Author + information
- Received June 17, 2003
- Revision received September 16, 2003
- Accepted September 23, 2003
- Published online February 18, 2004.
- Michael J. Racz, MA*,∥∥,
- Edward L. Hannan, PhD*,* (, )
- O.Wayne Isom, MD†,
- Valavanur A. Subramanian, MD‡,
- Robert H. Jones, MD, FACC§,
- Jeffrey P. Gold, MD∥,
- Thomas J. Ryan, MD, FACC¶,
- Alan Hartman, MD#,
- Alfred T. Culliford, MD**,
- Edward Bennett, MD††,
- Robert A. Lancey, MD‡‡ and
- Eric A. Rose, MD§§
- ↵*Reprint requests and correspondence:
Dr. Edward L. Hannan, Professor and Chair, Dept. of Health Policy, Management, and Behavior, School of Public Health, State University of New York, University at Albany, One University Place, Rensselaer, New York 12144-3456, USA.
Objectives This study was designed to compare in-hospital mortality and complications and three-year mortality and revascularization for off-pump and on-pump coronary artery bypass graft (CABG) surgery after adjusting for patient risk.
Background The use of off-pump CABG surgery has increased tremendously in recent years, but little is known about its long-term outcomes relative to on-pump CABG surgery, and most studies have been very small.
Methods Short- and long-term outcomes (inpatient mortality and complications, three-year risk-adjusted mortality, and mortality/revascularization) were explored for patients who underwent off-pump CABG surgery (9,135 patients) and on-pump CABG surgery (59,044 patients) with median sternotomy from 1997 to 2000 in the state of New York.
Results Risk-adjusted inpatient mortality was 2.02% for off-pump versus 2.16% for on-pump (p = 0.390). Off-pump patients had lower rates of perioperative stroke (1.6% vs. 2.0%, p = 0.003) and bleeding requiring reoperation (1.6% vs. 2.2%, p < 0.001) and higher rates of gastrointestinal bleeding, perforation, or infarction (1.2% vs. 0.9%, p = 0.003). Off-pump patients had lower postoperative lengths of stay (median 5 days vs. 6 days, p < 0.001). On-pump patients had higher three-year survival (adjusted risk ratio [RR] =1.086, p = 0.045) and higher freedom from death or revascularization (adjusted RR = 1.232, p < 0.001). When analyses were limited to 1999 to 2000, the two-year adjusted hazard ratio for survival was not significant (adjusted RR = 0.99, p = 0.81).
Conclusions On-pump patients experience better long-term survival and freedom from revascularization than off-pump patients. However, the survival benefit from on-pump procedures was no longer present in the last two years of the study.
Although the first coronary artery bypass graft (CABG) procedures in the 1960s were performed on a beating heart, refinement of the cardiopulmonary bypass pump led to the widespread use of this technology for CABG surgery beginning in 1968. However, various complications of CABG surgery, including adverse cerebral outcomes (1,2), perioperative renal dysfunction (3,4), myocardial dysfunction (5), and the systemic inflammatory response (6)attributed to the cardiopulmonary bypass pump in the past few years have caused numerous surgeons to re-examine the safety and efficacy of CABG surgery performed on a beating heart (7–35).Please verify all reference numbers in reference list and text are correct.> Off-pump CABG surgery has been safely performed for many years in two centers in South America (7,8), and this early success has been reconfirmed by more recent experience in The Netherlands (9–11). The major question that remains about off-pump surgery is whether the need to contend with heart motion and more blood in the operative field compromises the quality of distal coronary graft anastamoses and results in a less durable or less complete revascularization.
The purpose of this study is to investigate and compare short-term outcomes (in-hospital mortality and complications) and long-term outcomes (three-year mortality and the need for repeat revascularization) between all patients undergoing on-pump CABG surgery with sternotomy and all patients undergoing off-pump CABG surgery with sternotomy in the state of New York from 1997 to 2000 whose long-term outcomes could be tracked.
The two primary databases used for the study are the Cardiac Surgery Reporting System (CSRS) for the state of New York and New York's vital statistics death file. The CSRS contains numerous demographic variables; patients' clinical risk factors and complications; dates of admission, surgery, and discharge; and discharge status on every patient undergoing CABG surgery in the state of New York. The 34 hospitals with cardiac surgery programs that are responsible for coding the CSRS forms are trained to capture relevant information. When data fields are found to be missing, hospitals are contacted and asked to complete the missing information. Also, comprehensive audits of approximately half of the hospitals in the CSRS are conducted each year to ensure coding accuracy. The vital statistics file identifies all residents of the state who die each year. Another database employed in the study was New York's percutaneous coronary intervention (PCI) registry, which was used to identify subsequent PCIs after CABG surgery.
Study group and end points
The total number of patients undergoing CABG surgery who were discharged between January 1, 1997, and December 31, 2000, in the 34 hospitals in New York certified to perform the procedures was 73,113. This group was limited by excluding non-New York residents (n = 2,106) and patients who underwent CABG surgery without sternotomy (n = 1,828). The remaining patients (59,044 with on-pump and 9,135 with off-pump CABG surgery) were included in the study.
End points included inpatient mortality, long-term (three-year) survival, and long-term (three-year) survival and freedom from subsequent revascularization (CABG or PCI). Deaths during the same admission as the procedure were identified using CSRS, and deaths after discharge after the procedure were identified using New York's vital statistics file. The time of occurrence of revascularization was derived from data in the CSRS and the New York PCI registries. Patients who did not die or require revascularization after the initial CABG surgery were censored at the termination of follow-up on December 31, 2000.
The prevalence of risk factors known to be associated with mortality were calculated for the two procedure categories. These variables included the number of diseased vessels; patient age, gender, race, and ethnicity; a variety of comorbidities; and measures of the patient's hemodynamic state and ventricular function. All variables, including age and ejection fraction, were separated into categories, and for each variable, differences between the distributions of categories between the two types of procedures were tested using Fisher exact tests for all 2 × 2 comparisons and chi-square tests for variables with more than two categories.
To compare risk-adjusted inpatient mortality rates for off-pump surgery versus conventional CABG, a stepwise logistic regression procedure was developed using the LOGISTIC procedure in SAS, version 8.2 (SAS Institute, Cary, North Carolina). Discharge status from the hospital after the procedure, with in-hospital death coded as “1,” was used as the binary dependent variable. Candidates for the independent variables included all the demographic and clinical variables available in CSRS. After the logistic regression model was developed, risk-adjusted mortality rates (observed/expected mortality rate ratios multiplied by the statewide mortality rate) were tested for significant differences between off-pump and on-pump patients.
Three-year survival and three-year absence of mortality and repeat revascularization were examined while controlling for differences in patient severity of illness, using stepwise Cox proportional hazards models and the SAS procedure for proportional hazards regression, PHREG. Off-pump surgery was coded as a binary independent variable. Other variables used as candidates in the two models included age, gender, ethnicity, race, ejection fraction, previous myocardial infarction (MI), number of diseased vessels, and numerous comorbidities. Ninety-five percent confidence intervals for the logarithm of the adjusted hazard ratios (HRs) were calculated to test for significant differences in outcomes between the two procedures. This was done for all patients and for seven selected subgroups of patients chosen on the basis of demographic and comorbid characteristics that were associated with a greater chance of undergoing off-pump surgery than other risk factors.
Cox proportional hazard models were also used to construct adjusted survival curves for the procedures, where the type of procedure was used as a stratification factor. The analyses were then repeated with time until revascularization (CABG or PCI) or death used as the dependent variable.
To control for selection bias, a propensity model was developed to find significant predictors of off-pump surgery versus on-pump surgery (36,37). The propensity score for each patient was obtained by fitting a logistic regression model with a binary dependent variable representing off-pump surgery. Independent variables consisted of all the demographic, risk factor, and coronary anatomy measures available in CSRS. The propensity score, ranging from 0 to 1, was subdivided into quintiles. Three-year survival rates for off-pump and on-pump surgery were compared within each quintile. Differences in survival curves for the two groups within each quintile were tested with the log-rank test.
Table 1presents the prevalence of each available risk factor for patients undergoing off-pump and on-pump CABG surgery. In general, patients who underwent off-pump surgery were older and sicker than those undergoing conventional bypass. The off-pump group had a significantly higher prevalence rate of older patients, women, African Americans, and patients with lower ejection fractions, one or more previous open-heart operations, stroke, carotid/cerebrovascular disease, aortoiliac disease, electrocardiographic evidence of left ventricular hypertrophy, congestive heart failure, extensively calcified ascending aorta, hepatic failure, and renal failure. The on-pump group had significantly higher percentages of patients with an MI within 24 h, cardiopulmonary resuscitation, and left main coronary artery disease. Also, on average, they had more diseased coronary vessels.
Table 2presents the unadjusted and adjusted inpatient mortality rates for on-pump and off-pump procedures. Although the unadjusted (observed) inpatient mortality rates (2.48% for off-pump and 2.09% for on-pump surgery) were significantly different (p = 0.017), the risk-adjusted rates (2.02% for off-pump surgery and 2.16% for on-pump) were not significantly different (p = 0.390). Also, 2.03% of off-pump patients were converted to on-pump procedures from 1998 to 2000, the only years in which this information was available.
Table 3indicates that off-pump patients had significantly lower rates for two perioperative complications: stroke (1.6% vs. 2.0%, p = 0.003) and bleeding requiring reoperation (1.6% vs. 2.2%, p < 0.001). Off-pump patients also had a significantly higher rate for gastrointestinal bleeding, perforation, or infarction (1.2% vs. 0.9%, p = 0.003). It should also be noted that off-pump patients had significantly lower postoperative lengths of stay (median 5 days vs. 6 days for on-pump patients, p < 0.0001).
Figure 1presents the logarithms of the adjusted HRs for survival for off-pump to on-pump surgery. These ratios are provided for all patients and for seven subgroups of patients. The logarithm of the HR for off-pump to on-pump surgery for all patients is 0.082, which translates by exponentiation into a HR of 1.086 (p = 0.045). Thus, after adjusting for differences in patient risk factors, patients undergoing on-pump surgery were 1.086 times as likely to be alive at any point in time as patients undergoing off-pump surgery.
Of the seven patient subgroups investigated in Figure 1, only patients with extensively calcified ascending aorta demonstrated significant survival differences by type of surgery. The logarithm for the HR for these patients was 0.249, which translates into a HR of 1.282 (p = 0.009), denoting a higher survival rate for the on-pump group.
Figure 1also presents the logarithms of the adjusted HRs for freedom from death and revascularization for all patients and for the same subgroups of patients. The logarithm of the HR for on-pump:off-pump surgery for all patients is 0.209, which translates into a HR of 1.232 (p < 0.001). Of the seven patient subgroups investigated, two (extensively calcified ascending aorta and renal failure) had significant HRs, denoting fewer adverse events for the on-pump patients. All seven subgroups had HRs >1.
Figure 2demonstrates three-year adjusted survival for patients undergoing off-pump and on-pump surgery and three-year adjusted freedom from death and revascularization for patients undergoing the two types of surgery. As indicated, patients undergoing on-pump surgery were more likely to survive (89.5% vs. 88.8%) and were more likely to be free from death and revascularization (84.7% vs. 82.1%) at the end of three years.
In the propensity analyses, Kaplan-Meier survival curves were computed for both procedural groups within quintiles of predicted probabilities of off-pump surgery, with I referring to the quintile within which the probability of a patient receiving off-pump surgery is the smallest. For patients in quintiles I, II, IV, and V, the three-year survival rates were all higher for on-pump surgery, and they were significantly higher for quintiles II and IV. For all propensity quintiles for three-year survival and freedom from revascularization, rates were higher for on-pump surgery, and the differences were significant for the first four quintiles. In conclusion, because the advantage of on-pump surgery was not limited to groups of patients with very low probabilities of undergoing off-pump surgery, there is no evidence of major selection bias, and the results stated above are not altered by the propensity analyses.
Off-pump surgery with sternotomy has become increasingly popular as an option for CABG surgery patients who have traditionally undergone surgery with a pump oxygenator. In New York, the number of isolated CABG surgery patients undergoing off-pump surgery with sternotomy has risen from 573 (3.0% of all cases) in 1997 to 4,539 (27% of all cases) in 2000.
The purposes of our study were to determine differences in patient characteristics, short-term outcomes, and long-term outcomes between among all 9,135 patients undergoing off-pump CABG surgery with sternotomy and all 59,044 patients undergoing on-pump CABG surgery with sternotomy in New York in the time period from 1997 to 2000. We found that off-pump patients were more likely to be older and women, had somewhat lower ejection fractions, and had higher prevalences of comorbidities such as stroke, carotid/cerebrovascular disease, congestive heart failure, and renal failure. They were also less likely to have left main disease and had fewer diseased coronary arteries. The observed in-hospital mortality rates were 2.48% for off-pump CABG patients and 2.09% for on-pump CABG patients, and this difference was statistically significant (p = 0.02). However, after risk adjustment, the rates were no longer significantly different (2.02% vs. 2.16%, respectively, p = 0.39). Other studies, including one randomized controlled trial (RCT) (9)and a large non-RCT (18), have also found no significant difference in short-term mortality, although two large non-RCTs found that off-pump patients had significantly lower risk-adjusted operative mortality (30,31).
Off-pump patients in our study had significantly lower rates for two perioperative complications: stroke (1.6% vs. 2.0%, p = 0.003) and bleeding requiring reoperation (1.6% vs. 2.2%, p < 0.001). The only reasonably large RCT to examine stroke rates found no difference between on-pump and off-pump patients (9), and the results from non-RCTs are mixed, with some studies having found no significant differences (32)and others having found lower rates in off-pump patients (31). Significantly lower postoperative bleeding, although not necessarily related to the need for reoperation, has been identified in RCTs (9,16)and in a few non-RCTs (33–35). A large non-RCT found no significant difference in return to the operating room for bleeding (32).
Our study also found that off-pump patients had significantly higher rates for one complication: gastrointestinal bleeding, perforation, or infarction (1.2% vs. 0.9%, p = 0.003), and that off-pump patients had significantly lower postoperative lengths of stay (median 5 days vs. 6 days for on-pump patients, p < 0.0001).
With respect to longer term outcomes, the adjusted HR for survival within three years for on-pump to off-pump surgery was 1.086 (p = 0.045), meaning that on-pump patients were 1.086 times as likely to be alive at any time in the period as off-pump patients. The adjusted HR for freedom from death or revascularization for on-pump to off-pump surgery was 1.232 (p < 0.001). Patients undergoing on-pump surgery were significantly more likely to survive (89.6% vs. 88.8%, p = 0.022) and were significantly more likely to be free from death and revascularization (84.7% vs. 82.1%, p < 0.0001) at the end of three years. These findings are the most important ones in the study because although other studies found no long-term differences in adverse outcomes, they were underpowered RCTs (10,11,13).
An advantage of our study is that it is very large, with more than 9,000 off-pump patients and more than 59,000 on-pump patients. Also, the results are based on recent data, and long-term outcomes were examined for the two groups.
However, there are caveats to the study. We were able to capture deaths after discharge only for patients who continued to reside in the state of New York. Consequently we limited the study to New York residents who underwent CABG surgery or PCI, but if these patients moved out of state and died or underwent revascularizations we would have missed those events in the long-term analyses. It is expected that these omissions would occur proportionately among patients undergoing off-pump and on-pump surgery so that no large biases would occur. Also, in a study conducted several years ago (1991 to 1992) with Medicare data that were compared to CSRS and CARS data from the same time period to identify deaths missed because of patients moving out of state, there was no significant difference in deaths lost by type of intervention, and the number of out-of-state deaths lost was minimal (88.4% to 87.6% = 0.8% missing).
Another possible drawback of the study is that it is an observational study, not a randomized controlled trial that has the advantage of minimizing selection bias. However, observational studies have the advantage that they are representative of what has actually occurred in practice, are more amenable to generalization, and are better powered to identify significant differences in outcomes (38,39). Furthermore, there is evidence to the effect that when randomized controlled trials and observational studies have been used to compare two or more treatments or interventions, the treatment effects are not qualitatively different for the two types of studies, and that the observational studies do not overestimate the magnitude of the effects of treatment compared to randomized controlled trials (40,41). Also, our propensity analyses did not suggest any substantial selection bias. Nevertheless, surgical selection bias related to the unavailability of surgeons to perform off-pump surgery could have been a factor in the results reported here.
An important caveat of the study is that the frequent use of off-pump surgery in New York is relatively new, so that although this study reflects all off-pump and on-pump procedures performed with sternotomy in 1997 through 2000, the majority of the off-pump procedures (7,476 of the 9,135, or 81.8%) in the four-year period were performed in 1999 and 2000, whereas only 43.2% of the on-pump procedures were performed in the last two years. Mechanical stabilization for off-pump CABG was introduced in late 1997 and perfected in early 1999. Cardiac positioning devices for access to posterior coronary target vessels were introduced only in early 2000. Although censoring in the proportional hazards analyses accounts for the differences mentioned earlier, it is still possible that a bias is introduced. When the analyses were limited to 1999 to 2000 and two-year freedom from death or repeat revascularization was examined, the HR for survival for on-pump to off-pump patients decreased to less than one and was not significant (HR = 0.99, p = 0.81), and when the analyses were limited to hospital/years with at least 40% of all CABG surgery patients undergoing off-pump procedures, the HR for survival dropped even further to 0.97, p = 0.71.
However, the respective HRs for freedom from death or revascularization for on-pump to off-pump surgery were still significant (HR =1.17, p = 0.0004 and HR = 1.24, p = 0.0002, respectively). It should be noted that the lower revascularization rates for on-pump patients are likely to be a result of some combination of better graft patency (superior anastomoses) and more complete revascularization, but that our database does not enable us to determine the extent of the contribution of these two factors.
It is possible that because off-pump surgery is more technically demanding, off-pump surgery outcomes may improve relative to on-pump surgery in the future as more surgeons master the learning curve for off-pump surgery and as methods for stabilization of the heart during off-pump surgery improve. Also, it should be noted that the results presented here may not be reflective of regions in which off-pump surgery has experienced earlier practice penetration. It will be critically important to monitor and compare future outcomes for on-pump and off-pump surgery throughout the world.
The authors thank Kenneth Shine, MD, the Chair of the Cardiac Advisory Committee (CAC) for the state of New York, and the remainder of the CAC for their encouragement and support of this study; and Donna Doran, Casey Joseph, Rosemary Lombardo, and the cardiac surgery departments and cardiac catheterization laboratories of the 34 participating hospitals for their tireless efforts to ensure the timeliness, completeness, and accuracy of the registry data.
☆ This work is supported in part by the New York State Department of Health. Dr. Subramanian is a Scientific Advisor for Guidant/CTS, Inc.
- coronary artery bypass graft
- Cardiac Surgery Reporting System
- hazard ratio
- myocardial infarction
- percutaneous coronary intervention
- randomized controlled trial
- Received June 17, 2003.
- Revision received September 16, 2003.
- Accepted September 23, 2003.
- American College of Cardiology Foundation
- Van Dijk D.,
- Nierich A.P.,
- Jansen E.W.L.,
- et al.
- Ascione R.,
- Caputo M.,
- Calori G.,
- et al.
- Jansen E.W.,
- Gründeman P.F.,
- Borst C.,
- et al.
- Borst C.,
- Jansen E.W.,
- Tulleken C.A.,
- et al.
- Ascione R.,
- Lloyd C.T.,
- Gomes W.J.,
- et al.
- Rosenbaum P.R.,
- Rubin D.B.