Author + information
- Received February 22, 2005
- Revision received May 15, 2005
- Accepted May 31, 2005
- Published online September 6, 2005.
- Duminda N. Wijeysundera, MD⁎,⁎ (, )
- W. Scott Beattie, MD, PhD⁎,
- George Djaiani, MD⁎,
- Vivek Rao, MD, PhD†,
- Michael A. Borger, MD, PhD†,
- Keyvan Karkouti, MD, MSc⁎,‡ and
- Robert J. Cusimano, MD†
- ↵⁎Reprint requests and correspondence:
Dr. Duminda N. Wijeysundera, EN 3-450, 200 Elizabeth Street, Toronto, Ontario M5G 2C4, Canada.
Objectives The purpose of this study was to assess the effects of off-pump coronary bypass surgery (OPCAB) on mortality and morbidity.
Background Despite its potential for reducing morbidity and mortality, OPCAB’s role in clinical practice remains controversial.
Methods A meta-analysis of 37 randomized controlled trials (RCTs) (n = 3,449) and 22 risk-adjusted (logistic regression or propensity-score) observational studies (n = 293,617) was performed. Two reviewers performed literature searches (MEDLINE, EMBASE, PubMed, reference lists), quality assessment, and data extraction. Treatment effects were calculated as odds ratios (ORs) with 95% confidence intervals (CIs).
Results In RCTs, OPCAB was associated with reduced atrial fibrillation (OR 0.59; 95% CI 0.46 to 0.77) and trends toward reduced 30-day mortality (OR 0.91 95% CI 0.45 to 1.83), stroke (OR 0.52; 95% CI 0.25 to 1.05), and myocardial infarction (OR 0.79; 95% CI 0.50 to 1.25). Observational studies showed OPCAB to be associated with reduced 30-day mortality (OR 0.72; 95% CI 0.66 to 0.78), stroke (OR 0.62; 95% CI 0.55 to 0.69), infarction (OR 0.66; 95% CI 0.50 to 0.88), and atrial fibrillation (OR 0.78; 95% CI 0.74 to 0.82). At one to two years, OPCAB was associated with trends toward reduced mortality, but also increased repeat revascularization (RCT: OR 1.75, 95% CI 0.78 to 3.94; Observational: OR 1.35, 95% CI 0.76 to 2.39).
Conclusions Randomized controlled trials did not find, aside from atrial fibrillation, the statistically significant reductions in short-term mortality and morbidity demonstrated by observational studies. These discrepancies might be due to differing patient-selection and study methodology. Future studies must focus on improving research methodology, recruiting high-risk patients, and collecting long-term data.
Conventional coronary artery bypass surgery (CCAB) is associated with mortality and morbidity (1). Components that might contribute to these adverse events include cardiopulmonary bypass (CPB) and aortic cross-clamping. Although CPB allows for construction of anastomoses in a bloodless and motionless surgical field, it generates a systemic inflammatory response and microemboli (2). Furthermore, cross-clamping of the aorta leads to atheromatous macroembolization, which is associated with adverse outcomes (3).
Increased awareness of CCAB’s adverse effects has renewed interest in off-pump coronary artery bypass surgery (OPCAB). In OPCAB, anastomoses are performed on a beating heart without CPB, thereby avoiding the latter’s inherent risks. In particular, OPCAB might prevent adverse cerebral events. Off-pump coronary artery bypass surgery, however, carries its own risks: intraoperative hemodynamic instability and inadequate revascularization (4). Given the paucity of large randomized trials showing OPCAB to cause significant effects on important outcomes, OPCAB’s role in clinical practice remains controversial (5).
Therefore, we performed a systematic review to address this uncertainty about OPCAB’s benefits. This review focused on randomized controlled trials (RCTs) and observational studies that statistically adjusted for differences between patients who underwent OPCAB as opposed to CCAB.
Our study had four objectives: to estimate OPCAB’s effects within RCTs and observational studies; to describe differences between the study types with regard to treatment effects; to identify deficiencies in published reports; and to make suggestions about the design of future studies.
This review adhered to the QUOROM recommendations (6).
Two reviewers searched MEDLINE (1966 to June 2004), EMBASE (1980 to June 2004), and PubMed (up to June 30, 2004) for RCTs comparing OPCAB against CCAB. The text words employed were off-pump, beating heart, OPCAB, OPCABG, MIDCAB, MIDCABG, without extra-corporeal circulation, and without cardiopulmonary bypass. Included trials had to report any of the following outcomes: death, stroke, myocardial infarction, atrial fibrillation, or acute renal failure.
The same reviewers searched MEDLINE, EMBASE, and PubMed for observational studies reporting risk-adjusted effects of OPCAB on any of the same outcomes. Acceptable risk-adjustment methods included multivariable logistic regression or propensity score techniques. The text words and Medical Subject Headings employed were off-pump, beating heart, OPCAB, OPCABG, MIDCAB, MIDCABG, without extra-corporeal circulation, without cardiopulmonary bypass, logistic models (explode), multivariate analysis (explode), risk adjustment (explode), regression analysis (explode), multivariable, multivariate, multiple logistic, adjust$, and propensity. Bibliographies of included articles were also searched. No language restrictions were applied.
We employed several strategies to avoid duplicate publications. If the same institution produced multiple studies, we only considered studies reporting recruitment time periods. If there was sample overlap between studies, we included only the largest study. In the case of observational studies, the possibility of including overlapping patients exists when centers analyze their own data and also contribute them to national databases. We assessed the influence of such duplicate reporting by analyzing results from small (<10 centers) and large (≥10 centers) databases separately.
Quality assessment and data abstraction
Two reviewers performed quality assessment and data abstraction. Randomized controlled trial quality was rated with regard to randomization, allocation concealment, blinded outcome assessment, and dropouts. Risk-adjusted observational studies were evaluated on the basis of suggested criteria (7).
Data were abstracted on death, stroke, infarction, atrial fibrillation, acute renal failure, inotropic support, low output syndrome, red cell transfusion, re-operation for bleeding, repeat revascularization (one year), and bypass graft number. We accepted the outcome definitions used by the original researchers. All disagreements were resolved by consensus.
We employed Review Manager 4.2.7 (Cochrane Collaboration, Oxford, United Kingdom) to combine treatment effects among studies with the same design. All analyses were performed on an intention-to-treat basis. Effects on dichotomous outcomes were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). For observational studies, adjusted ORs were logarithmically transformed and combined with the weighted inverse variance method. Continuous outcomes were expressed as weighted mean differences with 95% CIs. Heterogeneity was assessed with the Q-statistic. In the absence of significant heterogeneity, treatment effects were pooled with the fixed-effects model. If there was significant heterogeneity (p ≤ 0.1), the random-effects model was used; in addition, we performed post-hoc analyses to explain the heterogeneity. Statistical significance was defined by p ≤ 0.05.
The sensitivity analyses assessed the influence of included studies and statistical models. The first analysis assessed the influence of study quality. In the case of RCTs, meta-analyses were repeated among studies reporting allocation concealment. Meta-analyses were also repeated among high-quality observational studies that employed prospective data, reported ORs with CIs, specified variables that were considered for inclusion, described the variable selection process, and had ≥10 outcome events per predictor variable. The second analysis examined the influence of the statistical model. Analyses that employed the fixed-effects model were repeated with the random-effects model.
The search yielded fifty-nine studies (Fig. 1).Thirty-seven RCTs (3,449 patients) were included (Table 1)(8–43). A single paper reported the results of two trials, which were analyzed separately (9). The short- and mid-term outcomes of two other RCTs were reported in separate publications (32,37,44,45).
Twenty-two observational studies (293,617 patients) were included (Table 2)(46–67). The short- and mid-term outcomes of the same sample were reported in separate publications (63,64). In the case of two other cohorts, OPCAB’s association with different outcomes was reported in separate publications (51,52,61,66). The quality of multivariable modeling was variable (Table 3).
Ten RCTs reported ≥1 death within 30 days of surgery, with an incidence of 1.7% (n = 29) (Table 4)(9,12,13,16,24–26,29,32,35). Off-pump coronary artery bypass surgery was associated with essentially no change in mortality (OR 0.91; 95% CI 0.45 to 1.83). Fourteen observational studies reported adjusted effects on short-term mortality (46,47,49,50,53–55,57,59–63,65). These studies showed that OPCAB was associated with significantly reduced short-term mortality (OR 0.72; 95% CI 0.66 to 0.78).
The number of RCTs reporting ≥1 stroke was 12, with an incidence of 1.3% (n = 27) (9,12,24–26,28,29,32,33,35,37). Off-pump coronary artery bypass surgery was associated with a reduction in stroke that bordered on statistical significance (OR 0.51; 95% CI 0.25 to 1.05). In the fifteen observational studies reporting risk-adjusted effects on stroke, OPCAB was associated with a significant reduction in stroke (OR 0.62; 95% CI 0.55 to 0.69) (47,48,50,51,53–61,63,65).
Nineteen RCTs reported ≥1 infarction, with an incidence of 2.7% (n = 70) (8–10,12,16,18,21,22,25,29,31–35,37,41,42). Off-pump coronary artery bypass surgery non-significantly reduced infarction (OR 0.79; 95% CI 0.50 to 1.25). The number of observational studies reporting risk-adjusted effects on infarction was six (49,53,55,59,61,63). These studies showed OPCAB to be associated with significantly reduced infarction rates (OR 0.66; 95% CI 0.50 to 0.88).
The number of RCTs reporting ≥1 episode of atrial fibrillation was 18, with an incidence of 22% (n = 557) (8,9,14,16,20,22,24–26,29,31,32,35,37,38,40,43). Off-pump coronary artery bypass surgery caused a significant reduction in atrial fibrillation (OR 0.59; 95% CI 0.46 to 0.77), albeit with significant heterogeneity (p = 0.09). This heterogeneity was explained by stratification, according to atrial fibrillation rates in the control arms (Table 5).Stratification also suggested that effects on atrial fibrillation varied with the baseline incidence of the arrhythmia. In the four observational studies reporting risk-adjusted effects on atrial fibrillation, OPCAB was associated with a statistically significant reduction in atrial fibrillation (OR 0.78; 95% CI 0.74 to 0.82) (53,57,59,67).
Off-pump coronary artery bypass surgery was also associated with favorable effects on other postoperative morbid events (Table 4). The only other statistically significant benefits that were demonstrated in RCTs were reduced inotrope requirements and red cell transfusion. As with atrial fibrillation, stratification suggested that OPCAB’s effect on inotropic support varied with the baseline incidence of inotrope use (Table 5).
Five RCTs reported ≥1 long-term (one to two years) mortality (Table 4) (9,29,44,45). Off-pump coronary artery bypass surgery was associated with trends toward reduced long-term mortality (OR 0.82; 95% CI 0.40 to 1.68) and infarction (OR 0.61; 95% CI 0.32 to 1.18). These improvements were, however, associated with a non-significantly increased need for repeat revascularization (OR 1.75; 95% CI 0.78 to 3.94). The two observational studies reporting risk-adjusted effects on long-term outcomes showed essentially no change in mortality (OR 1.01; 95% CI 0.74 to 1.40), non-significantly reduced infarction (OR 0.91; 95% CI 0.55 to 1.49), and trends toward increased repeat revascularization (OR 1.35; 95% CI 0.76 to 2.39) (53,64).
Adequacy of revascularization
Twenty-four RCTs (2,284 individuals) reported the mean number of bypass grafts (8–11,14–16,20,21,24–30,32–34,36–38,40,43). Despite random allocationto OPCAB or CCAB, the mean graft number was 0.19 lowerin the OPCAB arm. This difference was statistically significant (95% CI: 0.25 lower to 0.13 lower; p < 0.00001), without statistically significant heterogeneity (p = 0.15).
Treatment effects were generally not affected by study quality or statistical model (Table 6).Effects on mortality (OR 0.76; 95% CI 0.69 to 0.83) and stroke (OR 0.66; 95% CI 0.57 to 0.75) in large databases were similar to smaller datasets (mortality: OR 0.63, 95% CI 0.54 to 0.74; stroke: OR 0.54, 95% CI 0.46 to 0.66) (47,50,60).
Treatment effects associated with OPCAB are generally more favorable in observational studies than in RCTs. The magnitude of OPCAB’s effects in observational studies exceeded that in RCTs for important short-term outcomes: mortality, infarction, low-output syndrome, re-operation, and acute renal failure. Observational studies showed OPCAB to have statistically significant effects on short-term outcomes, aside from re-operation for bleeding. In contrast, RCTs demonstrated statistically significant benefits for only two outcomes (atrial fibrillation, red cell transfusion).
Importantly, RCTs did not show OPCAB to reduce mortality; even the observed trend was not impressive. All-cause mortality is an important outcome: as a “hard” outcome, it was less sensitive to the absence of blinded outcome assessment in most RCTs. Although the trend toward reduced stroke in RCTs is promising, this potential must be tempered by the general absence of blinded outcome adjudication and standardized stroke assessment protocols.
In contrast, OPCAB’s mortality benefit in observational studies was statistically significant and 20% greater in magnitude. All-cause mortality is likely the most robust outcome in observational studies, given that these studies lacked blinded outcome assessment and systematic outcome surveillance. There are several possible reasons for this discrepancy between RCTs and observational studies. First, statistical significance in observational studies is driven by the much larger sample size. Second, risk-adjustment techniques might have failed to adequately adjust for the differences between patients who underwent OPCAB as opposed to CCAB. Third, up to 13% of attempted OPCAB procedures are converted intraoperatively to CCAB (68). Given that urgent conversion is associated with mortality and morbidity, OPCAB’s benefits might be exaggerated if converted procedures were coded as CCAB in clinical registries (68). Finally, mortality benefits seen in RCTs might have been diminished because most recruited individuals were low-risk. The 1.7% mortality rate reported in the CCAB arms of RCTs is approximately one-half the average North American rate (1). As suggested by OPCAB’s effects on atrial fibrillation and inotrope requirement, its effect on mortality might be enhanced among higher-risk patients.
Both RCTs and observational studies raise important questions about OPCAB’s long-term effects. Both study types demonstrated that OPCAB was associated with trends toward increased repeat revascularization. In the context of significantly fewer grafts with OPCAB, future studies must address the frequency of repeat revascularization with OPCAB. This end point is likely more relevant to patients than assessments of graft patency.
Several previous systematic reviews have addressed OPCAB outcomes (54,69–74). Most focused only on RCTs or specific outcomes (69–72,74). Except for Cheng et al. (75), our study included more RCTs than did previous reviews. Reston et al. (73) was the only previous study to consider data from RCTs and observational studies for several outcomes; however, observational studies were not restricted to risk-adjusted ones. Although Reston et al. (73) suggest that including non-adjusted studies did not exaggerate OPCAB’s benefits, our results suggest otherwise. Their estimated effect on short-term mortality (OR 0.64; 95% CI 0.54 to 0.75) is larger than our estimate for both RCTs (OR 0.91) and observational studies (OR 0.72).
This review identifies important deficiencies in the literature that must be addressed by future RCTs. Trials remain underpowered to detect clinically important treatment effects. Assuming a 2% baseline stroke rate, approximately 5,000 patients would be required to detect the effect suggested by our study (alpha 0.05, power 80%) (60). In addition, patients enrolled in trials thus far were generally low-risk. This patient profile reduces the generalizability of the results, necessitates larger sample sizes, and might underestimate OPCAB’s benefits among higher-risk patients. Further improvements remain needed in the quality of RCTs. Most previous trials did not adequately conceal allocation or blind outcome assessors, both of which can influence estimated treatment efficacy (76). Blinding of outcome assessors is also integral to confirming our finding that OPCAB might prevent non-fatal morbid events (stroke, atrial fibrillation, inotropic support). Finally, more RCTs must address longer-term outcomes. The importance of these data cannot be overstated.
In the absence of large RCTs, risk-adjusted observational studies constitute the majority of the evidence supporting OPCAB’s benefits. Observational research within cardiac surgery has been facilitated by prospective databases with rigorous quality control and follow-up. Observational research has important advantages. The results might be more generalizable: patients and surgeons included in observational studies are more similar to those encountered in clinical practice. Randomized controlled trials generally recruit low-risk patients who undergo surgery performed by a few experienced surgeons. Given that most perioperative adverse events are uncommon, adequately powered RCTs remain large and expensive. Observational studies are a more economical method for evaluating OPCAB’s effectiveness.
These advantages, however, are counter-balanced by important limitations. Unmeasured confounders might bias even well-analyzed cohort studies; therefore, meta-analyses of observational data might produce precise but spurious results (77). In addition, interpretation of results from observational studies should be limited to associations; no causal inferences should be drawn. Despite these limitations, there remains a role for observational research in establishing the effectiveness of OPCAB. Future studies must report more transparent analytic plans, account for misclassification of converted OPCAB procedures, and incorporate longer-term outcomes.
Randomized controlled trials did not find, aside from atrial fibrillation, the significant reductions in short-term mortality and morbidity demonstrated by observational studies. These discrepancies are explained, in part, by differing patient selection and study methodology. This review identifies three important goals for future studies. First, research methodology must be improved to clarify OPCAB’s potential for preventing non-fatal morbid events. Second, long-term data are required to understand OPCAB’s association with repeat revascularization. Finally, as opposed to recruiting low-risk individuals, studies should focus on patients at increased risk for perioperative morbid events.
Drs. Wijeysundera, Rao, and Karkouti are supported in part by the Canadian Institutes of Health Research.
- Abbreviations and Acronyms
- conventional coronary artery bypass surgery
- confidence interval
- cardiopulmonary bypass
- off-pump coronary artery bypass surgery
- odds ratio
- randomized controlled trial
- Received February 22, 2005.
- Revision received May 15, 2005.
- Accepted May 31, 2005.
- American College of Cardiology Foundation
- Bruins P.,
- te Velthuis H.,
- Yazdanbakhsh A.P.,
- et al.
- Baker R.A.,
- Andrew M.J.,
- Ross I.K.,
- Knight J.L.
- Gulielmos V.,
- Eller M.,
- Thiele S.,
- et al.
- Krejca M.,
- Skiba J.,
- Szmagala P.,
- Gburek T.,
- Bochenek A.
- Legare J.F.,
- Buth K.J.,
- King S.,
- et al.
- Motallebzadeh R.,
- Kanagasabay R.,
- Bland M.,
- Kaski J.C.,
- Jahangiri M.
- Puskas J.D.,
- Williams W.H.,
- Duke P.G.,
- et al.
- Selvanayagam J.B.,
- Petersen S.E.,
- Francis J.M.,
- et al.
- Tang A.T.M.,
- Knott J.,
- Nanson J.,
- et al.
- van Dijk D.,
- Nierich A.P.,
- Jansen E.W.,
- et al.
- Velissaris T.,
- Tang A.,
- Murray M.,
- El Minshawy A.,
- Hett D.,
- Ohri S.
- Wehlin L.,
- Vedin J.,
- Vaage J.,
- Lundahl J.
- Zamvar V.,
- Williams D.,
- Hall J.,
- et al.
- Hirose H.,
- Amano A.,
- Takahashi A.
- Patel N.C.,
- Grayson A.D.,
- Jackson M.,
- et al.
- Tamis-Holland J.E.,
- Homel P.,
- Durani M.,
- et al.
- Legare J.F.,
- Buth K.J.,
- Hirsch G.M.
- Athanasiou T.,
- Aziz O.,
- Mangoush O.,
- et al.
- Van Der Heijden G.J.,
- Nathoe H.M.,
- Jansen E.W.,
- Grobbee D.E.
- Egger M.,
- Schneider M.,
- Davey Smith G.