Continuing DSMB concerns
A year later, the DSMB was still troubled. Dr. Wallentin emailed Dr. Stone in September 2017 asking for an updated analysis of the mortality data without any capping in time.
Dr. Wallentin added that he didn’t think that unblinding the mortality results would be an issue at that stage because these were late deaths in a trial where the interventions were long completed. But, he warned, “it might be very concerning if, in the future, suspicions were raised that already available information on mortality was withheld from the cardiology and thoracic surgery community.”
The investigators took a month to respond. They declined the request, saying that the trial was not statistically powered to measure mortality. In his email to Dr. Wallentin, Dr. Stone stressed that they were committed to complete disclosure of all of the EXCEL data and that the responsible time point to unblind was after 4 years. His coprincipal investigators (Dr. Serruys, Dr. Sabik, and Dr. Kappetein) as well as EXCEL statistical committee chair Stuart Pocock, PhD, and Mr. Simonton were all copied on the email.
Dr. Wallentin deferred to the principal investigators’ arguments.
Missing MI data
Death was not the only outcome of the EXCEL trial to draw scrutiny.
The EXCEL investigators used a unique definition of MI that was almost exclusively based on a rise in the cardiac biomarker CK-MB. This protocol definition of MI was later adapted into the Society for Cardiovascular Angiography and Interventions definition in a paper coauthored by Dr. Stone. The investigators agreed to also measure MIs that met the more commonly used Third Universal Definition as a secondary endpoint. The Third Universal Definition of Myocardial Infarction uses a change in biomarkers – preferably troponin or alternatively CK-MB – coupled with other clinical signs.
It is standard practice to report secondary endpoints in any analysis of the main findings of a study. Yet, the EXCEL investigators did not report the universal definition of MI in either the 3-year or 5-year publications.
This is critical because MI according to one definition may not count according to the other, and the final tally could tip the trial results positive, negative, or neutral for coronary stents.
In Dr. Taggart’s opinion, the protocol definition puts CABG at a disadvantage because it uses the same biomarker threshold for procedural-related MI for both PCI and CABG. Because surgery involves more manipulation of the heart, cardiac enzyme levels will naturally be higher after CABG than PCI. These procedure-related enzyme elevations are not “true clinical MIs,” according to Dr. Taggart and others.
Late last year, a dataset containing the 3-year follow-up of EXCEL, including the information on the universal definition of MI, was leaked to the BBC. Working with biostatisticians, the BBC confirmed that according to this definition, there were more MIs in the stent group.
Originally, the investigators disputed the finding, calling the BBC data “imaginary.” They claimed that they were unable to calculate a rate of MI according to the universal definition because they lacked routine collection of troponins, although the universal definition also allows use of CK-MB. They have since published an analysis of 5-year MI data according to the universal definition, which showed twice the rate of MI in the PCI group.
From the leaked data, the BBC calculated the main composite endpoint of death, stroke, and MI using the universal definition of MI. Now the results swung in favor of CABG.