Former EXCEL Investigator Alleges Trial Manipulation, Prompting Vehement Denials

LISBON, Portugal (UPDATED)—A leading academic cardiovascular surgeon has broken from the EXCEL trial over his interpretation of the recently published 5-year results comparing PCI versus CABG surgery for the treatment of patients with left main coronary artery disease (LMCAD).

David Taggart, MD, PhD (University of Oxford, England), the chairman of the EXCEL surgical committee during the design and recruitment phase of the trial, believes the investigators downplayed the increased risk of all-cause mortality with PCI and oversold the reduced risk of the study’s primary composite endpoint of death, stroke, and MI, particularly since the benefit was largely driven by a higher risk of periprocedural MIs in the CABG arm.

“I found it worrying that there seemed to be a strong signal here and I don’t think that was accurately reflected in the New England Journal of Medicine paper,” Taggart told TCTMD. “We’re not talking about two tablets for a headache. We’re talking about people dying. The data are the data, and that’s what the data show.”

The 5-year results, as presented at the recent TCT 2019 meeting in San Francisco, CA, showed that the risk of death, stroke, or MI—the study’s primary endpoint—was 22.0% in the PCI arm and 19.2% in the CABG-treated patients, a difference that was not statistically significant (P = 0.13). As a result, the EXCEL investigators concluded there was no significant difference between revascularization with PCI or CABG in patients with LMCAD of low or intermediate anatomical complexity. The 5-year outcomes were published simultaneously in the NEJM on September 28, 2019.

A little more than a week later, Taggart took dead aim at the EXCEL trial and its investigators, insisting that their conclusions are “at complete odds” with the data in the paper. Speaking during the European Association for Cardio-Thoracic Surgery (EACTS) 2019 meeting, Taggart said he attempted to “correct” their view of the results, particularly the lack of focus on the higher risk of death with PCI, but when unsuccessful ultimately pulled his name from the publication.

“In over 300 publications, I have never withdrawn my authorship before,” he said.

In his interpretation of the trial, which he called “clinical reality,” Taggart pointed out that death from any cause significantly favored CABG surgery at 5 years (9.9% vs 13.0% with PCI; OR 1.38; 95% CI 1.03-1.85) and appeared to be accelerating in favor of surgery. In terms of myocardial infarction, there was no significant difference between treatments at 5 years, but Taggart said CABG was set up to fail in this regard given that the definition of MI included periprocedural MIs that disadvantaged surgery.

“If you look at death itself, arguably the most important outcome of this trial, there is a strong benefit of bypass grafting,” said Taggart. “If you look at nonperiprocedural MI, in other words real myocardial infarction, a strong benefit in favor of CABG.”

Gregg Stone, MD (Icahn School of Medicine at Mount Sinai, New York, NY), EXCEL’s co-principal investigator, disagreed with Taggart’s interpretation of the trial, noting that while there were more deaths in the PCI arm, the study was not powered for mortality. “Also, when you have multiple endpoints that are not corrected for multiplicity, it’s very possible some of them are false positives,” Stone told TCTMD. “You have to ask yourself, one, is there a plausible underlying mechanism and two, is it consistent with prior data?”

With respect to all-cause mortality, 18 of the 30 excess deaths at 5 years were deemed noncardiovascular, said Stone, and there was no significant difference in the risk of cardiovascular death, which was 6.8% in the PCI arm and 5.5% in the CABG group (OR 1.26; 95% 0.85-1.85). These rates are consistent with the MI rate in the trial, which makes sense given that MI is the likely cause of cardiovascular death, said Stone.

Multiple meta-analyses of DES versus CABG surgery, including numbers culled from the EXCEL trial, have shown almost identical mortality rates between patients treated with PCI and those treated with CABG surgery, Stone added.

If you just look at one nonpowered endpoint from one trial, [that] is the ultimate in cherry-picking. Gregg Stone

“If you just look at one nonpowered endpoint from one trial, [that] is the ultimate in cherry-picking,” he said, referring to the all-cause mortality finding in EXCEL. He also pointed to data from the SYNTAX Extended Survival study, which was presented by Daniel Thuijs, MD (Erasmus University Medical Center, Rotterdam, the Netherlands), at EACTS—and last month at the European Society of Cardiology Congress, as reported by TCTMD—showing that while CABG was superior to PCI in patients with multivessel disease, there was no significant difference in all-cause mortality at 10 years among patients with left main lesions treated with PCI or surgery (HR 0.90; 95% CI 0.68-1.20).

Given all this, Stone said he believes the 38% higher risk of death in patients with LMCAD treated with PCI in EXCEL is likely “a false positive since it was never seen in any other study.”

Periprocedural vs ‘Real’ Myocardial Infarctions

During the EACTS presentation, Taggart took particular issue with ascribing equivalence between PCI and surgery in LMCAD since the primary endpoint included nonperiprocedural and periprocedural myocardial infarctions. In fact, the NOBLE trial, another study of PCI versus CABG surgery for patients with LMCAD, avoided including periprocedural MI as part of the primary composite endpoint. Unlike EXCEL, the NOBLE study showed that surgery was significantly better than PCI for reducing the risk of all-cause mortality, MI, repeat revascularization, and stroke in patients.

Breaking EXCEL down further, Taggart noted that the only endpoint in favor of PCI at 30 days was the reduction in myocardial infarction (3.9% with PCI vs 6.3% with surgery; HR 0.63; 95% CI 0.42-0.94), which was driven by periprocedural events. At 30 days to 1 year, there was no advantage with either procedure, but outcomes from 1 to 5 years clearly favored CABG surgery. From 12 months onward, rates of death, myocardial infarction, and ischemia-driven revascularization were significantly higher with PCI. If periprocedural myocardial infarction is excluded, there is a significant advantage with surgery over PCI at 5 years in terms of EXCEL’s primary composite endpoint of death, stroke, and myocardial infarction, said Taggart.

“If you look at real myocardial infarction, the ones that occur 30 days after an operation, then there was a strong benefit of CABG over PCI,” said Taggart. “In the EXCEL trial, 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.”

We’re not talking about two tablets for a headache. We’re talking about people dying. David Taggart

Surgeon Stephen Fremes, MD (Sunnybrook Health Sciences Center, Toronto, Canada), who reviewed a range of relevant left main PCI versus CABG comparisons in the same EACTS session, also pointed to the EXCEL primary endpoint event curves, noting there was a wide separation early in favor of PCI. The time-to-event curve for the primary endpoint did not cross until roughly 3 years. In other trials comparing PCI versus CABG surgery, such as FREEDOM, the time-to-event curves show an early benefit of PCI over CABG, but the curves tend merge around 1 year before separating in favor of CABG surgery.

Like Taggart, Fremes said the large early advantage with PCI in EXCEL was “entirely based on the difference in periprocedural myocardial infarction.” He pointed out that the rate of perioperative MI in EXCEL was 6.2%, which was higher than the rate observed in FREEDOM (1.7%) and SYNTAX (2.9%). “The question is in EXCEL if there is a true difference in the incidence of periprocedural MI or is it simply related to differences in the definition of periprocedural MI,” said Fremes.  

Stone disputes the characterization of these periprocedural MIs as clinically unimportant, noting the primary endpoint was agreed upon in advance by surgeons and interventional cardiologists. He also pointed out that that the EXCEL investigators previously published data—for which Taggart was one of the co-authors—showing that periprocedural events were strongly correlated with survival at 3 years. As such, these periprocedural MIs are prognostically important, possibly more so after bypass surgery than PCI.

“The primary endpoint of EXCEL was death, stroke, or large myocardial infarction,” Stone said. “It would be totally artificial and nonsensical to exclude periprocedural MI as it would be to exclude any other complication, or strokes, deaths, or spontaneous MIs. I don’t even follow that argument.”  

Changing Definitions or No Change at All?

While researchers and physicians may debate how much weight to place on secondary endpoints, such all-cause mortality, or whether periprocedural MIs bias the trial in favor of PCI over surgery, the most contentious aspect of the EACTS presentation stems from Taggart’s allegations that the definition of MI was changed midway through the trial to favor PCI.

For this reason, Taggart said he believes the trial was rigged to show an advantage of PCI over CABG surgery in patients with LMCAD.

“What happened in EXCEL was a disgrace that halfway through the trial the definition of myocardial infarction was changed,” he said. “Interestingly, if you take out periprocedural myocardial infarction, CABG is a very clear winner. I believe the data was manipulated using a changed definition of myocardial infarction to try to prove for the composite endpoint that there was no difference [between PCI and CABG].”

Stone adamantly disputes this.

“The definition was never changed,” he told TCTMD. “It always used a modified version of the [Society for Cardiovascular Angiography and Interventions] definition for periprocedural myocardial infarction and the Third Universal definition for spontaneous myocardial infarction. I have no idea where [Taggart] is getting this idea from. He was involved with the trial from the beginning, and we have published the first and final drafts of the protocol. You can see that the myocardial definitions have not changed.”

Pieter Kappetein, MD, a former cardiothoracic surgeon at Erasmus Medical Center, the Netherlands, and now chief medical officer/vice president of Medtronic, also grew quite heated over allegations the definition of myocardial infarction changed during the course of the trial. “You are misleading the audience here,” he said angrily following Taggart’s presentation. “The definition was not changed during the study. The only thing that happened is that there were CK-MB [values] missing and they were collected afterwards.”

Kappetein said that all other authors remain on the paper because they “back up the conclusions” of the EXCEL investigators.

For his part, Taggart told TCTMD that he absolutely stands by his statement that the definition of MI changed during the course of the study. In fact, he and others criticized the EXCEL investigators for the endpoint tweaks back in 2018 in Circulation. Taggart alleges that the final MI definition was developed near the end of the trial’s recruitment phase, not at the beginning, and that the definition is not aligned with the second and third universal definition of MI as it uses an “exclusively biochemical” threshold for PCI and CABG and one favoring the use of CK-MB over cardiac troponin.

Finally, Taggart also took issue with the involvement of industry in EXCEL, noting that the chief medical officer of Abbott Vascular, the study’s sponsor, was included as an author, as was the chief medical officer of Medtronic (Kappetein, who joined Medtronic in 2017, began involvement in EXCEL as a surgeon). Taggart added that one-third of study authors received personal fees from Abbott Vascular, while more than 40% received fees from other stent manufacturers.

“In my 30 years of being involved in authorships and publications, I have never seen witnessed such an attempt to distort what the actual data in this paper showed,” he said.   

Rita Redberg, MD (University of California, San Francisco), the editor-in-chief of JAMA Internal Medicine and a discussant during the contentious session at EACTS, said the overly rosy interpretation of EXCEL raises questions, particularly given the all-cause mortality findings and the industry involvement in the study. She also laid some of the responsibility on the NEJM, stating that reviewers have an obligation to hold researchers to account, particularly in this instance when the all-cause mortality endpoint was downplayed.

“I can’t help but see it as a failure of the review and editorial process to allow a final paper which is clearly a disservice to patients,” she said. “We’re talking about a 38% difference in mortality [between PCI and CABG at 5 years].” 

For his part, Stone stressed that EXCEL was designed by an equal number of academic cardiothoracic surgeons and interventional cardiologists, including Taggart. “It’s probably one of the most transparent and collaborative international efforts ever been done in cardiology. Dr. Taggart and the other surgeons were involved every step of the way. All of the surgeons and interventionalists save one are aligned with the message from the final 5-year outcomes,” he commented.

The “healthy debate” over EXCEL, as Stone called it on Twitter, is far from finished, with Taggart noting that he and other organizers are hosting the International Coronary Congress in New York in December, where he and Stone are scheduled to debate the results of EXCEL and the role of PCI for LMCAD. For now, in light of EXCEL and NOBLE, coronary stenting should be downgraded in the European Society of Cardiology/EACTS guidelines from its current 1A recommendation (in patients with a Syntax score 0-22) and that a much more cautious approach for PCI should be adopted in this setting, Taggart urged at EACTS.

Taking the Conversation to Twitter

The EXCEL debate following Taggart’s presentation took off on social media, drawing in surgeons, interventionalists, general cardiologists, clinical trialists, and statisticians. Their reactions to the unfolding story were varied, with some calling for a retraction of the NEJM paper, the EXCEL investigators stating unequivocally that the MI definition was never changed, and at least one statistician stating that it would be reasonable, and not unheard of, for trial leaders to tweak a trial to use a newly updated definition of a clinical endpoint, such as MI.

Dr. Taggart is mistaken or intentionally misleading the audience. The definition of MI NEVER changed in EXCEL. This is a delusion.

— Gregg W. Stone MD (@GreggWStone) October 6, 2019

The frustrating thing about the EXCEL “controversy” (for me, on Twitter) is watching people without full understanding of “how a trial happens” sling the word “misconduct” around & call for NEJM to retract the paper because...there were changes to the protocol

— Andrew Althouse (@ADAlthousePhD) October 8, 2019

Others focused on the clinical endpoint of all-cause mortality, which they viewed as a major win for CABG in the setting of LMCAD.

EXCEL trial 5 years: we have been discussing several issues, great to open this discussion #EACTS2019

One major issue not to forget: all-cause mortality was better in CABG (9.9%) than PCI (13%, P = 0.002). Full stop.@GreggWStone @AKappetein @tamaranihici @tssmn @ctsnetorg
1/8 pic.twitter.com/sfObKPFGpU

— Patrick Myers (@pomyers) October 7, 2019

I haven't been keeping up w/ details of EXCEL controversy, but to me, most important points are:

* CABG wins: death long term (3.1% ARR)
* CABG wins: death/stroke/MI/revasc (6.5% ARR)
* CABG prob. wins: death/MI (~4% ARR)
* Even primary outcome: 2.8% ARR for CABG vs. PCI (P=NS)

— Venk Murthy (@venkmurthy) October 7, 2019

Stone, however, only reiterated that relying on EXCEL alone was misguided when other meta-analyses did not show a benefit. Finally, others expressed disappointment with the EACTS program planners, stating that the controversy will only confuse patients and undermine their confidence in health providers. Including more diverse perspectives in sessions such as these would be the better approach, one tweeter suggested.

I find what is happening at @EACTS very sad. False and inappropriate accusations to major RCTs leaders, and unbalanced interpretation of the data. These things do not happen at IC or gen cardiology meetings. This is harmful for the cardiovascular community and for patient care

— Gennaro Giustino, MD (@g_giustinoMD) October 6, 2019

How these guys have interpreted the recent trials (EXCEL, etc) is factually incorrect and misleads the layman, considering that their opinion is endorsed by EACTS. For ex., ‘5-y survival with CABG dramatically better (1/3) than with PCI’?? C’mon!! How can you claim that?!

— Lorenzo Azzalini (@lorenzo2509) October 6, 2019

Note: Stone is Co-Director of Medical Research and Education at the Cardiovascular Research Foundation, the publisher of TCTMD.