dard" therapy that render trials "outdated". The results of a randomized trial do not necessarily apply to populations that were systematically excluded from that trial. Because most trials, especially procedural trials, exclude patients with complex and high-risk conditions, their results provide the most guidance for the simplest decisions. Almost all CABG trials have systematically excluded patients within 7 to 30 days of an acute MI, patients with LVEF less than .35, hemodynamically unstable patients, patients with anatomy deemed unfavorable for CABG (no anatomic contraindication is harder to define than "diffuse disease"), patients older than 70 years of age, patients with one or more prior heart surgeries, and patients with severe comorbidities, including but not limited to chronic obstructive pulmonary disease, pulmonary hypertension, prior stroke, cancer, severe liver disease, and severe renal failure.
The results of randomized trials do not necessarily apply in the same way if either the control (standard) therapy or the active therapy has undergone major improvement since completion of the trial. With highly significant improvements in survival resulting from the use of aspirin, statins, P-blockers, and ACE inhibitors/ARBs, the likely long-term mortality of medical therapy arms in 2008 is substantially better than it was in the 1960s, when medical therapy was compared with CABG. Although CABG has also improved since the 1960s, there is no body of large randomized trials showing superiority of current techniques to those of 40 years ago. For example, although it is believed that routine use of the left internal mammary or thoracic artery as a conduit is an important clinical advance, there is not one randomized trial comparing use of the internal mammary artery with saphenous vein graft implantation into the same recipient vessels, under the same conditions, with the same adjuncts and follow-up. This dearth of prospective randomized trials also applies to cardioplegic techniques, and it stands in marked contrast to PCI advances such as bare metal stents, drug-eluting stents, thienopyridines, direct thrombin inhibitors, and glycoprotein IIb/IIIa inhibitors, each of which has been compared with "standard" therapy in thousands of randomized subjects.
Beyond the generic limitations of randomized trials, there are additional limitations on trials of procedures. In revascularization trials, blinding of subjects or caregivers is impossible. Because caregiv-ers provide medications and patients choose whether to use them, undergoing a procedure can bias the subsequent use of medications known to alter survival among patients with CAD. Put another way, it can be said that any good clinician is trying to bias the outcome of his or her patients toward a favorable result. The clinician does this by prescribing and encouraging compliance with evidence-based medical therapy, among many other ways.
Additionally, the sample sizes of revascularization trials are an order of magnitude less than those of the megatrials of pharmacologic agents, and operators tend to select low-risk patients. Low risk means low event rate, and smaller numbers of subjects with lower event rates translates into significantly less power to detect meaningful differences in trials of procedures. The power of a study to detect differences is also reduced by large numbers of crossovers, which are expected in any trial of a procedure that is effective.
Was this article helpful?