Speakers at the European Society of Cardio-Thoracic Surgery annual meeting (EACTS 2021; 13–16 October; Barcelona, Spain and virtual) addressed the thorny topic of left main revascularisation, with discussion focusing on the controversy surrounding the EXCEL trial and the reflection of the data in current clinical guidelines.
EACTS attendees were shown previously unpublished data from EXCEL during a presentation by Nick Freemantle, director of the Comprehensive Clinical Trials Unit at University College London (London, UK), pointing towards a higher incidence of death, stroke or myocardial infarction (MI) among patients randomised to percutaneous coronary intervention (PCI), contradicting the trial’s initial finding, that was ultimately used to inform guidelines on left main revascularisation.
During his presentation, Freemantle discussed methodological challenges with composite primary outcome measures in revascularisation trials in general, but highlighted EXCEL as a “beautiful textbook example” of a composite endpoint “going spectacularly wrong”.
EXCEL was an international, open-label, multicentre, randomised trial that compared PCI using an everolimus-eluting stent (Xience, Abbott) and coronary artery bypass grafting (CABG) for the composite endpoint of death, myocardial infarction (MI), and stroke in patients with left main disease. The trial concluded that there was no significant difference between the procedures, though the study’s findings have been subject to intense debate, in particular due to questions over the definition of MI used in the trial.
During his presentation at EACTS 2021, Freemantle singled out the primary composite endpoint used in EXCEL as being “not fit for purpose” and showed a comparison of the published trial results in which PCI and CABG exhibited comparable rates of a primary composite endpoint of death, stroke and MI, based upon the study’s definition of procedural MI, contrasted with an analysis in which the same outcomes were assessed based upon the third universal definition of MI (UDMI). In the data shown by Freemantle, PCI was shown to be non-inferior to CABG for the composite primary endpoint, putting it at odds with the trial’s initial findings.
Freemantle told Cardiovascular News that the data had been date stamped contemporaneously with the three-year publication of the EXCEL trial results, containing a complete entry of the UDMI data which had been listed in the protocol as a secondary outcome of the trial, but not published alongside the initial results in the New England Journal of Medicine.
These data did not come to light until after an investigation by the journal and the London School of Hygiene & Tropical Medicine (LSHTM), and were eventually published in 2020, albeit long after the publication of joint guidelines from EACTS and the European Society of Cardiology (ESC) concerning myocardial revascularisation.
“I think this is a real problem,” Freemantle commented in his presentation at EACTS. “I would say that the EXCEL composite outcome is not fit for purpose. The UDMI is a well-established definition [that] is used in many trials. If the authors’ version of the outcome measure of MI, which is new for this trial, gives us a completely different result, then what should we believe?”
Concerning the delayed publication of the secondary outcome data including the UDMI analysis, Freemantle said he was “saddened” that these had not been shared by the EXCEL authors earlier.
“I am also saddened to discover that they truncated the publication of the results at three years, when they had recruited over a period of time, and they had patients with follow-up out to beyond five years,” Freemantle added. “It is fine for them to do that in their primary analysis, but they should also, surely, be sharing with us and with clinicians and patients everything they have got to inform decisions, and I think that the results of that have been really, frankly, very serious.”
The EXCEL data shared by Freemantle were brought into focus a second time during the meeting, on the second occasion by the president of the Latin American Association of Cardiac and Endovadcular Surgery (LACES) Victor Dayan (Universidad de la República, Montevideo, Uruguay) in a session concerning evidence and trial updates. In his presentation covering new evidence in left main coronary artery disease, Dayan argued in favour of new guidelines incorporating the latest data.
ESC and EACTS initiated a review of the available evidence for the treatment of patients with left main coronary artery stenosis in September 2020 after EACTS withdrew its support for the left main recommendations of the societies’ joint guidelines on myocardial revascularisation.
Making the case in favour of new guidelines covering left main coronary artery disease, Dayan made a number of points including arguing that there is a lack of evidence supporting the use of the SYNTAX Score to support decision-making in left main revascularisation as well as commenting that left main disease should not be considered as a separate clinical or statistical entity to multivessel disease. Furthermore he argued that five-year data from the EXCEL trial points to a higher rate of mortality for PCI, and a higher risk for the primary outcome.*
On the SYNTAX Score, Dayan argued that both EXCEL and NOBLE found no interaction between the SYNTAX Score and the primary outcome. “How reproducible is the SYNTAX Score if we are to use it as a way to make recommendations for left main revascularisation?” Dayan posited.
“This was evaluated by the EXCEL authors, and they found that in 50% of cases the SYNTAX Score reported at site was lower than reported at core lab, which suggests that interventional cardiologists at site underestimate the SYNTAX Score. Furthermore, in 24% of cases the risk score according to core lab was high, which actually is a class III recommendation for revascularisation in the left main. Such heterogeneity is score assignment along with risk of harm should not be guiding clinical practice for LM revascularisation”.
Turning to the EXCEL data presented by Freemantle, Dayan commented: “When you use the universal definition of MI, you see that not only is non-inferiority rejected for PCI, but also PCI is found to be worse than CABG.”
Furthermore, Dayan noted that the five-year results of the NOBLE trial have shown that PCI was worse for the primary outcome. “PCI was associated with more MI and new revascularisation, and this was especially true in the low SYNTAX Score group, which was a matter of concern for the authors that they are given an equal level of recommendations either for PCI or CABG,” he said.
Speaking after Dayan, interventional cardiologist Patrick Serruys (National University of Ireland, Galway, Ireland) offered a perspective on left main randomised controlled trials at five- and 10-years follow-up. Serruys commented that regarding randomised trials of left main revascularisation there are essentially “three debates in one”, including on the definition of peri-procedural MI (PMI) and its impact on composite endpoints; all-cause mortality or major adverse cardiovascular events (MACE) in PCI versus CABG in patients with left main coronary artery disease; and appropriateness of PCI or CABG in the treatment of patients with left main and three-vessel disease.
“Definition of PMI has a major theoretical impact on time to event curve, and a composite endpoint at five years,” said Serruys.
Serruys also commented on the 10-year PCI versus CABG data from the perspective of average treatment effect analysis. He argued that from this standpoint, the decision to opt for either treatment strategy should be viewed on an individualised basis.
“Based on the average treatment effect should you send all your patients to surgery? The answer is of course: no. We are moving from average treatment effect to individualised prognosis. We are creating decision tools to improve personalised care. We are moving the heart of medicine towards science,” he commented.
Presenting data detailing the average treatment effect assessed in clinical trials, he noted: “You have a mixed bag of patients, some expected to derive benefit from the alternative treatment, [some] expected to have equivocal response, and [some] expected to be harmed by the treatment, and an estimation of the average treatment effect.
“We have to identify the heterogeneous response to the treatment. We have to segregate the patient population based on treatment response, see who is going to be harmed in the trial, who is going to benefit in the trial, and who is going to be equivocal.”
Serruys, who was an investigator on the EXCEL trial, was later questioned by Freemantle on the reasons for the delayed publication of the UDMI data for the trial, however Serruys refused to be drawn on the issue.
*Amended 2/11/2021: Original text incorrectly stated that Dayan had suggested NOBEL pointed to a higher rate of mortality at five years with PCI.