Author + information
- Joanna Chikwe, MDa,b,∗ (, )@IcahnMountSinai@SUNY,
- Timothy Lee, MDa,
- Shinobu Itagaki, MD, MSca,
- David H. Adams, MDa and
- Natalia N. Egorova, PhDc
- aDepartment of Cardiovascular Surgery, Icahn School of Medicine at Mount Sinai, New York, New York
- bDepartment of Surgery, The State University of New York, Stony Brook, New York, New York
- cDepartment of Population Health Science and Policy, Icahn School of Medicine at Mount Sinai, New York, New York
- ↵∗Address for correspondence:
Dr. Joanna Chikwe, Department of Cardiovascular Surgery, Icahn School of Medicine at Mount Sinai, 1190 Fifth Avenue, New York, New York 10029.
Background Long-term benefits of off-pump versus on-pump coronary artery bypass grafting (CABG) are controversial.
Objectives The authors sought to compare long-term survival and morbidity after on-pump versus off-pump CABG.
Methods Mandatory clinical and administrative registries from New Jersey Department of Health were linked to identify patients who underwent CABG between 2005 and 2011, by surgeons who had performed at least 100 off-pump or on-pump CABG operations. Survival, stroke, myocardial infarction, repeat revascularization, and new dialysis requirement were compared using Cox modeling, propensity scores, and instrumental variable analysis. Median follow-up was 6.8 years (range: 0 to 11.0 years); last follow-up date was December 31, 2015.
Results Among 42,570 CABG patients, 6,950 who underwent off-pump CABG and 15,295 who underwent on-pump CABG met study criteria. Off-pump CABG was associated with higher mortality (33.4% vs. 29.6% at 10 years; hazard ratio [HR]: 1.11; 95% confidence interval [CI]: 1.04 to 1.18; p = 0.002) compared with on-pump CABG. Off-pump CABG was associated with a higher risk of incomplete revascularization (15.7% vs. 8.8%; p < 0.001), which was a predictor of late mortality (HR: 1.10; 95% CI: 1.03 to 1.17; p = 0.006); and higher rates of repeat revascularization (15.4% vs. 14.0% at 10 years; HR: 1.17; 95% CI: 1.01 to 1.37; p = 0.048). There were no significant differences in the rate of stroke, myocardial infarction, or new dialysis.
Conclusions In this mandatory clinical registry, off-pump was associated with increased incomplete revascularization, repeat revascularization, and mortality at 10 years compared with on-pump CABG, suggesting that on-pump CABG may be the appropriate choice for most patients undergoing surgical revascularization.
Coronary artery bypass grafting (CABG) is an effective treatment for patients with extensive coronary artery disease (1,2). This operation is one of the most frequently performed worldwide, and is most commonly conducted using cardiopulmonary bypass (on-pump CABG), which enables coronary anastomoses to be carried out on the arrested heart (3,4). Techniques to facilitate CABG on the beating heart without cardiopulmonary bypass (off-pump CABG) were developed to reduce complications associated with cardiopulmonary bypass and manipulation of the aorta (5). The strategies of off-pump and on-pump CABG have been compared in randomized clinical trials (6–10). Randomized surgical trials are the most effective way to control for unmeasured confounders and selection bias, but usually they lack power to analyze patient subgroups, involve highly selected patients and surgeons, and provide limited follow-up, which reduces the generalizability of their findings. Complementary data can be obtained from nonrandomized, clinical registries with long-term follow-up in larger populations more representative of clinical practice. Our objective was to compare long-term outcomes of off-pump and on-pump CABG in patients identified from a contemporary mandatory clinical registry.
This retrospective cohort analysis includes patients with coronary disease who underwent off-pump or on-pump CABG between January 1, 2005, and December 31, 2011, in New Jersey. The study was approved by institutional review boards at the New Jersey Department of Health and the authors’ institution, including a waiver of informed consent.
Patients were identified from a clinical database, the Open Heart Surgery Registry, a mandatory New Jersey registry of cardiac surgery started in 1994. Maintained by the New Jersey Department of Health, the registry collates data collected by trained clinical staff at each participating hospital, conducts data audits quarterly, and employs external medical auditors to verify data annually. All licensed state hospitals are required to submit data on each cardiac surgery patient without exception, and therefore, this clinical registry captures all CABG surgeries occurring in the state.
To obtain long-term clinical outcome data, this clinical registry was linked with 3 databases, each managed by the New Jersey Department of Health. These were the New Jersey Cardiac Catheterization registry, a mandatory clinical database of catheterizations; the New Jersey Discharge Data Collection System, a state-mandated administrative database that prospectively collects data on inpatient visits from 1994 and outpatient visits from 2004; and the New Jersey Vital Statistics deaths registry, which contains all deaths known to any state or federal agency. These databases were linked using probabilistic matching, and patients were matched with a 98% success rate. Additional details are included in the Online Appendix.
Patients were eligible for inclusion if they underwent isolated CABG. Exclusion criteria were other concomitant or any previous cardiac surgery, hemodynamic instability (pre-operative shock, cardiopulmonary resuscitation, or inotrope requirement), and emergency status. Additionally, patients with missing personal or surgeon identifiers, and non-New Jersey residents, were excluded to maximize capture of late outcomes.
Off-pump and on-pump CABG was determined using the clinical registry assignment, which provided an indicator of off-pump to on-pump conversions, thus identifying the subgroup of patients intended for off-pump CABG but who underwent on-pump CABG. The study was designed on an intention-to-treat basis, and the off-to-on pump conversion subgroup was considered as off-pump CABG. Incomplete revascularization was defined, using the clinical registry variables, as a greater number of diseased territories than the number of grafts placed. Hybrid procedures were identified as single-vessel off-pump coronary bypass procedures using a minimally invasive approach in patients with more than 1 diseased territory, who underwent elective percutaneous coronary intervention (PCI) within 90 days after surgery.
To reduce the effect of differential expertise bias, we added surgeon-specific criteria used in expertise-based randomized controlled trials (8,11). Patients were only included if their surgeon had performed more than 100 of the relevant procedure. These qualifying criteria were based on procedural data from 1994 onwards. A total of 49 of 83 surgeons performing off-pump CABG, and 71 of 86 surgeons performing on-pump CABG met the surgeon qualifying criteria. The median number of CABGs completed by these surgeons at the time of the index procedure was 905 for off-pump CABG, and 953 for on-pump CABG, respectively, with a median of 79 off-pump and 137 on-pump CABG cases completed during the 365 days immediately before surgery.
The primary endpoint was all-cause mortality. Patients were followed from the date of the index CABG until December 31, 2015. Deaths were identified from the New Jersey Vital Statistics death registry. Secondary outcomes included stroke, myocardial infarction, repeat revascularization, and new renal failure requiring dialysis. Stroke was defined as a hemorrhagic or ischemic cerebrovascular event during the index admission (but was not present at the time of the index admission) or the primary diagnosis of subsequent admissions; transient ischemic events were excluded. Myocardial infarction was defined as myocardial infarction that occurred during the index admission (but was not present at the time of the index admission), or the primary diagnosis of subsequent admissions. Repeat revascularization was defined as a post-operative PCI or repeat CABG. New renal failure requiring dialysis was defined as post-operative dialysis in patients with no history of pre-operative dialysis.
Continuous variables are reported as mean ± SD. Categorical variables are expressed as proportions. Baseline differences between patients undergoing off-pump and on-pump CABG were detected using the Student's t-test for normally distributed continuous variables and Pearson chi-square test for categorical variables as well as standardized differences for both continuous and categorical variables.
Cumulative event rates for mortality were estimated using Kaplan-Meier analysis. Cumulative event rates for myocardial infarction, stroke, repeat revascularization, and new dialysis-dependent renal failure were estimated using competing risk analysis with the competing event of death. For each outcome, the effect of off-pump surgery was determined using Cox proportional hazard models (12), adjusting for the following covariates: patient age, sex, race, insurance status, body mass index, previous myocardial infarction, previous PCI, previous stroke, renal disease, peripheral vascular disease, diabetes, smoking history, chronic obstructive pulmonary disease, heart failure, hypertension, atrial fibrillation, liver failure, cancer, left ventricular ejection fraction, left main stem stenosis, number of diseased vessels, year of surgery, use of multiple arterial grafts, and the use of pre-operative beta-blocker, aspirin, statin, glycoprotein IIb/IIIa inhibitors, adenosine diphosphate inhibitors, and warfarin (Online Tables 1 and 2 contain additional details). To control for the effect of surgeon and hospital, patients were clustered by surgeon and hospital using a marginal Cox approach with a robust sandwich variance estimator (13).
Significant differences in treatment effects across subgroups were analyzed by adding an interaction term in the Cox models between off-pump treatment and covariates. Subgroup analyses included cohorts stratified by patient age, sex, body mass index, diabetes status, renal disease, history of stroke, ejection fraction, presence of left main disease, number of diseased vessels, completeness of revascularization, and surgeon volume within the year before surgery. To assess the validity of the study findings, we conducted several sensitivity analyses (Online Appendix). First, we eliminated the minimum threshold for surgeon experience and examined the effect of off-pump treatment within the overall cohort of patients. Second, to account for the impact of off-to-on pump conversion, the analysis were repeated with the off-to-on pump conversion subgroup treated as on-pump CABG (as-treated basis), or as a separate cohort (conversion subgroup), or excluded. Third, propensity score matching was used to create comparison groups with similar baseline characteristics (14). The propensity score was calculated using logistic regression fit for off-pump CABG, adjusting for all measured covariates, with patients clustered within hospital. The C-statistic for the model was 0.86. Matching was performed one-to-one with a caliper width of 0.10 of the logit of the propensity score. After propensity score matching, differences in baseline characteristics between patients in each group were detected using a paired t-test for normally distributed continuous variables and the McNemar test for categorical variables as well as standardized differences. To quantify the effect of off-pump surgery on long-term mortality among the propensity-matched cohort, we fit a single Cox model for mortality, adjusting only for off-pump surgery, with patients clustered within matched pairs. We also utilized the same propensity scores to perform survival analysis using inverse probability weighting (15,16).
Finally, instrumental variables were used to address potential unmeasured confounding variables (17–21). In this study, the instrument was defined as the surgeon’s rate of off-pump versus on-pump CABG during the 365 days immediately before the index procedure. We found this instrument to be strongly associated with the treatment, with a first-stage F-statistic of 5,950 (an F-statistic >10 is generally considered to the threshold for a viable instrument) (20). We then identified cohorts of patients whose surgeons displayed a strong preference for off-pump or on-pump surgery (≥90% of cases in the last year performed with 1 procedure type), and among this subset of patients, we calculated the impact of off-pump treatment on long-term mortality using a 2-stage regression (20,21). This analysis was repeated in patient cohorts treated by surgeons with ≥80% preference. Additional details on the statistical analyses performed are provided in the Online Appendix.
All tests were 2-tailed, and an α level of 0.05 was considered statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, North Carolina).
A total of 42,570 patients underwent CABG between January 1, 2005, and December 31, 2011. After applying exclusion criteria, data from 15,295 patients who underwent on-pump CABG and 6,950 patients who underwent off-pump CABG were included in the primary analysis (Online Figure 1). Table 1 shows selected characteristics of the study patients (all variables are provided in Online Table 2). The main difference between groups was in severity of coronary disease, with on-pump patients more likely to have multivessel disease and decreased left ventricular function, and off-pump patients more likely to have single-vessel disease and preserved ventricular function (Table 1). Propensity score matching yielded 3,975 matched pairs of patients who underwent on-pump CABG and patients who underwent off-pump CABG, resulting in standardized differences <5% for each variable (Table 1). Operative characteristics for both cohorts are listed in Online Table 3.
Off-pump CABG was associated with a higher risk of death at 10 years than on-pump CABG (33.4% vs. 29.6%; hazard ratio [HR]: 1.11; 95% confidence interval [CI]: 1.04 to 1.18; p = 0.002) (Central Illustration, Table 2). The effect was similar using propensity score matching and weighting (Online Table 4, Online Figure 2); when analyzing a patient cohort without excluding patients on the basis of surgeon experience (Online Figure 3); in subgroup analyses based on characteristics including age ≥70 years, left ventricular ejection fraction <35%, and 3-vessel coronary disease (Online Table 5).
Conversion from off-pump to on-pump CABG occurred in 250 of 6,950 patients (3.6%), and was associated with increased mortality compared with patients who underwent on-pump CABG without conversion (HR: 1.41; 95% CI: 1.15 to 1.72; p < 0.001), and off-pump CABG without conversion (HR: 1.28; 95% CI: 1.04 to 1.58; p = 0.021) (Online Figure 4A). Off-pump CABG remained associated with higher long-term mortality whether the off-to-on pump conversion subgroup was treated as on-pump CABG (HR: 1.09; 95% CI: 1.02 to 1.16; p = 0.012) or was excluded from the analysis (HR: 1.10; 95% CI: 1.03 to 1.17; p = 0.006) (Online Figures 4B and 4C).
Incomplete revascularization was more common with off-pump compared with on-pump CABG (15.7% vs. 8.8%, respectively; p < 0.001). Incomplete revascularization was associated with increased long-term mortality (HR: 1.15; 95% CI: 1.06 to 1.23; p = 0.006). Off-pump CABG was associated with higher rates of repeat revascularization at 10 years (15.4% vs. 14.0%; HR: 1.17; 95% CI: 1.01 to 1.37; p = 0.048), and similar risks of myocardial infarction (7.5% vs. 7.3%; HR: 1.10; 95% CI: 0.95 to 1.26; p = 0.20), stroke (5.6% vs. 5.7%; HR: 0.92; 95% CI: 0.81 to 1.06; p = 0.26) and new renal failure requiring dialysis (4.7% vs. 5.0%; HR: 0.88; 95% CI: 0.73 to 1.06; p = 0.17) (Figures 1A–1D, Table 2). Hybrid procedures constituted 3.1% of the 1,090 off-pump coronary bypass patients who were under-revascularized: excluding these patients did not significantly change the rate of under-revascularization (15.7% vs. 15.4%).
To control for potential unmeasured confounding, we conducted instrumental variable analysis based on surgeon preference for off-pump surgery (Online Figure 5). The instrument was defined as the surgeon’s rate of off-pump versus on-pump CABG during the 365 days immediately before the index procedure. Among patients treated by volume-qualified surgeons with a ≥90% preference for off-pump surgery, off-pump surgery was still associated with increased long-term mortality (HR: 1.36; 95% CI: 1.04 to 1.78; p = 0.025) (Online Table 4).
In a contemporary cohort of patients undergoing CABG, the long-term risk of death associated with off-pump CABG was higher than that associated with on-pump CABG. Off-pump CABG was associated with increased risk of incomplete revascularization and increased rates of repeat revascularization. Incomplete revascularization was an independent predictor of late mortality. There was no significant difference in the rates of myocardial infarction, stroke, or new dialysis between the 2 groups. The survival effect with off-pump surgery was consistent, irrespective of whether the analysis was performed on an as-treated or intention-to-treat basis; across surgeons, irrespective of previous off-pump experience and preference; and across patient subgroups, including elderly and low-ejection fraction.
Our findings should be evaluated in the context of results from other studies. There have been 3 large, multicenter, randomized controlled trials comparing off-pump with on-pump coronary artery bypass grafting, and more than 80 smaller, mostly single-center trials (6–10). These trials have not typically been powered to detect differences in mortality rate, instead they have been based on composite outcomes including death, stroke, myocardial infarction, repeat revascularization, and renal failure requiring dialysis. For example, the CORONARY trial (Coronary Artery Bypass Surgery [CABG] Off or On Pump Revascularization Study), a large randomized trial comparing on-pump and off-pump CABG, reported 5-year mortality of 13.5% in the on-pump group and 14.6% in the off-pump CABG arm (p = 0.30) (8). It would have required >37,000 patients to have an 80% power of detecting this difference with an α of 0.05. The findings of the ROOBY (Randomized On/Off Bypass) trial, in which a significant difference in 5-year mortality was recently reported favoring on-pump CABG, have been attributed to the unrepresentative sample of patients who were predominantly male veterans, and the relative lack of experience of the primary surgeons (who had a minimum requirement of only 20 cases per surgeon, and who were residents in 58% of cases), which likely led to differential expertise bias favoring the less technically demanding on-pump procedure (6). Our current study was designed to minimize the impact of these limitations.
Our observation that off-pump surgery was associated with poorer long-term outcomes than on-pump surgery supplements information from shorter-term follow-up provided by 3 meta-analyses and 3 large observational studies (9,10,22–25). These include a Cochrane Database meta-analysis of 86 trials including 10,716 patients that showed superior long-term survival with on-pump compared with off-pump CABG, but this analysis defined long-term follow-up as >30 days, and did not include data from the 2 largest randomized trials (6,8,22). Two more recent meta-analyses that incorporated both these trials, including a systematic review conducted according to American College of Cardiology and American Heart Association standards for the development of practice guidelines, reported a mortality benefit with on-pump compared with off-pump CABG at 5 years (9,10). Similar findings in shorter follow-up have been reported in large observational studies (23–25).
The survival advantage consistently associated with on-pump over off-pump CABG has been attributed to the higher rates of incomplete revascularization, and worse graft patency with off-pump compared with on-pump CABG observed in randomized trials and retrospective studies (26). Patients undergoing off-pump CABG have repeatedly been shown to receive fewer bypass grafts either than planned or than the number of diseased territories, in comparison with patients undergoing on-pump CABG (6–8). In a meta-analysis of 76 randomized trials reporting the number of grafts performed, off-pump CABG was associated with fewer grafts compared with on-pump CABG (2.6 vs. 2.9; p < 0.001) (9). The incidence of graft occlusion within 30 days was also higher in patients who underwent off-pump compared with on-pump CABG in this meta-analysis (7.3% vs. 4.4%; p = 0.04), and the rate of repeat revascularization within 1 year was higher after off-pump CABG (2.2% vs. 1.5%; p = 0.01) (9). Our data confirm the higher rates of incomplete revascularization with off-pump CABG. These differences have been attributed to differential expertise bias in randomized and observational studies, due to the greater technical challenges of anastomosing a coronary artery on a beating heart, compared with the arrested heart in on-pump CABG. To address this, our study inclusion criteria specified surgical proficiency (experience of at least 100 on-pump or off-pump cases) for inclusion in each treatment arm. In this pool of relatively expert surgeons, off-pump CABG was still associated with fewer anastomoses and greater likelihood of incomplete revascularization, which we found to be an independent risk factor for late mortality in all patients. Finally, it has been argued that advances in technology and clinical practice, including optimal medical therapy, intraoperative epiaortic assessment, and cardiopulmonary bypass, have addressed more limitations of on-pump CABG surgery than off-pump CABG (26). We believe that these findings have clear implications for the optimal choice of procedure in the majority of patients undergoing surgical revascularization who do not have contraindications to cardiopulmonary bypass.
Study strengths and limitations
The advantage of linking clinical and administrative databases to compare treatment outcomes is the opportunity to evaluate large, representative populations over long follow-up periods, rather than the selected practice evaluated in surgical randomized trials. However, this observational study has specific limitations. Information on coronary anatomy is insufficiently detailed to fully adjust for differences between the treatment groups. The baseline characteristics of patients in the off-pump and on-pump CABG groups were different, and although this was addressed with multivariable adjustment and propensity scores, it is possible that the study findings could be explained by the presence of unmeasured confounding variables. For example, variables that may affect the decision to use or not to use cardiopulmonary bypass, such as aortic calcification, small or diffuse target vessels, and intraoperative hemodynamic instability, were not available for inclusion in the risk model. For the same reason, it was not possible to adjust for the use of “no-touch” aortic techniques, because although information on aortic clamp strategy is reported, the location of proximal anastomoses and use of in-situ versus free internal mammary grafts was not. Although instrumental variable analysis indicates that the effect of unmeasured confounding variables is likely to be small, it is conceivable that selection bias contributed to our study findings. Additionally, data were missing for post-operative variables that may impact long-term outcomes, including compliance with secondary prevention. Finally, patients and secondary clinical endpoints were limited to hospital visits in New Jersey, and our findings may not be generalizable to patients from other regions.
This study used data from a clinical registry of coronary bypass surgery with linkage to mandatory state claims records to evaluate the comparative effectiveness of off-pump and on-pump CABG. We found that off-pump surgery was associated with increased incomplete revascularization, repeat revascularization, and mortality compared with on-pump CABG. These findings suggest that on-pump CABG may be the most appropriate choice for patients without contraindications to cardiopulmonary bypass undergoing surgical revascularization.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In a mandated state-wide registry of CABG, off-pump procedures, even when performed by surgeons who completed >100 operations, were associated with incomplete revascularization, more frequent need for repeat revascularization, and higher mortality compared to on-pump CABG.
TRANSLATIONAL OUTLOOK: Future studies should evaluate outcomes of procedures performed using contemporary technology and revascularization strategies to better inform the selection of the optimum approach to cardiopulmonary bypass for patients undergoing CABG.
The Icahn School of Medicine at Mount Sinai receives royalty payments from Edwards Lifesciences for intellectual property related to the development of 2 mitral valve repair rings, and from Medtronic for intellectual property related to the development of a tricuspid valve repair ring. Dr. Chikwe has received speaker honoraria from Edwards Lifesciences. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- coronary artery bypass grafting
- confidence interval
- hazard ratio
- percutaneous coronary intervention
- Received June 17, 2018.
- Accepted July 2, 2018.
- Hillis L.D.,
- Smith P.K.,
- Anderson J.L.,
- et al.
- Windecker S.,
- Kolh P.,
- Alfonso F.,
- et al.
- ↵Weiss AJ, Elixhauser A. Trends in Operating Room Procedures in U.S. Hospitals, 2001–2011: Statistical Brief #171. Agency for Healthcare Research and Quality. March 2014. Available at: https://www.hcup-us.ahrq.gov/reports/statbriefs/sb171-Operating-Room-Procedure-Trends.pdf. Accessed July 17, 2017.
- Shroyer A.L.,
- Hattler B.,
- Wagner T.H.,
- et al.
- Lamy A.,
- Devereaux P.J.,
- Prabhakaran D.,
- et al.
- Smart N.A.,
- Dieberg G.,
- King N.
- Devereaux P.J.,
- Bhandari M.,
- Clarke M.,
- et al.
- Cox D.R.
- Bagiella E.,
- Karamlou T.,
- Chang H.,
- Spivack J.
- Brookhart M.A.,
- Schneeweiss S.
- Moller C.H.,
- Penninga L.,
- Wetterslev J.,
- Steinbruchel D.A.,
- Gluud C.
- Hannan E.L.,
- Wu C.,
- Smith C.R.,
- et al.
- Kim J.B.,
- Yun S.C.,
- Lim J.W.,
- et al.
- Adams D.H.,
- Chikwe J.