Off-pump vs. on-pump CABG: are we any closer to a resolution?

Afilalo and colleagues have presented an updated meta-analysis and meta-regression of randomized trials of off-pump vs. on-pump coronary artery bypass grafting (CABG) surgery. Their two major findings are that while there was a reduction in mortality, myocardial infarction, and stroke in the off-pump group, this only reached significance for the latter, and that the effect of off-pump CABG was similar on all clinical outcomes regardless of age, gender, number of grafts, and trial publication date. The two key questions are: (i) Are these conclusions justifiable based on the data presented? (ii) How relevant are the findings to routine clinical practice? Before answering these questions it is worth addressing (i) the current situation with off-pump CABG and (ii) the appropriateness of the nature and conduct of the meta-analysis and the meta-regression analysis. Off-pump surgery was initially proposed almost three decades ago to allow CABG in developing countries without the need for cardiopulmonary bypass that was beyond the economic reality of vast numbers of potential patients. Despite initial scepticism about its technical feasibility, off-pump CABG was gradually adopted by some surgeons in developed countries who believed that the elimination of cardiopulmonary bypass could potentially and substantially eliminate the adverse clinical consequences of extracorporeal circulation. In developed countries this was particularly relevant in a progressively elderly population undergoing cardiac surgery in view of their greater burden of existing co-morbidities and their greater susceptibility to the potentially deleterious consequences of cardiopulmonary bypass. On the other hand, critics maintained that off-pump surgery led to inferior revascularization through a reduction in both the number and quality of bypass grafts and that no study appeared to demonstrate any substantial clinical advantage of the off-pump technique over the ‘gold standard’ conventional technique using cardiopulmonary bypass. These differing views have remained essentially unchanged after 25 years of debate. Off-pump CABG plateaued in Europe and the USA a decade ago, and since then has remained at 15–20% of all CABG operations. However, it is important to realize that these numbers encompass the practices of a small number of surgeons who do virtually all CABG off pump and the majority who essentially do none using this technique. In contrast to Europe and the USA, in Asian countries off-pump CABG is performed in 60–100% of the whole CABG population. From a statistical perspective, to determine the overall effect of off-pump surgery the trial data were pooled and a meta-analysis performed using a random effects model. One of the major strengths of the meta-analysis of Afilalo et al. is that by including more recent studies, the pooled sample size is almost three-fold larger than previously published meta-analyses, thereby giving far greater statistical power and narrower confidence intervals (CIs), with a reduced level of uncertainty regarding treatment effects. Secondly, a unique feature of the study is use of meta-regression for the first time that allowed the investigators to examine individual trial characteristics such as mean age, sex, and number of grafts on the effects of off-pump surgery on outcomes. In summary, the analyses were performed in almost 9000 patients from 59 randomized trials. Post-operative mortality was 1.6% in the off-pump group and 1.9% in the on-pump group [risk ratio (RR) 0.9; 95% CI 0.63–1.3] and for myocardial infarction the respective figures were 3.4% and 3.9% (RR 0.8; 95% CI 0.69– 1.13). The most important finding was a one-third reduction in the incidence of stroke from 2.1% in the on-pump group to 1.4% in the off-pump group (RR 0.7; 95% CI 0.49–0.99). For each of these outcomes meta-regression did not identify significant variation in the off-pump vs. on-pump techniques in relation to mean age, proportion of females, trial date, and the number of bypass grafts. This is not the last word, however, as meta-regression based on study-level aggregate data has low power for patient characteristics. So we strongly support the authors’ suggestion of a meta-analysis using individual rather than pooled participant data.