Correlation of Peak Systolic Velocity and Angiographic Measurement of Carotid Stenosis Revisited
Background and Purpose Recent observations from the North American Symptomatic Carotid Endarterectomy Trial (NASCET) questioned the reliability of peak systolic velocity (PSV) criteria for grading carotid stenosis. We compared PSV and angiographic measurements at our center together with known physiological relationships to investigate the accuracy of ultrasound.
Methods Consecutive patients who underwent both color-coded duplex ultrasound and intra-arterial digital subtraction angiography were studied. PSV was determined with angle correction at the site of the tightest internal carotid artery narrowing. Carotid stenosis was measured on angiograms with the North American (N) and common carotid (C) methods. Variables for the stepwise multiple linear regression analysis were selected from an axisymmetrical flow model.
Results Eighty bifurcations were imaged in 40 patients. PSV did not exceed 140 cm/s in normal vessels. In diseased arteries, PSV increased proportionally with increasing stenosis and decreased to 0 cm/s at occlusion. In stepwise selection of polynomial terms, the linear, quadratic, and cubic correlations of .38, .17, and .22 for N and .45, .24, and .03 for C were found to be significant (P<.02). When only stenosed vessels were evaluated, PSV increase was found with greater scatter for the N measurement: r2=.73 for N and r2=.85 for C (n=50; P=.03 for the difference between two correlated correlation coefficients).
Conclusions At our laboratory PSV consistently correlates well with N and C angiographic measurements, as determined with a simple flow model. The complex nature of these correlations and greater variability of the N measurement should be taken into account when data from different centers are compared.
Measuring PSV is the most important component of the carotid Doppler examination.1 As a function of the area of the residual lumen, PSV increases with the narrowing of an artery, implying its usefulness for grading carotid stenosis.1 Although factors such as ICA-CCA velocity ratio and end-diastolic velocity provide additional valuable information for quantifying the stenosis,1 2 3 4 5 6 7 PSV data remain the best single velocity parameter for detecting an operable carotid stenosis by ultrasound.2 However, both ultrasound (screening test) and angiography (gold standard) are operator dependent, with considerable variability in the equipment used and the criteria for interpretation.5 8 9 Therefore, it is now widely accepted that individual laboratories should validate the velocity criteria instead of adopting reported criteria from other laboratories.5 To rely on the accuracy of ultrasound screening, further standardization is necessary before initiation of a multicenter study.10
With no standard ultrasound protocol, the NASCET collaborators retrospectively compared the velocity data obtained before February 1991 at 50 North American centers in 1011 symptomatic patients11 with the angiographic stenosis measurement based on the distal ICA; this is termed the N method.12 Despite methodological simplifications,13 ultrasonography still had moderate sensitivity and specificity, ranging from 65% to 71%, in grading severe carotid stenosis.11 The NASCET scatterplots in Fig 1⇓ in Reference 11 could be explained by the lack of comparable methods between ultrasound laboratories and by wide individual PSV variations produced by various circulatory conditions as well as variability of the N measurement itself.3 9 10
If, as suggested, ultrasonography can only be used as a screening test to rule out carotid artery disease,11 one would expect a relatively random scatter of data points above the normal PSV values. However, the NASCET scatter fits the area under a polynomial curve of the third order (Fig 1⇑) in which the velocity is proportional to the linear, quadratic, and cubic components. This type of curve was described by Spencer and Reid for an axisymmetrical stenosis flow model represented in Fig 6 in Reference 1. This nonrandom distribution of scatter suggests a systematic source of variability in the PSV measurement, which will be proposed here. It is known that different PSV values could be obtained from the same degree of stenosis determined by the N method at laboratories that use different machinery,2 3 4 and the sources of this error have been described.3 Thus, a lack of quality control before multicenter data collection could have contributed to the NASCET scatter.11 13 In addition, factors such as an asymmetrical stenosis, turbulent flow, operator's failure to detect the highest PSV, diminished flow volume or cardiac output, collateralization, and collapse of the distal ICA may affect this correlation.3
Eliasziw et al11 simplified these physiological relationships by retrospectively applying rigid 250-cm/s PSV criteria to the NASCET data despite the fact that flow rates decrease in severe occlusions1 because of a variety of physiological parameters that are often unpredictable.13 Eliasziw et al11 concluded that the use of ultrasonography should be restricted to excluding patients with no carotid artery disease. This “historical approach”13 could lead to the dangerous and regressive step of depriving patients of a noninvasive and reliable procedure while increasing the number of invasive intra-arterial angiograms. It is possible that the variability in the relationship between PSV and angiographic grading of carotid stenosis is due, at least in part, to angiographic methods used in the NASCET study.9 11 13 We therefore compared PSV with angiographic measurement of carotid stenosis in our laboratory and found that a modified definition of angiographic grade both yields a better correlation with PSV and conforms with a physical model of flow in axisymmetrical stenosis.
Subjects and Methods
Consecutive patients who underwent both color-coded duplex ultrasound and intra-arterial digital subtraction angiography were studied prospectively. Patients were referred by neurologists, surgeons, and general practitioners based on history of stroke, transient ischemic attack, or carotid bruit or a high-risk profile for cardiovascular diseases. Both vascular technologists and radiologists obtained the data independently from each other.
Ultrasound testing was performed with an ATL HDI Ultramark-9 unit. In color-coded duplex ultrasound, the flow velocity spectrum represents a color overlay superimposed on a gray-scale image (B-mode). A linear L7-4 probe with an emitting frequency range of 4 to 7 MHz was used. The pulsed-wave Doppler frequency was 4 MHz. The transducer was placed longitudinally parallel to the carotid artery on the neck. The carotid plaque was identified on B-mode scan. With the use of color-coded frequency shifts as a guide, the angle-corrected pulsed-wave Doppler sample volume was placed at the CCAs, external carotid arteries, and ICAs consecutively to measure the PSV. The angle-corrected velocity measurement was then performed to obtain the stenotic jet velocity spectrum.
Intra-arterial digital subtraction angiography was performed within a month of ultrasound screening in all cases by the femoral route with 1024×1024 matrix acquisition (Philips Integris V3000, Philips Medical Systems). Selective bilateral CCA injections were performed in the anteroposterior and lateral projections. The measurements were made on a printed hard copy with the N and C methods and expressed as percent diameter reduction of the vessel.12 The diameter of the residual lumen was measured at the view with the tightest stenosis. The N denominator was determined at the first segment of the far distal ICA with parallel walls beyond the poststenotic dilatation. The C denominator was obtained by measuring the diameter of the proximal CCA 3 to 5 cm below the bifurcation. Unlike previous methods, the C denominator is obtained from the CCA to avoid underestimation of the stenosis and greater variability inherent in the N method.14 15 16 The proximal CCA diameter must be multiplied by 1.2 to estimate the diameter of a normal ICA bulb, or a distal disease-free CCA diameter can be taken as a denominator to calculate percent stenosis. Nevertheless, both CCA measurement sites produce similar results that are close to the values obtained by the European technique; hence, the C method removes any guesswork in estimating the normal carotid bulb diameter.15 16
Least squares multiple linear regression was then used to select the significant variables in the model to explore whether the velocity is a combination of linear, quadratic, and cubic functions of the degree of stenosis. These functions were taken from the axisymmetrical flow model.1 The normal and occluded vessels were then excluded from the analysis to compare PSV and angiographic measurements in the stenosis range reported by NASCET.11 The scatter was plotted and the regression analysis performed with the use of Cricket Graph software, version 1.3.2. To compare which method, N or C, allows closer correlation with PSV, the difference between correlated coefficients was tested with the method suggested by Dunn and Clark.17 The best-curve fit was used to demonstrate graphically the correlation of PSV and angiographic measurements.
Eighty carotid bifurcations were imaged in 40 patients. On angiography 25 vessels (31%) were normal, ie, had no stenosis with both the N and C measurements. Five vessels (6%) were occluded, and 50 vessels (63%) had mild to severe stenoses, including 5 vessels with no stenosis with one method but stenosed with another measurement. In the stenosed vessels, mean (±SD) stenosis was 52±28% (median, 53%; range, 95%) with the N method and 64±27% (median, 68%; range, 96%) with the C method, and mean PSV was 326±153 cm/s (median, 355 cm/s; range, 510 cm/s).
PSV did not exceed 140 cm/s in normal vessels, increased proportionally with increasing degree of carotid stenosis, reaching a maximum of 550 cm/s at 70% to 96% stenosis, and decreased to 0 cm/s at complete occlusion. To ascertain the contribution of each function to overall correlation, stepwise selection in a regression model was performed. The following equations represent the regression model: PSV=aN+bN2+cN3, and PSV=dC+eC2+fC3. The results are represented in the Table⇓. All three variables for each angiographic measurement were significant at the P<.02 level. The steps show that the greatest contribution to the correlation between PSV, N, and C was that of the linear variables (.39 to .45).
The normal and occluded vessels were excluded to evaluate PSV increase with increasing degree of stenosis. A total of 50 measurements were analyzed, and the linear and quadratic correlations were good for both N and C measurements (Figs 2 and 3⇓⇓). However, greater scatter was noted for the N measurement, and the regression coefficient was higher for the C method: r2=.733 for N and r2=.850 for C (P=.03).
Our prospective study has confirmed the strong correlation between angiographic measurement and PSV assessed at a single laboratory with the use of color-coded duplex equipment that was documented in many other studies.1 2 3 4 5 6 7 Our study confirmed that all components of this correlation predicted from the flow models1 are significant, a finding that supports the accuracy of carotid ultrasonography compared with the N angiographic measurement of carotid stenosis.
PSV data correlate with angiographic measurements since both ultrasound and angiography reflect physiological flow changes that occur with increasing degrees of carotid stenosis.1 9 13 Blood flow velocity usually does not exceed 125 to 140 cm/s in normal or mildly stenosed vessels.3 However, increased PSVs were seen in N stenoses in the range of 5% to 30% (ie, PSV=220 cm/s; angiographic stenosis by the N method=18%) (Fig 2⇑). Similar discrepancies were also present in the original NASCET observation11 13 and are due to underestimation of the stenosis by the N method: 0% stenosis by the N method equals up to 50% bulb reduction when the bulb is filled with atheroma up to the straight walls; 30% stenosis by the N method equals 30% to 55% bulb reduction measured by the European technique.9 Conversely, when normal PSV is detected in the presence of moderate stenosis (ie, PSV=120 cm/s; angiographic stenosis by the N method=50%), this may be caused by the failure of the sonographer to detect the PSV at the site of maximal narrowing (usually because of shadowing artifact). On the other hand, asymmetrical stenosis could be overestimated by angiography since angiographic measurements are made at the view of the tightest residual lumen.
When atheroma produces approximately 30% to 40% bulb diameter reduction, it corresponds to more than 50% area stenosis, and PSV increases most abruptly.1 3 This has been shown in our study (Figs 2 and 3⇑⇑), consistent with simple flow models.1 In axisymmetrical constant-flow stenosis, PSV is proportional to the squared radius of the residual lumen.1 However, in our study the linear, quadratic, and cubic components of the flow model were all significant, with the greatest contribution from the linear components. This was observed because most stenoses are axiasymmetrical, and flow volume decreases with moderate to severe carotid artery disease. This is particularly evident at near occlusion when PSVs decrease to near zero values.3 11
From a clinical perspective, NASCET showed an increase of stroke risk with increasing angiographic stenosis.11 However, the NASCET investigators found no relationship between the multicenter PSV data and risk of stroke,11 probably because of individual PSV variations and differences in equipment used. Another possible reason why PSVs per se did not predict the risk of stroke in NASCET could be the decrease in flow volume and technical difficulties in assessing the highest PSV in the presence of high-grade stenosis produced by calcified plaques.1 3 4 18 However, the ICA/CCA velocity ratios, through which one can avoid PSV differences between patients or repeated measurements made on the same patient, showed a decline in benefit of surgery similar to that observed for angiographically defined stenosis.11 The further physiological decrease in blood flow velocities at near-occlusive and occlusive states parallels a decreased risk of stroke found in natural history studies19 20 and the NASCET data.21 Thus, the simple ultrasonographic parameter (PSV) represents an important criterion that should be used carefully in conjunction with other flow and imaging data.1 2 3 12
However, the question of why false-negative results dominate in the NASCET scatterplot remains.11 Several explanations may be offered, including the blinding of the observers and selection bias, which would greatly reduce the number of patients undergoing angiography without high PSVs.22 However, if this is true for an individual center, not all laboratories use 250 cm/s as a cutpoint for 70% stenosis as determined by the N method, as quoted in the NASCET analysis.11 A more fundamental bias is introduced by ultrasound equipment and the Doppler mode of velocimetry, by which the same individual would have different PSV values if evaluated with different machines. This systematic error of up to 50% in absolute PSV values is attributed to transducer geometry alone.23 However, this does not mean that this individual would have different degrees of carotid stenosis diagnosed. The degree of stenosis would be considered the same if different diagnostic criteria (specific to each laboratory or machine) were applied. Because these “errors” are systematic, as long as measurements are made on the same machine or different PSV criteria are used, the results in a given population of patients are consistent and reproducible. For this reason it is crucial to establish local criteria for grading carotid stenosis, and it is meaningless to compare data from different laboratories with the use of a rigid PSV threshold either with or without prior standardization.24 Recommendations for the development of local criteria and quality control were endorsed by the Intersocietal Commission for Accreditation of Vascular Laboratories.25
The scatter of measurements found in our study reflects deviations in both PSV and angiographic measurements from the correlation expected from the axisymmetrical model. This is largely due to asymmetrical residual lumen, which is particularly common in mild and moderate carotid stenosis,16 operator-dependent velocimetry, and variability of angiographic measurements.9 However, the least scatter and the closest correlation were obtained by the C method, since CCA diameter measurements avoid variability of the distal ICA diameter used in the N technique.9 16 The scatterplots represented in Figs 2 and 3⇑⇑ demonstrate that variability comes in large part from the definition of angiographic narrowing rather than the measurement of PSV.
In conclusion, we found that PSV and angiographic measurements obtained within the same institution do correlate, in agreement with known physiological phenomena.1 All the linear, quadratic, and cubic components are significant parts of this model. The complex nature of these correlations, limitations of the N angiographic measurement, and lack of standardization in interpreting ultrasound data before NASCET have led to underestimation of the role of ultrasound. Once identified, these issues should alert ultrasonographers to thorough validation of their own criteria.
Selected Abbreviations and Acronyms
|C||=||angiographic measurement according to the common carotid artery method|
|CCA||=||common carotid artery|
|ICA||=||internal carotid artery|
|N||=||angiographic measurement according to the North American method|
|NASCET||=||North American Symptomatic Carotid Endarterectomy Trial|
|PSV||=||peak systolic velocity|
- Received September 16, 1996.
- Revision received November 18, 1996.
- Accepted November 18, 1996.
- Copyright © 1997 by American Heart Association
Spencer MP, Reid JM. Quantitation of carotid stenosis with continuous wave (C-W) Doppler ultrasound. Stroke. 1979;10:326-330.
Strandness DE. Duplex Scanning in Vascular Disorders. New York, NY: Raven Press Publishers; 1990:92-120.
Neale ML, Chambers JL, Kelly AT, Connard S, Lawton MA, Roche J, Appleberg M. Reappraisal of duplex criteria to assess significant carotid stenosis with special reference to reports from the North American Symptomatic Carotid Endarterectomy Trial and the European Carotid Surgery Trial. J Vasc Surg. 1994;20:642-649.
Moneta GL, Edwards JM, Chitwood RW, Taylor LM, Lee RW, Cummings CA, Porter JM. Correlation of North American Symptomatic Carotid Endarterctomy Trial (NASCET) angiographic definition of 70% to 99% internal carotid artery stenosis with duplex scanning. J Vasc Surg. 1993;17:152-159.
Bladin CF, Alexandrov AV, Murphy J, Maggisano R, Norris JW. Carotid Stenosis Index: a new method of measuring carotid artery stenosis. Stroke. 1995;26:230-234.
Eliasziw M, Rankin RN, Fox AJ, Haynes RB, Barnett HJM, for the North American Symptomatic Carotid Endarterectomy Trial (NASCET) Group. Accuracy and prognostic consequences of ultrasonography in identifying severe carotid artery stenosis. Stroke. 1995;26:1747-1752.
Nicolaides AN, Shifrin E, Dhanjil S, Griffin M. Duplex grading of internal carotid stenosis. In: Caplan LR, Shifrin EG, Nicolaides AN, Moore WS, eds. Cerebrovascular Ischemia: Investigation and Management. London, England: Med Orion; 1996:101-110.
Ringelstein EB. Skepticism toward carotid ultrasonography: a virtue, an attitude, or fanaticism? Stroke. 1995;26:1743-1746.
Vanninen R, Manninen H, Koivisto K, Tulla H, Partanen K, Puranen M. Carotid stenosis by digital subtraction angiography: reproducibility of the European Carotid Surgery Trial and the North American Symptomatic Carotid Endarterectomy Trial measurement methods and visual interpretation. AJNR Am J Neuroradiol. 1994;15:1635-1641.
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.
Bladin CF, Alexandrova NA, Murphy J, Alexandrov AV, Maggisano R, Norris JW. The clinical value of methods to measure carotid stenosis. Int Angiol. In press.
Steinke W, Kloetzsch E, Hennerici M. Carotid artery disease assessed by color Doppler flow imaging: correlation with standard Doppler sonography and angiography. AJNR Am J Neuroradiol. 1990;11:259-266.
Norris JW, Zhu CZ. Stroke risk and critical carotid stenosis. J Neurol Neurosurg Psychiatry. 1990;53:235-237.
Morgenstern L, Fox AJ, Sharpe B, Eliasziw M, Barnett HJM, Grotta JC. The risk and efficacy of carotid endarterectomy for near-occlusion: results from NASCET. Neurology. 1996;46:A280. Abstract.
Chang YJ, Golby AJ, Albers GW. Detection of carotid stenosis: from NASCET results to clinical practice. Stroke. 1995;26:1325-1328.
Daigle RJ, Stavros AT, Lee RM. Overestimation of velocity and frequency values by multielement linear array Dopplers. J Vasc Tech. 1990;14:206-213.
Harbison J, Eliasziw M. Response. Stroke. 1996;27:1436. Letter.
Katanick SL. Accreditation of vascular ultrasound laboratories. In: Tegeler CH, Babikian VL, Gomez CR, eds. Neurosonology. St Louis, Mo: CV Mosby Co; 1996:484-488.