| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2003;76:428-434
© 2003 The Society of Thoracic Surgeons
a Departments of Medicine, Surgery, and Community and Family Medicine, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire, USA
b Department of Surgery, Fletcher Allen Health Care, Burlington, Vermont, USA
c Department of Surgery, Beth Israel Deaconness Medical Center, Boston, Massachusetts, USA
d Department of Neurology, Beth Israel Deaconness Medical Center, Boston, Massachusetts, USA
e Department of Surgery, Catholic Medical Center, Manchester, New Hampshire, USA
f Department of Surgery, Maine Medical Center, Portland, Maine, USA
g Department of Surgery, Eastern Maine Medical Center, Bangor, Maine USA
Accepted for publication March 4, 2003.
* Address reprint requests to Dr Likosky, Clinical Research Section, Department of Medicine, Dartmouth-Hitchcock Medical Center, Lebanon, NH 03756, USA.
e-mail: donald.s.likosky{at}dartmouth.edu
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
METHODS: We conducted a regional prospective study of 11,825 consecutive patients undergoing coronary artery bypass graft surgery surgery from 1996 to 2001. Data were collected on patient and disease characteristics, intra- and postoperative care and course, and outcomes. Stroke was defined as "a new focal neurologic deficit which appears and is still at least partially evident more than 24 hours after its onset." Logistic regression identified significant predictors of stroke.
RESULTS: The incidence of stroke was 1.5%. The regression model significantly predicted the occurrence of stroke. As compared with cardiopulmonary bypass for less than 90 minutes, cardiopulmonary bypass for 90 to 113 minutes, odds ratio = 1.59, p = 0.022), cardiopulmonary bypass for 114 minutes or more (odds ratio = 2.36, p < 0.001), atrial fibrillation (odds ratio = 1.82, p < 0.001), and prolonged inotrope use (odds ratio = 2.59, p = 0.001) significantly improved our ability to predict stroke. Nearly 75% of all strokes occurred among the 90% of patients at low or medium preoperative risk.
CONCLUSIONS: The inclusion of factors associated with intra- and postoperative care and course significantly improved the prediction model. Most strokes occurred among patients at low or medium preoperative risk, suggesting that many of these strokes may be preventable. Reduction in stroke risk may require modifications in intra- and postoperative care and course.
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
Although some advances have been made in understanding the contribution of preoperative factors to a patient’s risk of stroke, much work lies ahead in understanding the added risk attributable to intra- and postoperative care and course, such as duration of cardiopulmonary bypass and atrial fibrillation. Few published reports focus on the association between a patient’s risk of stroke and intra- and postoperative factors while accounting for patient and disease characteristics. Although most researchers would agree that a patient’s risk of stroke might increase after surgery, few if any studies have tested this hypothesis. A revised probability of stroke would be useful for clinicians, to enable a better understanding of how intra- and postoperative factors contribute to a patient’s risk of stroke. Many studies focusing on stroke in the setting of CABG surgery are too small to detect associations reliably. This study of nearly 12,000 patients quantifies the association between intra- and postoperative care and course and the risk for developing a stroke, by using a preoperative stroke risk prediction model to adjust for a patient’s preoperative risk.
We identified and quantified the association between intra- and postoperative factors and a patient’s subsequent risk of stroke after isolated CABG surgery in northern New England, while adjusting for that patient’s preoperative risk. The goal of this study was to quantify the contribution of intra- and postoperative factors to a patient’s risk for developing a stroke secondary to CABG surgery.
|
Patients and methods |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
This study takes advantage of the work conducted by the Northern New England Cardiovascular Disease Study Group (NNECDSG). The NNECDSG is a voluntary regional consortium representing all medical centers in Maine, New Hampshire, Vermont, and one medical center in Massachusetts that perform CABG surgery. Since its inception in 1987, the NNECDSG has maintained prospective surgical registries on all patients undergoing CABG, heart valve replacement, and percutaneous coronary interventions. The group fosters continuous improvement in the quality of care for patients undergoing these procedures through pooling of process and outcome data, as well as timely feedback of this data to clinicians through group meetings and regional, center, and surgeon-specific reports [10].
Previous publications by the NNECDSG have discussed in detail our data collection methodology and definitions [11]. We have prospectively collected the following intra- or postoperative outcome variables: use of an intraaortic balloon pump (IABP) [1], length of cross-clamping, length of cardiopulmonary bypass, route of cardioplegia, type of pericardiotomy, use of two or more inotropic agents at 48 hours, status at hospital discharge (dead or alive), atrial fibrillation, low cardiac output syndrome (return to cardiopulmonary bypass, use of more than two inotropic agents at 48 hours, or the use of an intra- or postoperative IABP), and reoperation for bleeding (yes or no).
We have looked at the directionality of cause and effect between postoperative atrial fibrillation and onset of stroke among a subset of our CABG patients. Among 75 patients having stroke secondary to CABG surgery between 1992 and 2000, 56 (75%) patients had atrial fibrillation on the same day or before the onset of their stroke.
Estimates of pre- and postoperative risk were divided into terciles based on the following percentile cutpoints: low 0 to 49th, medium 50 to 89th, and high 90 to 100th.
We defined a stroke as a new neurologic deficit which appears and is still at least partially evident more than 24 hours after its onset, occurring during or following the CABG procedure and established before discharge.
Statistical analysis
Logistic regression was used to analyze the association between patient and disease characteristics, as well as intra- and postoperative factors, and the occurrence of stroke, using the STATA 7.0 statistical software package [12]. Logistic regression allows the analyst to model the log-odds of a dichotomous (yes/no) event’s occurrence.
Univariate analyses were conducted to test the association between intra- and postoperative factors and a patient’s risk of stroke. Variables found to be significantly associated with stroke were tested in a multivariate logistic regression model. Nested models were compared to ascertain whether added variables significantly improved the model’s performance. The revised model contained the estimated preoperative risk of stroke as well as intra- and postoperative variables.
The performance of the revised model was assessed in several ways. First, a receiver operating characteristic curve (ROC) was constructed. The ROC area may vary between 0.5 and 1.0. A useless test would have an area of 0.5, ie, a positive test would be as likely to be a false-positive as a true-positive at every threshold. A perfect test would have an area of 1.0. The ROC area is equivalent to a C-statistic [13]. Second, a likelihood ratio test (
2LR) was used to compare nested models. The parameterization of the revised model was determined through bivariate plots of the observed versus expected number of strokes in each decile of risk.
Following development of this revised multivariate prediction model, we calculated each patient’s predicted risk of stroke, and internally validated the model using bootstrapping techniques [14]. Bootstrapping allows the internal validation of a prediction model, by taking a large number of samples with replacement from the original dataset. Bootstrapping provides a nearly unbiased estimate of the predictive accuracy of the model [15]. We developed the multivariate model on the entire data set, and randomly drew with replacement 200 samples of 100%. We calculated the ROC area and standard error of the mean for each sample.
We assessed the calibration of the prediction model by applying it to the individual patients in the data set, and compared the observed and expected numbers of strokes by decile of predicted risk. We calculated the Lemeshow-Hosmer goodness of fit statistics. We used the relative contribution of each of the intra- and postoperative care and event variables to the prediction of risk of stroke as a measure of the total 2 uniquely associated with each factor [16].
|
Results |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
2 mg/dl, and vascular disease.
Assessment of the association between intra- and postoperative factors and a patient’s risk of stroke was determined through the calculation of odds ratios and 2 tests; results from our uni- and multivariate analyses are summarized below.
Incidence of stroke
There were 177 strokes in the dataset of 11,825 (1.5%) patients undergoing isolated CABG surgery.
Intra- and postoperative factors
Table 1
summarizes the association between stroke and intra- and postoperative factors. The use of cold cardioplegia, intra- or postoperative IABP, longer duration of cardiopulmonary bypass, return to cardiopulmonary bypass after initial separation, low cardiac output syndrome, prolonged inotrope use, and atrial fibrillation were all significantly associated with an increased risk of stroke. Patients having cold cardioplegia relative to other strategies had a 52% decreased risk of stroke (odds ratio [OR] = 0.48, p = 0.05). Patients undergoing an intra- or postoperative IABP insertion had a threefold increase in risk of stroke (OR = 3.0, p < 0.001). Patients in the highest quartile of cardiopulmonary bypass duration had a threefold increase in risk of stroke versus those in the lowest quartile (OR = 3.0, p < 0.001 for p trend). Patients returned to cardiopulmonary bypass after initial separation had twofold increased risk of stroke (OR = 2.3, p = 0.00). Patients with low cardiac output syndrome had nearly a threefold increased risk (OR = 3.1, p < 0.001), whereas patients needing prolonged inotropic agents had a fivefold increase in risk (OR = 5.0, p < 0.001). Patients developing atrial fibrillation had a twofold increase in risk of stroke (OR = 2.3, p < 0.001).
|
2 [6 d.f.] = 128.84, p < 0.001) predicted the occurrence of stroke in this data set. This model was then compared to the full model through the likelihood ratio test (2 [3 d.f.] = 4.27, p = 0.2335). The discrimination did not change appreciably between the full and revised models (full model = 0.73, revised model = 0.73).
|
2 [7 d.f.] = 11.19, p = 0.1303), indicating that there was no statistically significant departure from a perfect fit (Fig 1).
|
The inclusion of intra- and postoperative information significantly improved our model’s performance (2LR [4 d.f.] = 53.06, p < 0.001). Table 2 summarizes the odds ratios and coefficients for the revised model.
The relative importance of each intra- or postoperative factor in predicting the risk of stroke is summarized in Figure 2. Estimated preoperative risk of stroke contributed 52.9% of the predictive ability of the regression model, whereas cardiopulmonary bypass duration contributed 23.1%. Prolonged inotrope use contributed 10.1%, whereas atrial fibrillation contributed 14.0% of the predictive ability of the model.
|
|
|
|
Comment |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
2 [7 d.f.] = 11.19, p = 0.1303). This prediction model had a larger mean ROC area than did a model using only preoperative patient and disease characteristics, suggesting that the inclusion of intra- and postoperative variables increased our ability to discriminate between individuals with and without strokes. Both the absolute and relative risk of stroke increased with category of preoperative risk. Of the 177 strokes, nearly 75% occurred among those at low or medium preoperative risk. Both the preoperative and revised risks of stroke, although to differing degrees, were greatest among those having prolonged inotrope use and atrial fibrillation, followed by prolonged inotrope use, longer durations of cardiopulmonary bypass and postoperative atrial fibrillation. Although our dataset is drawn from the experience of six medical centers in a single region of the United States, it contains information from nearly 12,000 isolated CABG surgeries. It is likely that our findings are representative of the experience at other medical centers around the country, as the NNECDSG definition for stroke is quite similar to the one used by the Society for Thoracic Surgeons in their CABG registry [17].
There are two limitations to the present study. First, we did not have data on the precise nature and timing of the neurologic deficits. This information will likely be important to future work, since global encephalopathies and focal deficits may have different etiologies [18]. Future use of biological markers for cerebral damage might aid in the identification of the timing of stroke. Some of the strokes may have preceded the intra- or postoperative factors that we have identified. However, any attributable effect would bias our findings toward the null hypothesis. Second, we did not have data on the extent of carotid and aortic disease [19, 20]. However, work conducted by Birkmeyer and colleagues [21] suggests that a diagnosis of vascular disease provides information regarding the extent of a patient’s carotid and lower extremity disease. He showed that patients with peripheral vascular disease had a twofold increase in mortality compared to patients without peripheral vascular disease. In addition, patients with cerebrovascular disease alone, and those with lower-extremity occlusive disease alone, had similar mortality risks. We have found that of those patients with vascular disease, most had cerebrovascular disease, followed by lower extremity disease. One may surmise from pathophysiology that patients with vasculopathy regardless of location have an increased risk of stroke.
To our knowledge, there have been no studies to date that have adjusted postoperative stroke risk for a patient’s prior probability of stroke. The current analysis is based on methodology stemming from work done by L’Italien and colleagues [22] in the area of vascular surgery. In our current study, more than 90% of patients at low risk preoperatively remained at low risk after surgery, whereas 9% were at medium risk and the remaining at high risk. Nearly 55% of patients at medium risk preoperatively remained at medium risk postoperatively, whereas 34% were at low risk and the remaining 10% of patients were classified into high risk. More than half of patients at high risk preoperatively remained at high risk subsequent to surgery, whereas the remaining patients were classified into medium risk.
Previous work has explored the associations between intra- and postoperative factors and the risk of stroke. Blossom and colleagues [1] conducted a medical record review of patients undergoing CABG surgery over a 3-year period. They found that among the 3,428 patients undergoing CABG surgery, 1.3% had a stroke. These investigators categorized strokes using information regarding mental status at 24 and 48 hours, neurologic evaluation, and computed tomographic scanning of the head. Using this information, strokes were classified as either embolic or hypoperfusion. Blossom and colleagues made associations between the occurrence of strokes and a variety of intraoperative factors. Among the patients categorized as having an intraoperative stroke (35%), 6 of 12 patients had a hypotensive event (mean arterial blood pressure 
60 mm Hg for 10 minutes). Hypotensive events may occur for a variety of reasons, such as inadequate delivery of oxygen to tissues due to diminished cardiac output. This clinical picture has been termed low cardiac output syndrome. In the present study, we have used prolonged inotrope use as a proxy for low cardiac output syndrome.
Chugh and associates [23] reviewed evidence regarding risk of stroke secondary to the onset of atrial fibrillation. Atrial fibrillation may be due to a dilated or structurally abnormal atrium or ventricle, which may also lead to stasis or thrombus formation. Stasis may cause a hypercoagulable state, and thrombi may break off from the atrium, embolize to the brain, and cause a stroke. Chugh and associates noted that atrial fibrillation may be a marker for other risk factors such as an atherosclerotic aorta, cerebrovascular disease, or mitral annular calcification. These investigators reported that stroke risk is four- to fivefold higher in patients with atrial fibrillation.
Gardner and colleagues [24], in a case-control study of patients with strokes after isolated CABG surgery, identified duration of cardiopulmonary bypass as a significant predictor of stroke. In their article, cases were defined as patients with strokes consequent to CABG surgery. Controls were selected as the 2 patients preceding the stroke victim. These investigators compared the rates and means of different variables between the cases and controls. Variables that were found to be significantly different in univariate analysis included age, history of cerebrovascular disease, severe arteriosclerosis of the ascending aorta, duration of cardiopulmonary bypass, and severe perioperative hypotension.
Frye and colleagues [25], in a multicenter randomized control study, also identified duration of cardiopulmonary bypass as a significant predictor of stroke after CABG surgery. This article arose from the Coronary Artery Surgery Study (CASS) trial, in which patients were randomized to either medical therapy or CABG surgery. In-hospital stroke was defined as "transient or residual central nervous system deficit." Significant variables in the model of Frye and colleagues included age, noncardiac surgery, duration of cardiopulmonary bypass, left atrium, or pulmonary vein as the vent site, angina, use of 
-adrenergic agents after bypass, other surgery besides myocardial resection or ventricular muscle plication surgery, and the number of grafts performed. Duration of cardiopulmonary bypass was categorized on the basis of durations of 100 minutes or less, 101 to 200 minutes, and more than 200 minutes. Frye found a four-and-one-half–fold increased risk of stroke for patients in the highest tercile of cardiopulmonary bypass duration versus the referent lowest tercile.
Implications
Multiple regression analysis with revision for intra- and postoperative variables is a useful tool for determining the contribution of these factors toward the occurrence of stroke. The model allows one to compare rates of strokes across different levels of pre- and postoperative risk. This is a useful feature as we strive to tailor care to a patient’s level of risk.
Most patients were at low and medium risk. These patients accounted for nearly three quarters of all strokes. The greatest benefit may occur by identifying mechanisms for preventing strokes among these patients. Information from this study may have a direct impact on clinical care. We have provided a model for stratifying a patient’s care and course based on a patient’s preoperative risk. This information may help to guide care specific to a patient’s level of risk. For instance, a portion of all strokes may be avoided through implementation of some of the following: (1) use of echocardiography to identify areas free of aortic disease for cannulation and cross-clamping; (2) prevention and management of postoperative atrial fibrillation; (3) and developing strategies for separating patients from the cardiopulmonary bypass machine.
|
Acknowledgments |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
|
References |
|---|
|
TopAbstractIntroductionPatients and methodsResultsCommentAcknowledgmentsReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 |
