OPCAB selection bias

Abstract

The article by Kshettry and associates [1Kshettry V.R. Flavin T.F. Emery R.W. Nicoloff D.M. Arom K.V. Petersen R.J. Does multivessel, off-pump coronary artery bypass reduce postoperative morbidity?.Ann Thorac Surg. 2000; 69: 1725-1731Abstract Full Text Full Text PDF PubMed Scopus (129) Google Scholar] represents an intriguing attempt to address the question of morbidity differences in contemporaneous cohorts of patients undergoing on-pump and off-pump coronary artery bypass grafting (CABG). Equivalent adjusted-risk cohort groups were identified, and a retrospective review demonstrated equal fatal and morbid endpoints, with the exception of reoperation for bleeding. The questions I raise are directed more to the discussants of the report than to the authors. Drs Mack and Novitzky, who are pioneering advocates of off-pump CABG, expressed pessimism with the authors’ conclusions that, save bleeding, there were no significant differences between the on-pump and off-pump group outcomes. In particular, Dr Mack’s claim that the mortality results, equivalent in both groups, were not risk adjusted, is not supported in the manuscript. The discussants premised that study bias selected “high-risk” patients into the off-pump cohort and concluded that this benefited the outcome in the on-pump group. “Risk” was determined by the National Society of Thoracic Surgeons STS Logistical Regression Risk Model. This model is based primarily on the presence or absence of defined comorbid conditions that substantially influence risk relative to cardiopulmonary bypass as well as to the cardiac procedure per se. Coronary artery anatomical features do not factor as importantly in this risk assessment model; however, as all of us who perform beating-heart CABG have discovered, the nuances of the coronary anatomy (vessel diameter, sequential lesions, degree of calcification, need of endarterectomy) predominantly determine outcome [2Wait M.A. Treatment of coronary heart disease with minimally invasive surgery.Baylor U Med Center Proc. 2000; 13: 121-127PubMed Google Scholar]. Because the practice pattern of many programs (appropriately so) is to perform off-pump beating-heart CABG on the patients with the most favorable coronary arteries [3Wait M.A. What is the role of minimally invasive surgery in revascularization of patient?.ACC Ed Highlights. 1998; 14: 6-8Google Scholar], a selection bias may very well exist that would have a negative impact on the outcomes in the on-pump CABG group; that is, smaller vessels, which are more difficult to bypass, may well be overrepresented in the on-pump group. This is analogous to analyzing the outcomes between contemporaneous cohorts of patients undergoing percutaneous transluminal coronary angioplasty and CABG and ignores the fact that selection bias favors patients with more advantageous coronary anatomy in the percutaneous transluminal coronary angioplasty group. Indeed, the most common reason patients are referred for CABG in many centers is simply because percutaneous revascularization procedures are technically not feasible. It would be instructive to learn if the discussants or authors considered whether this premise, that study bias would have selected more difficult coronary anatomical features in the on-pump group, had an influence on the findings of this particular study. OPCAB selection bias: ReplyThe Annals of Thoracic SurgeryVol. 71Issue 5PreviewI appreciate the opportunity to respond to Dr Wait’s letter regarding my discussion of the study by Kshettry and associates [1]. Dr Wait raises multiple issues including risk-adjusted mortality, selection bias in the on-pump or off-pump group affecting outcomes, and the role of coronary artery anatomical features not factoring into risk assessment models. The implication of the last is that by default, selecting patients with adverse coronary artery anatomy (eg, small vessels, distal disease, and calcium) into the on-pump surgical group may have led to a selection bias against favorable outcomes in the on-pump group. Full-Text PDF OPCAB selection bias: ReplyThe Annals of Thoracic SurgeryVol. 71Issue 5PreviewAs stated in the presentation and in the published article [1], risk stratification was reported in both populations according to the National Society of Thoracic Surgeons Logistical Regression Risk Model and resulted in a nonsignificant difference between groups. Our review did not consider coronary anatomy as a predictor of patient selection. Appropriately argued, selection bias was and will prevail at some level in a retrospective review, and therefore, prospective, randomized trials are in their purest description a true method of eliminating preoperative selection bias. Full-Text PDF