In patients with coronary artery disease and heart failure, the addition of coronary-artery bypass grafting to medical therapy has no effect on the overall risk of death from any cause, but assessment of suitability for surgery may predict survival benefit in subgroups
- Division of Cardiovascular Surgery, University Health Network, Department of Surgery, University of Toronto, Canada
- Correspondence to: Terrence M Yau
4N-470, Toronto General Hospital, 200 Elizabeth Street, ON M5G 2C4, Canada.
More than 5.8 million Americans and 500 000 Canadians are currently living with heart failure (HF). The 5-year survival rate is a dismal 40–50%. Coronary artery disease (CAD) is the most common cause of HF, but the role of coronary artery bypass surgery (CABG) in HF has never been definitively established.
The Surgical Treatment for Ischaemic HF trial aimed to determine whether CABG improves survival in patients with CAD and impaired left ventricular (LV) function (hypothesis 1). A total of 1212 adults with CAD amenable to CABG and left ventricular ejection fraction (LVEF) ≤35% were enrolled from 127 sites across 26 countries. The original intended enrolment of 2000 patients was revised due to slow enrolment rates, necessitating an increase in the duration of follow-up from 3 to 5 years. Patients without significant left main (LM) stenosis and having Canadian Cardiovascular Society (CCS) class I–II symptoms could be enrolled into either stratum A or B. Patients who were judged to be candidates for surgical ventricular remodelling (SVR) were placed into stratum B and randomised to one of three arms: medical therapy, medical therapy+CABG or medical therapy+CABG+SVR. Otherwise, they were assigned to stratum A and randomised to either medical therapy or medical therapy+CABG.
In this trial, 17% of the medical group underwent CABG, while only 9% of CABG patients crossed over to medical treatment. An intention-to-treat analysis demonstrated no significant difference between the two groups (CABG, no CABG) in the primary outcome of all-cause mortality. When as-treated and per-protocol analyses were performed, the HRs for the primary outcome with CABG were 0.70 and 0.76 respectively (both were statistically significant). All secondary outcome measures, except for 30-day mortality, favoured CABG.
The investigators are to be applauded for having conducted an extremely large and complex trial in a difficult patient population. The randomisation process within each stratum was stringent, reflected in balanced demographic and clinical characteristics. Although this study may be interpreted as a trial to assess the benefit of CABG in HF patients, it is important to note that a large group of patients with HF and preserved systolic function, accounting for as many as 30–40% of HF hospital admissions, were not captured.
In addition, LVEF was measured as far as 3 months pre-enrolment. The 17% of medical patients who crossed over to surgery at a median interval of 142 days may have benefited from medical optimisation, resulting in interval improvement in LVEF and a reduced risk profile at the time of surgery. This would cause a bias in the results in favour of the medical group. The imbalance in crossover rates leading to a diminishment of the treatment effect of CABG by intention-to-treat analysis has been well recognised by the trial investigators and various editorialists.
Patients were excluded from the study by virtue of LVEF, LM stenosis and CCS class, whose measurements are subject to interpretation. Without a registry and screening-to-enrolment ratios, it is difficult to know what proportion of those excluded in fact underwent surgery. This potential selection bias may be further compounded by stratum assignment. Patients assigned to stratum A had a 50:50 chance of receiving either medical treatment alone or with CABG, while those assigned to stratum B had a 67% chance of undergoing CABG. The decision to assign a patient into stratum B was based on ‘dominant akinesis/dyskinesis of the anterior LV wall amenable to SVR’. If a surgeon believes that the patient is likely to benefit from and survive the surgery, assignment into stratum B ensures that the patient is more likely to undergo an operation. An assignment bias of this sort would be expected to result in a greater advantage of CABG in stratum B versus stratum A. Indeed, the HRs for death from any cause were 0.94 for CABG in stratum A, but 0.48, favouring CABG, in stratum B (p value 0.01). If the number of patients enrolled into stratum B (n=216) rivalled that of stratum A (n=1061), the conclusions of the trial might well have been different.
As a result of this landmark trial, the ACCF/AHA guidelines were modified from their 2004 recommendations to consider CABG for improving survival in patients with severe LV dysfunction1. The decision to operate resides evermore so on the judgement of clinicians in their comprehensive review of the individual patient. If one judges a patient to be a good surgical candidate, CABG will likely provide a mortality benefit. If one judges a patient to be a poor surgical candidate, a decision to provide best medical therapy alone is now rooted more in evidence and less in uncertainty.