Issuu on Google+

Deep Hypothermic Circulatory Arrest: II. Changes in Electroencephalogram and Evoked Potentials During Rewarming Mark M. Stecker, MD, PhD, Albert T. Cheung, MD, Alberto Pochettino, MD, Glenn P. Kent, BS, Terry Patterson, PhD, Stuart J. Weiss, MD, PhD, and Joseph E. Bavaria, MD Department of Neurology, Department of Anesthesia, and Division of Cardiothoracic Surgery, Hospital of the University of Pennsylvania, Philadelphia, Pennsylvania

Background. Electrophysiologic studies during rewarming after deep hypothermic circulatory arrest probe the state of the brain during this critical period and may provide insight into the neurological effects of circulatory arrest and the neurologic outcome. Methods. Electroencephalogram (EEG) and evoked potentials were monitored during rewarming in 109 patients undergoing aortic surgery with hypothermic circulatory arrest. Results. The sequence of neurophysiologic events during rewarming did not mirror the events during cooling. The evoked potentials recovered first followed by EEG burst-suppression and then continuous EEG. The time to recovery of the evoked potentials N20-P22 complex was significantly correlated with the time of circulatory arrest even in patients without postoperative neurologic defi-

cits (r ⴝ 0.37, (p ⴝ 0.002). The nasopharyngeal temperatures at which continuous EEG activity and the N20-P22 complex returned were strongly correlated (r ⴝ 0.44, p ⴝ 0.0002; r ⴝ 0.41, p ⴝ 0.00003) with postoperative neurologic impairment. Specifically, the relative risk for postoperative neurologic impairment increased by a factor of 1.56 (95% CI 1.1 to 2.2) for every degree increase in temperature at which the EEG first became continuous. Conclusions. No trend toward shortened recovery times or improved neurologic outcome was noted with lower temperatures at circulatory arrest, indicating that the process of cooling to electrocerebral silence produced a relatively uniform degree of cerebral protection, independent of the actual nasopharyngeal temperature. (Ann Thorac Surg 2001;71:22– 8) © 2001 by The Society of Thoracic Surgeons

T

circulatory arrest. The initial intent of this article is to examine the neurophysiologic events which occur after circulatory arrest is complete and antegrade cerebral perfusion is restored. A second goal of this investigation is to define the range of “normal” recovery after hypothermic circulatory arrest as well as the factors which affect neurophysiologic recovery. Finally, the relationship between the neurophysiologic recovery and neurologic outcome is explored.

he effects of rapid cooling on electroencephalogram (EEG) and evoked potentials (EP) prior to deep hypothermic circulatory arrest were studied in the first article of this series [1]. In that communication, it was demonstrated that the temperatures at which specific neurophysiologic events occurred were not dependent on any patient specific or surgical factors examined. However, the time elapsed between the initiation of cooling and certain neurophysiologic events was dependent on a variety of physical factors which influence the dynamics of the cooling process. These include hemoglobin concentration, which can change blood viscosity, PaCO2, which can change peripheral vascular resistance, and body surface area which affects the exchange of heat with the external environment. Both EEG [2– 4] and evoked potentials [5, 6] can be sensitive markers of cerebral injury during cardiac operations, and so the time course of EEG and evoked potential recovery after circulatory arrest may be indicators of ischemic injury that may have occurred during Presented at the Poster Session of the Thirty-sixth Annual Meeting of The Society of Thoracic Surgeons, Fort Lauderdale, FL, Jan 31–Feb 2, 2000. Address reprint requests to Dr Stecker, Department of Neurology, Geisinger Medical Center, 100 N Academy Rd, Danville, PA 17822; e-mail: mark_stecker@yahoo.com.

© 2001 by The Society of Thoracic Surgeons Published by Elsevier Science Inc

Patients and Methods One hundred and nine patients undergoing thoracic aortic surgical procedures with circulatory arrest form the core group for this study [1]. All surgical procedures were performed between January 1995 and December 1998 at the Hospital of the University of Pennsylvania. Data were obtained prospectively for analysis and entry into a database in accordance with a protocol approved by the Institutional Review Board. The following parameters were determined for each patient: age, body surface area (BSA), hemoglobin concentration at circulatory arrest, the maximum isoflurane concentration during rewarming, the nasopharyngeal and central temperatures at circulatory arrest, the dura0003-4975/01/$20.00 PII S0003-4975(00)02021-X


Ann Thorac Surg 2001;71:22– 8

tion of circulatory arrest (absence of antegrade cerebral perfusion), the “no-flow” time (duration of absence of both antegrade and retrograde cerebral perfusion), and the degree of stenosis in the most affected carotid artery. Each chart was reviewed to determine whether the patient suffered postoperative confusion or an intraoperative stroke, and patients were classified as having a postoperative neurologic impairment if either of these occurred. Patients were evaluated prospectively in the postoperative period by a single neurologist. A full formal neurological examination was performed and documented on patients with any significant impairments. The total number of postoperative days for which each patient was intubated in the hospital, and the number of days each patient spent in the hospital, were also determined.

EEG/Evoked Potential The EEG and evoked potential methodology used in this article are the same as in our previous study [1] except for the following definitions of the neurophysiologic events during the recovery period after antegrade cerebral perfusion is reestablished. The time after reinstitution of antegrade cerebral perfusion at which the burstsuppression first returned to the EEG was designated TR and CTRBS. The nasopharyngeal and central (rectal or bladder) temperatures at this time are designated NTRBS and CTRBS, respectively. The time at which continuous EEG activity first reappears is designated TRCont with the corresponding nasopharyngeal and central temperatures NTRCont and CTRCont. Similarly, the time elapsed between the reinstitution of antegrade cerebral perfusion before the reappearance of the N20-P22 complex was designated TRN20 with the corresponding temperatures NTRN20 and CTRN20. The time at which N13 reappeared was designated TRN13.

Surgery/Anesthesia All patients were anesthetized with fentanyl, midazolam, isoflurane, and pancuronium. Cardiopulmonary bypass was instituted using standard bicaval venous cannulation and arterial cannulation of either the left femoral artery, ascending aorta, or the aortic arch. The left ventricle was vented via the right superior pulmonary vein. Patients were cooled on cardiopulmonary bypass (CPB) for a minimum of 30 minutes [1]. Antegrade cerebral perfusion was not interrupted until the EEG became isoelectric. At that time, the patient was partially exsanguinated, the superior vena cava was snared between the right atrium and the azygous vein, and retrograde cerebral perfusion (RCP) was administered. RCP was oxygenated blood was adjusted to maintain a right internal jugular venous pressure of 25 mmHg with the patient in an approximately 10-degree Trendelenberg position. RCP was interrupted for variable periods of time during deep hypothermia. After completion of aortic arch anastomoses, air was removed from the aorta and graft. Rewarming was initiated a variable period after antegrade cere-

STECKER ET AL EEG AND EP DURING REWARMING

23

bral perfusion was restored. After “deairing” a crossclamp was placed across the ascending aorta and standard cardiopulmonary bypass with antegrade cerebral perfusion was reinstituted for the final repair. The process of rewarming involved two distinct stages. Immediately after the end of circulatory arrest, antegrade cerebral perfusion was initiated without active rewarming. The duration of this phase was variable and determined by surgical variables such as the time required to complete the proximal aortic reconstruction or the number of patch sutures at the arch. However, the goal was to initiate active rewarming 5 minutes after the initiation of antegrade graft perfusion. The duration of this phase of the surgical procedure was estimated as the time from the end of circulatory arrest until the nasopharyngeal temperature rises 3°C above the temperature at the end of circulatory arrest and was called T3deg. During rewarming, the gradient between the warming bath and the blood temperature was maintained less than 10°C at all times. Rewarming continued until nasopharyngeal temperatures of 36.5° to 37°C were reached. In order to quantify the rewarming rate, the time required to warm from 3°C above the nasopharyngeal temperature at circulatory arrest to a temperature 3°C below the maximum temperature achieved during rewarming was computed. The warming rate is the ratio of the change in temperature over this range to the time required for warming between these two temperature points.

Statistics Analysis of variance (ANOVA; Statistica, Statsoft, Tulsa, OK) was used to test for differences between the measured parameters in 3 groups of patients: neurologically normal patients, patients with preoperative strokes, and patients with postoperative strokes. Multiple t tests were used to determine whether the times or temperatures of any of the electrophysiologic events were different in the neurologically normal patients and the patients who suffered new postoperative neurologic impairment. Because of the multiple t tests, p less than 0.01 was selected for statistical significance. A Spearman rank correlation analysis was used to determine which factors influenced the time or temperature of the EEG and EP events during rewarming. Because of the use of multiple comparisons, statistical significance was taken at p less than 0.005. A relationship was considered a statistical trend when p less than 0.02. Factors with a p less than 0.02 in this univariable analysis were selected for entry into a forward stepwise multiple linear regression analysis in order to determine whether any of the associations seen in the univariable analysis remained significant in the multivariable analysis. The F value to enter the regression was chosen as 2.0. Because of the small number of patients on whom carotid ultrasound studies were available, this factor was not entered into any of the multiple regression analyses. A forward stepwise logistic regression analysis was carried out using the BMDP-LR program (SPSS, Chicago, IL) to isolate the factors associated with postoperative


24

STECKER ET AL EEG AND EP DURING REWARMING

Ann Thorac Surg 2001;71:22– 8

Table 1. Basic Demographic Data Neurologically Normal Both Pre and Postop Factor Age (years) Maximum carotid stenosis Body surface area (BSA) (Meters2) Circulatory arrest time (min) No flow time (min) (no antegrade or retrograde) Nasopharyngeal temperature at circulatory arrest (°C) Rectal temperature at circulatory arrest (°C) Hemoglobin concentration at circulatory arrest (g/dl) Time to rewarm 3°C above temp at end of CA (T3deg) Rewarming rate (°C/min) (nasopharyngeal) Maximum isoflurane during rewarming (%) Days intubated after surgery Days in hospital after surgery

Intraop Stroke or Postop Confusionb

Preop Strokea

No.

Mean (SD)

No.

Mean (SD)

No.

69 32 68

62.7 (14) 19.2 (27.5) 1.96 (0.24)

18 6 18

68.5 (11.4) 54.8 (38.2) 1.86 (0.22)

27 8 27

69.3 (12) 30.6 (35) 1.9 (0.3)

69 68

36.6 (11.8) 3.8 (5.5)

18 18

42.3 (10.1) 4.5 (6.4)

27 27

51.6 (21.3)c 4.9 (6.4)

69

14.5 (1.8)

17

14.2 (2.1)

26

14.2 (2.3)

66

18.7 (2.6)

16

18.1 (2.6)

25

18.7 (2.6)

67

7.8 (1.5)

18

7.0 (1.7)

27

7.6 (1.6)

67

15.8 (7.8)

15

16 (13)

24

14.3 (8.25)

67

0.42 (0.19)

16

0.42 (0.15)

26

0.44 (0.15)

62

0.52 (0.51)

18

0.56 (0.84)

24

0.55 (0.73)

65 62

2.1 (4.9) 14.1 (9.9)

16 14

6.4 (7.1) 16 (11.3)

21 21

6.4 (5.8)d 33.6 (52)

a Independent of whether or not the patient had an intraoperative stroke or postoperative confusion. c a preoperative stroke; p ⫽ 0.006 main effect of intraoperative stroke or postoperative confusion; or postoperative confusion.

neurologic impairment. Postoperative neurologic impairment was the dependent variable and the independent variables pool included all of the patient/surgical factors in Table 1 and all of the electrophysiologic factors in Tables 2 and 3, except for the central temperature data. Only variables in the pool whose correlation with impaired neurological outcome by the Spearman rank correlation analysis was associated with p less than 0.2 were used in this analysis and a p value for entry into the logistic regression was chosen at 0.05. Receiver operator characteristics and the relative risk of postoperative neurologic impairment associated with significant factors were computed.

Results Demographics Basic data on the cohort of 109 patients (88% of whom underwent surgery via a median sternotomy and 12% of whom underwent a left thoracotomy) in this study is contained in Table 1. In particular, this table demon-

b

Mean (SD)

Independent of whether or not the patient had p ⫽ 0.022 main effect of intraoperative stroke

d

strates mean circulatory arrest times of 36.6 (⫾ 12) minutes for the neurologically normal patients and 51.6 ⫾ 21 minutes in the group of patients with postoperative neurologic impairment (p ⫽ 0.006). This indicates a trend for longer circulatory arrest times in the patients with postoperative neurologic impairment. The nasopharyngeal temperature at circulatory arrest averaged 14.4 ⫾ 2°C. T3deg averaged 15.9 ⫾ 8.9 minutes with a range of 3 to 57 minutes. Gradual rewarming following the phase of initial reperfusion occurred at an average rate of 0.42 ⫾ 0.19°C per minute. There was a trend for patients with postoperative neurologic impairment to remain intubated for longer (6.4 ⫾ 5.8 days) periods than patients without postoperative neurologic impairment (2.1 ⫾ 4.9 days; p ⫽ 0.02). The number of postoperative days in hospital was also longer for the neurologically impaired patients (34 ⫾ 52 days) than the neurologically normal group (14 ⫾ 10 days), although this difference did not reach statistical significance.

Table 2. Summary of Times and Temperatures for Events During Rewarminga Description Reappearance Reappearance Reappearance Reappearance a

Factor of of of of

burst-suppression continuous EEG N20-P22 N13

R

T BS TRCont TRN20 TRN13

No. 48 47 66 37

Mean (SD) 19.0 (9) 47.1 (26) 14.2 (7) 12.6 (6)

Factor R

NT BS NTRCont NTRN20 NTRN13

Includes only patients without either preoperative, intraoperative strokes, or postoperative confusion.

Mean (SD)

Factor

Mean (SD)

21.2 (5) 30.1 (5) 18.6 (3) 17.2 (2)

R

20.6 (3) 26.8 (4) 20.0 (3) 19.1 (2)

CT BS CTRCont CTRN20 CTRN13


Ann Thorac Surg 2001;71:22– 8

25

STECKER ET AL EEG AND EP DURING REWARMING

Table 3. Summary of Times and Temperatures for Events During Rewarminga Description Reappearance Reappearance Reappearance Reappearance

Factor of of of of

burst-suppression continuous EEG N20-P22 N13

R

T BS TRCont TRN20 TRN13

No. 9 9 16 10

Mean (SD) 26.2 (8) 80.5 (28) 20.3 (9) 16.5 (9)

pb 0.02 0.0008 0.004 0.12

Factor R

NT BS NTRCont NTRN20 NTRN13

Mean (SD) 24.8 (6) 36.2 (0.8) 22.3 (4) 18.8 (2)

p

Factor

Mean (SD)

p

0.07 0.0007 0.0002 0.07

R

22.2 (5) 32.8 (4) 20.0 (2) 19.7 (3)

0.23 0.0006 0.95 0.45

CT BS CTR,Cont CTRN20 CTRN13

a b Includes only patients with intraoperative strokes or postoperative confusion; Probability that corresponding factors in the neurologically normal and postoperative neurologic impairment groups are the same by Student’s t test.

Rewarming During rewarming, in the neurologically normal patients (Table 2), the N13 wave reappeared on average 12.6 ⫾ 6 minutes after initiation of antegrade cerebral perfusion, just prior to the reappearance of the N20-P22 complex which recovered on average 14.2 ⫾ 7 minutes after reinstitution of antegrade cerebral perfusion. The reappearance of the N20-P22 complex was followed in the next 4.8 minutes by the reappearance of burstsuppression on the EEG. However, on average, another 28.1 minutes elapsed before the EEG showed continuous electrical activity. The temperatures associated with these events followed a similar pattern with N13 recovering at the lowest temperature and continuous EEG returning at the highest temperature. The TRCont, NTRCont, TRN20, and NTRN20 were all significantly longer in the patients with new postoperative neurologic impairments than in the patients without postoperative neurologic deficits (Table 3). The order of reappearance of the various EEG and EP waveforms during rewarming was the same in the patients with and without postoperative neurologic impairment but was different than the sequence of changes during cooling (Fig 1). During cooling [1], the lowest temperature at which continuous EEG appears is 24.4°C ⫾ 4°C, but, during rewarming, the lowest temperature at which continuous EEG appears is 30.1°C ⫾ 5°C. During cooling [1], the highest temperature at which electrocerebral silence (ECS) was seen was 17.8°C ⫾ 4°C, while during rewarming this appears at 21.2°C ⫾ 5°C, N20-P22 complex disappears prior to ECS during cooling at 21.4°C ⫾ 4°C, but reappears prior to the return of electrocerebral activity at 18.6°C ⫾ 3°C. Rewarming is a more complex process than cooling since the physiologic changes occurring during circulatory arrest can have important effects on the recovery of the EEG and evoked responses. When data from the entire group of patients is subjected to univariable analysis with a Spearman rank correlation test, the following results were obtained. The occurrence of preoperative stroke, carotid stenosis, age, hemoglobin concentration at circulatory arrest, and rewarming rate did not affect any of the electrophysiologic variables significantly. TRCont is prolonged in patients with procedures performed via a left thoracotomy approach (r ⫽ 0.36, p ⫽ 0.003), although no other electrophysiologic variables were dependent on the surgical approach. There is a tendency for lower temperatures at circulatory arrest to be associated with higher NTRCont (r ⫽ 0.34, p ⫽ 0.005). However, the strongest effects were higher temperatures at the return

of N20-P22 (NTRN20) and continuous EEG (NTRCont) in the patients with postoperative neurologic impairment (NTRN20, r ⫽ 0.41, p ⫽ 0.00003; NTRCont, r ⫽ 0.42, p ⫽ 0.0004) or long circulatory arrest times (NTRN20, r ⫽ 0.35, p ⫽ 0.0003; NTRCont, r ⫽ 0.30, p ⫽ 0.01). Because postoperative neurologic impairment is associated with increased circulatory arrest times, it is possible that these effects may be due to the neurologic impairment itself or specifically due to circulatory arrest. In order to elucidate these possibilities, the univariable analysis was repeated using the group of neurologically normal patients. In this analysis, a significant prolongation in TRN20 with increasing duration of circulatory arrest was seen (r ⫽ 0.37, p ⫽ 0.002), and a trend toward elevations in NTRN20 with prolonged circulatory arrest times was seen (r ⫽ 0.29, p ⫽ 0.02). No significant effects of any parameter on the recovery to continuous EEG were noted in the group of neurologically normal patients. This suggests that there may be some specific effect of circulatory arrest on the recovery of the evoked potentials independent of the occurrence of neurologic injury. Multivariable linear regression is another method that may separate the effects of circulatory arrest and neurologic impairment. This analysis confirms the results of the univariable analysis. A significant effect of the circulatory arrest time on TRN20 was seen in both the group of neurologically normal patients (slope ⫽ 0.20, p ⫽ 0.002) and in the group of all patients (slope ⫽ 0.22, p ⫽ 0.0001). A trend toward increased NTRN20 with increased duration of circulatory arrest was seen in both the group of neurologically normal patients (slope ⫽ 0.08°C/min, p ⫽ 0.05) as well as in the entire group of patients (slope ⫽ 0.05°C/min, p ⫽ 0.03). The multivariable analysis also confirms the fact that lower nasopharyngeal temperatures at circulatory arrest result in both slower return to continuous EEG (slope ⫽ 6.7 min/°C, p ⫽ 0.002) and higher NTRCont (slope ⫽ ⫺1.0, p ⫽ 0.009) even in neurologically normal patients. Other effects seen in the multivariable analysis included increased TRCont and NTRCont with increased isoflurane during rewarming. Neither the duration of circulatory arrest nor postoperative neurologic impairment influenced the recovery of the N13 potential; however, significant effects of the time to warm by 3°C T3deg were noted in both the univariable and multivariable analyses.

Outcome Three outcome measures were studied: the number of postoperative days in hospital, the number of days intubated, and the occurrence of new postoperative neuro-


26

STECKER ET AL EEG AND EP DURING REWARMING

Fig 1. (A) Dependence of the N20-P22 amplitude on temperature during cooling. The line represents a sigmoidal curve fit to data obtained from 31 neurologically normal patients. The temperature at which various electroencephalogram (EEG) and evoked potential events are illustrated on the graph as are their standard deviations. (B) Dependence of the N20-P22 amplitude on temperature during rewarming after circulatory arrest. The line represents a sigmoidal curve fit to data obtained from 32 neurologically normal patients. The temperature at which various EEG and evoked potential events are illustrated on the graph as are their standard deviations. The difference in sequence of electrophysiologic events in these two figures is clear. Only 31 and 32 patients, respectively, were randomly chosen out of the 69 possible neurologically normal patients to provide an illustration of the changes in amplitudes as a function of temperature.

Ann Thorac Surg 2001;71:22– 8

increased duration of circulatory arrest correlated with postoperative neurologic impairment (r ⫽ 0.34, p ⫽ 0.0004). Prolonged TRCont and NTRCont were both associated with postoperative neurologic impairment (TRCont, r ⫽ 0.42, p ⫽ 0.0004; NTRCont, r ⫽ 0.44, p ⫽ 0.0002) and a longer duration of postoperative intubation (TRCont, r ⫽ 0.39, p ⫽ 0.002; NTRCont, r ⫽ 0.38, p ⫽ 0.003). Prolonged NTRN20 was associated with postoperative neurologic impairment (r ⫽ 0.41, p ⫽ 0.00003). To confirm the results of the univariable analysis, a forward stepwise logistic regression analysis was performed. This analysis (Table 4) indicates that only the NTRCont and NTRN20 make statistically significant contributions toward predicting the occurrence of postoperative neurologic impairment. In particular, the relative risk for postoperative neurologic impairment increases by a factor of 1.56 (95% CI 1.1 to 2.2) for every degree increase in NTRCont and by 1.27 (95% CI 1.02 to 1.56) for every degree increase in NTRN20. For example, the relative risk for a patient whose EEG first becomes continuous at a nasopharyngeal temperature of 37°C is 22-fold greater than a patient whose EEG first becomes continuous at 30°C! The performance of outcome prediction using NTRN20 and NTRCont is shown in the receiver operating characteristic of Figure 2. It can be seen that if an 80% sensitivity is required, the predictor has only 80% specificity, although using the best possible cutoff points, this procedure can correctly classify the neurologic outcome in 89% of cases. For the continuous outcome variables (hospital days and days intubated), a multivariable linear regression analysis was performed. The duration of intubation was a significantly related to the NTRCont (slope ⫽ 0.42 ⫾ 0.17 days intubated/°C, p ⫽ 0.02) but not directly to the circulatory arrest time. No significant relationships between the number of days in hospital and any of the explanatory variables were found.

Comment logic impairment. The mean number of days intubated (Table 1) was 2.1 in the neurologically normal patients and the mean number of hospital days was 14.1 in that group. All 8 patients with apparent intraoperative strokes were confused postoperatively. Nineteen patients experienced transient postoperative confusion without evidence of an intraoperative stroke. Thirty-three percent of the patients with postoperative confusion had experienced a preoperative stroke. There was no significant difference in the presence of postoperative confusion in the patients operated on through a median sternotomy and the patients operated on through a left thoracotomy. Multiple Spearman rank correlation analyses were performed to determine which surgical, patient-related and electrophysiologic factors influenced these outcome measures. The temperature at circulatory arrest, warming rate, degree of carotid stenosis, hemoglobin concentration, body surface area, and age did not significantly correlate with any outcome measure. The presence of a preoperative stroke was strongly correlated with an increased duration of intubation (r ⫽ 0.29, p ⫽ 0.004) and

Rewarming/Outcome The pattern of recovery of brain electrical activity during rewarming is different from the pattern of disappearance during cooling (Fig 1). The somatosensory evoked response N20-P22 complex generally returns during rewarming while the EEG continues to demonstrate ECS although it typically disappears during cooling while the EEG shows continuous activity or burst-suppression. This is a dramatic example of the “hysteresis” of evoked potentials described by Markand and colleagues [8] related to cooling and rewarming during cardiopulmonary bypass. This phenomenon is most likely due to the fact that, during rapid cooling and rewarming, the body is not in temperature equilibrium. As a result, different neural structures are at different temperatures which are not necessarily reflected by nasopharyngeal or central temperatures [7]. Since the evoked response depends on conduction in peripheral nerve, brachial plexus, spinal cord and brainstem, as well as cerebral cortex, while the EEG depends predominantly on cortical and thalamic


Ann Thorac Surg 2001;71:22– 8

STECKER ET AL EEG AND EP DURING REWARMING

activity, differential cooling or warming of these structures will produce different patterns of recovery. If the brainstem cooled more quickly and warmed faster than the cerebral cortex, one might expect quicker disappearance and quicker reappearance of the somatosensory evoked responses than the EEG activity. Analysis of the factors that affect the various EEG and evoked potential events is difficult because of patient heterogeneity and the sheer number of patient and surgical factors, many of which could not be systematically considered in these articles. However, two conclusions are clear. First, although there is no direct effect of circulatory arrest on the time or temperature to recovery of continuous EEG in the absence of neurologic injury, the recovery of EEG activity is influenced by intraoperative brain injury. Thus, prolonged time to recovery of continuous EEG, and elevated temperature at which the EEG first becomes continuous, are associated with an increased risk of postoperative neurologic injury. On the other hand, there is a direct effect of circulatory arrest on the recovery of the N20-P22 component of the evoked responses even in patients with no postoperative neurologic impairment. This is expected since the evoked responses return before the EEG and, hence, probe brain function at an earlier point after circulatory arrest and may be more sensitive to transient cerebral dysfunction caused by circulatory arrest. With this in mind, it is interesting to note that there have been two previous univariable analyses of the time to recovery of the N20-P22 complex in patients undergoing deep hypothermic circulatory arrest without retrograde cerebral perfusion [6, 9]. In a study involving 32 adult patients, Guerit and colleagues [9] demonstrated that the time to return of N20 prolonged by 0.45 minutes for each minute of circulatory arrest which is much larger than the estimate of 0.22 for the patients in our study. In a study involving 9 infants, Coles and coworkers [6] analyzed the time to return of P22 as a function of the product of the circulatory arrest time and the nasopharyngeal temperature at circulatory arrest. This analysis indicated that the time to return of P22 increased by roughly 0.089 minutes for each 1 minute°C increase in the time-temperature product. Repeating this analysis in our patient group, demonstrated only a 0.014 minute prolongation in ecovery for each 1 minute°C increase in the time-temperature product. The significantly shorter recovery times of the N20-P22 complex after circulatory arrest in the present study may be due to many factors including differences in rewarming protocols, differences in the use of anesthetic agents during the rewarming Table 4. Multivariable Analysis of Factors Associated With Either Intraoperative Stroke or Postoperative Confusion by Stepwise Logistic Regressiona Factor NTRCont NTRN20 a

All patients.

Relative Risk (per °C)

95% CI

p

1.56 1.27

1.1–2.2 1.02–1.56

⬍ 0.001 0.015

27

Fig. 2. Receiver operator characteristic curve demonstrating the sensitivity and specificity of predicting postoperative neurologic impairment based on NTRCont and NTRN20 using a logistic regression analysis (Table 4).

period, or other differences in patient population. It is also possible that this could be related to the fact that retrograde cerebral perfusion was employed in the present study and not in the previous studies. This study does not provide sufficient evidence to answer this question, and future studies would be needed. The temperature of first return to continuous EEG, rather than the time of return to continuous EEG, is the better predictor of outcome. This conclusion is supported by the results of the previous article in this series [1], where it was demonstrated that the timing for various events was more dependent on the various physical factors, such as hemoglobin and cooling rate, than were the temperatures. Thus, a greater part of the variability in factors such as NTRCont is likely to be due to biologic factors, such as brain injury, than for factors such as TRCont. In fact, using both the time to recovery of continuous EEG activity and the time to recovery of the N20P22 complex after circulatory arrest allows a correct prediction of postoperative neurologic impairment in 89% of cases. A greater predictive accuracy would not be expected based on the type of data submitted for analysis since, in the current study, focal or unilateral changes in EEG and evoked potentials were not considered.

Best Temperatures for Circulatory Arrest Although the nasopharyngeal temperature at the return of the N20-P22 complex was strongly dependent on the duration of circulatory arrest, there was no significant effect of the temperature of circulatory arrest on the recovery of the N20-P22 complex. This is one indication that the circulatory arrest temperatures chosen by the appearance of ECS during cooling provide a physiologically uniform level of cerebral protection. Furthermore, if the circulatory arrest temperature arrived at by the appearance of ECS were above the ideal temperature, lower temperatures would be associated with shorter times to recovery of the N20-P22 complex. If the temperatures were so low that they caused cerebral injury in and of themselves, a prolonged N20-P22 recovery might also be seen. Neither is the case, and no effect of the circulatory arrest temperature on the appearance of postoperative neurologic injury is noted. There were, however,


28

STECKER ET AL EEG AND EP DURING REWARMING

significant prolongations in TRCont and NTRCont with lower circulatory arrest temperatures even in the neurologically normal patients. This result can be explained either by brain injury occurring as a result of the low temperatures or by the fact that it takes longer to rewarm a colder brain (with brain temperature being cooler than the nasopharyngeal temperature [7]). The fact that the recovery of the N20-P22 has been shown to be more highly affected by the transient effects of circulatory arrest, but is not affected by the circulatory arrest temperature, argues against the first possibility. It can therefore be argued that the circulatory arrest temperature chosen by the appearance of ECS is consistent from patient to patient [1] and is neither “too low” or “too high.”

References 1. Stecker MM, Cheung AT, Kent GP, et al. Deep hypothermic circulatory arrest: I. Effects of cooling on EEG and evoked potentials. Ann Thorac Surg 2001;71:14 –21. 2. Salerno TA, Lince DP, White DN, Lynn RB, Charrette EJP. Monitoring of electroencephalogram during open-heart surgery. J Thorac Cardiovasc Surg 1978;76:97–100.

Ann Thorac Surg 2001;71:22– 8

3. Witoszka MM, Tamura H, Indeglia R, Hopkins RW, Simeone FA. Electroencephalographic changes and cerebral complications in open-heart surgery. J Thorac Cardiovasc Surg 1973; 66:855– 64. 4. Mezrow CK, Midulla PS, Sadeghi AM, et al. Quantitative electroencephalography: a method to assess cerebral injury after hypothermic: circulatory arrest. J Thorac Cardiovasc Surg 1995;109:925–34. 5. Stecker MM, Cheung AT, Patterson T, et al. Detection of stroke during cardiac operations with somatosensory evoked responses. J Thorac Cardiovasc Surg 1996;112:962–72. 6. Coles JG, Taylor MJ, Pearce JM, et al. Cerebral monitoring of somatosensory evoked potentials during profoundly hypothermic circulatory arrest. Circulation 1984;70(Suppl 3 Part 2):I96 –102. 7. Stone JG, Young WL, Smith CR, et al. Do standard monitoring sites reflect true brain temperature when profound hypothermia is rapidly induced and reversed? Anesthesiology 1995;82: 344–51. 8. Markand ON, Warren C, Mallik GS, et al. Temperature dependent hysteresis in somatosensory and auditory evoked potentials. Electroencephalogr Clin Neurophysiol 1990;77: 425–35. 9. Guerit JM, Verhelst R, Rubay J, et al. The use of somatosensory evoked potentials to determine the optimal degree of hypothermia during circulatory arrest. J Card Surg 1994;9: 596 – 603.

INVITED COMMENTARY Stecker and coauthors have presented an elegant description of the natural history of brain cooling, assessed by a variety of neurophysiologic parameters infrequently applied in a clinical setting. In Part I, the variability of cooling to achieve electrocerebral silence (ECS) is striking. In Part II, the “hysteresis” of warming and cooling seen in neural tissue is beautifully illustrated by the disappearance of sensory evoked potentials (SEP) prior to ECS during cooling, with reappearance of SEP while ECS persists during warming. An interesting subanalysis suggests that retrograde cerebral perfusion produces a more rapid rate of rewarming, when comparing results from this series to previous published experience. Not surprisingly, the duration of circulatory arrest appears to correlate with the rate of SEP recovery, and a lower temperature during circulatory arrest is associated with a trend toward later return of neurologic function. No doubt the authors hoped to connect observation to practice by linking conduct of circulatory arrest to outcome. This is something of a Holy Grail in the treatment of conditions requiring circulatory arrest. Specifically, is there a “best temperature” for circulatory arrest? The conventional answer to this question would be 16° to 18°C, clouded by uncertainty about the relationship between brain temperature and the temperature recorded at various peripheral monitoring sites. One implication of this pair of articles is that a significant number of patients arrested at 18°C will not have achieved ECS. Does this mean that EEG monitoring, to allow confirmation of ECS, is essential for safe conduct of circulatory arrest? Alternatively, should all patients be cooled to 12°C, or cooled for 50 minutes? The authors are careful to warn against

© 2001 by The Society of Thoracic Surgeons Published by Elsevier Science Inc

leaping to these conclusions, which their data do not support. They can only conclude that “appearance of ECS is consistent from patient to patient and is neither ‘too low’ or ‘too high’.” The difficulty linking observation to practice lies in isolating the variables. “Best temperature” variables such as ECS are dwarfed by other clinical variables. Where neurologic outcome is concerned, the most striking finding in this series is that mean circulatory arrest time was 36.6 ⫾ 12 minutes in neurologically normal patients, versus 51.6 ⫾ 21 minutes in patients with postoperative neurologic impairment ( p ⫽ 0.006). Therefore, duration of circulatory arrest is the most important predictor of outcome. Assuming reasonable technical facility, duration of circulatory arrest is determined by the complexity of the surgical pathology, not by the management of warming and cooling. The authors make no claims to have assessed the relationship between surgical complexity and outcome, which is difficult to accomplish, and was not the goal of this study. As a result, the findings fall short of a mandate for a change in practice, while still contributing to our understanding of warming and cooling. Craig R. Smith, MD Division of Cardiothoracic Surgery College of Physicians & Surgeons of Columbia University 177 Ft Washington Ave Suite 7-435 New York, NY 10032

0003-4975/01/$20.00 PII S0003-4975(00)02169-X


hypothermia_EEG_EP_part2