To supplement the anatomic conclusions derived from the aforementioned 1960s trials, the practice guidelines and other authoritative sources cite the results of several large prospective registries. The Duke database is a long-term, single-center registry that is often cited, and the State of New York, VA, and Society of Cardiovascular Surgeons all maintain large, multicenter registries. These registries have the advantages of consistent definition of variables. Nevertheless, they all are prone to the epidemiologic triad of alternative explanations for their results; specifically, selection bias, confounding, and/or information bias may be a possible or even a probable alternative explanation for many of their findings.
With regard to the possibility of selection bias, the conventional wisdom, during the entire period that most registries have existed, has been that patients with left main and/or three-vessel disease have better survival with CABG.3,39-41 This conclusion has been promulgated by the ACC/AHA Guidelines.3 Under these circumstances, what kinds of patients with left main and/or three-vessel CAD would not be offered CABG at Duke or in the State of New York? Among the categories of patients likely to be denied CABG are the following groups: patients with shock, cancer, "diffuse disease," or life-threatening comorbidities; moribund patients; demented patients; patients with recent cardiac arrest; and so on. At the end of 10 years, the favorable patients with left main and/or three-vessel disease who underwent CABG may have more survivors than the medical therapy group, but this may be as much a function of the inclusion of such high-risk patients only in the medical arm as of a difference between CABG and medical therapy. This sampling problem cannot be adequately dealt with by any multivariable method unless all of the significant groups have been defined and measured during the period of the registry. This limitation also applies to multicenter registries and is not made less important by simply having a larger registry.
Confounding is the mixing of effects. Because of the multiplicity of anatomic and physiologic associations, as well as treatment with clinical associations, it is another extremely common form of systematic bias or alternative explanation for study results. The major advantage of the randomized trial is that an adequately powered randomized study tends to balance both known and unrecognized confounding factors; therefore, the observed results of the trial can be safely inferred to result from differences in treat ment rather than differences in the patient groups. A simple example is that most patients who undergo CABG go through some form of postprocedure rehabilitation, during which they are likely to get risk factor and medication reinforcement that they would not get if they were in the medical arm of the registry. Sophisticated attempts to "correct" for inadvertent imbalances between compared groups (e.g., propensity analysis) are dependent on the validity and objectivity of the measurements used to define clinical and angiographic and procedural characteristics. Clearly, one cannot "correct" for undefined and unmeasured variables. Perhaps the most common anatomic factor cited for "turning down" patients for CABG, and a commonly cited reason for procedural MI or other adverse outcome of CABG, is "diffuse disease" or "poor targets." This factor is so subjective that no trial or registry has defined it or attempted to measure it.
The third major alternative is information bias, which means misclassification or systematic differences in the way outcomes are measured. Specific examples include the detection of MI or stroke after PCI versus after CABG versus among medically treated patients. The PCI Guideline recommends highly sensitive troponin measurement for every PCI patient with chest pain and even gives a IIa ("ought to consider") recommendation to routine troponin measurement after PCI. In contrast, CABG patients with chest pain or change in mental status are often assumed to have surgery-related or intensive care unit-related symptoms, and measurement of enzymes or electrocardiographic recording is usually eschewed. Medical patients are probably in the middle, getting electrocardiograms and/or enzyme analysis primarily based on their presentation with new symptoms.
A fundamental result of reviewing randomized trial comparisons of two treatment options is the tendency to see them as "either-or" rather than complementary options.1,2 In particular, the fact that the major rationale for CABG, historically, has been anatomic has fostered an emphasis on the anatomic differences (advantages/disadvantages) between CABG and PCI. Nevertheless, the physiologic differences between the two revascularization techniques turn out to be every bit as important.
The major advantage for CABG, which has a structural or anatomic basis, is the ability to achieve "complete revascularization," even in the face of one or more chronic total occlusions. This allows for the application of CABG to anatomic subsets that are not approachable with PCI. A second major advantage of CABG, namely the superior durability of its results (at least compared with bare metal stents), is at least partly structural, in that bypass conduits protect territories, whereas stents treat lesions. But another major advantage of CABG has been its near-simultaneous treatment of multiple vessels at the same procedure, which has been possible because the patient's physiologic state is "controlled" by general anesthesia, intubation/ventilation, heart-lung bypass, sternotomy exposure, cardioplegia, and so on. It is from these control factors that two of the major disadvantages of CABG compared with PCI derive.2 First, it takes finite time to initiate these control features, and this time can translate into necrosis if the patient is suffering from an acute MI. Second, each of these control features contributes to the overall morbidity of CABG. This morbidity is also enhanced if CABG is undertaken during an MI. Other patient comorbidities, such as severe cerebrovascular disease or severe pulmonary disease, can compound the interaction of MI, time, and the invasive control features of CABG.2,42
In contrast, the major advantages of contemporary PCI (which includes stents, thienopyridines, and gly-coprotein IIb/IIIa inhibitors, among other adjuncts) are speed of achieving normal or near-normal perfusion and relatively mild morbidity. Because of these factors, patients with an acute MI can often be rapidly stabilized by a single-lesion procedure, with more complete revascularization achieved by either PCI or CABG during subsequent procedures, which they undergo at much lower risk (because their ischemic, hemodynamic, and/or electrical instability has been relieved). All of these advantages apply to acute coronary syndromes (STEMI and NSTEMI), which are purely physiologic subsets.
The major disadvantages of PCI include inability to treat chronic total occlusions and inability to protect territories rather than lesions. These are structural issues that apply primarily to stable patients.
Medical therapy and risk factor modification have been shown to provide clinical benefit for stable and unstable patients.43-46 Furthermore, stability implies time to document ischemia that is medically refractory, whereas instability implies ischemia that is refractory. Acute coronary syndrome is a reason to consider emergent or urgent revascularization, most often by PCI. Stable coronary syndromes provide time to optimize medical management and document ischemia, before considering either CABG or PCI. Neither PCI nor CABG provides either a "cure" or prevention of future events. Both PCI and CABG (like acute MI) can be viewed as opportunities to help patients change their lifestyles to become more cardioprotective.43-46
Was this article helpful?