Quantitative common carotid artery flow: Prediction of internal carotid artery stenosis

Transcription

1 Quantitative common carotid artery flow: Prediction of internal carotid artery stenosis blood Nigel Ackroyd, M.R.C.P., F.R.C.S., F.R.A.C.S., Robert Gill, M.Sc., Ph.D., Kaye Griffiths, S.R.N., D.M.U., George Kossoff, D.Sc., and Michael Appleberg, F.C.S.(S.A.), F.R.A.C.S., F.A.C.S., Sydney, Australia Common carotid artery (CCA) blood flow was measured noninvasively with a pulsed Doppler duplex scanner modeled after the Octoson (Ultrasonics, Inc., No. Yonkers, N. Y.). The aim of the study was to determine normal values and to assess the accuracy of CCA flow as a predictor of internal carotid artery (ICA) stenosis. One hundred one people who did not have disease were studied; the overall mean flow rate was ml/min (mean -+ S.D.). There was no significant correlation with age, height, or body surface area but there was with body weight (p < 0.05). A statistically significant difference was evident between men (424 ~- 88 nil/rain) and women ( ml/min) (p < 0.001). The intrasessional variation (S.D./mean) was 13% and the intersessional variation, 16%. No significant difference was seen between the sides. Ninety-two patients who had carotid angiography were studied and the flow rates compared with the degree of ICA stenosis on each side. The flow rate for mild ICA stenosis (1% to 39%) was ml/min, for moderate stenoses (40% to 69%), ml/min, and for severe stenoses (70% to 99%), ml/min. There was a significant difference in flows only between mild and severe grades of stenosis of the ICA (p < 0.01). With unilateral stenosis, comparison of flow values in the normal and affected sides showed the greatest discriminatory power when the absolute difference of flow values was taken (p < 0.005). In terms of clinical use, a flow criterion of less than 350 ml/min gave an overall sensitivity of 59% and a specificity of 64% for distinguishing greater than 50% ICA stenosis. We conclude that the individual variation of CCA flow gives the technique only moderate accuracy as a noninvasive test in the prediction of ICA stenosis. (J VAsc SURG 1986; 3: ) With a combination of imaging and Doppler ultrasound, it is now possible to measure blood flow noninvasively in deep and superficial vessels, ~3 including the common carotid artery (CCA).4 Therefore, quantitative flows can be measured in conscious patients. An important clinical application of this technique would be to determine whether internal carotid artery (ICA) stenosis could be diagnosed by the measurement of CCA blood flow. If this were so, then the technique could be added to the noninvasive methods of the vascular laboratory. It was with this purpose in mind that the present study was undertaken. Carotid stenosis is said to be hemodynamically significant when it causes a reduction in blood flow. From the Vascular Surgical Unit, Royal North Shore Hospital, and Ultrasonics Institute, Millers Point. Reprint requests: Mr. Nigel Ackroyd, Dept. of Surgery,, Royal North Shore Hospital, Sydney, 2065, Australia. 846 It is generally held that flow reduction begins at approximately 50% stenosis of arterial diameter but does not become pronounced until stenosis has reached 80%. At higher grades ofstenosis than these, blood velocity increases in the stenotic segment but total flow decreases markedly. Although the diameter of the residual lumen is the most important factor governing flow, the length of the stenosis, its luminal irregularity, and blood viscosity are also considerations. Finally, stenoses have greater effects on high rates of flow than on slow rates. These hemodynamic considerations are discussed more fully by Strandness and Sumner. ~ Therefore, it is expected that a high-grade ICA stenosis would reduce total flow in the CCA and so provide a means to detect these lesions. However, angiograms have demonstrated that the external carotid artery (ECA) may contribute collateral flow intracranially in the presence of ICA stenosis; therefore, as ICA flow decreases, ECA flow would, up to

2 Volume 3 Number 6 June 1986 Quantitative carotid blood flow 847 Table I. Summary of CCA flow in normal subjects as measured by different systems for quantitative flow measurement No. of Variation Imager Doppler Author Flow SD subjects (%) System (MHz) (MI-Iz) Current study Octoson 3 3 Muller et al. ~ QFM 6 5 Uematsu et al QFM 5 5 Fujishiro and Yoshimura QFM 5 5 Furuhata s 500 * * * QFM 6 5 Fitzgerald, O'Shaughnessy, Alvar 8 8 Keaven 9 Payen et al. ~ Alvar 8 8 Simon et al.n MGPD * * Keller et al. n 390 * 22 * MGPD NIL 7 Borodzinski et al)3 530 * 1 * UDIMP Olson * 7 * A-mode 8 8 Fish ~s * * * * Mavis 5 5 QFM = quantitative flow meter; MGPD -- multigated pulsed Doppler. *Information not stated. a point, increase. This collateral flow would have the effect of maintaining CCA flow when ICA stenosis is present and perhaps detract from the usefulness of the technique, although the procedure does provide valuable information regarding the maintenance of total cerebral blood flow. We also thought the technique was worthy of assessment, with a view to determine the physiologic norms for CCA flow. Ultrasonic flowmeters need to obtain information about the cross-sectional area of the vessel, the mean Doppler frequency shift, and the angle of insonation of the ultrasound beam to measure flow rate. These parameters are related to the flow rate as shown in equations 1, 2, and 3. Fd x 2C Velocity - Fo cos 0 (1) in which Fd is the Doppler-shifted frequency; Fo, the Doppler frequency; C, 1540 m/see; and 0, angle of insonation. Area = ~rr 2 (2) in which R is the radius of the artery. Finally, Flow = area x velocity (3) The unique feature of the Doppler configuration on the Octoson is that it insonates the whole vessel in three planes. Hence, the instantaneous mean velocity of flow can be measured regardless of the velocity profile of blood flow; this fact gives it a considerable advantage over other systems, s The area is readily determined from the diameter or can be measured directly in cross section from the B-mode image. Likewise, the angle of insonation of the Doppler probe can be deduced from this image and used to convert the Doppler frequency shift into a velocity. The determination of mean velocity does pose some problems and several configurations of pulsed-wave Doppler (PWD) have been designed for its measurement (Table I). Blood velocity varies with the phase of the cardiac cycle and also the center streamlines move at a faster rate than those at the boundary. Thus a flowmeter system would have to sample all the velocities at all phases of the cardiac cycle to calculate flow/some systems take repeated small sample volumes across the diameter of the vessel and then calculate the mean velocity (multigated PWD) on the assumption of axisymmetric flow. The Octoson system samples velocity from the whole vessel in one step and therefore has considerable advantages in terms of simplicity. MATERIAL AND METHODS Control group. One hundred one volunteers (46 men, 55 women; age range, 20 to 80 years) were studied as control subjects. None had a carotid bruit or history of vascular disease or previous vascular surgery. Twenty-two patients who had nonischemic neurologic symptoms and had a completely normal cerebral angiogram were included in the control group. Twenty normal subjects had repeat studies approximately 1 month later to study the intersessional variation of flow readings. Study group. Angiograms of 92 patients showed abnormalities. All had biplanar carotid and cerebral angiography performed by the Seldinger technique via a groin puncture. The percentage of diameter stenosis of all the major branch vessels from the aortic arch to the intracerebral circulation were recorded. The diameter of the normal ICA was taken as the

3 848 Ackroyd et al. Journal ot VASCULAR SURGERY Fig. 1. Longitudinal section of CCA showing 10 cm length of vessel. Calipers measured diameter of vessel as 7.25 mm. Angle bar has been placed along axis of flow at point of measurement. Note angulation in vessel (arrow). diameter at a normal point nearest the area of stenosis, and diameter reduction was calculated according to this reference point. Equipment and technique. The Octoson combines B-mode ultrasound and a pulsed Doppler system to quantitate flow rate) s The B-mode component can use up to eight 3 MHz transducers to scan simultaneously, although four transducers are all that are usually required. This feature allows the whole length of the vessel from the manubrium to the bifurcation to be visualized, which facilitates accurate orientation (Fig. 1). Next, the transducer that gives the best image (usually the one most perpendicular to the vessel wall) is chosen and the image magnified. The imaging rate during this phase is approximately 4 Hz and final orientation is made along the three orthogonal coordinates in addition to tilt and rotation. The internal diameter of the CCA is measured with calipers and an angle bar is visually lined up with the longitudinal axis of the vessel (Fig. 1); this allows correction of the Doppler-shifted frequencies for the angle of incidence of the Doppler beam. The pulsed Doppler mode has a frequency of 3 MHz and a pulse repetition frequency of 3.4 khz. Any one of the eight transducers may be used to obtain Doppler information (Fig. 2). Unidirectional Doppler is used to avoid interference from the counterflowing jugular vein. The range gate of the pulsed Doppler is such that it samples the whole width of the vessel both in the x-y plane and, since the beam is 14 mm thick, also in the z-axis. With this arrangement uniform insonation is obtained and the mean velocity calculated regardless of any radial asymmetry of flow that is present during the cardiac cycle. Signals obtained with an angle of incidence of 50 to 64 degrees are preferentially chosen to measure the blood velocity. Smaller angles are liable to produce frequency aliasing, whereas angles greater than 64 degrees are attended by potentially unacceptable angle errors. The sampling area may require minor positional adjustment to obtain the maximum audio signal while the average velocity waveform is observed on the monitor screen. The Doppler trace is frozen and the flow averaged over five to eight cardiac cycles. The measurements are repeated four to five times on each side (Fig. 3). RESULTS Control group. CCA flow in the 101 subjects was ml/min (mean _+ S.D.) (Table II). There was a significant difference between blood flows of men and women (424 and 371 ml/min, respectively [p < 0.001]). There was no age difference between the men and women but marked differences in weight, height, and body surface area were evident (all p < ) (Table II), indicating that the women were considerably smaller than the men. The correlation coefficients for flow against age,

4 Volume 3 Number 6 June 1986 Quantitative carotid blood flow 849 Fig. 2. Doppler sample volume has been placed over whole width of CCA. Fig. 3. Picture is split in three parts during Doppler acquisition phase. Top left is focusing frame. Top tight is B-mode frame. Along bottom is mean velocity signal, which has been measured over four cardiac cycles to give absolute flow rate of 519 ml/min. weight, and body surface area show no significant association with flow (Table Ill); however, there was a significant association with body height (p < 0.05). When the flows were normalized with regard to sex, then no correlation with weight, height, and body surface area was present (Table III). The 95% confidence interval (mean + 2 S.D.) for all normal flows was 237 to 553 ml/min, which indicates the wide range of normality. When repeat studies were made in 20 normal subjects, the intersessional coefficient of variation was found to be 16%, whereas the intrasessional variation was 13% (Table IV). Study group. The ICA stenosis ranges were classifted as normal, mild (1% to 39%), moderate (40% to 69%), and severe (70% to 99%); the flow rates

5 S0 Ackroyd et al. Journal of VASCULAR SURGERY Table II. Overall data of control group Normal values (mean +_ SD) Overall Men Women Parameter (N = 101) (N = 46) (IV = 55) Sex difference* (r, Value) Age (yr) ± Weight (kg) ± Height (cm) 169 ± Surface area (m 2) _ Right to left ratio 1.06 ± ± ± Right to left difference ± (ml/min) Right to left % 4.1 -! ± difference Right flows (ml/min) 402 ± _ Left flows (rnl/min) 387 ± ± Average flow (ml/min) ± *Student t test results between men and women. Note large difference in size between sexes. Table IlL Correlation coefficients between CCA and age, weight, height, and body surface area in normal subjects Table IV. Coefficients of variation (S.D./mean) during same session between sessions and between individuals Parameter Correlations with flow (r values),nonnormalized Sexnormalized Intrasessional 13 % Intersessional 16% Interindividual 20% Sex * Age Weight 0.13 Height 0.24* Surface area *p = When normalized for sex differences the correlation with body size disappears. were 404 _+ 109 ml/min, ml/min, and 351 _+ 109 ml/min, respectively (Table V). These stenosis ranges were selected to avoid the problems associated with the overlap of angiographic interpretation at approximately 50% ICA stenosis. The moderate range represents, in effect, a borderline hemodynamically significant group. The difference in flow rates between mild and severe ICA stenosis was statistically significant (p < 0.01), but there was no difference between mild and moderate and moderate and severe stenosis. When the ICA was totally occluded, the flows reduced to 213 +_ 58 ml/min and this was significantly different from normal, mild, moderate, and severe stenoses (all, p < ). With arch or main stem stenosis or occlusion, the flow rates were reduced to 67 +_ 58 ml/min; but this was a small and heterogeneous group. The correlation coefficient between the percentage of stenosis of the ICA for the 170 nonoccluded angiographically measured vessels and the flow rates was -0.23, which was significant (p < 0.01). The Table V. Summary of all cases ICA stenosis Flow Group (%) No. of cases (ml/rain) Normal ± 79 Mild _+ 109 Moderate Severe Occlusion Main stem ± 58 NOTE: Significant differences exist between main stem and occlusion and all other groups. Significant difference seen between normal or mild and severe (p = 0.008). regression equation was (constants and S.D. indicated): Flow = (412 12) - ( ) (4) (%ICA stenosis) With a CCA flow value of <350 ml/min used as the criterion for the prediction of disease, >50% ICA stenosis gave a specificity of 64% and a sensitivity of 59%; for >70% ICA stenosis, a specificity of 66% and a sensitivity of 72% were determined. Unilateral stenoses. Unilateral stenoses represented a subset of the study group and consisted of those patients who had <40% ICA stenosis on at least one side. This subset was defined to remove the effect that a contralateral stcnosis may have on the

6 Volume 3 Number 6 June 1986 Quantitative carotid blood flow 851 Table VI. Normal to abnormal comparisons for unilateral stenosis No. of Normal/ Normal/ Difference ICA stenosis cases Raw flows abnormal abnormal (%) Normal and mild Moderate Severe _ Occlusion F test NOTE: All cases had one side with <4096 ICA stenosis and the other side normal or stenosed to the extent indicated. Normal and mild stenosis groups have been grouped tog.ether to increase subset size. Excluding the occlusion group became of the small numbers, F values from the one way analysis of variance test indicated that normal - abnormal difference has the greatest discriminatory power. ipsilateral flow and to allow valid normal to abnormal comparisons to be made. The CCA flow values of the left to fight side were compared, giving the normal to abnormal comparisons shown in Table VI. Four modes of comparison are indicated: CCA flow in the affected side only, the ratio of flows in the affected and unaffected sides, and the absolute and percentage difference in flows between the two sides. When the F value from the one-way analysis of variance was applied to these four modes of normal to abnormal comparison (occlusions were excluded became of their small number), the difference in flows between the two sides appeared to have the greatest discriminatory Power followed by the percentage difference in flow and the normal to abnormal ratio. The receiver operating characteristic (ROC) curves (Fig. 4) show the greater discriminatory power of normal to abnormal comparisons over raw flow values for decision making regarding individual patients as opposed to grouped data to which the F value refers. The data points for flow were <400, <350, and <300 ml/min; for normal to abnormal ratios >1.1, >1.2, and >1.3, for normal minus abnormal difference, >0, >50, and >100 ml/min; and for normal to abnormal percentage of difference, >10%, >20%, and >30%. The most satisfactory values for decision making are <350 ml/min for flow, > 1.2 for ratios, > 100 ml/min for difference between normal and abnormal, and >20% for percentage of difference. DISCUSSION The flows in this study have shown a marked intrasessional (13%), intersessional (16%), and individual (20%) variability. The 95% confidence interval for all normal subjects ranged from 237 to 553 ml/min. This large scatter has limited the usefulness of raw flow data in the determination of normality or disease and has been noted by other authors. 4,6-~s (Table I). The sources of this variation are threefold: the equipment, the observer, and the patient. A re- cent review of the accuracy of ultrasonic flow measurements has shown that with appropriate equipment the systematic error can be expected to be 6%, with a random error component of 12% to 15%. 26 The latter component is reduced, of course, when several readings are taken and averaged. The relatively low frequency of the B-mode scanner (3 MHz) is not ideal for detailed carotid bifurcation scanning but is sufficient for imaging the CCA. The observer variation can be explained partly by slight variations in determining the diameter and orientation of the vessel and placement of the Doppler sample volume. The effects of inter- and intraobserver variation in similar carotid studies reported by other groups are not inconsiderable and have been addressed by Knox et al. ~7 and Lally, Johnston, and Cobbold.lS The intraobserver sources of error can be reduced by practice and training and the interobserver errors by the adoption of common criteria for the diagnosis of disease. However, the patient is the greatest source of variation. It is known that cardiac output can vary over a considerable range with posture and exercise. 19 Although cerebral autoregulation is expected to keep cerebral blood flow constant, there may be increases in flow through the ECA with exercise or indeed any other stimulus to increased cardiac output such as anxiety. To assess the effect of this variation adequately the cardiac output would have to be measured concurrently and its relationship to CCA flow defined. Moreover Uematsu et al.6 have shown that CCA and ICA blood flow varies directly with the amount of inspired carbon dioxide, a factor which may be important in the anxious patient. The marked sex difference is a feature described by other groups. Muller et al.4 noted that blood flows of women were 5% less than those of men. The only significant correlation in the present study with body size parameters was with height, but it is notable that the men were highly significantly heavier and taller and, therefore, larger in terms of body surface area

7 852 Ackroyd et al. Journal of VASCULAR SURGERY 100 IL ROC CURVE : UNILATERAL LESIONS SENSITIVITY 60 ]- zr, e f FLOW RATIOS T o 20 DIFF = ~DtFF # SPECIFICITY Fig. 4. Receiver operator curve (ROC) illustrates greater discriminatory power of normal to abnormal comparison parameters for detecting unilateral >50% ICA stenosis. Test criteria for "FLOW" ROC (400, 350, 300 ml/min) are shown. Corresponding values on normal to abnormal ROC (RATIO) are 1.1, 1.2, and 1.3, for normal minus abnormal difference ROC (DIFF) 0, 50, and 100 ml/min, and for normal to abnormal percentage difference ROC (%DIFF) 10%, 20%, and 30%, respectively. than the women. When the flows were normalized according to sex, then the effect of body size was insignificant (Table III). It is likely, then, that the flow differences between men and women are due, at least in part, to differences in body size. Fujishiro and Yoshimura 7 have noted a reduction in flow, an increase in vessel diameter, and a reduced mean flow velocity with age. Both Fujishiro and Yoshimura 7 and Uematsu et al.6 use ultrasound flowmeters that have tracking A-mode to follow the vessel wall motion and both have noted an average 6% variation in diameter with the cardiac cycle. Indeed, the pulsatility was 11% in the second decade of life and 4% in the seventh decade illustrating the loss of elasticity with age. This also has a bearing on the ultimate accuracy of diameter measurements, since the screen image could be frozen in systole or diastole in imaging systems such as the Octoson. In a comprehensive study at the time of operation Archie and Feldtman 2 showed that a diameter stenosis of 75% in the ICA produces a 40% reduction in flow. However, as they point out, this underestimates the amount of hemispheric blood flow because of the collateral supply from the ECA and the contralateral side via the circle of Willis. Indeed the finding of a modest reduction in CCA flow with increasing ICA stenosis attests to the capacity of the ECA to accommodate to the extra flow imposed on it. Nevertheless Uematsu et al., 6 using a quantitative Doppler flowmeter with an angiographic criterion of >40% of ICA stenosis and including total occlusions, found a sensitivity, of 96% for normal to abnormal ratios >1.20 and a sensitivity of 66% for flows under 360 ml/min. The corresponding specificities were 71% and 85%. Our figures have not shown this degree of accuracy because of our apparently greater degree of variability. We will be interested in the accuracy data of other groups as they become available. The measurement of CCA flow is an indirect test of the hemodynamic significance of the degree of stenosis. It does not offer any morphologic information on the carotid bifurcation. Indeed, the current trend is for a greater emphasis on the identification of the structural characteristics of the plaque and to rely on Doppler information to confirm the severity of a stenosis and to aid vessel identification. Moreover, flow measurements as a noninvasive test would be in competition with high-resolution duplex scanners and Doppler-imaging equipment, which directly interrogate the site of disease. Unless flow measurement can achieve accuracy, specificity, and sensitivity in the order of 90% or more, then it will be an inferior method of noninvasive testing and could only be considered in a complementary role to be used in conjunction with other more accurate methods. There are other areas where CCA flow measurements may be of use, such as in monitoring changes in cardiac output. Since flow measurement is entirely noninvasive, it would not be attended with the same problems as the dye-dilution or catheter methods. Portable ultrasound units that can be used in the operating room are becoming available and it should be possible to monitor changes or trends in flow in a major artery as a reflection of total cardiac output. Also, Payen et al. 1 have shown a close correlation between cerebral blood flow as measured by xenon

8 Volume 3 Number 6 June 1986 Quantitative carotid blood flow and CCA flow in a study of 11 anesthetized patients undergoing intracranial surgery. These may be fruitful areas for further development. This study has shown that the measurement of CCA flow is not as accurate as current noninvasive methods in the prediction of ICA stenosis, although it can distinguish total occlusion; this fact is due to the wide range of normal flow, which does not allow clear separation from abnormal flow. Because of the natural variability of flow, any studies of this nature will be designed to use the patient as their own control and to concentrate on grouped flow data rather than individual measurements. We believe the principal use of this technique will be for physiologic or pharmacologic studies of blood flow rather than as a tool to make clinical decisions about individual patients. However, one may speculate on the increased accuracy that would be expected to accrue from the measurement of flows in the ICA itself with the higher resolution duplex systems, which are fitted with the appropriate software to make the flow calculations. REFERENCES 1. Gill RW, Manoharan A, Picker RH, Ellard KT, Lunzer MR, Kossoff G, Griffiths KA. Portal and splenic blood flow measurements. In: Kurjak A, KossoffG, eds. Recent advances in ultrasound diagnosis, 4. Amsterdam: Elsevier Science Publishers BV, 1984: Gill R, Warren PS, Garrett WJ, KossoffG. Umbilical venous blood flow, IUGR, and fetal risk. In: Kurjak A, Kossoff G, eds. Recent advances in ultrasound diagnosis, 4. Amsterdam: Elsevier Science Publishers BV, 1984: Gill RW. Pulsed Doppler with B-mode imaging for quantitative blood flow measurement. Ultrasound Med Biol 1979; 5: Muller HR, Radue EW, Pallotti C, Gratzl O. Common carotid CW Doppler flow measurement in neurovascular surgery. In: Kuoak A, Kossoff G, eds. Recent advances in ultrasound diagnosis, 4. Amsterdam: Elsevier Science Publishers BV, 1984: Strandness Jr DE, Sumner DS. Hemodynamics for surgeons. New York: Grune & Stratton Inc, 1975: Uematsu S, Yang A, Preziosi TJ, Kouba R, Toung TJ. Measurement of carotid blood flow in man and its clinical application. Stroke 1983: Fujishiro K, Yoshimura S. Haemodynamic changes in carotid blood flow with age. Jikeikai Med J 1982; 29: Furuhata H. Noninvasive and quantitative measurement of volume flow-rate at internal and external carotid and vertebral. In: Lerski RA, Morley P, eds. Ultrasound '82. Oxford: Pergamon Press, Fitzgerald DE, O'Shaughnessy AM, Keaveny VT. Pulsed Doppler: Determination of blood velocity and volume flow in normal and diseased common carotid arteries in man. Cardiovasc Res 1982; 16: Payen DM, Levy BI, Menegalli DJ, Lajat YI, Levenson JA, Nicolas FM. Evaluation of human hemispheric blood flow based on noninvasive carotid blood flow measurements using the range-gated Doppler technique. Stroke 1982; 13: Simon A, Levenson J, Safar M, Diebold B, Peronneau P. Noninvasive pulsed Doppler measurement of blood flow: Investigation of internal carotid stenosis (abstr). Ultrasound Med Biol 1982; 8(Suppl 1): Keller HM, Meier WE, Anliker M, Kumpe DA. Noninvasive measurement of velocity profiles and blood flow in the common carotid artery by pulsed Doppler ultrasound. Stroke 1976; 7: Borodzinski K, Eilipczynski I, Nowicki A, Powalowski T. Quantitative transcutaneous measurements of blood flow in carotid artery by means of pulse and continuous wave Doppler methods. Ultrasound Med Biol 1976; 2: Olson RM. Human carotid artery wall thickness, diameter, and blood flow by a noninvasive technique. J Appl Physiol 1974; 37: Fish PJ. A method of transcutaneous blood flow measurement--accuracy considerations. In: Kurjak A, Kratochwil A, eds. Recent advances in ultrasound diagnosis, 3. Amsterdam: Elsevier Science Publishers BV, 1981: Gill RW. Measurement of blood flow by ultrasound: Accuracy and sources of error. Ultrasound Meal Biol (In press.) 17. Knox RA, Phillips DJ, Breslau PJ, Lawrence R, Primozich J, Strandness Jr DE. Empirical findings relating sample volume size to diagnostic accuracy in pulsed Doppler cerebrovascular studies. J Clin Ultrasound 1982; 10: Lally M, Johnston KW, Cobbold RS. Limitations in the accuracy of peak frequency measurements in the diagnosis of carotid disease. J Clin Ultrasound 1984; 12: Epstein SE, Beiser GD, Stamper M, Robinson BF, Braunwald E. Characterisation of the circulatory response to maximal upright exercise in normal subjects and patients with heart disease. Circulation 1967; 35: Archie JP, Feldtman RW. Critical stenosis of the internal carotid artery. Surgery 1981; 89:67-72.