Cerebral injury is among the most common and disabling complications of open heart surgery. Attempts to provide neuroprotection have yielded conflicting results. We assessed the potential of propofol-induced burst suppression during open heart surgery to provide cerebral protection as determined by postoperative neuropsychologic function.
Two hundred twenty-five patients undergoing valve surgery were randomized to receive either sufentanil or sufentanil plus propofol titrated to electroencephalographic burst suppression. Blinded investigators performed neurologic and neuropsychologic testing at baseline, postoperative day (POD) 1 (neurologic testing only), PODs 5-7, and PODs 50-70. Neuropsychologic tests were compared with the results of 40 nonsurgical patients matched for age and education.
Electroencephalographic burst suppression was successfully achieved in all 109 propofol patients. However, these patients sustained at least as many adverse neurologic outcomes as the 116 controls: POD 1, 40% versus 25%, P = 0.06; PODs 5-7, -18% versus 8%, P = 0.07; PODs 50-70, -6% versus 6%, P = 0.80. No differences in the incidence of neuropsychologic deficits were detected, with 91% of the propofol patients versus 92% of the control patients being impaired at PODs 5-7, decreasing to 52 and 47%, respectively, by PODs 50-70. No significant differences in the severity of neuropsychologic dysfunction, depression, or anxiety were noted.
Electroencephalographic burst suppression surgery with propofol during cardiac valve replacement did not significantly reduce the incidence or severity of neurologic or neuropsychologic dysfunction. The authors' results suggest that neither cerebral metabolic suppression nor reduction in cerebral blood flow reliably provide neuroprotection during open heart surgery. Other therapeutic approaches must be evaluated to address this important medical problem.
This article is accompanied by an Editorial View. Please see: Hindman BJ, Todd MM: Improving neurologic outcome after cardiac surgery. Anesthesiology 1999; 90:1243–7.
CENTRAL nervous system dysfunction after cardiac valve replacement surgery continues to pose a significant challenge, with stroke occurring in up to one fourth of patients undergoing open heart procedures and detectable neurologic abnormalities in up to one half. [1–3] The reported incidence of neuropsychologic dysfunction has not been extensively studied in this patient population, although the incidence is believed to be at least as high as that for coronary artery bypass graft (CABG) surgery, with dysfunction occurring in up to 79–88% of patients within a week of surgery. [4–6] Perioperative neurologic and neuropsychologic dysfunction are of profound significance because they lead to increases in mortality rate and resource use, i.e., prolongation of intensive care unit and hospital stays and increased need for long-term rehabilitation. [2,7]
Despite widespread recognition of the problem, few advances have been made in preventing or treating cerebral complications precipitated by cardiac surgery. Efforts have been focused on reducing cerebral embolization with the use of arterial filters and membrane oxygenators, alterations in management of the ascending aorta, and surgical air-evacuation maneuvers.  A main-stay of neuroprotective therapy has been systemic hypothermia, although recent reports differ regarding its effectiveness in the setting of cardiopulmonary bypass. [8–11] Pharmacologic cerebral protection has been proposed, including calcium-channel antagonists  and electroencephalographic (EEG) burst suppression with thiopental. [13,14] However, the former has not been proven to be effective, and the latter approach remains controversial. Although one study of thiopental showed significant protection in patients undergoing open-chamber procedures,  a subsequent study failed to show neuroprotection in CABG patients.  Furthermore, the doses of barbiturates necessary for EEG burst suppression have been associated with significant myocardial depression, increased need for vasopressors, and prolonged time to tracheal extubation. [13,14] Thus, the use of barbiturates for cerebral protection in this setting currently is not widespread, with attention focusing on other methods of attaining EEG burst suppression.
Propofol has been shown to have effects similar to thiopenal on cerebral metabolism and blood flow.  It has a favorable pharmacokinetic profile, having been used successfully in CABG surgery without significant myocardial depression or prolonged sedation. [16,17] In addition, it also has been shown to have antioxidant properties  and perhaps calcium-channel antagonism  that may reduce the impact of cerebral injury during cardiac surgery. Therefore, we designed a prospective, randomized clinical trial to determine whether propofol titrated to EEG burst suppression reduces the incidence or severity of cerebral injury associated with valve surgery using neuropsychologic function 2 months postoperatively as the primary endpoint. To assess possible effects on a broad spectrum of cerebral injuries, we used a strategy of testing for neurologic and neuropsychologic abnormalities with validated testing instruments at repeated intervals. Finally, to assess safety, we ascertained the effects of burst suppression doses of propofol on hemodynamics and time to extubation.
After institutional review board approval at each site and obtaining informed consent, we studied 225 patients undergoing aortic or mitral valve surgery at the San Francisco Kaiser-Permanente Medical Center, the Duke University Medical Center, and the University Hospital, University of Western Ontario, with central analysis performed by the Ischemia Research and Education Foundation. To be eligible for the study, patients had to be between 21 and 79 yr old, hemodynamically stable, and scheduled to undergo elective valve repair or replacement (with or without CABG). Exclusion criteria included pregnancy, history of cerebral infarction or significant neurologic illness, recent seizure history, an ejection fraction less than 0.40, cardiogenic shock recent history of drug abuse, recent participation in another investigational trial, and inability or unwillingness to comply with the protocol. Patients were randomized for each surgeon at each medical center in balanced blocks.
Patients were premedicated with lorazepam, 0.03–0.06 mg/kg orally, and morphine sulfate, 0.10–0.15 mg/kg 60–90 min before surgery. All routine cardiovascular medications were continued through the morning of surgery. Before induction of anesthesia, catheters were inserted into the radial and pulmonary arteries. EEG monitoring was performed using a Neurotrac II system (Moberg Medical, Ambler, PA) and consisted of 10 leads affixed in a standard parasagittal bipolar block montage using 10-mm gold cups with collodion. The raw EEG was monitored and recorded continuously with an amplitude adjusted to 5 [micro sign]V/mm. Impedance was maintained below 5,000 Omega throughout the study.
For all patients, anesthesia was induced with sufentanil, 5 [micro sign]g/kg, followed by infusion at 1 [micro sign]g [middle dot] kg-1[middle dot] h-1. If necessary, additional boluses of 1 [micro sign]g/kg were given for indications of light anesthesia, with incremental increases in the infusion rate of 0.5 [micro sign]g [middle dot] kg-1[middle dot] h (-1) with each bolus. After induction of anesthesia, patients randomized into the propofol treatment group received propofol, via computer-assisted continuous infusion titrated to achieve EEG burst suppression (60 s between bursts from previous aortic cannulation through chest closure). Heart rate was maintained at less than 120% of baseline, and systolic blood pressure was maintained within 20% of baseline (determined as the mean of three preoperative determinations) by adjusting the level of anesthesia, administering volume, or administering vasopressors, [Greek small letter beta] blockers, or vasodilators as clinically indicated.
Cardiopulmonary bypass was managed with membrane oxygenators, [Greek small letter alpha]-stat blood gas management, 20-[micro sign]m arterial blood filters, nonpulsatile perfusion at 2.0–2.61 [middle dot] m-2[middle dot] min-1, and systemic hypothermia to 25–28 [degree sign]C (nasopharyngeal). Mean arterial blood pressure was maintained between 50 and 90 mmHg, using phenylephrine or sodium nitroprusside as indicated. Hematocrit was maintained at more than 18%, and glucose was maintained less than 200 g/dl with insulin, if necessary.
Neurologic and Neuropsychologic Assessments
Patients underwent a battery of neurologic and neuropsychologic tests preoperatively administered by trained nurse specialists with intra- and interobserver validity ensured. Standardized neurologic testing consisted of the National Institutes of Health Stroke Scale  and the Western Perioperative Neurologic Scale.  The neuropsychologic battery consisted of the Digit Symbol subtest of the Wechsler Adult Intelligence Scale-Revised, the Paired Associated Learning Subset of the Wechsler Memory Scale, the Trials A and B test, the Groove Pegboard test, and the Letter Cancellation test, all administered using a computer installed with Bowman Gray's Automated Behavioral Assessment System software program (courtesy of Dr. David A. Stump, Wake Forest University School of Medicine, Winston-Salem, NC). In addition the Spielberger State Anxiety Inventory, the Spielberger Trait Anxiety Inventory and the Beck Depression Inventory were administered to all patients.
Neurologic testing was repeated on postoperative days (PODs) 1 and 2, PODs 5–7, and PODs 50–70. Changes in neurologic function were assessed by determining the frequency and severity of the score change from baseline at the test periods on PODs 1 and 2 and PODs 50–70. Patients were considered to have adverse neurologic outcomes if they showed a change of more than three points on either scale (National Institutes of Health Stroke Scale, Western Perioperative Neurologic Scale). Severity of neurologic outcome was determined by combining the change score of the National Institutes of Health and the Western Perioperative Neurologic Scale examinations. The summation of scores has the effect of emphasizing factors common to both scales and deemphasizing factors common to both scales and deemphasizing but still accounting for factors used on only one scale. Frequency of adverse neurologic outcome was determined by the number of test periods in which the patient was classified as positive for neurologic outcome.
Neuropsychologic testing was performed before surgery (baseline) and repeated on PODs 5–7 and PODs 50–70. Data obtained from the neuropsychologic battery were assigned to one of four domains: verbal learning, motor dexterity, visual scanning, and psychomotor speed. Each test was scored and the scores of the two testing periods (PODs 5–7 and PODs 50–70) were compared to the preoperative baseline test scores. When the change score from baseline of any test was below the fifth percentile of normative change scores (based on a substudy of 40 nonsurgical volunteers matched for age, gender, and education and tested at identical intervals to the study population), the patient was considered positive for neuropsychologic outcome, with deficits in two or more domains considered to be a severe neuropsychologic outcome. To be considered calculable at each testing period, the patient had to complete testing in each of the four domains or test positive in at least one domain. Severity of neuropsychologic outcome in each testing period was determined by the number of domains in which the patient had a positive outcome. Frequency of adverse neuropsychologic outcome was determined by the number of times the patient tested positive across the two testing periods.
For categoric variables, the homogeneity of the odds ratios was evaluated and the general association between treatment and primary endpoints was derived, after controlling for center and coronary artery disease using the Cochran-Mantel-Haenszel method. For continuous variables, a general linear model test (adjusted for center and grouped by center) was used and, if the results were not consistent, the Kruskal-Wallis median test was used. All tests were two tailed, with P <or= to 0.05 considered significant.
A total of 225 patients were enrolled in the study and randomized to either group A (propofol; 109 patients) or group B (sufentanil only; 116 patients). Demographic medical history and surgical data were similar between groups (Table 1). There were no differences between groups with respect to neurologic history, type or duration of surgery, or time to extubation (Table 2).
Patients in the sufentanil (control) group received a total of 1,040 +/- 583 [micro sign]g of sufentanil per patient, compared with 767 +/- 280 [micro sign]g of sufentanil in the propofol group (P = 0.0004). To achieve initial burst suppression, 150 +/- 132 mg propofol was used, with a total dose of 3,062 +/- 2,040 mg necessary for maintenance of EEG burst suppression (mean time 3.2 +/- 0.9 h).
Neurologic and Neuropsychologic Outcomes
Patients in the propofol group tended to have a higher incidence of adverse neurologic outcomes at PODs 1 and 2 (40 vs. 25%, P = 0.06), and PODs 5–7 (18 vs. 8%, P = 0.07), but these differences had resolved by PODs 50–70 (6.2 vs. 6.2%, P = 0.80)(Table 3). There were no significant differences in severity or frequency of neurologic outcome between the two groups at any time period.
Neuropsychologic deficits were present in 91% of the propofol patients versus 92%(P = 0.73) of the sufentanil patients at PODs 5–7 decreasing to 52 and 47%(P = 0.58), respectively, by PODs 50–70. Likewise, neither the severity nor the frequency of neuropsychologic outcomes differed significantly between treatment groups at any postoperative test period (Table 4), nor were there differences for the depression or state-trait anxiety scales.
Intraoperative Hemodynamic and Surgical Factors
There were no significant differences between groups for aortic cross-clamp time, cardiopulmonary bypass time, or duration of operation. Hemodynamic analysis revealed that the propofol group had a significantly higher incidence of hypotension prebypass (74 vs. 43%, P < 0.001), during bypass (78 vs. 50%, P < 0.001), and postbypass (56 vs. 41%, P < 0.02). This coincided with a twofold difference in use of phenylephrine intraoperatively (5.0 +/- 5.3 mg vs. 2.5 +/- 5.6 mg, propofol vs. sufentanil, P = 0.002). The sufentanil group had a significantly higher incidence of hypertension during bypass (28 vs. 15%, P = 0.01). Use of an intraaortic balloon pump was similar (6 vs. 2%, P = 0.11), as were the need for a pacemaker (71 vs. 68%, P = 0.89), use of inotropes or vasopressors, and median time to extubation (19.4 vs. 18.3 h)(Table 2). Finally, the in-hospital mortality rate from all causes was 3.1%, without difference between groups (propofol 4.5%, sufentanil 1.7%, P = 0.27).
This multicenter study is the first report of the neurologic and neuropsychologic effects of propofol-induced EEG burst suppression during cardiac surgery. Our randomized clinical trial of 225 patients undergoing valve surgery showed that propofol-induced EEG burst suppression did not reduce adverse neurologic or neuropsychologic outcomes up to 2 months after surgery. Neurologic outcomes persisted in 6.2% of patients in each group at 2 months without significant differences in severity. Approximately half of the patients in each group showed neuropsychologic deficits at 2 months, also without significant differences in severity.
Because neuroprotective agents generally limit the size of cerebral infarct rather than the incidence of infarction,  we designed this trial to assess the severity of insult and the incidence. Because of the logistic difficulties and expense associated with perioperative imaging studies, we used neurologic scales that have been shown to be sensitive in detecting neurologic injury  and to correlate with infarct size.  Neuropsychologic dysfunction over time also correlates with the extent of injury ; hence, we repeated testing at several intervals up to PODs 50–70. Testing several months postoperatively also minimizes potential confounders such as pain, concomitant medications, and environmental stimulation that may interfere with perioperative neuropsychologic assessments. 
Animal Studies. Studies of the neuroprotective effects of propofol in animal models have yielded inconsistent findings. Kochs et al.  demonstrated improvement in neurologic outcomes and neuronal damage in propofol-treated rats subjected to right common carotid artery occlusion. However, Ridenour et al.  detected no improvement in neurologic outcomes or cerebral infarct volumes after ligation of a middle cerebral artery in rats given propofol compared to rats given halothane. Propofol protected against transient forebrain ischemia-induced delayed hippocampal neuronal death in a gerbil model, although it did not improve survival. 
Studies in Patients. Many changes in surgical or cardiopulmonary bypass techniques have been thought to diminish cerebral injury associated with cardiac surgery. Although there is evidence to suggest arterial filters,  membrane oxygenators,  and surgical evacuation of intracardiac air may decrease cerebral embolism, little evidence supports claims of improved central nervous system outcomes associated with these techniques. Currently, a great deal of attention has been focused on management of the diseased proximal aorta. Although some studies have shown promise, no prospective randomized study has convincingly shown the effectiveness of any of the suggested techniques. [7,29]
In contrast, both arterial blood-gas management and temperature and have been studied in prospective, randomized trials. Murkin et al.  demonstrated an improvement in neuropsychologic outcomes with [Greek small letter alpha]-stat blood-gas management. Regarding hypothermia, some investigators have demonstrated improved cerebral outcomes with hypothermia [9,11,30]; other investigators have detected no differences between normothermic and hypothermic patients. [8,10] Interpretation of these studies is confounded by differences in study design, including untreated hyperglycemia during bypass [9,30] and the use of bladder rather than nasopharyngeal temperatures. [8,9,30]
Pharmacologic protection from cerebral ischemia has been an elusive goal. Barbiturates, calcium-channel blockers, and other agents, including adenosine-regulating agents, have been tested in patients undergoing cardiac surgery. Calcium-channel antagonists may block the flow of calcium into ischemic cells, interrupting the ischemic cascade and improving postischemic hypoperfusion. However, only L-channel blockers, such as nimodipine, are clinically available, and these do not prevent central excitatory neurotransmitter release. An investigation of nimodipine in cardiac surgical patients was discontinued early because of excess mortality and bleeding without evidence of neuroprotective effects. 
Acadesine, an adenosine-regulating agent, was found in one study to be associated with a decreased incidence of stroke after CABG from 4.5% to 0.5%.  A meta-analysis of other studies (n = 2013) showed a trend toward reduced incidence of stroke (P = 0.10).  These studies suggest potential for adenosine-regulating drugs to reduce strokes after cardiac surgery.
Although animal studies with thiopental have shown smaller infarct size after transient focal ischemia, only one study in patients undergoing open chamber cardiac surgery has shown any benefit in humans.  Although this report initially was received with enthusiasm, the side effects of hemodynamic depression and prolonged sedation coupled with a subsequent report showing no benefit in coronary artery bypass surgery  have led to skepticism of the role of thiopental in providing neuroprotection in the setting of cardiac surgery.
The differences in results of the Nussmeier et al.  and Zaidan et al.  studies may be explained by a number of factors. Perhaps most notably, the studies were performed in different patient populations-open-chamber versus coronary artery bypass procedures. Although there may be considerable overlap in the causes of cerebral injury between the two types of procedures, open-chamber procedures are associated with a much higher incidence of air embolism and valve debris; CABG procedures are more likely to result in embolism from aortic atheromas.  It has been postulated that air emboli lead to temporary cerebral arterial occlusion, which may be more likely to benefit from intraoperative therapy than atheromatous emboli, which form permanent occlusions.  In addition, management of cardiopulmonary bypass in the Nussmeier et al.  study used several techniques that may increase cerebral embolization or worsen neurologic outcomes, i.e., bubble oxygenators, [28,34] relative normothermia (temperature >or= to 34[degree sign]C), [8,11,30] pH-stat blood-gas management,  no arterial filters,  and glucose priming solutions.  If one or more of these techniques worsen neurologic outcomes, a neuroprotective effect would be easier to detect.
Although it remains unclear whether thiopental provides cerebral protection in the setting of cardiac surgery, it clearly limits cerebral damage caused by focal ischemia in animal models. [36,37] Despite its benefits in animal models, the mechanisms of protection remain uncertain. Classically, the cerebral protectant effects of barbiturates were attributed to their ability to depress cerebral metabolism.  However, it has been shown that the amount of energy used by the cells for functional metabolism is significantly less than the total amount of energy necessary for maintenance of cellular stability.  In addition, other drugs that depress metabolic function have not yielded consistent cerebral protection. [40–43] Other proposed mechanisms include scavenging of free radicals,  membrane-stabilizing effects,  attenuation of free fatty acid accumulation,  and the possibility of decreased numbers of cerebral emboli because of decreased cerebral blood flow with thiopental. 
Current Study Findings and Clinical Implications
Propofol has effects similar to thiopental on cerebral metabolism and cerebral blood flow, although it maintains flow-metabolism coupling better than thiopental.  Propofol also appears to possess antioxidant properties  and calcium-channel antagonist effects,  which may be beneficial in limiting cerebral damage resulting from focal ischemic insults. Thus, we hypothesized that propofol may provide cerebral protection similar to thiopental. Therefore, we designed our trial using a similar patient population and degree of burst suppression and controlled for other potential confounders.
Despite the theoretical advantages of propofol, we were unable to detect any cerebral protectant effects at our primary 2-month endpoint, as shown by an extensive battery of neurologic and neuropsychologic testing. The lack of neuroprotection may have been caused by several factors, including the possibility that propofol lacks neuroprotective effects. Other potential reasons for the lack of a protective effect include management techniques believed to minimize cerebral injury such as [Greek small letter alpha]-stat blood-gas management, hypothermia, use of arterial filters and membrane oxygenators, and relatively tight control of glucose. If these techniques in fact decrease the likelihood of cerebral injury, they also would diminish the power to discern a protective effect of propofol. It is also possible the trend toward worse neurologic outcomes at PODs 1 and 2 and PODs 5–7, with improvement by PODs 50–70, represents a neuroprotective effect over the long-term. More likely, the worsened in-hospital dysfunction represents residual drug effects that had dissipated by PODs 50–70, or possibly the fact that some unforeseen action of propofol actually may lead to a detrimental effect, as recently has been shown with nimodipine. 
The findings of this study have several important implications for application of drugs thought to have cerebral protectant properties because of suppression of cerebral metabolism. Although propofol quite effectively provided EEG burst suppression for more than 3 h, it did not provide meaningful neuroprotection. This finding supports the hypothesis that profound cerebral cortical metabolic depression does not, in itself, confer protection from focal insults. This is in keeping with a recent study finding that EEG burst suppression is not necessary for maximal neuroprotection with pentobarbital.  In addition, it has been postulated that the reduction in cerebral blood flow accompanying EEG burst suppression with agents such as thiopental may reduce the delivery of emboli into the cerebral circulation, and thereby reduce cerebral injury. If this were the case, it would be expected that propofol would provide similar benefit because it also significantly reduces cerebral blood flow. More likely, other properties are responsible for any neuroprotective capabilities of drugs such as barbiturates.
Although we used an extensive battery of neurologic and neuropsychologic tests, the tests may not be sensitive enough to reliably detect differences in cerebral infarct size. In addition, this study was powered to detect differences in neuropsychologic outcome-the primary endpoint-and not neurologic outcomes, per se, which would have required many more patients. However, because we detected no difference in neurologic outcomes at 2 months, increasing our sample size is unlikely to yield different results. Finally, although hypothermia clearly has been shown to be neuroprotective, use of hypothermic cardiopulmonary bypass necessitates rewarming before discontinuation of bypass. Because most emboli occur during clamp removal and resumption of pulsatile flow  when brain temperature may actually exceed 37[degree sign]C, the brain may be even more susceptible to injury. However, this effect would have been similar between groups and should not affect our conclusions.
In summary, we were unable to detect any neurologic or neuropsychologic benefit from using propofol-induced EEG burst suppression during cardiac valve surgery. Approximately one half of these patients demonstrate abnormal cognitive functioning at 2 months postsurgery, with 6.2% demonstrating neurologic deficits. Burst suppression with propofol is associated with vasodilation but is otherwise well-tolerated, with similar times to extubation. This study shows that drugs that decrease cerebral metabolism and cerebral blood flow do not necessarily provide cerebral protection during open-chamber cardiac surgery. Future investigations should focus on other potential mechanisms of pharmacologic neuroprotection.
The following institutions and persons coordinated the McSPI EPI-I study. Study Chairman-D. Mangano; Coordinating Center: Ischemia Research and Education Foundation-D. M. Boisvert, C. Dietzel, V. Katseva, E. Kwan, A. Herskowitz, C. Ley, L. Ngo; Outcome Validation Committee-S. Graham, C. Mora Mangano, N. Nussmeier, G. Roach, R. L. Wolman; Editorial/Administrative Group-D. Beatty, I. Deem, B. Xavier, W. von Ehrenburg.
The following institutions and persons participated in the McSPI EPI-I study. Centers and investigators: University of Alabama at Birmingham-W. Lell; Baylor College of Medicine-S. Shenaq, R. Clark; Cedars-Sinai Medical Center, California-A. Friedman; University of Chicago-M. Trankina, W. Ruo; Cleveland Clinic Foundation-C. Koch, N. Starr; Cornell University-O. Patafio, R. Fine; Duke University-T. Stanley, M. Newman; Emory University-C. Mora Mangano, J. Ramsay; Harvard University, Beth Israel Hospital-M. Comunale; Harvard University, Brigham and Women's Hospital-S. Body, R. Maddi; Harvard University, Massachusetts General Hospital-M. D'Ambra; University of Iowa-A. Ross; Kaiser-Permanente Medical Center, San Francisco-G. Roach, W. Bellows; University of Michigan-J. Wahr; New York University-M. Kanchuger, K. Marschall; University of Pennsylvania-J. Savino; Rush Presbyterian, St. Luke's Medical Center-K. Tuman; Stanford University-E. Stover, L. Siegel; Texas Heart Institute-S. Slogoff; M. Goldstein; Milwaukee Veterans Administration Medical Center-A. Aggarwal; San Francisco Veterans Administration Medical Center-G. Ozanne; D. Mangano; Medical College of Virginia, Virginia Commonwealth University-J. Fabian, R. L. Wolman, B. Spiess; Yale University-J. Mathew.