Long-Term Survival and Freedom From Reintervention After Off-Pump Coronary Artery Bypass GraftingClinical Perspective
A Propensity-Matched Study
Background: The long-term outcomes of off-pump coronary artery bypass grafting (CABG) are the subject of speculation. Our institution has >15 years of experience performing CABG both off-pump (OPCAB) and on cardiopulmonary bypass (CPB). Our null hypothesis was that there would be no difference in a long-term composite of death and revascularisation between the 2 methods.
Methods: We performed a retrospective cohort study of all isolated CABG at our institution from 2001 to 2015. We used an intention-to-treat analysis, performing risk adjustment with adjustment for and matching to propensity score. In total, 13 226 patients had CABG: 5882 had OPCAB and 7344 had CPB, with a median follow-up of 6.2 years.
Results: Of the 5882 OPCAB, 76 (1.3%) converted to CPB. One-, 5-, and 10-year survivals in each group were similar (OPCAB vs CPB: 96.7%, 87.9%, 72.1% vs 96.2%, 87.4%, 72.8%). No difference was found in long-term survival (adjusted hazards ratio [HR] 1.03; 95% confidence interval [CI]: 0.94–1.11 for OPCAB vs CPB; P=0.56) or freedom from death and reintervention (HR 0.98; 95% CI: 0.92–1.06 for OPCAB vs CPB; P=0.23). Patients receiving OPCAB had higher EuroSCOREs (median [quartiles]: 2.81 [1.53–5.57] vs 2.73 [1.51–5.22]; P=0.01), fewer grafts (mean±SD: 3.0±0.9 vs 3.3±0.9; P<0.001), but more total arterial grafting (45.9% vs 8.4%; P<0.001). OPCAB also had more trainee first operators (15.3% vs 12.5%), lower cardiac enzyme rise, shorter length of stay, and fewer complications (such as myocardial infarction).
Conclusions: OPCAB is associated with similar long-term outcomes to CABG performed on CPB in our institution. Our low conversion rate to CPB, while training junior surgeons, demonstrates that OPCAB can be taught safely. The number of grafts performed between the 2 approaches is clinically comparable, if statistically different, and appears to provide equal benefits to survival and freedom from reintervention as on-pump CABG.
Editorial, see p 1221
The first coronary artery bypass graft (CABG) was performed on a beating heart.1 After the advent of cardiopulmonary bypass (CPB), however, the majority of surgeons turned to cardioplegia and extracorporeal support to perform anastomoses to the coronary arteries. As the deleterious effects of CPB became better understood,2 efforts to minimize these risks led to a resurgence in beating heart surgery. Currently, ≈20% of surgical coronary revascularization performed in the United Kingdom is off-pump coronary artery bypass (OPCAB) grafting.3
Numerous trials and observational studies have compared OPCAB with conventional surgery performed with CPB. These have been the subject of systematic reviews and meta-analyses,4,5 which suggest that long-term survival is impaired with OPCAB. Many of the trials and cohort studies that make up these systematic reviews were potentially biased, however, by inexperienced off-pump surgeons attempting beating heart techniques. For example, the ROOBY trial (Veterans Affairs Randomized On/Off Bypass) was noted to have a conversion rate 6 times the national average, with OPCAB surgeons defined as those who had performed only ≥20 off-pump cases.6 The second-largest trial in the meta-analysis had experienced surgeons performing OPCAB, but excluded patients <75 years of age and only followed up patients for 12 months.7 The CORONARY study (CABG Off or On Pump Revascularization Study) has addressed some of the issues of surgeon experience and presented data for 30-day follow-up, but is still in the process of collecting data on 5-year follow-up.8 The largest observational study of propensity-matched patients in the literature included 8911 OPCAB patients from 42 centers over a 14-year period.9 This equates to <2 OPCAB cases per center per month. The authors noted that the absence of surgeon experience in the data set was a limitation, and they were unable to cite the number of conversions from OPCAB to CPB. Another propensity-matched study using national registry data suffered from similar limitations,10 as did the largest institutional review of OPCAB.11
With extensive experience and long-term follow-up of OPCAB at our institution, we sought to identify the long-term freedom from death and reintervention at our center compared with CPB, when patient demographics were adjusted for.
Materials and Methods
We conducted a single-center retrospective cohort study with propensity-matched statistical analysis. The Institutional Review Board waived the need for informed consent.
All consecutive, isolated, first-time coronary artery bypass operations performed at our institution between December 2001 and October 2015 were included. The start date was chosen as this was beyond the institutional learning curve from which time patients were no longer specially selected for beating heart surgery. No patients were excluded from the study. Patients due to undergo CABG without the aid of CPB were classified as off-pump. Conversion to CPB was analyzed in an intention-to-treat manner. A prospectively collected database, validated to the Society for Cardio-Thoracic Surgery in the Great Britain and Ireland data set,3 was interrogated for preoperative patient characteristics, operative data, and in-hospital mortality and morbidity. Long-term mortality was established by data from the Office for National Statistics. Reintervention was defined as subsequent revascularization procedures by percutaneous or surgical means. The definitions for all variables were taken as those used in the EuroSCORE calculator.12 Although the EuroSCORE II calculator has been shown to have good discrimination, the calibration is poor.13 For benchmarking and quality assessment, we therefore routinely still utilize the original calculator with risk modifiers.
All patients underwent general anesthesia with endotracheal intubation and ventilation. OPCAB was performed with deep pericardial sutures and steep Trendelenburg maneuvers to enucleate the heart, intracoronary shunts, CO2 mister-blowers, and a variety of cardiac stabilization devices, predominantly various generations of Octopus device (Medtronic Inc., Minneapolis, MN). Silastic coronary slings were occasionally used to facilitate placement of intracoronary shunts.
Patients undergoing CPB had ascending aortic arterial cannulation unless aortic calcification prohibited this, in which case femoral cannulation was used. Venous drainage was achieved by direct right atrial cannulation with 2-stage cannulae. Either antegrade alone or a combination of antegrade and retrograde cardioplegia was used to establish diastolic arrest, with either cold blood or St. Thomas’ crystalloid cardioplegia solutions.
There was a bimodal distribution in preference of surgical technique among 17 surgeons. Nine surgeons performed ≥95% of their CABG operations with CPB. Six surgeons performed ≥95% of their surgery off-pump. Two surgeons had mixed practice, with ≈30% of patients performed on-pump by one and ≈30% off-pump by the other. No statistically significant differences in the outcomes were found between surgeons.
Some cases (1808, 13.7%) were performed by trainees. The definition of training was adopted from the Society for Cardiothoracic Surgery in Great Britain and Ireland. Where the majority of the proximal and distal anastomoses were performed by the trainee, this was recorded as a “trainee first-operator” case. The consultant surgeon may have been scrubbed as a first (or, rarely, a second) assistant, unscrubbed but supervising from within the operating theater, or remotely supervising from within the hospital depending on the trainee’s seniority.
Our primary outcome measures were long-term survival and long-term freedom from death and reintervention. Secondary outcome measures included inpatient mortality, length of hospital stay, number of grafts, training, inotrope use, postoperative blood loss, cardiac enzyme rise, and postoperative complication rates.
All variables were analyzed using R version 3.1.3 (R Core Team, Vienna, Austria) with Amelia and MatchIt packages14–16 and JMP version 11.0.0 for Mac (SAS Institute Inc., Cary, NC). Univariate analyses were performed as Wilcoxon signed rank for all continuous variables and X2 for categorical variables. Survival was computed by use of the Kaplan-Meier curve, and the 2 groups were compared using a log-rank test for unadjusted variables and Cox-regression analysis for adjusted survival.
Imputation of Missing Data and Derivation of Propensity Score
Data were missing for <1% of most preoperative variables with the exception of spirometry, where there was ≈10% missing data. Multiple imputation methods (where m=5) were used to account for the missing data before propensity matching. Each imputation data set was used to calculate a propensity score based on covariates selected for clinical and/or statistical association with the treatment group. After assessment and removal of collinear variables, 39 variables were used to develop the final propensity model. Extent of coronary disease was matched exactly to optimize the matching balance. The propensity scores from each of the 5 imputed data sets were averaged to create a final pooled score used for matching.
Matching and Adjusting for Propensity Score
We used 2 approaches to using the propensity score data. First, to maximize inclusion of all patients, we generated quintiles of the propensity score and adjusted for this trait in outcomes for raw (unmatched) data, using multiple regression or Cox-regression analyses. Second, we performed 1:1 propensity score matching to select patients receiving OPCAB or CPB with comparable preoperative characteristics. Matching was performed using greedy, nearest neighbor matching without replacement and a caliper of 0.2. The full code for the analysis is available in online-only Data Supplement Appendix I. Matched data were assessed using paired nonparametric tests where data were continuous. Survival data were presented as both unmatched and matched, as well as unadjusted (log-rank) or adjusted (Cox regression).
The preoperative patient characteristic data of the unmatched (ie, raw data, before matching) and matched groups are presented in Table 1. After matching, fewer differences were found between the groups. Covariate balance was assessed using jitter plots of the propensity scores, dotplots of the standardized mean differences (SMDs), and a histogram of the propensity score distribution (Figures 1–3). These tests show good distribution of propensity scores, with wide recruitment from both groups. The SMD of each of the covariates was between –0.1 and 0.1, indicating good matching.
Patients in the OPCAB group had statistically less renal impairment but more left ventricular impairment. In the OPCAB group, there was a higher proportion of unstable angina and patients in CCS class IV but fewer patients with triple-vessel coronary artery disease. A greater proportion of the OPCAB group required preoperative nitrates, but fewer required preoperative inotropes. Slightly more off-pump operations were performed on an urgent basis. Overall, the OPCAB patients had a higher risk profile than the CPB patients (EuroSCORE median [quartiles]: 2.81 [1.53–5.57] vs 2.73 [1.51–5.22]; P=0.011).
After propensity matching, 5539 patients were included in each group. The rate of preexisting renal dysfunction was adequately adjusted for, as was the difference in preoperative left ventricular function. Significantly more patients had single-vessel coronary artery disease in the OPCAB group (5.5% vs 1.7%; P<0.001, SMD=0.285) after initial attempts at matching; hence, exact matching was used for this variable. Preoperative EuroSCORE was significantly higher in the matched OPCAB group when compared with nonparametric tests but demonstrated adequate balance with an SMD ≈0.1.
Statistically significant differences were found in the grafting strategy between the OPCAB and CPB groups (Table 2). Fewer grafts were performed off pump than on pump (3.03±0.83 vs 3.28±0.91; P<0.001), but mean and median number of grafts was ≥3 in both groups. Quartiles and ranges were identical. Pedicled left internal mammary use was higher in OPCAB for both matched and unmatched groups. Bilateral mammary usage was also higher in OPCAB both before and after matching. The use of total arterial revascularization (ie, nonuse of vein graft) and sequentialization was substantially higher in off-pump surgery than CPB both before and after matching. A higher proportion of OPCAB cases were performed by trainees (unmatched: 15.1% vs 12.5%; P<0.001; matched: 14.7% vs 12.3%; P<0.001).
Primary Outcome Measures
Maximum follow-up approached 14 years, and median follow-up was 6.2 years. One-, 5-, and 10-year survivals in each group were similar (CPB vs OPCAB: 96.2%, 87.4%, 72.8% vs 96.7%, 87.9%, 72.1%).
In unadjusted analysis including all individuals, there was no statistically significant difference in long-term survival in unmatched patients over follow-up (Figure 4, log-rank P=0.96). In an analysis of all individuals adjusted for quintile of propensity score, the adjusted hazard ratio (HR) of long-term survival was 0.96 (95% confidence interval [CI]: 0.89–1.04; P=0.96 for OPCAB vs CPB). After propensity matching, using a univariate Cox-regression model, the result persisted (HR 1.04 [95% CI: 0.96–1.13]; P=0.30 for OPCAB vs CPB).
Similarly, there was no long-term difference in freedom from death or coronary reintervention in unadjusted or adjusted analysis of all individuals (HR of death or coronary reintervention 0.98 [95% CI: 0.92–1.06]; P=0.23 for OPCAB vs CPB) (Figure 5). In the propensity-matched groups, this equivalence persisted, with no statistically significant difference in long-term death or reintervention (HR 1.07 [95% CI: 0.99–1.15]; P=0.10).
Secondary Outcome Measures
OPCAB recipients had lower in-hospital complications compared with CPB, both before and after matching (Table 3 shows P values for unmatched data adjusted for propensity score quintiles). In unmatched patients, those in the OPCAB group had less postoperative inotrope use, shorter overall ventilator times, less postoperative blood loss, and smaller rises in cardiac enzymes at 24 hours. Fewer overall complications occurred in the OPCAB group, with less postoperative myocardial infarction, but no difference in need for renal replacement or neurological outcomes. Length of hospital stay was shorter in the OPCAB group. Neither in-hospital mortality (OPCAB 1.7% vs CPB 2.1%; P=0.135) nor CVA (0.6% vs 0.8%; P=0.289) reached statistical significance. Mortality for the 76 patients who required conversion from OPCAB to CPB was 10.5%, which was significantly higher than the mortality for OPCAB cases that were not converted (1.6%; P<0.0001).
In the propensity-matched group, the rates of complications, ventilation times, and cardiac enzyme were lower in the OPCAB group and remained statistically significant (Table 3). The mortality was also lower in the OPCAB group (1.8% vs 2.2%; P=0.198), although this did not reach statistical significance. The odds ratio for in-hospital mortality with OPCAB was 0.83 (95% CI: 0.64–1.08). The rate of stroke remained nonsignificantly lower (0.6% vs 0.8%; P=0.18) with matching.
This is the largest observational, single-institution European study to compare long-term survival after OPCAB versus CPB. Our results find no significant differences between the 2 approaches, suggesting that they may have similar short- and long-term outcomes. Our results were robust to matching for propensity scores, which suggests that our findings are not due to negative confounding.
These findings are different to previous studies, which were potentially limited by poor operator experience, lack of long-term follow-up, and high numbers of conversions.
We have extensive experience in off-pump surgery, performing 44.5% of isolated CABGs without the aid of CPB. Surgical single-vessel coronary artery disease is much more likely to be performed off pump here (71.8% vs 28.2%; P<0.0001), suggesting a clear preference for off-pump surgery even among on-pump surgeons given appropriate circumstances. Analysis of unmatched patients undergoing CABG for single-vessel disease shows a marked benefit for OPCAB (Figure 6a).
Off-pump patients typically had more comorbidities but also less diffuse coronary artery disease, with 10% fewer in the unadjusted OPCAB group suffering from triple-vessel coronary disease. In the propensity-matched analysis, these characteristics were well balanced, and there was no difference in the long-term survival or freedom from reintervention. The number of grafts performed for triple-vessel disease was statistically significantly higher for CPB but with comparable distributions in both groups. However, the proportion of arterial grafts in the OPCAB group was significantly higher.
Our overall rate of conversion from OPCAB was low, with only 1.3% of cases requiring a change of strategy to on pump. Data are not available for numbers of conversions from proposed on-pump to off-pump surgery (eg, previously unknown porcelain aorta), but this number is low, probably <20 cases per year per discussion with colleagues. The mortality of 10.5% in patients requiring conversion from OPCAB requires further interrogation. Although the numbers are small, with just 8 deaths as a result of failed conversions over the last 15 years, it may reflect the dangers of persisting with off-pump techniques in patients who do not have the capacity to cope with this strategy. In patients with valvar insufficiency or poor ventricular function, we have a low threshold for early conversion or intraaortic balloon pump support to mitigate the need for conversion in these higher risk individuals. Our strategy for OPCAB is not selective, however, and mortality rates are not different from cases performed on pump. The overall risk of death from failed conversion in OPCAB in this series was 0.1%. It is impossible to predict whether these deaths could have been prevented by adopting an on-pump strategy from the outset or whether they represent high-risk patients with contraindications to surgical revascularization as a whole. At our center, we minimize these risks by using inotropic and intraaortic balloon pump support prophylactically in patients with poor left ventricular function and having low threshold for conversion in patients with mitral regurgitation who might be at risk of intolerance to enucleation of the heart. With the institutional experience that we have, conversions are rarely unheralded as the struggling heart is readily noted and the surgical strategy modified accordingly. We contend that, despite a mortality rate of 10.5% in our conversions that detractors to OPCAB will highlight, the overall outcomes in our expert center justify our use of this technique. Off-pump surgery is indeed inappropriate for a small proportion of patients; however, when one has assessed the function of the heart by initial manipulation to view the coronary targets, it is usually clearly evident which hearts should not proceed on to an OPCAB approach. Beating heart with CPB support is the preferred technique in cases where off-pump surgery must be abandoned, as this minimizes ischemia and aortic manipulation.
The mean number of grafts performed was lower in the off-pump group, but with comparable median, interquartile range, and maximum numbers of grafts (6 in each group). This does not appear to affect the rate of reintervention in either the unadjusted or propensity-matched groups. This may reflect, therefore, a tendency to overgraft in the CPB group rather than incomplete revascularization in the OPCAB group. Other studies have had similar findings.17 It has already been demonstrated that aggressive revascularization with stents is detrimental,18 and the functionality of the coronary stenosis is also likely to be of importance in surgical revascularization. This finding may explain why the outcomes were no different despite fewer grafts. Off-pump surgeons may have a tendency to graft only more severely stenosed vessels for other reasons, such as the use of radial artery conduit or more perceived risk-benefit balance. The effects of grafting on the physiological response to revascularization are also more evident at the time of procedure with OPCAB, and this may have affected the number of grafts. Another possible explanation is that in the early experience with OPCAB, fewer grafts were considered; but as the technique has become more established, the grafting strategy is similar to that of on pump. All targets are readily accessible; indeed, lateral and inferior targets are not more likely to not be grafted in our hands. The greater proportion of single-vessel disease in the OPCAB group does not appear to be a contributing factor in this difference, because the outcomes for matched patients with triple-vessel disease are identical (Figure 6c). This does not account for the myriad differences in severity within the umbrella of triple-vessel coronary artery disease. One possible way to account for these might have been to use an anatomic stenosis scoring schema, such as SYNTAX, but >50% of our data precedes the development of this tool and would be beyond the scope of this study. Other differences in the patient population are relatively well matched after propensity scoring, with statistically significant but clinically low impact differences.
There was a substantially higher proportion of total arterial grafting in the off-pump group, including the use of bilateral mammary and radial arteries, which did not translate to a survival benefit. These findings appear to correlate well with a similar, smaller recent study from the US.19 Total arterial revascularization with sequential and radial T-grafting from the left internal mammary artery meant that proximal aortic management were reduced in the OPCAB group. The use of “side-biting” partial occlusion clamps in both groups, however, might explain the similarity in neurological outcomes between the two.
The preferences of surgeons in our institution for either on- or off-pump CABG were virtually binary, with most performing predominantly one or the other technique. Two surgeons with mixed practice had comparable outcomes to the others. We acknowledge the impact of the learning curve in OPCAB, but we have demonstrated that with adequate supervision, this can be rapidly overcome. For surgeons currently performing conventional CPB-supported CABG, the need for structured supervision is paramount. In addition, the immersion of anesthetic, scrub, perfusion, and nursing colleagues into the technique (rather than occasional forays) means that off-pump surgery is not considered a special case. It is a second default for surgical revascularization at our institution for CABG.
Our study has the typical limitations of a retrospective single-center study. We have attempted to account for this in part by performing a propensity score-matched analysis to adjust for differences in risk stratification and account for selection biases. However, as with all observational analyses, bias and confounding may persist despite our attempts to control for them.
Propensity score matching is a contentious method of adjusting for covariate imbalances in observational studies. The matching process discards large numbers of patients from analysis, which may affect the generalizability of the results. Although proponents argue that it has benefits over regression adjustments in nonrandomized studies,20 care must be taken when interpreting the results of such statistical manipulations. To overcome this limitation, we also conducted analyses that included all individuals and adjusted for quintile of propensity score; results from these analyses were similar to the matched approach, which provides greater confidence in our analytic approach.
We assume that as the regional tertiary referral center for complex reintervention, patients undergoing revascularization procedures after CABG would return to us. If patients have migrated, there is an unknown loss to follow-up for this outcome measure.
As an observational study, our data do not include angiographic evidence of long-term patency. Although this may demonstrate the longevity of anastomoses performed by each technique, it is a surrogate marker to the clinical effect of graft stenoses, which is reflected in the mortality and reintervention rates we cite.
OPCAB can be performed safely and effectively with equivalent short- and long-term outcomes to surgical revascularization on CPB. Although the grafting strategies are different, with statistically fewer grafts and more arterial conduit used in the OPCAB group, these do not appear to have a bearing on the long-term survival or freedom from reintervention. Patients selected for OPCAB tend to have a higher risk profile; but even when adjusting for this, the outcomes are not significantly different. We found advantages in OPCAB on several postoperative measures of performance, including perioperative myocardial infarction but not in rates of renal impairment or stroke, and we have observed a trend toward lower mortality in the short term. In our expert center, OPCAB demonstrates excellent long-term results.
Presented at the Society for Cardiothoracic Surgery in Great Britain and Ireland annual meeting, March 13–15, 2016, Birmingham, United Kingdom.
Sources of Funding
Sources of Funding, see page 1219
The online-only Data Supplement is available with this article at http://circ.ahajournals.org/lookup/suppl/doi:10.1161/CIRCULATIONAHA.116.021933/-/DC1.
Circulation is available at http://circ.ahajournals.org.
- Received February 7, 2016.
- Accepted September 13, 2016.
- © 2016 American Heart Association, Inc.
- Ascione R,
- Angelini GD
- 3.↵The Society for Cardiothoracic Surgery in Great Britain & Ireland. SCTS Blue Book Online: Cardiac Surgery Audit Data [Internet]. Cited August 13, 2015; available from: http://bluebook.scts.org/
- Deppe AC,
- Arbash W,
- Kuhn EW,
- Slottosch I,
- Scherner M,
- Liakopoulos OJ,
- Choi YH,
- Wahlers T
- Shroyer AL,
- Grover FL,
- Hattler B,
- Collins JF,
- McDonald GO,
- Kozora E,
- Lucke JC,
- Baltz JH,
- Novitzky D
- Diegeler A,
- Börgermann J,
- Kappert U,
- Breuer M,
- Böning A,
- Ursulescu A,
- Rastan A,
- Holzhey D,
- Treede H,
- Rieß FC,
- Veeckmann P,
- Asfoor A,
- Reents W,
- Zacher M,
- Hilker M
- Lamy A,
- Devereaux PJ,
- Prabhakaran D,
- Taggart DP,
- Hu S,
- Paolasso E,
- Straka Z,
- Piegas LS,
- Akar AR,
- Jain AR,
- Noiseux N,
- Padmanabhan C,
- Bahamondes JC,
- Novick RJ,
- Vaijyanath P,
- Reddy S,
- Tao L,
- Olavegogeascoechea PA,
- Airan B,
- Sulling TA,
- Whitlock RP,
- Ou Y,
- Ng J,
- Chrolavicius S,
- Yusuf S
- Nashef SA,
- Roques F,
- Hammill BG,
- Peterson ED,
- Michel P,
- Grover FL,
- Wyse RK,
- Ferguson TB
- Kirmani BH,
- Mazhar K,
- Fabri BM,
- Pullan DM
- 14.↵R Development Core Team. R Development Core Team (2013). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0. http://www.R-project.org. Accessed October 4, 2016.
- Honaker J,
- King G,
- Blackwell M
- Ho DE,
- Imai K,
- King G,
- Stuart EA
- Tonino PA,
- De Bruyne B,
- Pijls NH,
- Siebert U,
- Ikeno F,
- van’ t Veer M,
- Klauss V,
- Manoharan G,
- Engstrøm T,
- Oldroyd KG,
- Ver Lee PN,
- MacCarthy PA,
- Fearon WF
- Emerson DA,
- Hynes CF,
- Trachiotis GD
What Is New?
This is the largest single-institution European study and demonstrates equipoise in long-term mortality and mortality plus reintervention between off- and on-pump surgery. This applies to all-comers for coronary artery bypass grafting and, separately, 3-vessel disease.
Additionally, subset analysis seems to demonstrate that single-vessel coronary disease is best treated with off-pump surgery.
The in-hospital outcomes between the 2 groups were broadly similar. These data, taken together, show that in the right hands, off-pump surgery is at least as good as on-pump surgery out to 15 years follow-up.
What Are the Clinical Implications?
In an institution such as ours with a high proportion of surgery undertaken by exclusively off-pump surgeons, the institutional knowledge base is such that these operations are the norm.
Our positive outcomes are reliant on the fact that off-pump surgery is ubiquitous, requires no special measures, and is actively trained from the outset of a trainee’s education. In such circumstances, off-pump coronary artery bypass and coronary artery bypass grafting on cardiopulmonary bypass have equivalent long-term outcomes.