Patient Selection and Current Practice Strategy for Off-pump Coronary Artery Bypass Surgery
Objective— Previous studies comparing off-pump coronary artery bypass surgery (OPCABG) to conventional techniques utilizing cardiopulmonary bypass (CABG-CPB) have failed to provide patient selection guidelines. We sought to determine guidelines, attempting to rectify the limitations of previous studies.
Methods and Results— A retrospective analysis of prospectively collected data from the Society of Thoracic Surgeons National Database, from January 1999 through December 2000, identified 204 602 multivessel coronary artery bypass (CABG) patients. Unadjusted and risk-adjusted odds ratios (OR) were calculated to compare OPCABG and CABG-CPB morbidity and mortality. A propensity model was developed to identify factors associated with selection for OPCABG. All off-pump patients were pair-matched with on-pump patients based on their propensity to receive an off-pump procedure. Off-pump patients, 8.8% of the total, had significantly different patient characteristics than the on-pump group. Characteristics associated with OPCABG selection included fewer diseased vessels, absence of left main disease, fewer bypass grafts, no previous CABG, older age, chronic lung disease, and renal failure. Unadjusted and risk-adjusted odds ratios indicate a significant off-pump survival benefit and decreased morbidity including stroke and renal failure in the overall group. Propensity matching also showed a significant OPCABG survival benefit [OR (95% CI) 0.83 (0.72, 0.96)]. Subgroup analysis of propensity-matched groups identified off-pump survival benefits in patients with previous CABG (OR=0.53), diabetics (OR=0.66), LVEF between 30% to 50% (OR=0.75), females (OR=0.79), and age 66 to 75 years (OR=0.80).
Conclusion— OPCABG imparts some survival benefit to most patient subgroups. Higher risk patients including those undergoing reoperative CABG, diabetics, and the elderly may gain the most benefit.
Localized stabilization of the beating heart allowed for the limited successful introduction of off-pump coronary artery bypass surgery (OPCABG) prior to widespread adoption of cardiopulmonary bypass (CPB) in the late 1960s.1 The still, bloodless field provided by cardioplegic arrest of the heart combined with the technological advances of cardiopulmonary bypass around the same time, overshadowed the off-pump approach to become the dominant modality in coronary surgery.2,3 Resurgence in beating heart surgery began in the early 1990s in an attempt to decrease the morbidity associated with CABG without jeopardizing benefits, and was spurred by the observed benefits of avoiding CPB and its associated deleterious effects.4–6 Early development of off-pump coronary artery bypass was hindered by crude instrumentation, as well as limited exposure through small incisions. Technological advancements have significantly facilitated the performance of beating heart surgery through a sternotomy in the past 8 years. Although numerous reports have demonstrated the safety and efficacy of off-pump coronary artery bypass and the associated benefits of avoiding CPB, large, randomized trials comparing the two techniques have been logistically difficult due to the reluctance of patients, their referring physicians, and surgeons to participate because of the perceived benefits of off-pump surgery.7–10
Through a large retrospective study utilizing prospectively collected data from the Society of Thoracic Surgery database, we sought first to analyze the contemporary use of off-pump surgery in patients undergoing multivessel coronary artery bypass. Second, we sought to determine the benefits in terms of mortality and morbidity associated with beating heart techniques and avoidance of CPB. Finally, we examined subsets of patients most likely to benefit from off-pump surgery.
The STS National Cardiac Database
The STS National Cardiac Database (NCD) was established as a voluntary database in 1989 for the purpose of outcomes assessment following adult cardiac surgery.11 STS NCD data are harvested semiannually from a majority of U.S. hospitals performing open heart surgery. Clinical patient data are entered at the sites using uniform definitions and certified software systems. Data includes 30-day, but no longer term, postoperative follow-up. Data quality standards must be met before a local dataset can be entered into the aggregate national dataset. Data are warehoused at the Duke Clinical Research Institute in Durham, NC, which produces semi-annual site-specific reports to STS participants for outcomes analysis and quality improvement efforts.
A retrospective review of prospectively collected data voluntarily submitted to the Society of Thoracic Surgeons (STS) National Cardiac Database identified 234 145 patients that underwent isolated coronary artery bypass (CABG) surgery from January 1999 through December 2000. All single vessel CABG procedures as well as any CABG combined with another cardiac surgical procedure, such as valve replacement, were excluded. The total number of multivessel isolated CABG procedures with complete data suitable for analysis was 204 602 of which 91.22% (186 663/204 602) were performed in a conventional fashion with cardiopulmonary bypass (CABG-CPB), whereas 8.78% (17969/204602) were performed without cardiopulmonary bypass (OPCABG). For purposes of this study, an OPCABG procedure was defined based on a reported cardiopulmonary bypass time of zero, and use of cardioplegia as “no”. Selection criteria utilized in determining which CABG procedure individual patients would receive, with or without CPB, was neither specified nor obtainable from the database. Treatment assignment of individual patients was at the discretion of each operating surgeon and varied accordingly by patient, surgeon, and institution; however medical comorbidities that increase the risks of CPB are often used to select patients for OPCABG. The CABG-CPB and OPCABG patients were contemporaneous and not sequential cohorts.
Patient data were collected and analyzed according to The Society of Thoracic Surgeons (STS) National Cardiac Database guidelines and definitions (http://www.ctsnet.org/doc/2167). Outcomes definitions in the STS NCD include operative mortality (determined as death within 30 days of surgery, regardless of location (in-hospital or out of hospital) or death in-hospital regardless of length of stay), stroke (new-onset CVA persisting >72 hours); renal failure (acute postoperative renal insufficiency with one or more of: (1) increase in serum creatinine >2.0; (2) 50% or greater increase in creatinine over baseline preoperative value; (3) new requirement for dialysis); re-operation (reexploration for any reason including bleeding, graft occlusion, other cardiac problem, other noncardiac problem), and prolonged ventilation (pulmonary insufficiency requiring ventilatory support for 48 hours or more). Preoperative data included diabetes, chronic pulmonary disease (COPD), prior cerebrovascular event (CVA), peripheral vascular disease (PVD), obesity, hypertension, hypercholesterolemia, family history of CAD, renal failure, renal failure on dialysis, current smoking, prior myocardial infarction, preoperative IABP use, preoperative cardiogenic shock, left main coronary disease, age, gender, left ventricular function, and prior CABG
Patients were grouped and compared according to surgical treatment, OPCABG versus CABG-CPB. First, preoperative patient characteristics and individual risk factors, intraoperative course, and operative outcomes including mortality were compared. Data are reported as a percentage for categorical variables or as the median and first and third quartile for continuous variables. Categorical variables were compared using a chi-square test and comparisons of continuous data were done using the Wilcoxon rank sum test. All tests are two-sided. probability values less than or equal to 0.05 were considered significant.
Unadjusted odds ratios (OR) and confidence intervals (CI) to compare off-pump OPCABG and on-pump CABG-CPB procedures were calculated for several outcome measures including operative mortality and four morbidities. Patients were also divided into three clinically meaningful risk groups based on the current STS mortality model, and odds ratios and confidence intervals were calculated for mortality within these risk groups.
To control for selection bias as a result of nonrandom treatment assignment, two methods were used: risk-adjustment and propensity analysis. For risk-adjustment, a hierarchical mixed-effects logistic regression model was used to determine the effect of off-pump surgery after simultaneously adjusting for site as well as up to 27 preoperative patient risk factors contained in the STS CABG mortality and morbidity models.
For the propensity analysis, a hierarchical mixed-effects logistic regression model, which included both site and patient factors, was used to create OPCABG-selection propensity scores (ie, probability of receiving an off-pump procedure). Patients whose status were classified as emergent or salvage were excluded from the propensity-matched analysis. After these exclusions, the 16,937 off-pump patients were all pair-matched (in a 1:1 ratio) to the on-pump patient with the most similar propensity to receive an off-pump procedure. Propensity scores were required to be within 0.01 of each other to be considered a match. This propensity score analysis allows for the comparison of outcomes between groups that have a similar likelihood of receiving off-pump surgery. Once patients were matched, conditional logistic regression was then used to determine the overall effect of off-pump surgery for the resulting 16 937 matched pairs. Subgroups were selected for further analyses based on results from previous studies of selected high risk patient populations. Odds ratios and confidence intervals for subgroups were calculated by adding interaction terms to the logistic regression model.
Overall, the on-pump and off-pump groups were widely disparate. Patients receiving off-pump surgery had significantly more single and double vessel rather than triple vessel coronary disease, less left main disease, less recent myocardial infarction, less pre-operative shock and intra-aortic balloon pump use, less emergent status, and less diabetes. The off-pump patient group also were older and more often female, had more class III and IV heart failure, more chronic lung disease, more history of stroke and cerebral vascular disease, renal failure, peripheral vascular disease, and previous percutaneous interventions (PTCA) (Table 1).
Characteristics associated with off-pump selection in the hierarchical logistic regression model included fewer diseased vessels, fewer bypass grafts, no prior cardiac surgery, older age, absence of left main coronary disease, chronic lung disease (COPD), and renal failure (Table 2).
Unadjusted odds ratios indicate an off-pump survival benefit in the overall group [OR=0.82, 95%CI=0.74, 0.91] that persisted and was more pronounced after risk adjustment [OR=0.76, 95%CI=0.68, 0.84]. Risk stratification of the entire group into 4 mortality risk groups indicates an increasing survival benefit of off-pump surgery as predicted risk increases (Table 3). Off-pump patients also had less morbidity that was more pronounced with risk-adjustment including a 40% reduction in risk of stroke, a 20% reduced risk of post-operative renal failure, a 40% relative risk reduction in the incidence of prolonged ventilation, and a 30% relative risk reduction in reoperation (Table 4).
There was a 100% match on propensity score between the off-pump and on-pump patients. Patient characteristics for the propensity-matched pairs were similar (Table 5). The survival benefit of off-pump surgery based on the conditional logistic regression model using the propensity-matched pairs was significant with a computed odds ratio of 0.833 (95% CI (0.724, 0.957)). Subgroup analysis of propensity-matched groups identified off-pump survival benefits in many patient groups including those with previous CABG [OR 0.53 (CI 0.32 to 0.87)], diabetics [0.66 (0.52 to 0.85)], LVEF between 30% to 50% [0.75 (0.59 to 0.95)], females [0.79 (0.64 to 0.99), and age 66 to 75 years [0.80 (0.70 to 0.93)] (Figure 1).
This analysis represents the largest contemporary multicenter comparison of outcomes in multivessel coronary artery bypass patients performed with and without cardiopulmonary bypass. As shown in previous studies, early operative mortality without adjusting for risk was less in the off-pump group compared with the on-pump group, 2.4% versus 2.9%. Morbidity was also decreased in the off-pump group including decreased incidence of stroke, renal failure, prolonged ventilation, and reoperation. The significance of these findings is limited by the markedly different patient characteristics in the off-pump and on-pump treatment groups as expected in any large nonrandomized comparison, however, risk adjustment helps to offset some of these differences. The mortality and morbidity benefits of off-pump surgery persisted with risk-adjustment and in addition were more significant.
We also identified those specific patient characteristics that influenced selection of patients for off-pump surgery. The factors associated with OPCAB selection included among others: no prior cardiac surgery, fewer numbers of diseased vessels and corresponding need for fewer bypass grafts, absence of left main coronary disease, absence of pre-operative intra-aortic balloon pump (IABP), older age, presence of chronic lung disease (COPD), and renal failure. These characteristics were used to develop propensity scores to computer match off-pump patients with patients in the on-pump group who had equivalent likelihood to receive off-pump surgery, thereby minimizing differences between the two treatment groups inherent in nonrandom selection. The resulting propensity-matched pairs represent comparable patient populations with similar characteristics for comparison of treatment. The survival benefit of off-pump surgery in the propensity-matched analysis was similar to that found in other studies (OR=0.83).
A prime objective of this study was to identify subgroups of patients who, based on projected improved outcomes, should be preferentially selected for off-pump or on-pump coronary artery bypass surgery. Previous studies have shown that high-risk coronary artery bypass patients may particularly benefit from avoiding cardiopulmonary bypass.12–15 Risk stratification into four risk groups based on predicted operative mortality revealed a significant survival benefit of off-pump surgery as predicted risk of mortality increased. Lower-risk patients as a group also appear to benefit from off-pump surgery but this comparison lacked statistical significance because of wide confidence intervals. This may be a result of greater diversity within the low-risk groups representing some patient characteristics that predispose benefit and others that do not. Note that no risk group had significantly better survival or trended toward better outcomes with on-pump surgery.
Subgroup analysis of the propensity-matched pairs, in further attempts to identify specific patients who might derive greater benefit from off-pump or on-pump surgery, revealed no group that failed to benefit from off-pump surgery, with the exception perhaps of dialysis patients. Because of the relatively small number of dialysis patients in this study, the benefit of off-pump surgery in this high-risk subgroup is inconclusive.
Patients that receive the most survival benefit from off-pump surgery include those with previous CABG (OR=0.53), diabetics (OR=0.66), mild to moderate left ventricular dysfunction (OR=0.75), females (OR=0.79), and age 66 to 75 years (OR=0.80).
The small number of patients in the off-pump group (8.8% of the total), relative to the on-pump group may be an additional limitation of this study, although any impact this may have on the conclusions is not apparent. A larger cohort of off-pump patients may be associated with different selection criteria that may in turn alter the propensity analysis. Factors not taken into account in this retrospective analysis due to limitations of the database and that may impact conclusions include (1) variations in coronary anatomy such as calcified, intramyocardial, or small coronary arteries, (2) variations in individual surgeon skill and experience, (3) long-term patency and event-free survival. These questions can only be answered in a large, multicenter, prospective, randomized trial with long-term follow-up. The limited adoption of off-pump surgery, despite apparent benefits as demonstrated clearly in this large, contemporary, retrospective, multicenter comparison with sophisticated methods to control for nonrandom treatment selection bias, suggests the need and timeliness for such a study.