Two randomised controlled trials have produced what appears to be conflicting results for how left main percutaneous coronary intervention (PCI) compares with coronary artery bypass grafting (CABG), with one (EXCEL; Evaluation of Xience vs. coronary artery bypass grafting surgery for effectiveness of left main revascularisation) suggesting that PCI is non-inferior to surgery and the other (NOBLE; Nordic-Baltic British left main revascularisation study) indicating that it is inferior to surgery. However the studies used different primary endpoints, devices, and follow-up points—which may explain the different outcomes.
Writing in The New England Journal of Medicine, the EXCEL investigators Gregg Stone (New York Presbyterian Hospital and Columbia University Medical Center, New York, USA) and others report that studies—specifically the SYNTAX trial—have indicated that PCI may be a suitable alternative to surgery for selected left main patients (such as those with low or intermediate SYNTAX scores). However, they add that data for contemporary metallic stents (SYNTAX used a first-generation drug-eluting stent) and for contemporary surgical techniques, in the context of left main disease, are “warranted”. Similarly the NOBLE investigators Timo Mäkikallio (Department of Cardiology, Oulu University Hospital, Oulu, Finland) report in The Lancet that the non-inferiority margin of the randomised trials assessing left main PCI “was wide” and thus the trials were not powered to “definitively determine the best treatment for unprotected left main coronary artery disease”
But while both sets of investigators concur on the need for further data for left main disease treatment, the methods that they use to address this need are different. For example, in EXCEL, 99.2% of patients randomised to undergo PCI received an everolimus-eluting stent (Xience, Abbott Vascular) whereas the first 73 patients randomised to undergo PCI in NOBLE received a first-generation drug-eluting stent and thereafter, a biolimus-eluting stent (Biomatrix Flex, Biosensors) was the recommended stent.
Furthermore, the primary endpoint in EXCEL was different from that of NOBLE. Speaking at the 2016 Transcatheter Cardiovascular Therapeutics (TCT) meeting (29 October–2 November, Washington, DC, USA), where he presented the data in a late-breaking trial session, Stone said that endpoint was the three-year rate of a composite of death, stroke, or myocardial infarction (including periprocedural myocardial infarction)—with revascularisation excluded. He explained that he and his fellow investigators wanted an endpoint that would not require too large a patient sample to show a difference and, therefore, excluded certain endpoints. “The EXCEL leadership of cardiac surgeons and interventional cardiologists jointly agreed that revascularisation is not nearly as important an endpoint as death, large myocardial infarction or stroke. Revascularisation is no more important than many other adverse events that can occur after either revascularisation procedure, such as arrhythmia, bleeding, renal failure, major infection, etc. Thus we relegated it to a secondary endpoint along with all the others,” he said. Stone added that they also wanted to use a definition of myocardial infarction that would be large, prognostically important, and suitable for evaluating both PCI and CABG. By contrast, the NOBLE primary endpoint was the five-year rate of major adverse cardiac or cerebrovascular events (MACCE)—a composite of all-cause mortality, non-procedural myocardial infarction, any repeat revascularisation, and stroke.
In EXCEL, of 1,905 eligible patients (those with low or intermediate anatomical complexity), 948 were randomised to receive PCI and 957 were randomised to receive surgery. At three years, the primary composite endpoint occurred in 15.4% of PCI patients and in 14.7% of surgery patients (p=0.02 for non-inferiority). Additionally, at 30 days, this rate was significantly lower in the PCI patients: 4.9% vs. 7.9% for surgery (p=0.008 for superiority) and Stone et al report that this finding was driven by the lower rate of myocardial infarction in the PCI group (3.9% vs. 6.2%, respectively; p=0.02 for superiority). However, they add: “In a post hoc landmark analysis of the period between 30 days and three years after randomisation to PCI or CABG, more primary endpoint events occurred in the PCI group than in CABG group. Ischaemia-driven revascularisation during follow-up was more frequent after PCI than after CABG (12.6% vs. 7.5%; p<0.001 for superiority).” Stone told TCT delegates: “There is big upfront hit to take with surgery but if you get through the first 30 days, then you have a more durable procedure”.
Responding to Stone’s presentation, TCT panellist Craig Smith (Columbia University Medical Center, New York Presbyterian Hospital, New York, USA) called the study “a tremendous achievement” but added “once you took primary revascularisation out of the primary endpoint and loaded it with early events, then this outcome was more or less preordained”. He also predicted that the continued follow-up (five years is planned) would continue to show an advantage for CABG after three months.
Of the 1,201 patients in the NOBLE study, 598 were randomised to PCI and 603 to surgery. Mäkikallio et al report that, at five years: “Kaplan-Meier estimates of MACCE by intention-to-treat after five years were 29% (121 events) for PCI and 19% (81 events) for CABG. The hazard ratio was 1.48, exceeding the limit for non-inferiority (1.35) and was significant for superiority for CABG compared with PCI (p=0.0066)”. However, they state: “Notably, one year rates of MACCE in the two groups were the same (7% vs. 7%; p=1).”
NOBLE investigator Evald H Christiansen (Department of Cardiology, Aarhus University Hospital, Skejby, Aarhus, Denmark), who presented the data at TCT, immediately after Stone presented the EXCEL study, commented that he and fellow investigators would have included periprocedural myocardial infarction in the primary endpoint if they were to design the study today. Responding to Stone’s comment that “myocardial infarctions were myocardial infarctions”, he explained that the cardiac marker CKMB was not measured in all hospitals and the Society for Cardiac Angiography and Interventions definition of large procedure-related myocardial infarction was not defined at the time that NOBLE was initiated (2008).
In The Lancet paper, Mäkikallio et al acknowledge that the stent primarily used in their study had “a strut thickness above most types of currently used permanent metallic stents”. However, they note that Biomatrix Flex is still used in clinical practice with very good results (SORT OUT VI, The Lancet)
Given their different results, not surprisingly, Stone et al and Mäkikallio et al had opposing conclusions. The former state “For the treatment of patients with left main coronary artery disease and low or intermediate SYNTAX scores, PCI with everolimus-eluting stents was non-inferior to CABG with respect to the composite of death, stroke, or myocardial infarction at three years” while the latter conclude: “The findings of this study suggest that CABG might be better than PCI for treatment of left main stem coronary artery disease.”
Speculating on why the trials had different outcomes, Marc Ruel (Division of Cardiac Surgery, University of Ottawa Heart Institute, Ontario, Canada) said: “It is the endpoints that is the huge difference. I think the non-inferiority of PCI in EXCEL is driven by the higher incidence of periprocedural myocardial infarction. Once you pass the 30-day endpoint, the non-inferiority was not met for PCI. I think the five-year data (of EXCEL) is going to be interesting.”
Stone told Cardiovascular News that there were “several reasons” why NOBLE and EXCEL had different outcome, stating: “First, Noble excluded periprocedural myocardial infarction, a very important endpoint which favours PCI. Second, the stents used in Noble were inferior to those used in Excel, with a four times higher definite stent thrombosis rate (3% vs. 0.7%). Third, there was an absolute 3% higher incidence of stroke after PCI compared to CABG, a finding which is not biologically plausible and has not shown up in any other comparative trial. While likely due to chance, this nonetheless contributed to the observed data in NOBLE. Finally, EXCEL was substantially larger, and thus did not require revascularisation to be in the primary endpoint. The larger size allowed greater precision for the point estimate of death, large myocardial infarction, or stroke.”
Christiansen attributed the difference in outcomes to there being “five year follow-up in NOBLE, three year in EXCEL, and different endpoints”