Transcranial Doppler Monitoring and the Causes of Stroke from
Carotid Endarterectomy

Merrill P. Spencer, M.D.
Sound Vascular Monitoring and
The Institute of Applied Physiology and Medicine
Seattle, Washington, USA

Acknowledgment: The author thanks Peter Mansfield M.D., Director of The Heart Center - Providence Seattle Medical Center, for assistance in statistical analysis.

Abstract

Background and Purpose: The value of carotid endarterectomy (CEA) depends on the safety of the operation. Transcranial Doppler (TCD) was used to evaluate the possibilities of hypoperfusion, hyperperfusion, and embolization as causes of stroke and evaluated the significance of Doppler microembolic signals (DMES).

Methods: Five hundred CEAs were monitored with TCD on the ipsilateral MCA during various phases of CEA observing for hemodynamic changes and incidence of DMES. Complications were graded according to their severity and their probable cause determined from TCD criteria and review of hospital charts.

Results: There were 24(4.8%) CVCs, including 7 with TIAs and 15(3%) with permanent deficits. Among all CVCs embolism was judged to be responsible in 13(54%) p< 0.02 compared to hypoperfusion, hyperperfusion in 7(29%) p<0.14 when compared to hypoperfusion, and hypoperfusion in 4(17%) p<0.08 compared to embolism plus hyperperfusion. The surgeons responded to TCD information by several strategies depending on the TCD information. The incidence of hospital permanent deficits diminished from 7.0% in the first 100 operations to 2% for the last 400 p<=0.01. Shunting was more associated with CVCs than non-shunting, however not significantly so, p=0.24. Intraoperative DMES numbers were strongly associated with CVCs p=0.02.

Conclusions: Embolism is the principal cause of CVCs from CEA with hyperperfusion and hypoperfusion also important causes. TCD provides information to allow prompt identification and treatment of these three major causes of stroke from this operation. The perioperative stroke rate can be reduced by appropriate measures, taken by the surgeons, based on findings of TCD monitoring.

Introduction

Surgical excision of atherosclerotic plaque at the bifurcation of the common carotid artery (CCA) is performed to prevent cerebrovascular complications (CVCs), however, carotid endarterectomy (CEA) itself can produce CVCs including disabling stroke and death. Many intraoperative monitoring techniques have been used to warn the surgeon of the possibility of an adverse outcome ^1,2,3,4,5 but these, including self monitoring under regional block, address only the possibility of brain hypoperfusion during the time of crossclamping of the carotid arteries. Other major etiologies recognized included hyperperfusion after release of the clamps ⁶ and embolism ⁷. Recently reports have appeared to expand on the basic causes of CVCs ^8,9,10. In a study of the cause of perioperative stroke after 3062 CEAs 20 different mechanisms were identified but mainly categorized into ischemia from carotid artery clamping, postoperative thrombosis and embolism, and intracerebral hemorrhage ¹⁰.

The hemodynamic and embolic information provided by TCD monitoring offers the opportunity to identify, in real time, all these major causes of stroke and give early warning so that preventive measures can be undertaken. The first reports on TCD in carotid endarterectomy appeared in the late 1980s^{11,12,13,14,15,16,17,18} and many authors further published for and against its usefulness ^{19,20,21,22,23} in measuring hemodynamic and embolic parameters. TCD with simultaneous EEG ^24,25,26,27 also proved useful indications of flow in the middle cerebral artery (MCA) and perfusion of the cortex. Although embolization is generally agreed to be the main cause of cerebral complications during CEA, it was not until 1990 ^28,29 that TCD was reported to detect particulate as well as bubble emboli in this operation. This report relates our experience with TCD monitoring in determining the hemodynamic and embolic causes and prevention of CVCs through alterations in surgical techniques and management.

Methods

Included in this report are the first 500 carotid endarterectomies (CEAs) of the internal carotid artery (ICA) performed between October 1985 and July 1994, in which an ipsilateral temporal bone ultrasonic window could be found, in which the technologist was able to obtain TCD data throughout the surgery, and in which complete hospital chart review was obtainable. All charts were completely reviewed and complications identified by the author. In addition observations were always made for complications in the recovery room were made by the author, the TCD technologist and the nursing staff. No patients were included who underwent CEA of the external carotid artery only and none underwent other concurrent surgical procedures. Indications for CEA were symptomatic and asymptomatic carotid artery stenosis. Several patients were operated with stenosis < 60% for reasons including repeating TIAs, the radiologists diagnosis of ulceration, radiologist overestimation of stenosis compared to micrometer diameter measurements of the angiograms e.g. area percentage vs. diameter percentage, and occlusion of the contralateral artery. Forty four patients underwent a previous CEA of the contralateral ICA. Eight underwent repeat CEA on the same ICA and one patient underwent contralateral CEA as well as a repeat CEA.

The preoperative symptoms of the patients were classified as:

Asymptomatic - patient denies all symptoms of cerebrovascular insufficiency over the past 6 months or complains of non-lateralizing cerebral symptoms such as dizziness, headaches, et cetera.

TAF - transient unilateral monocular blindness.

TIA or RIND - transient lateralizing symptoms of cerebrovascular insufficiency and speech deficits but not including unilateral monocular blindness.

Stroke - recent persistent unilateral symptoms of cerebrovascular insufficiency.

Twenty one operations were performed under sedation and regional anesthesia. The remainder were conducted under general anesthesia using nitrous oxide in oxygen and Isoflurane or Enflurane after induction with thiopental. The operations were performed by 23 vascular surgeons in seven community hospitals of three western counties of Washington State. Shunting of the endarterectomy site was performed in 297 (59%) of operations including nine incomplete attempts. One hundred twenty-seven (25%) operations were performed without angiography with the severity of stenosis based on the highest frequencies found with 5 MHz CW Doppler. The following scale was verified in 222 of the cases where both angiograms and Doppler examinations were carried out. <3Khz,0% stenosis; 3,10%; 4,20%; 5-6,30%; 7-8,40%; 9-10,50%; 11-13,60%; 14-15,70%; 16-18, 80%; >18,90%.

The middle cerebral artery (MCA) flow velocity signals were monitored with a 2 MHz pulse Doppler probe * placed over the temporal bone, on the operated side, and focused on the MCA at a depth of 45 or 50 mm with a focal length of 20 mm. A special headband was used to hold a ball-shaped transducer with a position locking mechanism allowing hands-off monitoring. Positive identification of the MCA was confirmed by responses on the Doppler spectrum to finger oscillations of the cervical CCA or intraoperative surgical manipulations of the carotid arteries.

* Transpect©, manufactured by Medasonics Inc.

TCD monitoring phases were defined as: Preoperative Before anesthesia, Dissection During surgical exposure of the carotid arteries up to the time of crossclamping of the common carotid artery, Crossclamping including shunting, if performed, Release 5 minutes following removal of the carotid artery clamps, Closure the period following release up to final closure of the skin incision, Recovery the following period until recovery from anesthesia, and Follow-up After recovery from the anesthetic including the following hospitalization day after surgery.

At the time of crossclamping of the CCA the decrease in mean MCA velocity was calculated as the percentage of the remaining velocity compared to the pre-crossclamp velocity ²³ after 10 seconds were allowed for autoregulation adjustment. Upon release of the crossclamp, following closing of the arteriotomy, the percentage increase in velocity was calculated as the velocity after 15 minutes compared to the pre-crossclamp velocity. The Doppler signals, voice annotations and spectral signals were recorded on audio-video tape for detailed analysis.

DMES in the MCA were recognized by combined audio and spectral criteria previously published ^29,30 and their numbers and the observation time were recorded for each phase of surgery. DMES occurring within the release phase were considered to primarily represent bubbles of air but particulate matter may have been present. If they occurred before opening of the artery or after the release phase they were considered to represent particulates, or formed element emboli. The surgeons were informed of relevant information during the surgery and their responses recorded. The tape recordings were analyzed postoperatively for velocity changes and DMES. The abstracted TCD measurements were entered into an analytical database. Statistical analysis used chi squared, two tailed T Students test and statistical significance was considered when p<0.05.

During the first 100 operations TCD criteria were developed to identify the risks of intraoperative hypoperfusion, hyperperfusion, and embolism as well as strategies for cerebral protection. Surgeons were continuously informed of TCD information throughout all operations. Causal criteria analyzed to determine the etiology of CVCs were:

HYPOPERFUSION - Intraoperative MCA velocities less than 30% of the pre-carotid crossclamp value for more than 5 minutes. Forty percent of the pre-crossclamp value can be tolerated indefinitely ²³ intraoperatively, as well as postoperatively, if carotid artery occlusion occurs;

HYPERPERFUSION - persistence of MCA velocities >1.5 times the pre-crossclamp values, during shunting or following final release of the carotid crossclamps, without adequate corrective measures;

EMBOLISM - failure to meet the previous two hemodynamic criteria and the persistence of DMES during the intraoperative and recovery phases, except for the release phase, or postoperatively without adequate preventive measures. No special limitations on the microembolic numbers or rates were established but the surgeons became sensitive to their occurrence and measures were frequently taken to minimize their numbers.

CVCs were identified during the post surgical hospitalization beginning in the recovery room and by examination of the following patient records: all hospital chart notes of the anesthesiologist, surgeon, and nurses and progress notes, consultation notes, and discharge summaries. CVCs were graded according to severity of clinically evident injury as follows:

Grade I:   TIA or RIND

Grade II:   Residual Deficits Not Impairing Ordinary Daily Activities

Grade III:   Impairment Necessitating Assistance With Some Ordinary Activities

Grade IV:   Requiring Assistance With All Ordinary Activities

Grade V:   Coma Or Death

CVCs were considered to occur intraoperatively if symptoms were noted upon recovery from the anesthetic. They were considered to occur postoperatively if symptoms developed after recovery from the anesthetic. The probable cause of each CVC was determined 1) from the TCD criteria, 2) reopening the arteriotomy and inspection for thrombus or occlusion, 3) postoperative ultrasound examination to confirm ICA and MCA patency, and 4) when available CCT findings and the opinions in a neurologist's consultation.

Results

Among the 500 CEAs there were no deaths but 24 CVCs occurred for a 4.8% rate with a 95% confidence interval from 2.9% to 6.7%. There were 15 permanent deficits, beyond the 9 TIAs, for a stroke rate of 3% with a 95% confidence interval from 1.5% to 4.5%. Fourteen complications occurred intraoperatively and 10 postoperatively. Table 1 lists all of the CVCs with the principal hemodynamic, embolic, and other data used to determine the probable cause of the complications. Among all CVCs embolism was judged to be responsible in 13(54%) p<0.02 compared to hypoperfusion, hyperperfusion in 7(29%) p<0.14 when compared to hypoperfusion, and hypoperfusion in 4(17%) p<0.08 compared to embolism plus hyperperfusion. More than one causal factor sometimes appeared to be involved, but the primary cause was usually identifiable.

During the crossclamp phase MCA velocities decreased below 30% of their pre-crossclamp values in 81 (16%) of operations. The surgeons responded to hypoperfusion threat by shunting or attempted shunting except in 3 cases when no shunt was used and no CVC resulted. The time without shunting, in these 3 cases, was 17, 21, and 21 minutes with the residual crossclamp MCA velocities 7, 18, and 28% respectively. The arterial pressure was increased in these cases to improve perfusion. Presumably cortical collateral played a part to provide adequate perfusion in all three. Other responses to hypoperfusion included adjustment of an occluded shunt, intermittent release of the clamps when extra time was needed for suturing of the arteriotomy, and reopening and re-inspecting the operative site for intimal flaps or an occluding thrombus. Surgeons whose policy was to routinely shunt began to rely on TCD information for selective shunting. Those surgeons who used a shunt selectively, based on stump pressure measurements, eventually abandoned the pressure technique and relied solely on the TCD velocity information to decide for or against use of a shunt.

Following release of the crossclamps MCA velocities abruptly increased above pre-clamp values and then usually decreased toward the pre-clamp values. In 73 (15%) of operations they persisted at values above twice the pre-clamp values. In 3 (4.1%) of these there were subsequent complications. In two of the three, hyperperfusion was considered the primary cause and in the third one it was considered a secondary cause. Measures to prevent hyperperfusion included intraoperative temporary partial occlusion of the CCA, lowering of the arterial pressure, induction of hypocarbia by increased pulmonary ventilation, and infusion of Nitroprusside or other hypotensive agent. Figure 1 illustrates the sequence of events in a patient with hyperperfusion and treatment by control of the arterial blood pressure. This patient was a 71 year male normotensive who preoperatively had a left hemispheric TIA. Angiograms were interpreted to represent 80-90% narrowing of the LICA but micrometer measurements of the ICA represented 50% diameter stenosis. Doppler and Duplex examinations diagnosed a 30% stenosis. When the surgeons were notified of hyperperfusion danger they lowered the arterial pressure and partially pinched off the CCA to decrease the MCA velocities. No postoperative headache or CVC occurred in this patient.

The prevalence and numbers of DMES during the perioperative phases are shown in Table 2. For this analysis all intraoperative phases including dissection, crossclamping, release, and closure were grouped together. The preoperative prevalence of one or more DMES in the MCA in all 500 cases was 16%. However, analyzing those monitored for more than 5 minutes, to eliminate trivial monitoring times, the preoperative prevalence in 443 cases increased to 19%. In this group the difference between the preoperative number of DMES / minute in CVC cases and non-CVC cases was not significant p=0.95. During the intraoperative phases the DMES prevalence was 93% and the difference in DMES between the CVC and non-CVC cases was significant, p=0.02. The difference during the recovery phase, between CVC and non-CVCs, was marginally significant, p=0.06. These statistics were computed from the 2 tailed T test.

When the surgeons released the ECA and CCA clamps, before release of the ICA, DMES were frequently detected in the MCA presumably passing retrograde through the ophthalmic artery which served as a collateral channel to the ICA. Though these were mainly considered bubbles and were not treated with the same seriousness as particulate DMES of other phases, the surgeons responded to release DMES by altering their techniques to more thoroughly eliminate intralumenal air before final closure of the arteriotomy. No decrease in MCA velocity was observed following any of the DMES. This observation suggests that they, individually or in clusters upon clamp release, did not greatly obstruct MCA flow. Surgeons responded to the presence of particulate DMES during surgical maneuvers with more care during the dissection, early crossclamping of the distal ICA prior to dissection of the CCA and ECA, and administration of extra heparin. Postoperative DMES were responded to by administration of heparin, intravenous Dextran, and by increased alertness to possible postoperative complications.

Table 3 indicates a tendency for embolism to be more associated with lesser grades of CVCs and for hemodynamic factors to be associated with more severe grades of CVCs. For the distribution of CVCs found among the 3 causes of Table 3 p<0.07. Considering only hypoperfusion and hyperperfusion incidence p<0.14. Considering only embolic and hypoperfusion cause the p<0.08 and p=0.17 by Bonferroni adjustment. Comparing only the emboli and hypoperfusion data, p<0.02 and therefore the distribution between these 2 causes is statistically significant.

Table 4 shows that shunting was more associated with CVCs than non-shunting, although the difference was not statistically different, p= 0.24. CVCs from embolization appeared equally distributed among the shunted and non-shunted cases. Hyperperfusion accounted for the major difference between shunted and non-shunted operations with seven CVCs in the shunted operations while none of the non-shunted operations sustained a hyperperfusion outcome. This finding suggests non-shunted cases were adequately perfused before surgery and also during crossclamping with their collateral and autoregulation mechanisms adequate or unchallenged. In nine of the shunted operations placement was only partially successful with the non-shunted duration of crossclamping ranging from 10 to 58 minutes. Among these nine partially shunted case there were 4 CVCs (44%). Embolism was considered the primary cause in three and a secondary factor in one. Table 1 indicates the secondary causes of CVCs when appropriate.

Post-surgical occlusion of the operated ICA was found in six of the CVC patients, three occurring intraoperatively and three postoperatively. All ICA occlusions were confirmed by reopening the arteriotomy with finding of thrombosis of the ICA. Three out of the five patients with ICA occlusion demonstrated microemboli in the recovery room prior to return to the operating room. Embolization from the occluding thrombus was concluded to be the cause of complications in five of the six occlusions because crossclamp velocities and stump pressures were considered adequate to maintain global perfusion even without anesthesia protection. In the sixth case, hypoperfusion was concluded to be the cause but large numbers of microemboli, detected in the recovery room, was thought to be a contributing factor.

There was one ipsilateral MCA occlusion (patient # 347, Table 1) thought to be due to primary thrombosis developing at the site of a 50% MCA stenosis detected by TCD and seen on the preoperative angiogram. This patient was initially referred to surgery for a RIND and a 90% stenosis of the ICA. Thrombosis of the MCA was completed in 6 hours postoperatively when the patient became comatose and Doppler examination of the cervical ICA and the MCA confirmed MCA occlusion. In this patient there were only seven microembolic signals occurring during a 100-minute dissection phase and no microemboli seen in a 60-minute recovery phase suggesting the occlusion was the result of primary thrombosis of the MCA and not embolization from the operative site. Three postoperative CCT examinations demonstrated old infarctions in the left posterior parietal-occipital area and no hemorrhage.

Table 5 discloses the CVCs by preoperative symptom category. There was no significant difference between the number of strokes in 239 asymptomatic patients and 261 symptomatic patients, p=0.30. It appears the most risk free symptomatic category included those patients referred for amaurosis fugax, and patients referred for hemispheric symptoms sustained the highest rate of CVCs, however, in 2 x3 cross tabulation analysis of the 3 symptomatic categories p= 0.20. Among the 44 patients with a prior contralateral CEA there were no complications. Among the 8 undergoing a repeat ipsilateral CEA there was 1 with a grade 1 complication from the first operation and one grade 3 complication with the second CEA. One of the 8 sustained a grade 3 stroke due to a thrombus developing at the site of the crossclamp of the CCA. There were no hospital complications among the 21 regional anesthesia patients, though only six were referred for hemispheric symptoms and the numbers of patients are insufficient for separate statistical analysis.

There were 499 cases with the percent stenosis quantitated by either angiography or Doppler using the method giving the greatest severity. Of the 499, 190 (38%) had stenosis <70% and 309 (62%) had stenosis >= 70%). Twelve CVCs occurred in the 190 cases with <70% stenosis for a 6.3% complication rate and 12 CVCs occurred in the 309 cases >=70% stenotic for a 3.9% rate p= 0.22. There was no statistically significant difference between symptomatic cases with >=70% and the cases with <70% stenosis where in 156 cases with >=70% there were 8 CVCs (5.1%) while in 105 with <70% there were 7 CVCs (6.7%) with p=0.61. Likewise there was no difference between asymptomatic cases with >= 60% stenosis and cases with <60% stenosis where in 200 cases with >= 60% stenosis there were 8 CVCs (4.0%) while in 38 cases with <60% stenosis there was 1 CVC (2.6%) with p=0.68.

The proportion of patients in asymptomatic vs. symtomatic categories was the same among the 1st 100 and the last 400 cases i.e. 48% were asymptomatic in both groups of cases. The incidence of permanent deficits diminished from 7% in the first 100 operations to 2% for the last 400, p<=0.012. For all CVCs the incidence diminished from 8% in the first 100 to 4% in the last 400, p<= 0.11.

Discussion

In this study it appears that embolization represents the main cause of hospital cerebral complications from CEA with hyperperfusion as the second major cause. Hypoperfusion is the least frequent cause of CVCs in this study where 59% of cases were shunted. Even so, this later finding, along with the higher complication rate for shunted operations, suggests that shunting may be advisable only in highly selected cases and should be particularly avoided when difficulties in placement are anticipated. This conclusion was also reached by a previous large group study ³¹.

Cortical collateral, may, account for tolerance to crossclamping with MCA velocities below the 30% level for extended periods in some patients. The residual MCA velocity after crossclamp may be reduced by cortical collateral whose effect would be in an opposite direction to the collateral from the circle of Willis. The presence of cortical collateral, however, does not invalidate the shunting criteria use here because any error from this effect is on the side of safety.

In this study no single embolus or combination of microemboli has been recognized which clearly produced the embolic complications but transient decreases in MCA velocity have been reported associated the embolic showers ³² in the recovery room which have led to CVCs. It was not possible in the present study to completely monitor the entire intraoperative and postoperative time periods and therefore a large embolus or group of particulate emboli may well have been missed, however, DMES are clearly more associated with CVC cases than in those without CVC. Controlling the numbers of microemboli appears to prevent the development of large, clinically significant emboli and saves brain tissue from many small infarcts as reported by others ³¹. The size of microemboli cannot, as yet, be measured but multifrequency techniques show great promise ^34,35.

The evidence here that TCD monitoring has reduced the incidence of cerebral injury is based on improvement of outcomes since our initial experience and also on the actions by the surgeons based on the TCD information. A randomized blinded study will be difficult to perform in this environment because of the convictions of many vascular surgeons that the technique represents good medical practice and what is learned from the monitored cases will be applied to those not monitored thus rendering comparison difficult between monitored and non-monitored patient groups. The work of others using TCD in CEA confirms the present results ^36,37.

Problems with TCD Monitoring of CEA patients mainly relate to finding and holding the MCA signal throughout the surgery because of the narrow, and sometimes non-existent, temporal bone window. Fifteen percent of CEA patients (13% of women and 5% of men) do not have an adequate transtemporal window to allow insonation of the MCA. Improved headbands and application techniques are now available to stabilize the probe and allow hands-off bilateral monitoring in most cases.

Conclusions

Nineteen percent of carotid endarterectomy patients demonstrate microemboli to the MCA.

Embolism from the operative site during carotid endarterectomy is the principal cause of CVCs exceeding hypoperfusion (p<0.02) with hyperperfusion and hypoperfusion also significant causes.

Post-arteriotomy occlusion of the ICA by thrombosis is a major cause of CVCs with the major mechanism being embolization from the thrombosed artery.

TCD monitoring provides relevant on-line information to allow prompt identification and treatment of three major causes of CVCs from CEA.

TCD information provided to the vascular surgeon perioperative to carotid endarterectomy leads to alterations in surgical techniques and patient management.

The perioperative stroke rate can be reduced by appropriate measures by the surgeons, based on findings of TCD monitoring, p<=0.01..

Doppler microembolic signals are produced intraoperatively in more than 90% of all carotid endarterectomies with or without cerebral complications. The numbers, however, of intraoperative Doppler microemboli signals are strongly associated with CVCs from CEA p=0.02.

TCD monitoring of CEAs provides important knowledge about the mechanisms of cerebral complications in carotid endarterectomy.

There is no statistically significant difference between the number of carotid endarterectomy strokes in asymptomatic and symptomatic patients, p=0.30.

Shunting of the operative site may be associated with more complications than non-shunting, particularly when difficulties are encountered with inserting the shunt.

Shunting is recommended in a small percentage of patients who can be selected on the basis of TCD detected velocity changes after crossclamping of the common carotid artery.

References

Legend: Table 1. Probable Causes and Relevant Information in 24 Cerebrovascular Complications From Carotid Endarterectomy. The primary cause for each complication is listed immediately following IO or PO in the last column. No.= patient number; Sx= Preoperative Symptom code where A= asymptomatic, D= dizzyness, T= hemispheric TIA, TAF= amaurosis fugax, R= RIND, and S= stroke; % Stn= percent stenosis; pSt= stump pressure; % vX= percent mean MCA velocity of pre-crossclamp value; Shu= Shunt Yes No; YN= attempted but withdrawn for more than 10 minutes during crossclamping; tXC= longest time in minutes the carotids were crossclamped without a shunt in place; vR%= MCA velocity after release / pre-crossclamp velocity; IOE= total number of intraoperative microemboli; RCE= number of microemboli detected in the recovery room; Gr= grade of CVC severity; Cause= probable cause of the CVC; IO= intraoperative; PO= postoperative; Emb= Embolism, Hyper= Hyperperfusion, Hypo= Hypoperfusion; += secondary cause, occ.= occlusion; hypten.= hypertension, hem.= hemorrhage, * = verified by surgical reopening of the artery; †= verified by postoperative CTT; ‡ = verified by post-operative Duplex or Doppler. A blank= no information recorded; § = no tape available for review but technologist reported no microemboli detected.

 Table 2. Microembolic Signals and Perioperative Phases

Legend Table 2. Prevalence and average numbers of DMES in perioperative phases of CEA. Prv5min= prevalence among cases monitored for more than 5 minutes, DMES / min= average number of DMES divided by the total minutes monitored, n= number of cases in each sample.

Table 3. Cerebral Complication Grade and Causes

Legend Table 3. Severity and Distribution of 24 Cerebrovascular Complications

According to Probable Cause EMBO= embolization, HYPO= hypoperfusion, HYPER= hyperperfusion, *= one patient had occlusion of the MCA.

Table 4. Shunting and Causes of Cerebral Complications

Legend Table 4. Cerebrovascular Complications in Shunted and Non-shunted Carotid Endarterectomies

Table 5. Cerebral Complications and Preoperative Symptoms

Legend Table 5. Cerebrovascular Complications (CVCs) from Carotid Endarterctomy According to Preoperative Symptoms.