| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Thorac Cardiovasc Surg 2002;124:698-707
© 2002 The American Association for Thoracic Surgery
Surgery for Acquired Cardiovascular Disease (ACD) |
From the Departments of Thoracic and Cardiovascular Surgery and Biostatistics and Epidemiology, The Cleveland Clinic Foundation, Cleveland, Ohio.
Partial funding of nursing personnel for data collection of the off-pump cases was provided by Guidant Corporation, which otherwise had no input into the analyses and did not review or endorse the results reported in this study.
Read at the Eighty-first Annual Meeting of The American Association for Thoracic Surgery, San Diego, Calif, May 6-9, 2001.
Received for publication May 15, 2001. Revisions requested July 31, 2001; revisions received Oct 12, 2001. Accepted for publication Oct 24, 2001. Address for reprints: Joseph F. Sabik, MD, The Department of Thoracic and Cardiovascular Surgery, The Cleveland Clinic Foundation, 9500 Euclid Ave, Desk F25, Cleveland, OH 44195 (E-mail: sabikj{at}ccf.org).
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
Cardiac surgeons have readopted off-pump coronary artery bypass surgery to decrease the morbidity and mortality of myocardial revascularization. However, does off-pump surgery truly reduce morbidity and mortality? Some clinical studies appear to indicate that it does.
1-6 This issue remains unresolved, however, because characteristics of on- and off-pump patient groups are dissimilar in these studies. At present, no randomized studies address this issue.
To reduce the effect of selection bias, we have compared outcomes of on- and off-pump coronary artery bypass surgery in propensity-matched pairs. Using this technique, we analyzed patients who were equally likely to be treated with on- or off-pump surgery and excluded patients from the analysis if there was a high likelihood of using only one of the revascularization strategies.
|
Patients and methods |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
All patients included in the study underwent revascularization through a median sternotomy. Patients undergoing coronary reoperations, combined operations, or coronary revascularization through a thoracotomy were excluded. However, patients who were converted from off-pump to on-pump bypass were retained (n = 15); thus, this represents an intent-to-treat study.
Methods
Surgical technique: Off pump
All operations were performed through a median sternotomy. Conduits were harvested and prepared, and patients were heparinized. Patients were placed in the Trendelenburg position and rotated to the right to assist in coronary artery exposure and prevent hemodynamic instability during construction of the distal anastomoses.
Our technique of coronary artery exposure evolved during this study. Early in our experience, 4 slings (2 through the transverse sinus and 2 behind the inferior vena cava) were manipulated to expose the coronary arteries. For most of the study, 2 to 4 deep pericardial sutures were placed in the posterior pericardium, at the level of the left superior and inferior pulmonary veins, and between the left inferior pulmonary vein and inferior vena cava. Traction on these sutures elevated and rotated the heart to the right and exposed the lateral, posterior, and inferior coronary arteries.
Commercially available mechanical stabilizers were used to perform distal anastomoses. During construction of these anastomoses, target vessel hemostasis was obtained with intravascular shunts, a proximal loop, or proximal and distal vessel loops. A humidified carbon dioxide blower was used.
Distal anastomoses were usually constructed before proximal anastomoses. The left internal thoracic artery to left anterior descending coronary artery anastomosis was constructed first. Proximal anastomoses were constructed to the aorta with a tangential clamp. After completion of all anastomoses, heparin was reversed, and aspirin was given 8 hours postoperatively.
Surgical technique: On pump
All operations were performed through a median sternotomy. Conduits were harvested and prepared, and patients were heparinized. Standard cannulation for CPB was performed with ascending aortic cannulation and dual-stage cannulation of the right atrium. Both antegrade and retrograde cardioplegia cannulas were placed.
Although there were some minor differences in CPB routine among surgeons, for the most part, the procedure was conducted as follows. Arterial flow was adjusted to 2.0 to 2.2 L · min-1 · m-2, and blood pressure was maintained between 50 and 70 mm Hg. The aorta was crossclamped, and cold blood cardioplegia was administered either antegrade or retrograde or in combination on induction and every 15 minutes. Patients' temperatures were either kept warm or allowed to drift to 34°C during CPB.
Distal anastomoses were usually constructed first, and the proximal anastomoses were constructed to the ascending aorta during a single crossclamp period, thus eliminating the need for tangential and multiple clampings of the aorta. After the patient was weaned from CPB and decannulated, heparin was reversed, and aspirin was administered 8 hours postoperatively.
Data collection
Patients were identified from the prospective computerized Cardiovascular Information Registry, and preoperative, operative, and postoperative variables (Appendix Table 1) were retrieved from this database and 2 other computerized databases whose use for research has been approved by The Cleveland Clinic Foundation Institutional Review Board.
Identifying the influence of selection
Patients operated on off pump differed in many clinical characteristics from on-pump patients. Off-pump patients were older than on-pump patients (66 ± 11.3 vs 64 ± 10.4 years, P < .001), more likely to be women (32% vs 25%, P = .001), less likely to have insulin-treated diabetes (10% vs 14%, P = .05), and more likely to have carotid artery disease (22% vs 14%, P < .001), peripheral vascular disease (18% vs 13%, P = .009), hypertension (75% vs 70%, P = .03), and renal disease (3.7% vs 1.1%, P < .001). They were less symptomatic (in New York Heart Association class IV: 17% vs 29%, P < .001) and had less extensive coronary artery disease (3-system disease: 53% vs 72%, P < .001).
To reduce the influence of selection on the comparison of outcome, we used propensity-score pairwise matching of off-pump patients to the pool of available on-pump patients. 7,8 This was a 2-step process. First, using known preoperative variables (Appendix Table 1) and multivariable logistic regression, we determined the factors associated with on- versus off-pump bypass (group membership). We initially analyzed the continuous and ordinal variables by means of decile analysis, selecting those transformations of scale that best calibrated the variable to group membership.
Variable selection, with a P value criterion for retention of variables in the model of .05, used bootstrap bagging. 9 This was a 4-step process. First, a patient was randomly selected from the original data set to begin a new data set. The original data set continued to be sampled until the new data set was 75% of the size of the original. Second, risk factors were identified with automated forward stepwise selection. Third, results of the variable selection were stored. These 3 steps were repeated 200 times. Finally, the frequency of occurrence of variables related to group membership was ascertained and indicated the reliability of each variable. 9 In addition, a guided analysis was performed of the original data set. 10 All variables with bootstrap reliability of 50% or greater were also in the guided analysis, but 2 variables identified by guided analysis, history of hypertension and insulin-treated diabetes, appeared in only 40% and 42% of bootstrap samples, respectively.
Having established a parsimonious model (Appendix Table 2), we added other variables representing groups of patient and coronary disease factors that might be related to unrecorded selection factors (saturated model). These variables were sex, body mass index to represent patient size variables, left ventricular function, history of arrhythmias (atrial fibrillation, ventricular fibrillation, and ventricular tachycardia), smoking history, peripheral vascular disease, chronic obstructive pulmonary disease, previous stroke, left main disease, and left anterior descending coronary artery disease. The C statistic was .72.
A propensity score was calculated for each patient by solving the saturated model for the probability of off-pump bypass. Using only the propensity score, on-pump patients were matched to off-pump patients by using a greedy matching strategy. 11 For this, matching to 5 decimal points was initially performed, followed by 4-, 3-, 2-, and 1-decimal-point matching. Off-pump patients whose propensity scores deviated more than 0.10 from those of on-pump patients were considered unmatched. This yielded 406 off-pump patients propensity matched to 406 on-pump patients.
Outcomes
In-hospital outcomes were recorded by respiratory therapists dedicated to recording postoperative complications occurring in the intensive care unit and by experienced nurses also dedicated to collecting patient information, operative details, and in-hospital complications. Operative details were collected within 48 hours of the operation.
Operative variables compared included bypass conduits, number and site of distal anastomoses, and completeness of revascularization. Incomplete revascularization was defined as lack of grafting into any system with 50% or more lesions or failure to graft into both the left anterior descending and circumflex systems for 50% or greater left main trunk disease.
Postoperative events, identified according to definitions of the Society of Thoracic Surgeons for Version 2.35 of the Adult Cardiac Surgery Database (see National Databases for details, www.sts.orq) included hospital death, stroke, encephalopathy (as determined by a staff neurologist), myocardial infarction, reoperation for bleeding, red blood cell use, sternal wound infection, renal failure requiring dialysis, acute respiratory distress syndrome, pancreatitis, and gastrointestinal bleeding.
Direct technical cost data, exclusive of professional salaries, were obtained from the Transaction Systems Incorporated cost accounting system (Eclipsys Corporation, Boston, Mass). Because of Cleveland Clinic policy, financial data are presented by relative statistics only.
Comparison of outcomes
Outcomes were compared between the 2 propensity-matched groups of patients. We also compared the outcomes in 75 unmatched off-pump patients with those matched.
These comparisons were supplemented by 2 other methods of using the propensity score: quintile substratification analysis and multivariable adjustment. The results were entirely consistent with the matched-pairs analysis and are not shown.
|
Results |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
.05) among 70 examined.
|
|
|
|
|
|
|
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
However, in recent years, the deleterious effects of CPB and aortic manipulation have become well known. Therefore, surgeons have attempted to reduce even further the morbidity of cardiac operations by turning once again to off-pump coronary artery bypass. The renewed interest in off-pump myocardial revascularization has generated technological advances that have improved methods of coronary artery stabilization and exposure, allowing access to all coronary arteries and complete revascularization.
Challenges of comparison
Off-pump coronary artery bypass surgery is now possible, but is it better? Multiple clinical studies have examined this question without arriving at a consensus. The problems are twofold. First, both on- and off-pump strategies yield good outcomes in experienced centers. Second, patient selection has had a large confounding effect on comparisons. The best way to eliminate bias is to perform a randomized trial. However, in this setting, with very low occurrence of postoperative events with on-pump coronary artery bypass surgery, a randomized study would require thousands of patients.
The standards of randomized trials can be achieved partially with clinical studies such as this one, by matching patients to obtain comparable groups. In this study we used propensity-score matching. 7,8,11
Principal findings
Death, stroke, and myocardial infarction
Hospital mortality, stroke, and myocardial infarction were similar in the propensity-matched pairs. The risk of each of these events with both revascularization strategies was low, about 1% (Appendix Table 3). Other clinical studies have likewise demonstrated no clear superiority in terms of mortality, stroke, or myocardial infarction when comparing on- and off-pump surgical strategies. 12-16
In unmatched off-pump patients, mortality was higher. This group of patients was selected for off-pump surgery because they were believed to be too high risk for CPB. Compared with on-pump patients, these patients were either very old or very young, had a large body mass index, had more preoperative renal failure, had a more recent operation, and had less extensive coronary artery disease. We believe risk factors accounted for higher mortality in these patients.
Encephalopathy, red cell transfusion, sternal wound infection, and dialysis
There was more postoperative morbidity in the on-pump patients. Encephalopathy, red blood cell transfusions, sternal wound infections, and renal failure requiring dialysis were all more common. We believe these morbidities in the on-pump patients can be attributed directly to CPB.
Neurocognitive impairment after CPB, reflected in this study by encephalopathy, is believed to be due to microemboli, nonphysiologic perfusion, and the inflammatory response. 17 These result in neuronal injury, increased permeability of the blood-brain barrier, and cerebral edema. 18 Studies have demonstrated increased levels of the biochemical marker for neuronal injury, S100ß, after on-pump surgery, with no significant increase after off-pump surgery. 19,20 However, the long-term importance of the neurocognitive changes and S100ß elevations after CPB is unknown. Some studies suggest that these neurocognitive deficits might be transient and resolve several months after the operation, 17,21-23 whereas a recent study by Newman and colleagues 24 suggests that patients exhibiting neurocognitive dysfunction after CPB are at risk for long-term neurocognitive dysfunction.
Our study might have underestimated the neurocognitive benefit of off-pump surgery. We measured encephalopathy, a clinically apparent event. Because much of the neurocognitive impairment after CPB is subtle, extensive preoperative and postoperative neurocognitive testing is required to detect a postoperative decrement. A limitation of our study is that these tests were not performed.
Findings of increased red blood cell transfusions in on-pump versus off-pump patients, similar to our own results, were recently reported by Ascione and coworkers. 25 They demonstrated increased red blood cell loss and higher red blood cell, platelet, and fresh frozen plasma transfusions in on-pump patients. CPB can lead to increased red blood cell transfusion requirement in multiple ways. Surface activation of platelets and coagulation proteins during CPB results in platelet dysfunction and coagulopathy, leading to postoperative bleeding. 26,27 The crystalloid solutions needed to prime the pump result in hemodilution, and the turbulence, cavitation, and osmotic stresses during CPB result in red blood cell membrane injury and hemolysis.
Renal injury during CPB might be due to altered renal perfusion during periods of hypotension or low flow, vasoconstrictors, or microemboli. Hemoglobinuria, as a result of hemolysis during CPB, might also lead to renal dysfunction.
CPB has multiple effects depressing the immunologic system. Complement activation resulting in complement factor depletion, lymphopenia, and natural killer cell impairment and pulmonary sequestrations of polymorphonuclear leukocytes and neutropenia have been shown to occur after CPB. 28,29 These might result in increased susceptibility to infections and explain our finding of increased sternal wound infections in on-pump patients.
Incomplete revascularization
A worrisome finding in our study was more incomplete revascularization in the off-pump patients. When the reasons for incomplete revascularization were given, they were similar in patients undergoing on- and off-pump bypass. The difference in incomplete revascularization occurred primarily in patients in whom the reason could not be determined, and only rarely for hemodynamic instability unique to off-pump revascularization. Although incomplete revascularization did not increase early risk, it might affect the long-term results of off-pump revascularization. 30-32 Arom and colleagues 16 found that at 1 year after the operation, off-pump patients had a higher risk of recurrent angina and reinterventional procedures compared with on-pump patients. Similar to our findings, the off-pump patients in their study received fewer bypass grafts than did the on-pump patients.
Cost
Total direct technical costs were increased in the off-pump patients. Much of this increase occurred in the operating room as a result of longer procedures and disposable stabilizers. Perfusion costs, as expected, were less in the off-pump patients, but this did not compensate for the increase in other operating room costs. There were only minor differences in intensive care unit and floor costs.
Does Off-pump surgery reduce mortality and morbidity?
The results of both on- and off-pump myocardial revascularization are excellent, with low risk of death, stroke, and myocardial infarction. The procedures, however, are not equivalent and appear to have different risks. On-pump surgery is associated with higher morbidity, which can be attributed directly to CPB. Thus, an advantage of off-pump revascularization is lower postoperative morbidity. However, there is more incomplete revascularization in off-pump surgery, and this might be a future disadvantage.
|
|
|
|
Appendix: Discussion |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
I think this is the best attempt so far to compare off-pump with on- pump coronary artery bypass grafting, and you have done a very nice job, short, of course, of randomization, which is yet to be done.
I did want to point out one thing. Your abstract is a declarative sentence, but the manuscript that you sent me and your introductory slide presented a question, and the data are slightly different. Could you clarify this when you discuss these questions?
I read all 55 pages of your manuscript, including the 15-page appendix. Obviously Dr Blackstone's touch is here, and like every surgeon that presents a new technique, every couple of years Dr Blackstone presents to us a new concept in statistical treatment, which obviously adds greatly to our knowledge. Propensity matching is excellent because it more carefully allows us to match patients.
Overall, in this 5-way analysis, off-pump grafting shows underrevascularization. In the old days, the 1980s and 1990s, we thought that complete revascularization was a bedrock of coronary bypass. Do you have a sense of how this is going to play out, or are these patients generally at such high risk that that is not really going to become a factor in the long term because these patients might not survive for other reasons?
With your analysis, you did show that the early events—encephalopathy, renal failure, wound infection, red cell transfusions—are better in the off-pump group. Encephalopathy is a very controversial area. Some of us took exception to a recent article in The New England Journal of Medicine because it was not matched according to age. Do you have a concept that the encephalopathy you saw in both groups was correlated with age? At Brigham, we noticed a remarkable decrease in encephalopathy and stroke, especially in patients older than 70 years of age, when we reverted to the single-clamp technique for coronary bypass about 10 years ago.
You noted that the cost is 13% greater with off-pump bypass grafting. Unlike other studies that purport to say that, you can do an off-pump graft and the patient goes home the next day, and you had exactly the same intensive care unit and hospital stay. You mentioned the costs being higher with off-pump grafting. Can you explain why that might be?
Finally, and you did not have a chance to go into it because it is such a complete statistical treatment of these data, in the highest risk quintiles of the propensity matching, the fourth and fifth quintiles, the operative mortality in the off-pump group is higher than in those patients who were treated with coronary artery bypass grafting on bypass. This raises an interesting paradox. On our service, we often use off-pump grafting in the highest risk patients, trying to decrease these encephalopathic findings, renal failure, et cetera, but are we trading a slightly lower incidence of encephalopathy for higher operative risk?
Finally, on the basis of these data, should we expand our indications for off-pump grafting, or should we continue to restrict our indications to the highest risk patients?
Dr Sabik. Thank you, Dr Cohn, for your kind and thoughtful remarks. The reason that the abstract we submitted is a little different from this presentation and the paper is that we slightly changed the patient population and statistical methods of the study. We believe these changes improved this study by improving our ability to match the off-pump and on-pump patients. The first change we made was to remove patients undergoing reoperations. We believe it is important to match the on- and off-pump patients in terms of the extent of coronary artery disease. Patients who are having coronary reoperations have both patent vein grafts and diseased native vessels, and this makes it difficult to classify them as to whether they have single-, double-, or triple-vessel disease. Therefore, to match the patients in regard to the extent of coronary artery disease, those who underwent reoperations were removed.
Our second change was to rematch the patients using a new statistical method involving propensity scores. This technique, called greedy matching, was not known to us at the time we submitted this abstract. With this technique, propensity scores are used to do one to one matching. We were able to improve on the matching of the on-pump and off-pump patients and obtain 2 groups of patients with very similar preoperative characteristics. Despite having changed the methods of the study slightly, I do not believe that our conclusions have changed significantly. Mortality and the major morbidities of stroke and myocardial infarction remained similar in the on-pump and off-pump patients. Our belief, from the conclusions of the first abstract, was that the benefits of off-pump surgery were related to the elimination of the transient effects associated with CPB. By improving the matching, I believe we were better able to demonstrate this. The complications that can be related directly to CPB, such as coagulopathy, immunosuppression, and brain edema, led to more morbidity in the on-pump patients, such as more red blood cell transfusions, more sternal wound infections, and more encephalopathy. Therefore, we believe that by improving the matching, we were able to strengthen the conclusions of our original abstract. Because we were able to identify more morbidity in the on-pump patients, we changed the title from a declarative statement to a question.
I do not know how the increased occurrence of incomplete revascularization in the off-pump patients will play out over time. In this study incomplete revascularization did not increase the early risk in the off-pump patients, but it might increase the late risk. If the risk of CPB is greater than the risk associated with incomplete revascularization in high-risk patients, then off-pump surgery with incomplete revascularization might be beneficial in these patients. We will have to follow up these patients and determine the effect, if any, of incomplete revascularization over time.
In the propensity-matched groups there were too few patients with the postoperative complication of encephalopathy to determine its preoperative risk factors. However, in the entire group of on-pump patients, there were enough individuals with postoperative encephalopathy to determine some preoperative risk factors. These include high preoperative creatinine, preoperative atrial fibrillation, and high New York Heart Association class; age was not among them.
We were surprised by the increased cost in the off-pump patients. Most of this increase occurred in the operating room. In the off-pump patients most of the costs associated with perfusion are eliminated. However, in our experience this did not compensate for the increased costs in the off-pump patients because of the disposable equipment necessary for off-pump bypass and because off-pump operations took longer.
When we compared the off-pump and on-pump patients by propensity-matched quintiles, mortality was higher in the off-pump patients in the fifth quintile. In the fifth quintile the off-pump patients were very dissimilar from the on-pump patients, and in this quintile are located the high-risk patients who underwent off-pump surgery because the risk of on-pump surgery was believed to be too high. We believe these high-risk off-pump patients account for the increased mortality in the off-pump patients in the fifth quintile.
|
Acknowledgments |
|---|
|
References |
|---|
|
TopAbstractIntroductionPatients and methodsResultsDiscussionAppendix: DiscussionReferences |
|---|
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 |
