Secondary Logo

Journal Logo

Comparing Doppler Ultrasonography and Cerebral Oximetry as Indicators for Shunting in Carotid Endarterectomy

Grubhofer, Georg, MD*; Plöchl, Walter, MD*; Skolka, Michael, MD*; Czerny, Martin, MD; Ehrlich, Marek, MD; Lassnigg, Andrea, MD*

doi: 10.1097/00000539-200012000-00006

To determine the thresholds of selective shunting in carotid endarterectomy during general anesthesia, we compared transcranial Doppler ultrasonography and cerebral oximetry (RSO2). During carotid cross-clamping, RSO2 and mean blood flow velocity in the middle cerebral artery (Vm, mca) was simultaneously monitored in 55 of 59 patients. A relative decrease in Vm, mca to <20% of preclamp velocity was the indication for selective shunting. Three patients were shunted, two because of criteria of Vm, mca and one in which Vm, mca measurements were impossible. No postoperative neurological deficits occurred. During cross-clamping, both Vm, mca (42 ± 16 vs 26 ± 12 cm/s;P < 0.001) and RSO2 (68 ± 7% vs 62 ± 8%;P < 0.01) decreased and a significant correlation between %Vm, mca and ΔRSO2 was found (R2 =0.40;P = 0.003). Decreases in RSO2 >13% identified two patients later shunted; however, this threshold would have indicated unnecessary shunting in seven patients (false positives = 17%). Transcranial Doppler ultrasonography identified patients at risk for ischemia more accurately than RSO2. Relying on RSO2 alone would increase the number of unnecessary shunts because of the low specificity. Accepting higher decreases in RSO2 does not appear reasonable because it bears the risk of a low sensitivity.

Implications Although cerebral oximetry was easy to apply but considerably unspecific (13% false positives), transcranial Doppler ultrasonography was more accurate in indicating the risk of cerebral hypoperfusion during carotid cross-clamping. Additionally, the improvement in cerebral blood flow velocity after inducing arterial hypertension might prevent cerebral hypoperfusion during cross-clamping.

Departments of *Cardiothoracic and Vascular Anesthesia & Intensive Care, and †Cardiovascular Surgery, University of Vienna, Vienna, Austria

August 8, 2000.

Address correspondence and reprint requests to Georg Grubhofer, MD, University Clinic of Anesthesia, Waehringer Guertel 18–20, A-1090 Vienna, Austria. Address e-mail to Georg.Grubhofer

The inherent risk of carotid surgery is perioperative stroke occurring at rates from 5% to 7.5% (1,2). A cause of perioperative stroke is hypoperfusion (3–5) or embolization during cross-clamping of the internal carotid artery. Although shunting prevents cerebral hypoperfusion, the routine use of an intraluminal shunt in all patients increases the rate of perioperative stroke resulting from embolic events (5,6). Thus, many authors advocate selective shunting guided by signs and measurements of cerebral hypoperfusion during cross-clamping.

Monitoring signs of neurologic complications in the awake patient during cross-clamping provides a maximum of specificity and avoids “false positive” results. In contrast, the best cerebral monitoring to detect critical limits of cerebral perfusion during general anesthesia is an issue of controversy. Multiple methods, including transcranial Doppler ultrasonography (TCD), near-infrared spectroscopy (NIRS), electroencephalography, and somatosensory evoked potentials have been recommended. TCD is capable of detecting clamp-related hypoperfusion, but it is technically demanding and not feasible in ∼20% of patients. In contrast, cerebral oximetry measured by NIRS has the advantage of being continuous, noninvasive, portable, and easy to use.

In this study, we compared TCD and NIRS monitoring in their ability to identify cerebral hypoperfusion during carotid cross-clamping. To determine the accuracy of NIRS, relative changes in NIRS values were compared with changes in mean blood flow velocity in the middle cerebral artery (Vm, mca). Furthermore, the effect of vasopressor-induced increases in arterial blood pressure on cerebral perfusion variables Vm, mca, pulsatility index, and NIRS values was examined.

Back to Top | Article Outline


After approval of the IRB and obtaining written informed consent, 59 consecutive patients (45 male; 14 female; age 67 ± 9 yr; weight 78 ± 18 kg) scheduled for elective carotid endarterectomy participated in the study. Patients with myocardial infarction within the last 6 mo were excluded from the study protocol. The indications for operation included severe (60%–80%;n = 15) and critical (>80%;n = 44) atherosclerotic internal carotid artery stenosis, either asymptomatic (n = 13) or accompanied with neurologic symptoms (n = 46). Contralateral stenosis >60% was found in 18 patients. Anamnestic neurological symptoms presented by patients were amaurosis fugax (n = 20), transient ischemic attacks (n = 27), progressive reversible neurologic deficits (n = 4), and previous stroke (n = 17). Associated pathologic conditions included hypertension (systemic arterial pressure >160 mm Hg; 21.3 kPa) in 33 patients and diabetes mellitus in 17 patients.

All patients were premedicated with midazolam 7.5 mg PO. Monitoring included invasive measurement of mean arterial pressure (MAP), pulsoxymetric hemoglobin saturation, five-lead electrocardiography, ETco2, and arterial blood gas analysis. General anesthesia was induced with IV fentanyl (3 μg · kg-1) and the slow administration of IV propofol (1.5–3 mg ·kg-1). Muscle relaxation was achieved with IV vecuronium (0.1 mg · kg-1). Ventilation was adjusted to normocapnia (Paco2 between 35 and 45 mm Hg; 4.7 to 6.0 kPa) with an air-oxygen mixture (inspired fraction of oxygen ∼0.30). Anesthesia was maintained with inspired isoflurane 0.2–0.4 Vol % and bolus doses of fentanyl and vecuronium as needed. Heparin (5000 U IV) was administered before carotid cross-clamping. MAP was maintained at the patient’s preoperative level by adjusting the inspired isoflurane concentration and by the administration of phenylephrine (20 μg bolus IV) if MAP declined by >20%.

Pre- and intraoperative TCD ultrasonography was performed by a pulsed emission, 2 MHz bidirectional Doppler apparatus (Multidop X, DWL; Elektronische Systeme, Sipplingen, Germany). Preoperative TCD signals of the middle cerebral artery (MCA) were obtained by a hand-held probe via the temporal window. Verification of the MCA signals included the direction of flow, the depth of emission, and the absolute flow velocity. The best MCA signal quality was obtained in the range between 46 and 54 mm (sample size 20 mm). After intubation, the Doppler probe was secured by using a metal ring adjustment (DWL; Elektronische Systeme, Sipplingen, Germany) and MCA velocity was monitored continuously during the entire procedure.

All TCD measurements included systolic (Vs, mca), diastolic (Vd, mca), and mean (Vm, mca) blood flow velocity and the calculation of pulsatility index (PI = {Vs, mca − Vd, mca }/Vm, mca). Relative changes in %Vm, mca were additionally calculated as the percentage of the velocity at stable conditions before carotid cross-clamping.

The NIRS monitor (INVOS 3100; Cerebral Oximeter, Somanetics, Troy, MI) uses two wavelengths (750 and 850 nm) and displays a percent value of regional cerebral oxygenation (RSO2), calculated from the proportion between oxygenated hemoglobin and total hemoglobin in brain tissue. The sensor (Somasensor No. 3100 S, Somanetics) consists of two light-collecting optodes, allowing removal of extracranial contributions of scattered light by the application of a subtraction algorithm. The sensor was placed on the forehead of the operative side with the light transmitters ∼3 cm away from midline to avoid distortion from the sagittal sinus and was secured with an opaque dressing. RSO2 was measured continuously at 15-s intervals and stored for later analysis on a disk drive.

To investigate the effect of MAP on cerebral perfusion and oxygenation before cross-clamping, we administered phenylephrine (10 to 100 μg/min) to increase MAP by approximately 20% immediately before and during carotid cross-clamping. Although the protocol provided the possibility to stop the vasopressor infusion on the occurrence of myocardial ischemia (ST segment depression, conduction disturbances), none of these criteria were present in any patient.

Within the first 3 min of cross-clamping, severe cerebral hypoperfusion was defined if the Vm, mca decreased to <20% of preclamp velocity (baseline). In that case, a shunt was inserted (ArgyleTM; Carotid Artery Shunt, Sherwood Medical, Tullamore, Ireland). The numeration of further timepoints where data were obtained is given below. All patients were extubated in the operation room and were examined for the development of new neurological deficits on awakening and periodically during hospitalization.

Data analysis and interpretation were based on MAP, MCA blood flow velocities, arterial blood gas values, and RSO2 values at the following five time points: 1) awake patient; 2) stable values during surgery approximately 10 min before carotid cross-clamping (baseline); 3) immediately before carotid cross-clamping, MAP was increased by vasopressor infusion to a target value of 20% greater than baseline; 4) 3 min after carotid cross-clamping; and 5) 5 min after cross-clamp release.

Differences in values between the five timepoints were calculated by analysis of variance followed by Bonferroni’s multiple comparison test. For comparing relative changes in Vm, mca (%Vm, mca) and RSO2 (ΔRSO2 from baseline), a linear regression analysis was performed. The relationship between MCA blood flow velocity, RSO2 data, and the preoperative characteristics of the patient (degree of carotid stenosis, presence of contralateral stenosis >80%, preoperative clinical symptoms) was calculated by Spearman’s coefficient of correlation. All calculations were performed by GraphPad Prism (GraphPad Software, Inc., San Diego, CA) on a personal computer. Data are given as mean ± sd; a P value < 0.05 was considered significant.

Back to Top | Article Outline


Intraoperatively, simultaneous recording of Vm, mca and RSO2 was possible in 55 patients; in 4 patients, we could not identify a sufficient MCA TCD signal. Reconstruction of the carotid artery was successfully completed in all cases. Duration of carotid clamping was 38 ± 12 min in patients without a shunt (n = 56) and 41 ± 7 min in patients with shunts (n = 3). No deaths or postoperative neurological deficits occurred during the hospital stay. Two patients were shunted according to the criteria defined by decreases in Vm, mca. In these two patients, Vm, mca as well as RSO2 increased to preclamp values with a shunt in situ (Table 1). In the third shunted patient, the detection of a satisfying TCD signal was not possible. Because the surgeon expected an increased risk of clamping ischemia as a result of the preoperative condition, shunting was performed (Patient 9 in Table 1).

Table 1

Table 1

Overall, cross-clamping resulted in a significant decrease in both Vm, mca and RSO2 values (Table 2). On average, the baseline value of RSO2 in patients with or without later TCD criteria for shunting (%Vm, mca <20% of baseline) was similar. Within the first minutes of cross-clamping, two patients later shunted presented a more pronounced decrease in RSO2 (P = 0.051) than patients without shunting (Table 3). ΔRSO2 values in four patients with unfeasible recordings of Vm, mca were 13, 6, 9, and 5%.

Table 2

Table 2

Table 3

Table 3

A significant correlation (R2 = 0.40;P = 0.003) between %Vm, mca and ΔRSO2 was calculated during cross-clamping (Figure 1). The plot shows a considerable scattering of values. The ΔRSO2 limit calculated by linear regression (Figure 1) to identify severe hypoperfusion as defined by TCD was a decrease of more than 13% (95% confidence interval 11% to 15%). Two patients with later shunting as a result of TCD criteria were correctly identified by RSO2 measurements (ΔRSO2: −14% and −16%). However, another seven patients presented RSO2 decreases >13% without %Vm, mca changes <20% of baseline (Figure 2). Thus, if criteria for shunting were based on RSO2 alone, unnecessary shunting would have been performed in these 7 of 53 patients (13% false positives). This increased rate of false positives was independent of the chosen TCD criteria. We calculated the specificity of RSO2 also for %Vm, mca changes to lower than 30% and 40% of baseline. Table 4 shows that the specificity and positive predictive values are equally decreased at these thresholds.

Figure 1

Figure 1

Figure 2

Figure 2

Table 4

Table 4

In contrast to cross-clamping, phenylephrine infusion before cross-clamping did not cause a uniform effect on Vm, mca and RSO2. Although increases in MAP did increase Vm, mca and Vs,mca (P < 0.001; paired t-test), no such effect was seen in RSO2 values. On the contrary, RSO2 values during phenylephrine infusion decreased significantly (ΔRSO2 mean, −2.1 ± 3.7%; range, 6 to −14%;P = 0.002 different vs zero; one-sample t-test). The preoperative neurologic condition of the patient was not correlated with measures of TCD or RSO2 values at any measurement point. The degree of carotid stenosis also did not correlate with the decrease in Vm, mca or RSO2 during cross-clamping. Five minutes after cross-clamp release, Vm, mca but not RSO2 exceeded preclamp values. Paco2 influences the accuracy of TCD measurements, but remained unchanged throughout the entire procedure. Pulsatility index did change at the beginning and after the release of carotid clamping (Table 2).

Back to Top | Article Outline


In this study, TCD identified patients at risk for ischemia during carotid cross-clamping more accurately than RSO2. Although decreases in RSO2 during cross-clamping correlated with decreases in Vm, mca, the number of “false positive” results was considerably large. All patients with later shunting because of Vm, mca values <20% of baseline were correctly identified by RSO2 measurements. However, if criteria for shunting were determined by RSO2 values alone, unnecessary shunting would have been performed in 13% of patients.

During general anesthesia, relative changes in Vm, mca are a useful measure to determine critical limits of cerebral perfusion during carotid cross-clamping (7,8). Changes in cerebral blood flow are correlated with Vm, mca (9,10) and a completely disappearing signal indicates regional cerebral blood flow <12 mL/100g/min (11). Unfortunately, the recommended value of Vm, mca indicating cerebral hypoperfusion and requiring insertion of a shunt varies considerably from an absent signal to a residual flow of <35% from baseline. Although the “safe limit” for reductions of Vm, mca has been suggested in early investigations to be 30% to 40% of the baseline value (12,13), recent studies propose a much lower threshold. Evidence for using lower limits is given in awake patients, who experience clamping ischemia not until Vm, mca decreases to lower than 22% (14) and 15% (8). Furthermore, Halsey (15) reported in a large retrospective study, that during general anesthesia patients have a clear benefit from shunt insertion if Vm, mca decreased <15%, whereas shunting at higher flow velocities increased the risk of inducing an embolic stroke. Therefore we set the limit at a residual Vm, mca <20% of baseline. By applying this threshold, our rate of shunting was low and no perioperative strokes were observed. We cannot prove, if the performed shunting in two patients based on Vm, mca criteria was really justified in respect to cerebral ischemia, because clear evidence has only been given for the awake patient. However, the zero flow phenomena in one patient (Patient 31), which was accompanied by a continuing decrease in RSO2 to values of 40%, was suspicious for severe cerebral hypoperfusion. In these two shunted patients, shunt insertion led to a prompt restoration of both Vm, mca and RSO2 values.

Although TCD is an excellent noninvasive measure of cerebral hemodynamics, its routine clinical use is restricted by the inability to find a signal in ∼20% of patients and by the need of a skillful investigator. Cerebral oximetry, in contrast, is easy to apply, offers a real-time and noninvasive measurement of cerebral oxygenation, and generates a number simple to interpret. Because of intersubject variability, RSO2 trends, rather than absolute numbers, provide an assessment of cerebral oxygenation (16). Presently, most authors consider a change of 10% in RSO2 as a reasonable, although arbitrary, threshold for shunting (17,18).

In our study, a relative decrease of >13% in RSO2 was found to be sensitive enough to identify patients with Vm, mca <20% of baseline. However, the major disadvantage of applying this threshold in clinical practice is the resulting increased rate of unnecessary shunting. The chosen Vm, mca criteria of 20% is probably the absolute minimum value for selective shunting, but Vm, mca thresholds of 30% and 40% would not improve the specificity of cerebral oximetry (see Table 4). Unnecessary shunting bears the risk of embolization (5) and especially in patients with sufficient cerebral collateralization, shunting per se results in an increase in strokes (6). Therefore, proper neuromonitoring needs to identify patients who will benefit from shunt placement not only with a high sensitivity but also with a high specificity.

The low specificity of RSO2 in the detection of severe cerebral hypoperfusion is most likely caused by the influence of extracerebral tissue oxygenation. Despite improvements in sensor technology, the compensation built in to eliminate the effect of extracerebral tissues is still inadequate (19). Evidence for this source of error is also given in our results. In patients without any change in Vm, mca during cross-clamping, RSO2 still decreased by 5% (see Figure 1). This may be because of the impaired perfusion of extracerebral areas supplied by the clamped external carotid artery. Furthermore, an increase in Vm, mca induced by hypertension before cross-clamping was not reflected in RSO2. Because phenylephrine acts exclusively at extracranial vessels (20), the converse course of Vm, mca and RSO2 values probably did reflect opposite effects of phenylephrine on intra- and extracerebral perfusion.

A further reason why we observed a discrepancy between changes in Vm, mca and RSO2 values relates to the different brain areas and different aspects of brain function determined by the two monitoring techniques. RSO2 determines the regional oxygenation state of the frontal lobe, whereas TCD determines the cerebral perfusion of the parietal lobe and deep areas of the brain. Studies in animals have shown, that a critical low flow and ischemia in deeper areas of the brain do not usually correlate with ischemia in more superficial regions (21). Therefore, RSO2 and other regional measures may not always accurately indicate the complete cerebral condition.

A limitation of this study is that the accuracy of TCD and NIRS in detecting cerebral hypoperfusion was not determined by additional electroencephalographic recordings. Thus, moderate cerebral hypoperfusion in regional areas might have escaped detection by means of TCD. In view of the absence of neurological deficits in our study, we believe that severe and clinically relevant cerebral hypoperfusion was not present in our patients during cross-clamping. Generally, the evaluation of any cerebral monitor for carotid surgery is difficult because the prevalence of postoperative neurological deficits is small. Therefore, the accuracy of TCD in definitively detecting cross-clamp-induced strokes or ischemia has to be proven in a larger study with sufficient statistical power. Despite the number of patients studied here being relatively small, the low specificity of RSO2 was already evident in our results.

Regarding the effect of blood pressure on Vm, mca, induced hypertension did augment cerebral perfusion before carotid clamping. The rationale for inducing hypertension in patients with carotid stenosis is the assumption that cerebral perfusion, and thus delivery of oxygen, is pressure-dependent with increasing degree of carotid stenosis (22). Clinical evidence that hypertension improves cerebral perfusion during cross-clamping in awake patients is given by Imparato et al. (23) and Walleck et al. (24), who reported the disappearance of neurologic symptoms without shunt placement when arterial pressure was increased by ephedrine. The effect of vasopressors during cross-clamping was not the primary aim of our investigation. Nevertheless, induced hypertension might result in less severe decreases in Vm, mca and therefore reduce the rate of shunting. This hypothesis is supported by our small rate of shunting and the measured Vm, mca values, which were considerably higher than those reported without induced hypertension (25). This finding would favor the technique of induced hypertension, but has yet to be confirmed in a separate study addressing this question.

In conclusion, TCD identified a small group of patients at risk for ischemia during carotid cross-clamping. The reviewed literature, as well as our results, suggests that a decrease to 20% of baseline Vm, mca is a reasonable threshold indicative for selective shunting. In contrast, RSO2 does not allow a satisfactory identification of patients requiring shunt insertion because of the low specificity, which would result in unnecessary shunting. Furthermore, induced hypertension by vasopressors markedly affected RSO2, which makes the interpretation of values difficult. Accepting lower limits of RSO2 for selective shunting does not appear to be reasonable, because it bears the risk of a low sensitivity. By improving NIRS technology, particularly if a refinement and better exclusion of extracranial oxygenation is possible, NIRS might become a more accurate monitor of cerebral hypoperfusion. In that case, it would offer advantages, especially if a TCD signal cannot be identified or other techniques of neuromonitoring are not available.

Back to Top | Article Outline


1. Ferguson GG, Eliasziw M, Barr HW, et al. The North American symptomatic carotid endarterectomy trial: surgical results in 1415 patients. Stroke 1999; 30: 1751–8.
2. Randomised trial of endarterectomy for recently symptomatic carotid stenosis: final results of the MRC European Carotid Surgery Trial (ECST). Lancet 1998; 351: 1379–87.
3. McKinsey JF, Desai TR, Bassiouny HS, et al. Mechanisms of neurologic deficits and mortality with carotid endarterectomy. Arch Surg 1996; 131: 526–31.
4. Finocchi C, Gandolfo C, Carissimi T, et al. Role of transcranial Doppler and stump pressure during carotid endarterectomy. Stroke 1997; 28: 2448–52.
5. Spencer MP. Transcranial Doppler monitoring and causes of stroke from carotid endarterectomy. Stroke 1997; 28: 685–91.
6. McDowell HA Jr, Gross GM, Halsey JH. Carotid endarterectomy monitored with transcranial Doppler. Ann Surg 1992; 215: 514–8.
7. Giannoni MF, Sbarigia E, Panico MA, et al. Intraoperative transcranial Doppler sonography monitoring during carotid surgery under locoregional anaesthesia. Eur J Vasc Endovasc Surg 1996; 12: 407–11.
8. Fiori L, Parenti G, Marconi F. Combined transcranial Doppler and electrophysiologic monitoring for carotid endarterectomy. J Neurosurg Anesthesiol 1997; 9: 11–6.
9. Larsen FS, Olsen KS, Hansen BA, et al. Transcranial Doppler is valid for determination of the lower limit of cerebral blood flow autoregulation. Stroke 1994; 25: 1985–8.
10. Kochs E, Hoffman WE, Werner C, et al. Cerebral blood flow velocity in relation to cerebral blood flow, cerebral metabolic rate for oxygen, and electroencephalogram analysis during isoflurane anesthesia in dogs. Anesth Analg 1993; 76: 1222–6.
11. Halsey JH, McDowell HA, Gelmon S, Morawetz RB. Blood velocity in the middle cerebral artery and regional cerebral blood flow during carotid endarterectomy. Stroke 1989; 20: 53–8.
12. Jorgensen LG, Schroeder TV. Transcranial Doppler for detection of cerebral ischaemia during carotid endarterectomy. Eur J Vasc Surg 1992; 6: 142–7.
13. Giller CA, Mathews D, Walker B, et al. Prediction of tolerance to carotid artery occlusion using transcranial Doppler ultrasound. J Neurosurg 1994; 81: 15–9.
14. Cao P, Giordano G, Zannetti S, et al. Transcranial Doppler monitoring during carotid endarterectomy: is it appropriate for selecting patients in need of a shunt? J Vasc Surg 1997; 26: 973–80.
15. Halsey JHJ. Risks and benefits of shunting in carotid endarterectomy: The International Transcranial Doppler Collaborators. Stroke 1992; 23: 1583–7.
16. Pollard V, Prough DS, DeMelo AE, et al. Validation in volunteers of a near-infrared spectroscope for monitoring brain oxygenation in vivo. Anesth Analg 1996; 82: 269–77.
17. Cho H, Nemoto EM, Yonas H, et al. Cerebral monitoring by means of oximetry and somatosensory evoked potentials during carotid endarterectomy. J Neurosurg 1998; 89: 533–8.
18. Carlin RE, McGraw DJ, Calimlim JR, Mascia MF. The use of near-infrared cerebral oximetry in awake carotid endarterectomy. J Clin Anest 1998; 10: 109–13.
19. Germon TJ, Evans PD, Barnett NJ, et al. Cerebral near infrared spectroscopy: emitter-detector separation must be increased. Br J Anaesth 1999; 82: 831–7.
20. Strebel SP, Kindler C, Bissonnette B, et al. The impact of systemic vasoconstrictors on the cerebral circulation of anesthetized patients. Anesthesiology 1998; 89: 67–72.
21. Ackerstaff RG, van de Vlasakker CJ. Monitoring of brain function during carotid endarterectomy: an analysis of contemporary methods. J Cardiothorac Vasc Anesth 1998; 12: 341–7.
22. White RP, Markus HS. Impaired dynamic cerebral autoregulation in carotid artery stenosis. Stroke 1997; 28: 1340–4.
23. Imparato AM, Ramirez A, Riles T, Mintzer R. Cerebral protection in carotid surgery. Arch Surg 1982; 117: 1073–8.
24. Walleck P, Becquemin JP, Desgranges P, Bonnet F. Are neurologic events occurring during carotid artery surgery predictive of postoperative neurologic complications? Acta Anaesthesiol Scand 1996; 40: 167–70.
25. Arnold M, Sturzenegger M, Schaeffler L, Seiler RS. Continuous intraoperative monitoring of middle cerebral artery blood flow velocities and electroencephalography during carotid endarterectomy. Stroke 1997; 28: 1345–50.
© 2000 International Anesthesia Research Society