WITH the rising costs of health care and the desire to demonstrate best outcomes, there is increasing interest in demonstrating that surgical procedures are both efficacious and cost effective. There is now sufficient justification to perform randomized trials to evaluate the appropriateness of procedures for specific indications.1 Coronary artery bypass grafting (CABG) continues to be one of the most commonly performed procedures that has a significant impact on overall health care resources. In this issue of the Journal, Cheng et al.  report on their meta-analysis of the randomized literature comparing off-pump (OPCAB) and conventional approaches to coronary revascularization.2 They utilized a comprehensive search strategy to identify the relevant randomized studies and therefore their work should represent the best evidence on which to base future treatment decisions and research directions. In an analysis of 37 trials involving 3,369 patients, they found no difference in their primary outcomes of 30-day and 2-yr mortality rates. For their secondary nonfatal outcomes, results were mixed, but with no significant difference in the incidence of major adverse clinical outcomes such as myocardial infarction, renal failure, or stroke.

There are significant limitations in how best to generalize their findings to current clinical practice and whether they have been able to effectively address the original hypotheses that drove the development of OPCAB. As this article documents, there are incidental benefits to OPCAB that may be important, such as decreased transfusion, lower incidence of atrial fibrillation, and even important economic benefits. However, we must not forget that the push to develop techniques for OPCAB can be traced to the plethora of reports documenting that CABG with cardiopulmonary bypass results in an alarmingly high incidence of postoperative cognitive dysfunction (POCD) and a lesser but important occurrence of stroke. These sequelae are the major adverse clinical outcomes that OPCAB was developed to ameliorate. Testing the hypothesis that these outcomes are diminished by OPCAB should be a priority in any randomized study. Furthermore, anyone who has witnessed the challenge of sewing coronary grafts on a beating heart would agree that one other primary outcome should test the hypothesis that OPCAB results in poorer graft survival, given the increased degree of difficulty.

Reports of the complication of POCD after CABG have reached the mainstream media3 and have been the focus of hospital advertising campaigns promoting OPCAB. The public has been convinced that coronary revascularization with cardiopulmonary bypass will likely lead to the syndrome known colloquially as “pump head.” As a result, many surgeons are faced with patients who present for CABG demanding OPCAB, thus compelling many surgery programs to offer OPCAB as an alternative or, in many cases, as the primary approach for CABG. This has resulted in a steady increase in the number of OPCAB procedures performed over the past several years; it has plateaued at approximately 20–30%. There is general agreement that the incidence of POCD early after cardiac surgery is high but that cognitive function probably returns nearly to baseline in 6–12 weeks.4,5 However, some studies have suggested that although patients may improve dramatically during the 6–12 weeks after cardiac surgery, a significant deficit remains which may last for years.6 Some suggest that POCD may herald further long-term decline,5 and still others deny the existence of any long-term evidence of POCD.7 

Despite the high level of interest in POCD, another, better, reason to support the development of OPCAB may be to prevent strokes. Although stroke after cardiac surgery occurs in only 2–3% of all patients, the rate may approach 35–70% in aged patients with multiple risk factors such as hypertension, diabetes mellitus, and previous stroke.8,9 Furthermore, although the overall incidence of stroke is much less than that of POCD, the consequences of stroke after cardiac surgery, including a fivefold to 10-fold increase in early mortality,10 are far more devastating than any sequelae of POCD.

Given this historical and clinical perspective on the origins of OPCAB, readers are left to question whether the data in this report are sufficient to adequately address the major hypotheses outlined above.

Cheng et al  . reported no consistent differences in the occurrence of POCD at the timepoints measured and, most importantly, demonstrated no benefits of OPCAB at 1 yr. The authors state that the lack of consistency in testing approaches, as well as the small number of studies meeting criteria, underpower this meta-analysis to draw conclusions regarding important clinical differences in the incidence of POCD.

This report also concludes that there is no significant difference in the incidence of stroke between OPCAB and conventional coronary artery bypass. This conclusion should also be a cautious one, as the quoted aggregate risk of stroke from this meta-analysis is only 1% (lower than the average of 2–3% quoted in most studies), and the mean age was only 63 yr (younger than the national average undergoing CABG).1As the authors suggest, this indicates that the subjects selected for these studies were at low risk of stroke to begin with. This report includes only one study11 that focused on 65 high-risk subjects. Furthermore, the authors of this meta-analysis note that “most studies stated that they excluded high risk subjects.”

To illustrate the difficulty in assessing the benefit, or lack thereof, of OPCAB versus  conventional coronary artery bypass in affecting the occurrence of stroke among a population such as that presented in this meta-analysis, we offer the following. In a population having an average risk of stroke is 1% and in which one was interested in demonstrating a 20% reduction in the rate of stroke, you would need 35,994 subjects in each study group, or a total of 71,988 subjects in the randomized study with 80% power. A total of 16,554 subjects would be needed to demonstrate a reduction of 40% in the rate of stroke. It is clear that this meta-analysis, which includes slightly more than 3,000 subjects, could not appropriately evaluate arguably the most important potential clinical benefit, reduction in the likelihood of stroke.

The OPCAB technique is more technically demanding, and long-term graft patency using this technique versus  CABG on a still heart has yet to be determined. Although Cheng et al.  report that all-cause mortality at 2 yr was not different, their conclusions regarding graft patency are not substitutes for a quantitative analysis of graft patency, nor can we be sure that 2 yr is an adequate period of follow-up to understand the potential impact of poor revascularization.

One should be aware that most reports of nonrandomized comparisons between OPCAB and conventional coronary artery bypass document less complete revascularization, as fewer average grafts are completed in the OPCAB series,12 and complete revascularization of the lateral wall of the left ventricle in the circumflex artery distribution is especially challenging. Two recent randomized trials have produced conflicting results.13,14 Using data from real-world experience rather than randomized trials, the 3-yr outcomes for survival, death, or revascularization slightly benefited the OPCAB approach in a large (>9,000 off-pump and 59,000 on-pump cases) analysis of the Cardiac Surgery Reporting System in New York State.15 

How does the current report help define where research efforts should be directed to give surgeons better direction as to when and where to use the OPCAB approach? The work by Cheng et al.  helps to frame the important questions that have yet to be answered, which is one of the goals of any good meta-analysis.

Given the disparate reports on graft patency, it is not hard to believe that success depends on the experience and innate ability of a surgeon, and it is far more likely that many cardiac surgeons have practice patterns with less robust results than the best groups.13 Therefore, it is crucial that graft patency be studied in a cross-section of centers with different practice patterns. It is also likely that any neurologic benefit of OPCAB in decreasing stroke in a group of patients without elevated risk for stroke would be small and difficult to prove. However, to dismiss the potential neurologic benefit of the OPCAB approach may also miss the mark. Any compromise in coronary outcome may be a reasonable trade-off in a group at high risk for stroke.8 Therefore, the hypothesis that stroke can be decreased in a high-risk population can and should be tested with appropriate controls for important confounders. Cheng et al.  have helped to define these questions.

* Assistant Professor of Anesthesia, † Robert D. Dripps Professor and Chair of Anesthesia Professor of Medicine, University of Pennsylvania School of Medicine Philadelphia, Pennsylvania.