A recent ISCHEMIA trial substudy is under scrutiny from surgeons for a data discrepancy, rekindling concerns about reliance on the landmark trial data in the latest coronary revascularization guidelines.
As previously reported, the main ISCHEMIA findings showed no significant benefit for an initial strategy of percutaneous coronary intervention (PCI) or coronary bypass graft surgery (CABG) over medical therapy in patients with stable moderate to severe ischemic heart disease.
The 2021 substudy by Reynolds et al. showed that coronary artery disease (CAD) severity, classified using the modified Duke Prognostic Index score, predicted 4-year mortality and myocardial infarction in the trial, whereas ischemia severity did not.
Dr. Joseph F. Sabik III
Cardiac surgeons Joseph F. Sabik III, MD, and Faisal Bakaeen, MD, however, spotted that only 40 patients are in the Duke category 6 group (three-vessel severe stenosis of at least 70% or two-vessel severe stenosis with a proximal left anterior descending lesion) in Supplemental tables 1 and 2, whereas 659 are in the main paper.
In addition, the Supplemental tables list the following:
- 659 patients in Duke group 5, not 894 as in the paper.
- 894 patients in Duke group 4, not 743 as in the paper.
- 743 patients in Duke group 3, not 179 as in the paper.
The surgeons penned a letter to Circulation early in April flagging the discrepancies, but say it was rejected April 15 because it was submitted outside the journal’s 6-week window for letters. They posted a public comment on the Remarq research platform, as advised by Circulation’s editorial office, and reached out directly to the authors and ISCHEMIA leadership.
“They just keep saying it’s a simple formatting error. Well, if it is a simple formatting error, then fix it,” Dr. Sabik, chair of surgery at University Hospitals Cleveland Medical Center, said in an interview. “But here we are now, a month later, and they still haven’t published our letter. Why? We’re the ones who identified the problem.”
Dr. Sabik said the accuracy of the data has important implications because the recent AHA/ACC/SCAI coronary revascularization guidelines used the ISCHEMIA data to downgrade the CABG recommendation for complex multivessel disease from class 1 to class 2B. Patients with a Duke 6 score are also typically the ones referred for CABG by today’s heart teams.
Several surgical societies have contested the guidelines, questioning whether the ISCHEMIA patients are truly reflective of those seen in clinical practice and questioning the decision to treat PCI and surgery as equivalent strategies to decrease ischemic events.
Dr. Bakaeen, from the Cleveland Clinic, told this news organization they don’t want a public battle over the data like the one that befell the EXCEL trial, and that it’s entirely possible the investigators might have inadvertently upgraded all the Duke score assignments by 1.
A systematic error, however, is more plausible than a formatting error, he said, because Supplemental tables 1 and 2 correspond exactly to the Duke 1 to Duke 7 sequence, suggesting the tables are correct and that the error might have occurred downstream, including in the manuscript.
The numbers should be consistent across all the ISCHEMIA manuscripts, Dr. Bakaeen added, but currently “don’t add up,” even after adjustment for different denominators, and especially for participants with left main disease.
They hope that publication of their letter, he said, will convince the authors to publicly share the data for patients in each of the seven modified Duke categories.
Lead author of the ISCHEMIA substudy, Harmony Reynolds, MD, New York (N.Y.) University Langone Health, told this news organization via email that as a result of a “formatting error in the transfer of data from the statistical output file to a Word document, data in Supplemental tables 1 and 2 were incorrect.”
Dr. Harmony Reynolds
She explained that they planned to present six, not seven, rows for the Duke score in the tables, collapsing the first two categories of nonobstructive disease (Duke 1-2), as they were in all other tables and figures. However, the Supplemental tables had incorrect row headings and because the Word program is designed to fill all available rows, it inserted the data from the output file into a seven-row table shell, duplicating the values for row 1 in the last row for left main disease of at least 50%.
“The data were correctly presented in the main manuscript tables and figures and in the remainder of the supplement, with a total of 659 patients in the subset with modified Duke prognostic index category 6 on coronary CT angiography,” Dr. Reynolds said.
She noted that Circulation will issue a correction. In addition, “we are in the process of preparing the data for public sharing soon. The data will include the Duke prognostic score at all levels.”
Circulation editor-in-chief Joseph A. Hill, MD, PhD, chief of cardiology at UT Southwestern Medical Center, Dallas, declined to be interviewed but confirmed via email that Dr. Bakaeen and Dr. Sabik’s letter and the correction will be published the week of May 16.
As for the delay, he said, “I received their reach-out just over 1 week ago, and per protocol, we conducted an internal evaluation of their allegations, which took a bit of time.”
A version of this article first appeared on Medscape.com.