| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2000;69:1077-1083
© 2000 The Society of Thoracic Surgeons
a Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina, USA
b Department of Surgery, Duke University Medical Center, Durham, North Carolina, USA
Address reprint requests to Dr Hill, Department of Anesthesiology, Duke University Medical Center, Box 3094, Durham, NC 27710
e-mail: hill0012{at}mc.duke.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMaterial and methodsResultsCommentReferences |
|---|
Methods. Outcome data from 2,862 patients undergoing coronary artery bypass grafting were merged with intraoperative physiologic data. A preoperative stroke risk index was calculated for each patient. Variables found significant by univariate logistic regression were tested in a multivariable model to determine association with outcome.
Results. The incidence of stroke or coma in the study population was 1.3%. After controlling for stroke risk and bypass time, only an index of low mean arterial pressure during bypass retained a significant inverse association with outcome (p = 0.0304).
Conclusions. This study found no evidence that glucose concentration or minimum hematocrit are associated with major adverse neurologic outcome. The association between lower pressure during bypass and decreased incidence of stroke or coma persisted in all risk groups. This points to mechanisms other than hypoperfusion as the primary cause of neurologic injury associated with cardiac surgery.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsCommentReferences |
|---|
Perioperative neurologic complications after CABG prolong intensive care unit and hospital stay, raising the cost in both dollars and quality of life substantially [1, 3]. Outcome data indicate that less than 33% of patients with perioperative stroke are discharged home after CABG, with the others either dying or being discharged to a long-term care facility. This compares with a home discharge rate of greater than 90% in patients who undergo CABG without new perioperative neurologic complications [3]. Controversy exists over the role that intraoperative management of clinical variables has on neurologic outcome [4–7].
Intraoperative physiologic data acquisition systems permit automated recording of the many physiologic variables measured during cardiac surgical procedures [8]. These data, when merged with prospectively collected clinical outcome data, allow us to determine the association of the various physiologic variables with outcome. The purpose of this observational database study was to evaluate the possible association between minimum hematocrit level, maximum glucose concentration, and mean arterial pressure during cardiopulmonary bypass (CPB) with perioperative neurologic injury. For the purpose of this study, adverse neurologic outcome is defined as perioperative stroke, coma, or TIA.
|
Material and methods |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsCommentReferences |
|---|
The physiologic variables studied include (1) minimum hematocrit (MINHCT), defined as the lowest recorded hematocrit recorded at 15-minute intervals during CPB; (2) maximum glucose (MAXGLC), defined as the highest recorded glucose concentration during the surgical procedure; (3) mean arterial pressure less than 50 mm Hg (MAP < 50), defined as the integrated area less than a MAP of 50 mm Hg at each minute during CPB; (4) MAP more than 50 mm Hg (MAP > 50), defined as the integrated area greater than or equal to a MAP of 50 mm Hg at each minute during CPB; and (5) CPB time (CPBTIME), defined as the CPB duration in minutes.
A MAP less than 50 mm Hg during CPB has historically been considered to represent significant hypotension [10] and falls below the low MAP group investigated in the study by Gold and colleagues [5]. The MAP < 50 calculation represents a measure of low pressure, which takes into account how low the MAP was and how long it remained lower than 50 mm Hg during CPB. If MAP never fell below 50 mm Hg for a particular patient, the MAP < 50 was zero. The MAP > 50 calculation represents a measure of normal to high pressure, which takes into account how high the MAP was and how long it remained greater than or equal to 50 mm Hg during CPB. If MAP never exceeded 49 mm Hg for a particular patient, the MAP > 50 was zero.
A preoperative stroke risk index (SRI) developed by Newman and colleagues [11] was used as the main preoperative risk variable to predict adverse neurologic outcome. Other variables not part of the SRI but included in this analysis as potential covariates were a clinical history of hypertension (HTN) and the presence of a carotid bruit (BRUIT) on preoperative physical exam. Hypertension was defined as a previously diagnosed condition requiring intervention with medication or dietary adjustment. These preoperative variables were also included in the analysis because they are considered by other researchers [3, 4] to be either possible markers for increased risk of stroke or factors that have an impact on management of the patient’s intraoperative blood pressure. Any variables found to be associated with adverse neurologic outcome by univariate analysis (p < 0.05) were entered into a multivariate analysis to assess true significance. Significant univariate predictors were first combined individually with the SRI in a multiple variable logistic regression model. Variables retaining a significant association with adverse neurologic outcome after controlling for SRI were added to the model.
Determination of stroke risk index
Preoperative risk factors for stroke identified by Newman and associates [11] have been used previously to establish probability of perioperative central nervous system injury in patients undergoing CPB. Variables included in the SRI are age as well as presence of unstable angina (angina at rest or occurring with a crescendo pattern), diabetes, history of symptomatic neurologic disease, prior CABG surgery, history of vascular disease, and history of pulmonary disease. Each variable present is assigned a weighted point value, and the point values are added for a total score. The total score is compared to an analog scale, which estimates percentage risk of adverse neurologic outcome.
Neurologic assessment
Physiologic data from the intraoperative database were merged with outcome data of the cardiac surgical patients maintained in the Duke Cardiovascular Database [12] using hospital record number and surgery date as unique identifiers. Adverse neurologic outcomes (stroke, coma, or TIA) were entered by the surgical team into the database at the time of the patient’s discharge (or death if the patient never left the hospital). Verification of outcome was retrospective using International Classification of Diseases, Ninth Revision, Clinical Modification discharge codes followed by individual chart and radiologic review for all patients identified with an adverse neurologic outcome. Individual reviews were performed by a board-certified neurologist blinded to intraoperative management.
Statistical methodology
Statistical analysis was performed using SAS system, version 6.12 (SAS, Inc, Cary, NC). Continuous variables were described by mean, range, and standard deviation. Relationships between continuous variables were investigated using correlations, with analysis of stroke performed in stages. Stroke risk index scores were calculated for the entire study population. Univariate logistic regression was performed to test eight potentially important perioperative variables: SRI, MAXGLC, MINHCT, CPBTIME, MAP < 50, MAP > 50, BRUIT, and HTN. Significant variables were tested in a model controlling for the SRI. Finally, potential covariates were identified and added to the model to identify important variables associated with adverse neurologic outcome.
Sensitivity analysis
Because our study is retrospective and not randomized, there exists the possibility that patients who are maintained at higher levels of MAP are somehow different from patients maintained at lower MAP, and that this difference also results in different rates of adverse neurologic outcome. In other words, there might be a confounding variable influencing both MAP and neurologic injury that may be the actual variable conferring risk for adverse neurologic outcome. Despite our attempts to control for all such variables, the possibility exists that we have missed it. Therefore, following the method proposed by Rosenbaum and Rubin [13], we conducted a sensitivity analysis to determine how large the effect of this unknown confounding variable would have to be to eliminate or reverse the direction of our findings with regard to MAP during CPB. Because the methodology of the sensitivity analysis requires a binary predictor, MAP was dichotomized so that patients could be described as being maintained at either ‘high’ or ‘low’ MAP during CPB. The validity of this dichotomous variable was tested in a multivariable model predicting stroke. The sensitivity analysis was then performed under a variety of assumptions about the unknown confounder with respect to its behavior, which was characterized by (1) its effect on MAP, (2) its effect on neurologic outcome under low MAP conditions, and (3) its effect on neurologic outcome under high MAP conditions. A relative risk was then calculated for each combination of assumptions.
|
Results |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsCommentReferences |
|---|
The preoperative risk of neurologic injury predicted by the SRI ranged from 0.11% to 80%. Stroke risk index effectively stratified the risk groups. Patients experiencing adverse neurologic outcome had a preoperative SRI (mean ± standard deviation, 0.077 ± 0.067) more than twice the group without major neurologic injury (0.034 ± 0.049). The incidence of HTN and BRUIT were 69% and 13%, respectively. The MAXGLC ranged from 79 mg/dL to 653 mg/dL (230 ± 69.0 mg/dL). The MINHCT ranged from 10% to 39% (19.0% ± 3.7%). The MAP < 50 ranged from 0 to -2,629 mm Hg · min (-131 ± 154 mm Hg · min). The MAP > 50 ranged from 8 to 6,966 mm Hg · min (1,185 ± 762 mm Hg · min). The CPBTIME ranged from 19 to 313 minutes (107 ± 34.3 minutes).
The SRI significantly predicted outcome in a logistic regression model (p = 0.0001; c-index = 0.751). The c-index (or, more formally, the probability of concordance) is a measure of a statistical model’s predictive ability; with a value of 1.0 indicating perfect prediction and a value of 0.5 indicating random prediction. Other univariate relationships between predictors and stroke are listed in Table 1. The uncontrolled association of MAP < 50 during CPB and neurologic outcome was significant (p = 0.03; c-index = 0.620), implying that lower MAP during CPB is associated with less risk of stroke or coma. The uncontrolled association of MAP > 50 was also significant (p = 0.0297; c-index = 0.561), with a higher MAP associated with poor neurologic outcome. BRUIT displayed a significant uncontrolled association with adverse neurologic outcome in the univariate analysis (p = 0.032; c-index = 0.562). MINHCT, MAXGLC, HTN, and CPBTIME displayed no significant association with adverse neurologic outcome in the univariate analysis.
|
|
|
|
Comment |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsCommentReferences |
|---|
|
Minimum hematocrit during CPB also did not correlate with stroke or coma in this patient population. Even when analysis was performed using a subset of patients with a minimum hematocrit less than 14% as well as a subset of patients at high preoperative risk for stroke, minimum hematocrit level was not found to correlate with increased stroke rate.
Unlike some previous studies that found an association between duration of CPB and stroke rate [4, 7, 19], our study found no such association by either univariate or multivariate analysis. We propose that atheromatous emboli resulting from ascending aortic plaque rupture or a genetic predisposition to neurologic injury is more likely to be responsible for adverse neurologic outcome than is either low hematocrit or bypass time.
The association between low MAP and decreased incidence of adverse neurologic outcome is an intriguing result of our analysis. Even when controlled for SRI and CPB duration, the association between low MAP and decreased stroke rate retained statistical significance. Although it would be difficult to draw definitive conclusions from this study about an association between high MAP and increased adverse neurologic outcome, the relationship between low MAP and decreased adverse outcome would tend to be understated by this analysis. When CPB time is considered, prolonged CPB time would tend to increase MAP > 50 because the longer duration of CPB increases the time available for the MAP to equal or exceed 50 mm Hg. Therefore, the longer and more difficult cases with higher incidence of adverse neurologic outcome would be expected to have a higher MAP > 50. Interestingly, this relationship did exist in the univariate analysis but disappeared when controlling for SRI even before CPB time was added to the model. Conversely, because MAP < 50 is a negative number, the longer the CPB duration, the longer the time available for the MAP to fall below 50 mm Hg and the more negative the number becomes. Therefore, the longer and more difficult cases would tend to have a lower (more negative) MAP < 50 and skew the data toward insignificance. The finding of a significant association between MAP < 50 and adverse neurologic outcome after controlling for SRI and CPB time strongly implies that low MAP (more negative MAP < 50) was not a risk factor for adverse neurologic outcome in our population. In fact, in our study, lower pressure was associated with improved outcome.
Our data seem to conflict with some earlier studies, which concluded that higher MAP yielded better neurologic outcome. In 1995, Gold and colleagues [5] published a randomized, prospective trial of blood pressure management strategies during CPB. In one of their treatment groups, patients were maintained with a MAP in the range of 80 to 100 mm Hg during CPB using vasopressor support if necessary. In the other group, MAP was maintained in a target range of 50 to 60 mm Hg using sodium nitroprusside to lower MAP if necessary. In this patient population, no difference in mortality was identified nor was there any difference between the groups in major morbidity end points. Only when the patients who experienced cardiac complications were added to those sustaining neurologic complications was there any statistically significant difference between the treatment groups. Because of small sample size, their study has been criticized for lacking sufficient power to draw valid conclusions about the optimal management of MAP during CPB [6]. With larger patient numbers available in the Duke database, the number of adverse neurologic outcomes tripled compared with the study by Gold and associates [5], with a resultant increase in study power and a very different conclusion with regard to low perfusion pressure on CPB. In a retrospective study of patients undergoing CABG, Gardner and associates [4] concluded that severe perioperative hypotension was a univariate predictor of stroke. Although the data collected in that study included hypotensive episodes occurring in patients during the normothermic intraoperative and postoperative periods, we were specifically concerned with management of alterable variables during hypothermic and normothermic CPB. Therefore, the results of that study are not strictly comparable with our results.
Our study found an inverse relationship between MAP during CPB and incidence of adverse neurologic outcome. There are several possible explanations for this finding. First, a substantial treatment bias may exist in this patient population toward treatment of perceived hypotension in a patient with known preexisting risk factors for stroke. As a result, the higher risk patients may be treated more aggressively with pressor agents while on CPB, with a higher MAP maintained. This patient group would then be expected to have a higher stroke rate because of preoperative risk. The relatively healthy patients would then be the patients whose MAP is allowed to fall below 50 mm Hg. However, because the statistical significance of our findings persisted in spite of adjustment for SRI, this explanation seems unlikely. To further test this hypothesis, we separated our patient population into risk quartiles on the basis of SRI (n = 708 to 731 patients per quartile). Within each risk quartile, we divided the patients into low MAP or high MAP groups on the basis of the median value for MAP < 50 in the entire population (86 mm Hg · min). This resulted in a nearly even distribution (±5%) of patients within each risk quartile to the low or high MAP group. We then calculated the stroke rate for the low and high MAP groups within each risk quartile. As can be seen in Figure 2, the trend toward higher stroke rates in patients with normal to high MAP persisted in each of the three highest risk quartiles. This also argues against treatment bias as the explanation for our findings.
The second possible explanation for the inverse relationship between MAP and rate of stroke or coma is that the process of neurologic injury may alter the patient’s physiology to produce systemic HTN. If inadequate cerebral blood flow develops because of emboli during CPB, and cerebral autoregulation is incapable of maintaining adequate blood flow to the ischemic region, the patient’s response may be one of systemic vasoconstriction in an attempt to increase perfusion pressure to the ischemic region. In this scenario, one may speculate that elevated MAP would be a marker for neurologic injury, but no causative relationship would exist. The true risk factor for neurologic injury may be the extent of arteriosclerosis in the ascending aorta, which is unrelated to perfusion pressure while undergoing CPB.
A third possible explanation for our findings may be the use of pharmacologic therapy to artificially raise MAP during CPB. Phenylephrine, which is the most common 
-adrenergic agent used during CPB in our institution, is a relatively pure 1 agonist. Because cerebral autoregulation is maintained during CPB, the use of vasoactive medication may alter the natural interaction between perfusion pressure and cerebral blood flow, resulting in cerebral ischemia. This hypothesis would seem contrary to the findings of Rogers and coworkers [20] on the effect of phenylephrine on cerebral blood flow. Their study used small groups of patients (n = 6 per group) to identify differences in cerebral blood flow between groups managed with -stat versus pH-stat blood gas management. Phenylephrine was used to raise MAP in all patients with the finding that cerebral autoregulation was not maintained during pH-stat management. However, numbers were insufficient to conclude that phenylephrine did not alter cerebral blood flow, nor did the authors draw this conclusion. If phenylephrine does alter cerebral blood flow, well-intentioned pharmacologic therapy may actually worsen neurologic outcome. Further study is needed to assess the impact of vasoconstrictive therapy on cerebral blood flow.
Obvious limitations exist in this study. First, this study is observational. However, the anesthesia information system contains data entered in real time that are not subject to manipulation bias. Therefore, this study should not be considered to be purely retrospective in design. However, the hypothesis was generated retrospectively. The data presented in this study provide association only and do not imply causation. Only a well-designed, randomized, prospective clinical trial with adequate predictive power could suggest causation. Because the questions posed in this study were conceived after data had been entered, tighter control of intraoperative variables by protocol was not possible.
Although the data reflect actual practice behaviors without manipulation bias, the variables ranged widely, increasing the possibility of missed correlations that are actually significant. The sensitivity analysis determined that a missing covariate with an effect equal to the strongest variable in our model would not change our results. Second, it demonstrated that this missing covariate would have to increase the likelihood of both high MAP and adverse neurologic outcome 15-fold to reverse the beneficial effect of low MAP (decrease the relative risk of high MAP from 2.31 to 0.8). This missing covariate would also have had to escape the notice of experienced clinicians and researchers as well as be independent of the variables already identified.
The outcome variable is fortunately an uncommon occurrence in this patient group. With 36 patients experiencing perioperative stroke or coma, it would take only 18 missed events to change the stroke and coma rate by 50% (1.3% to 2.0%). Even though stroke or coma at time of discharge is a relatively clear end point, it is less definite than the end point of mortality used in part 1 of this study, and it introduces more uncertainty into the results. The incidence of stroke in our population mirrors other similar experiences [7, 19, 21, 22]. Because the initial input of outcome data and neurologic complications were performed by individuals blinded to the intraoperative management, investigator bias is extremely unlikely. The inclusion of preoperative and postoperative examination data by a neurologist would have undoubtedly increased the number of neurologic events identified and added to the power of the study, inasmuch as it is well known that the number of observed events is related to the sensitivity of the examination. The absence of documented TIA events is likely related to the method of data collection in the surgical database, which tended to document events present at the time of death or discharge. Therefore, the database may miss complications such as a resolved TIA.
Transesophageal echocardiography and epiaortic scanning were not used routinely during CABG procedures at the Duke Heart Center during the period of study. This would have provided additional information about the contribution of aortic atheroma to perioperative stroke had it been available. Although the SRI is heavily weighted toward factors that contribute to generalized atherosclerosis and would be expected to correlate closely with the extent of ascending aortic atheroma, direct assessment of the extent of aortic atherosclerosis by transesophageal echocardiography may allow stratification of the patient population into subgroups that would benefit from perfusion pressure manipulation during CPB. The data in this study support the conclusion that MAP less than 50 mm Hg during CPB does not result in major neurologic injury in the cardiac surgical patient population as a whole, but do not exclude the possibility that a subset of patients may be identified in the future who may benefit from elevated MAP during CPB.
The most significant limiting factor of this study is that not all types of adverse neurologic outcome could be included as end points. Because the surgical database did not universally include end points such as incidence of encephalopathy, new-onset seizure, or neurocognitive dysfunction, these less definitive markers of central nervous system injury could not be studied. The study by Gold and coworkers [5] found no significant difference in cognitive or functional outcome in patients with high versus low MAP during CPB. However, when studied in a subset of patients older than 58 years, an association between low MAP during CPB and cognitive decline was suggested [23]. The incidence of neurocognitive dysfunction after CPB approaches 75% [24]. Although not considered to be as severe an adverse outcome by clinicians as stroke or coma, persistent neurocognitive decline after CPB can be devastating to the patient and family. Future studies can hopefully address the more subtle end points of neurologic injury.
In conclusion, we were unable to demonstrate a significant association between the intraoperative variables of MINHCT or MAXGLC and the incidence of stroke or coma. The SRI did accurately predict the development of stroke or coma. When adjusted for SRI score and CPB duration, a significant inverse association remained between MAP during CPB and incidence of stroke or coma. Even considering the limitations of this study, the data collected provide strong evidence that MAP less than 50 mm Hg during CPB does not correlate with increased rate of stroke or coma. This study also provides evidence that global hypoperfusion secondary to inadequate cerebral perfusion pressure is unlikely to be the primary cause of neurologic injury after cardiac surgery, further demonstrating the need for a large-scale, multicenter, randomized study to definitively identify the causes of adverse neurologic outcome in this large patient population.
|
References |
|---|
|
TopAbstractIntroductionMaterial and methodsResultsCommentReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. Puskas, H. P. Grocott, W. D. White, J. P. Mathew, M. F. Newman, and S. Bar-Yosef Intraoperative Hyperglycemia and Cognitive Decline After CABG Ann. Thorac. Surg., November 1, 2007; 84(5): 1467 - 1473. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Schoof, W. Lubahn, M. Baeumer, R. Kross, C.-W. Wallesch, A. Kozian, C. Huth, and M. Goertler Impaired cerebral autoregulation distal to carotid stenosis/occlusion is associated with increased risk of stroke at cardiac surgery with cardiopulmonary bypass J. Thorac. Cardiovasc. Surg., September 1, 2007; 134(3): 690 - 696. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Selim Perioperative Stroke N. Engl. J. Med., February 15, 2007; 356(7): 706 - 713. [Full Text] [PDF]
|
|
|
|
|
|
G. J. Murphy and G. D. Angelini Indications for Blood Transfusion in Cardiac Surgery Ann. Thorac. Surg., December 1, 2006; 82(6): 2323 - 2334. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. W. Hogue Jr, C. A. Palin, and J. E. Arrowsmith Cardiopulmonary bypass management and neurologic outcomes: an evidence-based appraisal of current practices. Anesth. Analg., July 1, 2006; 103(1): 21 - 37. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. A. Baker, L. J. Hallsworth, and J. L. Knight Stroke After Coronary Artery Bypass Grafting Ann. Thorac. Surg., November 1, 2005; 80(5): 1746 - 1750. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Karkouti, G. Djaiani, M. A. Borger, W. S. Beattie, L. Fedorko, D. Wijeysundera, J. Ivanov, and J. Karski Low Hematocrit During Cardiopulmonary Bypass is Associated With Increased Risk of Perioperative Stroke in Cardiac Surgery Ann. Thorac. Surg., October 1, 2005; 80(4): 1381 - 1387. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. A. Jonas, D. Wypij, S. J. Roth, D. C. Bellinger, K. J. Visconti, A. J. du Plessis, H. Goodkin, P. C. Laussen, D. M. Farrell, J. Bartlett, et al. The influence of hemodilution on outcome after hypothermic cardiopulmonary bypass: results of a randomized trial in infants J. Thorac. Cardiovasc. Surg., December 1, 2003; 126(6): 1765 - 1774. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Dworschak, M. Czerny, M. Grimm, G. Grubhofer, and W. Plochl The impact of asymptomatic carotid artery disease on the intraoperative course of coronary artery bypass surgery Perfusion, January 1, 2003; 18(1): 15 - 18. [Abstract] [PDF]
|
|
|
|
 |
|
 L. K. Ti, J. P. Mathew, G. B. Mackensen, H. P. Grocott, W. D. White, J. G. Reves, and M. F. Newman Effect of Apolipoprotein E Genotype on Cerebral Autoregulation During Cardiopulmonary Bypass Stroke, July 1, 2001; 32(7): 1514 - 1519. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
