UPDATED November 5, 2019 // LISBON, Portugal ― Many cardiovascular clinical trials have been controversial, but few have had their controversies explode on an international stage in quite this way.
A noted cardiothoracic surgeon who is an early member of the EXCEL trial's executive committee took a conference podium to charge that trialists "manipulated" data in a way that tilted the results in favor of percutaneous coronary intervention (PCI) compared to coronary bypass surgery (CABG).
At the heart of the accusations: an alleged midtrial alteration in the definition of a key part of the composite primary endpoint.
The trial's primary investigators just as vehemently deny there was any such change, and they point to publicly available protocols that appear to back them up.
The EXCEL trial's 5-year follow-up analysis, recently published in the New England Journal of Medicine (NEJM), yielded no significant difference in the primary endpoint of death from any cause, myocardial infarction (MI), or stroke between patients who had undergone PCI or CABG for left-main (LM) coronary lesions of low to intermediate anatomic complexity.
The findings, which were recently covered by theheart.org | Medscape Cardiology, were similar to those at 3 years, which the same journal reported in 2016.
However, the definition of MI was changed during the course of the trial, skewing the eventual primary outcome in a way that diminished what would have been a clear advantage for CABG, alleged David Taggart, MD, PhD, University of Oxford, United Kingdom, before a crowd here at the 33rd European Association for Cardio-Thoracic Surgery (EACTS) Annual Meeting.
Taggart also questioned consideration of periprocedural MI in the primary endpoint along with spontaneous MI, which he sees as more prognostically important.
EXCEL defined periprocedural MIs as those occurring within 72 hours of the procedure, largely in terms of levels of the MB isoform of creatine kinase (CK-MB), supplemented by electrocardiographic and imaging criteria.
The definition is consistent with one for revascularization-associated MI proposed in 2013, after the EXCEL protocol was published, by the Society of Cardiovascular Angiography and Interventions (SCAI).
"To equate a periprocedural biochemical definition of myocardial infarction, and give it the same weight as a nonprocedural myocardial infarction, was an absolute outrage in my opinion," Taggart said. The statement drew some applause from the audience.
"So I believe the data was manipulated, using a changed definition of myocardial infarction, to try and prove that for the composite endpoint, there's no difference," Taggart said at the EACTS session. "What happened in EXCEL was a disgrace."
Denials from several of the study's four principal investigators (PIs) have been just as fervent.
"The definition was not changed during the study," said one of the PIs, Arie Pieter Kappetein, MD, PhD, who spoke from the audience directly to Taggart after his EACTS presentation.
"It's a lie, it's misleading the audience here," he said.
Kappetein is a Medtronic vice president and the company's chief medical officer, Structural Heart business, and a surgeon at Erasmus University Medical Center, Rotterdam, the Netherlands.
(Editors note: This article has been corrected to note Kappetein's executive position at Medtronic, which he assumed in 2017.)
No change was made to the periprocedural-MI definition either before or after the EXCEL data were unblinded, agreed another PI, interventional cardiologist Gregg W. Stone, MD, Icahn School of Medicine at Mount Sinai, New York City.
"No one has any idea what he's talking about, and it is an outright falsehood," Stone told theheart.org | Medscape Cardiology by email, referring to Taggart's allegations.
A Tale of Two Protocols
Stone referred to two EXCEL protocols, one labeled version 4.0, dated July 22, 2011, and another labeled version 10.0, dated February 8, 2019, that are available to all at NEJM.org. There are no apparent differences in their enzymatic definitions for periprocedural MI included in the appendixes of the two documents.
The main bodies of both protocols specify "protocol-defined MI" and "MI adjudicated per Universal MI Definition" as distinct secondary endpoints. The "universal definition" does not include enzymatic criteria such as those in the SCAI definition.
Taggart also alleged that an elevated all-cause mortality risk for patients who had received PCI was framed in a way that hid a critical trial message in Stone's presentation of the 5-year outcomes at the recent Transcatheter Cardiovascular Therapeutics (TCT) 2019 conference.
With respect to the primary endpoint — the only endpoint for which the trial was adequately powered at 5 years — the odds ratio (OR) for PCI vs CABG was 1.19 (95% confidence interval [CI], 0.95 – 1.50). The difference was apparently driven by an all-cause mortality OR of 1.38 (95% CI, 1.03 – 1.85), given that ORs for MI and stroke were not markedly increased.
But in the EXCEL publication, figure 3 shows all-cause mortality as essentially equal between PCI and CABG for the first 1 to 2 years, after which it continuously diverges such that their rates spread to 13.0% vs 9.9%, respectively, by 5 years.
Although the 5-year rates are not significantly different, "it's the fact that mortality is accelerating. And that wasn't even shown at TCT," Taggart said.
"In my 13 years of being involved in authorships and publications, I have never witnessed such an attempt to distort what the actual data in this paper showed," he said.
"He is not justified in focusing on secondary endpoints that lack power," Stone said. It's also wrong to ignore evidence from other trials and meta-analyses that suggest mortality after PCI and CABG in left-main disease is 'almost identical,' " he said — for example, in the NOBLE trial at 5 years and SYNTAX after 10 years.
"Plus in EXCEL, there was no significant difference in cardiovascular mortality, which is what would be expected between two cardiac procedures if one was to improve survival," he said.
"Dr Taggart is a master of cherry picking, selectively choosing one data point from one study and conveniently ignoring all other data in the field," he said.
Taggart, who declined on two occasions to be interviewed for this story, told theheart.org | Medscape Cardiology only that he "stands by every word" in his EACTS presentation.
Some of his arguments are also documented in a December, 2018, article in Circulation, for which he is senior author.
The corresponding author of that article, Marc Ruel, MD, MPH, University of Ottawa, Ontario, Canada, pointed out for theheart.org | Medscape Cardiology that the primary endpoint as defined in the EXCEL protocols doesn't specify enzymatic periprocedural MI.
But, "the term 'MI adjudicated per Universal MI Definition' is specified and subsequently used repeatedly in the protocol, with every reference to adjudication of MI being followed by 'per Universal MI Definition,' " Ruel said by email.
"Notably, each of the 2007, 2012, and 2018 versions of the Universal Definition of Myocardial Infarction includes consensus diagnostic criteria pertaining to MI around PCI and CABG, which the SCAI definition used in the EXCEL trial does not corroborate or adhere to," he said.
"Why was the Universal MI Definition, to which every mention of MI adjudication in the 2011 protocol refers, not used to calculate the MI component of the primary endpoint?" Ruel asked.
Taggart is listed in the EXCEL publications as a member of its executive committee and its "CABG committee" and as a country leader for the United Kingdom. But he said he refused to be listed as a coauthor on the 5-year follow-up report.
That report's conclusions, he said at EACTS, "are at complete odds with the actual data presented in the paper. And having attempted to have this corrected, and having been unsuccessful, I withdrew my name."
At the EACTS session, Kappetein said that Taggart, as a trial officer, was aware of the protocol's periprocedural-MI definition throughout the study.
The protocol was twice submitted to and accepted by NEJM, "and you also accepted it during the steering committee meetings," he said. "And there are other surgeon authors of this paper who remained authors on this paper because they back up the conclusions from this trial."
"The change in myocardial infarction definition was never discussed with me," Taggart fired back. "Never."
Taggart was "intimately involved in all the decisions regarding the primary endpoint definitions, including periprocedural MI, and agreed to them," Stone said.
"He never raised any concerns about this issue in the past. His claim that these events are meaningless is thus groundless, and to selectively exclude these clinically relevant events is nonsensical."
"Real" Myocardial Infarctions
But Taggart pressed his case against periprocedural MI as not nearly as important an endpoint as spontaneous MI and that its inclusion favored PCI.
By his analysis, he said, with removal of periprocedural MI from the primary endpoint, so that it consists of death, stroke, and nonperiprocedural MI ― that is, "real myocardial infarctions" — the composite endpoint shows a "strong advantage for CABG."
His own view, which is based on EXCEL and other trials, is that "there should be a more cautious approach to the use of stents in patients with low- and intermediate-severity left main disease," he said.
EXCEL and other studies "confirm what we're currently doing" with regard to selecting patients with LM disease for either PCI or CABG, said Stephen E. Fremes, MD, Sunnybrook Health Sciences Center, Toronto, Canada.
Backing up Taggart's claim at the EACTS forum that the primary endpoint's MI definition was altered midstream, Fremes speculated about a potential reason: perhaps it was meant to increase the number of endpoints throughout the follow-up.
"In other words, to meet the sample size, with a smaller number of patients, they wanted to have more events, and this was one way of doing it," Fremes said.
Of note, at least one other EXCEL trialist, the chair of its "statistical committee," also expressed reservations about the inclusion of periprocedural MI within the composite endpoint.
The "real challenge" in interpreting EXCEL is that "the components of the primary endpoint vary in different ways," said Stuart J. Pocock, PhD, London School of Hygiene and Tropical Medicine, at the recent TCT sessions after Stone's presentation of EXCEL.
The PCI and CABG groups were of similar size, and there were about 20 fewer periprocedural MIs but 28 more spontaneous MIs in the PCI group, he noted.
"Both of those are significant, and they get a little bit lost when you combine them. They're going in opposite directions. So I'm not entirely favoring the combining of the two," said Pocock, an expert on clinical trial design.
Ruel took EXCEL to task for using "a custom periprocedural enzymatic MI definition, one favorable to the group treated with a technology provided by the sponsor," that is, stents from Abbott Vascular, "leaving out the other, 'universally' approved definition, which enabled a claim of noninferiority" ― and that, he said, "despite a 38%, highly statistically significant excess in the odds of death in the PCI over the CABG group."
Stone said that all of the surgeons and interventional cardiologists in EXCEL's leadership, including Taggart, "agreed on a single powered primary endpoint of greatest importance to the patient, the composite of death, stroke, or large MI."
Given the results over 5 years, a time frame that "is meaningful for most patients and practitioners," he said, "both procedures are acceptable for the types of patients that were enrolled in EXCEL."
EXCEL was sponsored by Abbott Vascular. Taggart, Stone, and Pocock have disclosed no relevant financial relationships. Disclosures for the authors of the EXCEL 5-year report are available at NEJM.com. Ruel has disclosed no relevant financial relationships.
33rd European Association for Cardio-Thoracic Surgery Annual Meeting. Da Vinci session and Trial Updates and Evidence Review. Presented October 5, 2019.
Transcatheter Cardiovascular Therapeutics 2019: Late Breaking Trials 3. Presented September 28, 2019.
Medscape Medical News © 2019
Cite this: Sparks, Accusations Fly Over Conduct of EXCEL PCI vs CABG Trial - Medscape - Oct 15, 2019.