Surgical decision-making and the ART of risk factor exploration

Despite near-universal acceptance of the use of a single internal mammary artery (IMA) graft to bypass the left anterior descending coronary artery, as well as considerable retrospective data supporting a survival benefit for bilateral IMA (BIMA) grafting [1], clinical acceptance has been guarded, with the most recent report from the Society of Thoracic Surgeons database recording BIMA use in only 5.5% of coronary artery bypass (CABG) operations [2]. However, non-randomized studies suggesting a benefit from bilateral IMA procedures may have also been biased towards patients with better prognosis [3]. Therefore, a clearer understanding of how multiarterial grafting may interact with specific patient factors might well help to resolve the controversy. In short, the question becomes not so much which operation is better, but which operation is better for which patient. It is specifically in this realm of inquiry that the post hoc analysis of the Arterial Revascularization Trial (ART) by Talukder et al. [4] in this issue of the journal provides potentially important information. The trial took place from 2004 to 2007 and enrolled 3102 patients from 27 hospitals in 7 countries (Australia, Austria, Brazil, India, Italy, Poland and the UK). Trial participants were randomized to receive either a single or bilateral IMA graft. Operating surgeons were allowed to exercise clinical judgement and modify the procedure as clinically indicated. As a result, 13.9% of patients randomized to the bilateral group ended up getting a single artery graft and 3.9% of those assigned to the single group ended up getting bilateral IMAs. In addition, 19.4% of patients in the bilateral group and 21% in the single groups received additional radial artery grafts [5]. In the intention-to-treat primary analysis, there was no survival difference between the single and bilateral IMA grafting, yet the BIMA group had a higher rate of sternal wound complications. In the current report, the authors leverage the clinically robust and carefully adjudicated information of a well-conducted clinical trial to analyse the specific factors associated with 10-year mortality. Complete follow-up was obtained in 3040 (98%) patients. At 10 years, 644 (20.8%) of the patients were known to have died, providing sufficient statistical power to examine factors that may have small but still meaningful effects [6]. The authors found that age, the serum creatinine concentration, diabetes mellitus, left ventricular ejection fraction, smoking status, prior cerebrovascular events and atrial fibrillation were associated with mortality following arterial revascularization. Most of these factors are known to be associated with increased overall mortality in the population. Of these factors, smoking may be the most amenable to intervention. Optimizing diabetic control and treating atrial fibrillation may have salutary effects as well, but the remaining factors are not easily modifiable. Although there is a certain wisdom to using the data from well-done prospective randomized control trials (RCTs), as there is a high degree of completeness, adherence to clear variable definition and data accuracy, there are clear potential limitations. As RCTs focus on the average treatment effect [7], there are multiple selection criteria that may impair the ability to generalize the findings. Comparing the ART participants with representative patient samples in large registry analyses [8], it is apparent that ART participants were generally healthier, with fewer comorbidities and underlying health conditions. Moreover, hospitals in North America, Africa, East Asia or Spanish-speaking South America were not included, and the population characteristics in these regions may differ significantly from those included in the trial. For example, 91.9% of the trial participants were of Caucasian ancestry, 4.8% were South Asian and only 0.5% African. The racial and ethnic mix of North American populations is significantly different. Since racial and ethnic minorities suffer disparities in susceptibility to cardiovascular disease and in the utilization of revascularization procedures [9], procedure-specific predictors of late mortality may vary considerably from those herein reported. Moreover, in the hospitals included in the study, referral patterns related to health insurance availability may differ substantially from those in other regions, potentially leading to differences in the severity of the disease at which a patient is considered a candidate for surgery. Therefore, population heterogeneity and sample size may help to account for the more robust list of risk factors emerging from registry data. Despite these limitations, there remain, even within the realm of the data available, several unanswered questions. Although the cause of death from registry and death certificate sources is notoriously inaccurate, the information from carefully performed RCTs can be highly informative. It would therefore be helpful to identify those factors specifically related to cardiac-related mortality. Perhaps more compelling, there were some unique features of the ART. Compliance with optimal medical therapy was extremely high (even for an RCT). In ART, after 10 years, compliance with angiotensin‐converting enzyme inhibitors was 71%, 90% for statins, 81% for aspirin and 74% for β‐blockers. Noncompliance with Guideline Directed Medical Therapy (GDMT) after CABG has been found to decrease long‐term survival, freedom from myocardial infarctions and the need for repeat coronary revascularization procedures [10]. Was there any interaction between compliance (or noncompliance) with GDMT and factors associated with mortality? For example, was the long-term risk for mortality in diabetics or patients with impaired ejection fraction ameliorated by high compliance with GDMT? Such information might directly inform the management essential to achieve optimal results following CABG. Similarly, was there any interaction between surgical approach and risk factors for late mortality? Authors have recently reported an as-treated analysis of the ART favouring multi-arterial grafting [11]. Are there any patient characteristics that can be specifically identified that particularly help to define the population most likely to benefit? Such information would help to address the long-unanswered question of how to best define the patients most likely to benefit from multi-arterial grafting. Ultimately, surgical decisions need to be carefully individualized, as should surgical risk estimates. Talukder’s report provides an insightful introduction; much work remains to be done.