET and least for CSI (NASCET ±30%, ECST ±19%, CSI ±15%).
For linear methods of measurement, there was no significant difference between the mean percent stenosis of CSI and ECST. There was, however, a significant difference between NASCET and ECST (36±24% versus 62±16%, P<.0001) and NASCET and CSI (36±24% versus 65±16%, P<.0001).
Comparison was then made of the area derivations of each of the three methods, duplex ultrasonography, and carotid plaque planimetry (Table⇓). This revealed no significant difference between carotid pathology results and CSI area, duplex area, and ECST area, but significant differences were seen between carotid pathology and NASCET area (P<.0001).
Technical factors limited the clinical usefulness of each method in consecutive angiograms. NASCET could be correctly used in only 89% and ECST in 95% of vessels because of the presence of overlapping vessels, poststenotic collapse or dilation, or inadequate views of the distal ICA or carotid bulb outline. Use of the NASCET method also resulted in an additional 9% of stenoses being recorded as “negative” stenosis, ie, the stenosis diameter is greater than the distal ICA diameter. The CSI method was applicable in 99% of cases, the only limitation being one CCA with extensive atheroma.
From these results a comparison was then made to ascertain the interrelationship between CSI and NASCET. Fig 2⇓ shows that a CSI stenosis of 42% is equivalent to 0% by NASCET; anything less than 42% becomes by definition a “negative” NASCET stenosis. The slope of the line in Fig 2⇓ shows that the relationship between NASCET and CSI data is described by the equation CSI=42%+0.6 NASCET.
To facilitate use of the CSI, a nomogram was devised (Fig 3⇓), allowing accurate measurement of percent stenosis by either linear or area methods. The nomogram consists of two parallel lines, one representing the CCA diameter, the other the ICA stenosis diameter, connected by a central diagonal on which linear and area percent stenosis markings are displayed. No calculations are required for its use, only the CCA diameter measurement at a site free of atheroma approximately 3 to 5 cm proximal to the carotid bifurcation and the ICA stenosis diameter. A line drawn across the nomogram then reads percent stenosis from the central markings.
The results of this study demonstrate that CSI provides a consistent and accurate angiographic measure of carotid stenosis that compares well with both duplex ultrasound and carotid plaque pathology. Of the current methods, NASCET had the most interobserver and intraobserver differences in results (and widest confidence intervals), reflecting the spectrum of individual interpretation possible in its use. The clinical significance of these differences can be judged from the means of each observer. The maximum difference in percent stenosis measurements was 10% for NASCET, 4% for ECST, and 2% for CSI.
There have been few published studies on the reliability of angiographic methods of measurement and none comparing different methods. Chikos et al16 examined 64 selected angiograms comparing the stenosis with the carotid bulb (ECST method). They found an overall intraobserver correlation coefficient of 0.98 with a mean difference in readings of 6±8%, although the validity of the statistical methods used is unclear. Murie and McKay17 compared 100 angiograms with the NASCET method using six categorical ratings. They found interobserver and intraobserver agreements of 74% and 83%, respectively (kappa analyses were not performed), with agreement improving with fewer categories used.
The results show variable applicability of NASCET and ECST to consecutive angiograms. For instance, 11% and 5% of vessels could not be technically evaluated using the NASCET and ECST methods, respectively, despite the use of biplanar angiography in all cases. All vessels could be assessed by CSI, since the CCA was always well visualized, and there were no instances of a “negative” stenosis (ie, a stenosis measurement <0%).
The technique of CSI depends on a measurable CCA to allow calculation of the proximal ICA diameter. The anatomic relationship between components of the CCA and the ICA has been established from angiographic studies based on normal, nonatheromatous bifurcations11 as well as Doppler studies comparing velocity ratios.9 Williams and Nicolaides11 analyzed 61 normal angiograms and found a ratio of 0.65 for distal ICA/CCA and 1.19 for proximal ICA/CCA. Others18 have compared distal CCA/proximal ICA for a ratio of 1.11; however, the distal CCA is less reliable because it is subject to more atheroma and more anatomic variation and is the only site to have a statistically significant intersex difference for the CCA/ICA ratio. Doppler velocity ratios for ICA/CCA are 0.4 to 0.7 for normal vessels9 10 ; in the assessment of carotid atheroma, a velocity ratio >1.5 is considered significant for stenosis >50%.10
In the development of CSI, we have adopted the ICA/CCA ratio of 1:1.2. The percent error with this numerical value is minor, and there are no significant interobserver differences and less variability compared with the other methods. The CCA is a good site for angiographic measurements because it is a straight artery with no overlying vessels and is least affected by extracranial atheromatous disease,11 19 although compensatory dilation is a possible confounding issue.20 Its proximal location ensures good angiographic contrast definition for measurement of vessel edges. Furthermore, for noninvasive methods, the CCA is readily identifiable on duplex ultrasonography, and its size and flow characteristics minimize potential errors with magnetic resonance angiography.
There is some controversy over whether the proximal or distal ICA should be used as the denominator in measuring carotid stenosis. Using the carotid bulb has been dismissed by some authors as “anatomic purism” with no hemodynamic relevance.8 However, as embolism (due to ulceration) may be the most common cause of carotid stenosis stroke,21 hemodynamic considerations may be of less importance.
There are other reasons for using the proximal ICA as the principal site of measurement.11 (1) The proximal ICA is where the atheromatous lesion commences and develops. Accurate measurement of minor grades of disease is only possible using this location as the basis for measurement. The gradation of stroke risk with increasing stenosis can then be correlated with morphology of the developing plaque.22 (2) Mild and moderate degrees of stenosis correlate with an equivalent risk of ischemic events. ECST has shown no benefit for surgery in patients with minor (0% to 29%) ICA stenosis. However, a 0% to 29% NASCET stenosis is more severe (equivalent to 42% to 60% stenosis ECST) and therefore has a greater likelihood of being symptomatic22 and responsive to surgery. There is transcranial Doppler evidence that with increasing stenosis there is more embolic activity (“high-intensity transient signals”).23 24 Using the proximal ICA therefore gives a more realistic assessment of the amount of atheroma and embolic potential of a stenosis and thus a patient’s stroke risk. (3) There is a close relationship with duplex sonography. Present duplex machines can detect minor degrees of stenosis with B-mode imaging.25 There is a poor correlation of ultrasound findings with angiographic methods such as NASCET, which are based on the distal ICA.6 (4) Correlation is possible with carotid pathology. The extent of atheroma and degree of stenosis found on carotid pathology are best described by angiographic methods that use the proximal ICA.6
The development of CSI has practical implications for patient management. NASCET clearly demonstrated the highest stroke risk per decile of symptomatic stenosis >70%, and the results of this and the other carotid surgery trials have led to many more patients with symptomatic severe carotid stenosis being referred for assessment by ultrasonography and angiography.26 However, an 80% stenosis measured by ECST is less than 70% by NASCET, even though pathologically it is the same stenosis. Depending on which angiographic technique is used, the patient may subsequently undergo surgery with its 2% to 3% major and 5% overall risk of mortality and morbidity.1
It has also been suggested26 27 that, with the recent increase in referrals for angiographic assessment of carotid stenosis, some clinicians are now giving patients with less severe degrees of stenosis the “surgical benefit of the doubt.” As shown in this study, the possible variations in selecting a distal ICA diameter (NASCET) or drawing a carotid bulb outline (ECST) make this subjective bias possible.
Until now there was no reliable way of relating the results of the two carotid endarterectomy trials. This was attempted recently by remeasuring ECST angiograms using the NASCET method,28 but our data suggest that the potential for error is greater with the NASCET method. CSI and ECST are similar, with both showing no significant differences from duplex or pathology. However, the CSI is more reliable, is less prone to technical error, and can be correlated with NASCET (Fig 2⇑), thus providing a bridge between current carotid surgery trials.
Dr Bladin was a postgraduate research fellow from the University of Melbourne (Australia). Dr Alexandrov is the recipient of the American Heart Association Robert G. Siekert Young Investigator Award in Stroke (Honorable Mention) for work related to this study.
Presented in part at the American Heart Association 19th International Conference on Stroke and the Cerebral Circulation, San Diego, Calif, February 17-19, 1994.
- Received June 10, 1994.
- Revision received October 25, 1994.
- Accepted October 25, 1994.
- Copyright © 1995 by American Heart Association
The Asymptomatic Carotid Atherosclerosis Study Group. Study design for randomized prospective trial of carotid endarterectomy for asymptomatic atherosclerosis. Stroke. 1989;20:844-849.
Haynes RB, Taylor DW, Sackett DL, Fox A, Rankin R, Barnett HJM. Poor performance of Doppler in detecting high grade stenosis. Clin Res. 1992;40:184a.
Alexandrov AV, Bladin CF, Maggisano R, Norris JW. Measuring carotid stenosis: time for a reappraisal. Stroke. 1993;24:1292-1296.
Bousser MG. Benefits from carotid surgery? Yes, but. . . Cerebrovasc Dis. 1992;2:122-126.
Blackshear WM, Phillips DJ, Chikos PM, Harley JD, Thiele BL, Strandness DE. Carotid artery velocity patterns in normal and stenotic vessels. Stroke. 1980;11:67-71.
Williams MA, Nicolaides AN. Predicting the normal dimensions of the internal and external carotid arteries from the diameter of the common carotid. Eur J Vasc Surg. 1986;1:91-96.
Schenk EA, Bond G, Aretz TH, Angelo JN, Choi HY, Rynalski T, Gustafson NF, Berson AS, Ricotta JJ, Goodison MW, Bryan FA, Goldberg BB, Toole JF, O’Leary DH. Multicenter study of real time ultrasonography, arteriography, and pathology: pathologic evaluation of carotid endarterectomy specimens. Stroke. 1988;19:289-296.
SAS Institute. SAS/STAT User’s Guide. 4th ed. Cary, NC: SAS Institute Inc; 1992.
Winer BJ, Brown DR, Michels KM. Statistical Principles in Experimental Design. New York, NY: McGraw-Hill Book Co; 1991.
Chikos PM, Fisher LD, Hirsh JH, Harley JD, Thiele BL, Strandness DE. Observer variability in evaluating extracranial carotid artery stenosis. Stroke. 1983;14:885-892.
Mohr JP, Gautier JC, Pessin MS. Internal carotid artery disease. In: Barnett HJM, Mohr JP, Stein BM, Yatsu FM, eds. Stroke: Pathophysiology, Diagnosis, and Management. 2nd ed. New York, NY: Churchhill Livingstone; 1992:302-305.
Bornstein NM, Norris JW. The unstable carotid plaque. Stroke. 1989;20:1104-1106.
Siebler M, Sitzer M, Rose G, Bendfeldt D, Steinmetz H. Silent cerebral embolism caused by neurologically symptomatic high-grade carotid stenosis. Brain. 1993;116:1005-1015.
Strandness DE. Duplex Scanning in Vascular Disorders. New York, NY: Raven Press Publishers; 1990.
Barnett HJM, Taylor DW. Clinical trials in stroke prevention: persisting uncertainties, firm answers. Neurology. 1993;43:2163-2166.
Barnett HJM. Progress in understanding the role of carotid endarterectomy. Can J Neurol Sci. 1994;2:85-87.
Barnett HJM, Warlow CP. Carotid endarterectomy and the measurement of carotid stenosis. Stroke. 1993;24:1281-1284.