Final EXCEL Results on Revascularization Choice in Left Main Disease: Interpret With Caution

Final results from the largest comparison of revascularization strategies for low-complexity left main coronary disease should not be viewed as the final word on the issue, contend two cardiac surgeons in an invited expert analysis of the EXCEL trial on ACC.org.

Findings at final five-year follow-up from the 1,905-patient randomized trial were recently published in the New England Journal of Medicine (2019;381:1820-1830), showing equivalence on the primary outcome — a composite of death, stroke or myocardial infarction (MI) — between coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI) with contemporary drug-eluting stents. Specifically, this composite outcome occurred in 22.0% of patients in the PCI arm and 19.2% of those in the CABG arm — a difference of 2.8 percentage points in favor of CABG (95% CI, –0.9 to 6.5 percentage points; P = 0.13)

Despite this statistical equivalence, the findings are more nuanced, argue the authors of the ACC.org analysis, Faisal Bakaeen, MD, of Cleveland Clinic, and Mario F.L. Gaudino, MD, PhD, of Weill Cornell Medicine.

Two limitations of the study design

In their analysis, Drs. Bakaeen and Gaudino highlight two features of the EXCEL trial’s design that they believe put CABG at a relative disadvantage:

• Definition of periprocedural MI. Periprocedural MI, which was significantly more frequent in the study’s CABG arm, was an important driver of the trial’s primary outcome at three and five years and a large reason for the two arms’ statistical equivalence on the primary outcome. The EXCEL investigators used a definition of periprocedural MI that permitted a purely enzymatic diagnosis, which “increases by 100% the enzymatic threshold in the PCI group but not in the CABG group, clearly disadvantaging surgery,” Drs. Bakaeen and Gaudino write. They point out that similar trials that used the generally adopted Third and Fourth Universal Definitions of Myocardial Infarction reported substantially lower rates of periprocedural MI after CABG — 1.7% in the FREEDOM trial and 2.9% in the SYNTAX trial, versus 6.2% in EXCEL.
• Important CABG practice variations. They also note that while therapy was homogeneous in the PCI arm — all patients received the same type of everolimus-eluting stent — treatment in the CABG arm varied in ways that could have compromised outcomes. Specifically, about 30% of surgical patients had off-pump CABG, which was associated with significantly higher three-year all-cause mortality relative to on-pump CABG. Additionally, only 24% of CABG recipients in EXCEL received bilateral internal thoracic artery grafts in spite of guideline support for multiarterial grafting, and fewer than 7% received radial artery grafts.

Secondary outcomes should not be ignored

The analysis authors also call attention to several noteworthy EXCEL findings beyond the primary outcome measure:

• There was statistically significant excess mortality in the PCI arm — 13.0% versus 9.9% (odds ratio = 1.38 [95% CI, 1.03-1.85]). While there was no significant difference in definite cardiovascular death, Drs. Bakaeen and Gaudino write that this “large difference in the most important outcome cannot simply be ignored.”
• All the cardiac outcomes in EXCEL (including non-periprocedural MI, “whose definition was not substantially modified,” they write) favor CABG over PCI.
• No excess deaths were observed in the CABG group in spite of the above-mentioned significantly higher incidence of periprocedural MI with CABG.

“The features of the study design that disadvantage CABG call for caution in interpreting the EXCEL trial’s finding of equivalence between the two revascularization strategies for low-complexity left main disease,” says Dr. Bakaeen. “Contemporary CABG that delivers complete revascularization with multiarterial grafting remains a highly compelling option for this patient population.”

The full expert analysis is available here.