Review of the EXCEL trial

PCI vs bypass surgery for left main coronary artery disease

N Engl J Med 2016;375:2223-2235

Background: Smaller randomized trials have shown that outcomes are not significantly different when patients with left main disease are treated with percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG). In the subgroup of patients with left main disease in the SYNTAX trial, outcomes were similar between PCI and CABG in patients with low or intermediate SYNTAX score but PCI was associated with worse outcomes in patients with high SYNTAX score.

The Evaluation of XIENCE versus Coronary Artery Bypass Surgery for Effectiveness of Left Main Revascularization (EXCEL) trial sought to assess if PCI was noninferior to CABG in patients with left main coronary artery disease.

Patients: Eligible patients had left main stenosis of 70% or more. Patients with stenosis of 50% to 69% were enrolled if the stenosis was hemodynamically significant as determined by non-invasive or invasive testing. Patients were also required to have low or intermediate SYNTAX score defined as a score of 32 or less.

Patients were excluded if they had prior PCI to the left main coronary artery, PCI to any other coronary artery within 1 year, prior CABG, a need for a concomitant cardiac surgery, elevated CK-MB, or life expectancy less than 3 years due to non-cardiac conditions.

Baseline characteristics: The trial randomized 1,905 patients – 948 randomized to PCI and 957 to CABG.

The average age of patients was 66 years and 77% were men. Approximately 74% had hypertension, 70% had hyperlipidemia, 29% had diabetes, 17% had prior myocardial infarction and 22% were current smokers. The average left ventricular ejection fraction was 57%.

The clinical presentation was myocardial infarction within 7 days in 14% of the patients, unstable angina in 24%, stable angina in 53%, and silent ischemia or other in 8%.

Distal left main bifurcation or trifurcation disease was present in 81% of the patients, and 2- or 3-vessel coronary artery disease was present in 51%. SYNTAX score based on a core laboratory evaluation was low (22 or less) in 36% of the patients, intermediate (23-32) in 40% and high (33 or more) in 24%. However, based on site assessment, SYNTAX score was low in 61% of the patients and intermediate in 39%.

Procedures: Patients were randomly assigned in a 1:1 ratio to undergo CABG or PCI using fluoropolymer-based cobalt–chromium everolimus-eluting stents (XIENCE, Abbott Vascular). Randomization was stratified based on the presence of diabetes, SYNTAX score (low vs intermediate) and study center.

Dual antiplatelets were given for at least 12 months following PCI.

CABG was performed with or without cardiopulmonary bypass based on the operator discretion. The use of arterial grafts was recommended.

Endpoints: The primary endpoint was a composite of death from any cause, myocardial infarction and stroke at 3 years. Secondary endpoints included the components of the primary endpoint as well as repeat revascularization.

Analysis was performed based on the intention-to-treat principle. Sample size was calculated based on non-inferiority. The sample size to provided 80% power with one-sided alpha of 0.025 was 1,900 patients. This calculation was based on an assumed 11% event rate in each study group and 4.2% absolute difference non-inferiority margin.

The original sample size was 2,600 patients which would have provided 90% power. However, both were adjusted due to slow enrollment.

Results: Among the 948 patients assigned to the PCI arm, 99% underwent the procedure. The mean number of stents implanted per patient was 2.4. Among the 957 patients assigned to the CABG arm, 96% underwent the surgery. The mean number of grafts per patient was 2.6. An internal mammary artery graft was used in 99% of the patients. The median follow up time was 3 years.

The primary composite endpoint was not significantly different between CABG and PCI (14.7% with CABG vs 15.4% with PCI, absolute difference: 0.7%, upper bound of the 97.5% CI: 4.0%; p= 0.02 for non-inferiority). There was no significant difference in death from any cause (5.9% with CABG vs 8.2% with PCI; p= 0.11), myocardial infarction (8.3% with CABG vs 8.0% with PCI, p= 0.64) or stroke (2.9% with CABG vs 2.3% with PCI; p= 0.37). Ischemia-driven revascularization was higher with PCI (12.6% vs 7.5%; p<0.001).

At 30-days, the primary outcome was higher in the CABG arm (7.9% vs 4.9%; p= 0.008) which was driven by higher myocardial infarction and stroke with CABG (6.2% vs 3.9%; p= 0.02) and (1.3% vs 0.6%; p= 0.15), respectively. From 30-days to 3-years, the primary outcome was lower in the CABG arm (7.9% vs 11.5%; p= 0.02).

The rate of stent thrombosis in the PCI arm was 0.7% and the rate of graft occlusion it the CABG arm was 5.4%.

Major bleeding according to the TIMI criteria was higher in the CABG group at 30-days (3.9% vs 1.2%; p< 0.001).

There were no significant subgroup interactions for the primary outcome.

Conclusion: In patients with left main coronary artery disease and low or intermediate SYNTAX score based on site assessment, PCI was non-inferior to CABG regarding the primary outcome of death from any cause, myocardial infarction or stroke over a median follow up of 3 years. The primary outcome was higher in the CABG arm at 30-days but lower from 30 days to 3 years.

A key difference from the SYNTAX trial is the use of everolimus-eluting stents here versus paclitaxel-eluting stents in SYNTAX. Everolimus-eluting stents have a lower incidence of stent thrombosis compared to paclitaxel-eluting stents, which was also reflected in this trial (0.7% stent thrombosis at 3-years in EXCEL vs. 3.2% at 1-year in SYNTAX).

Consistent with prior trials, CABG is associated with increased post-procedural risk, which is an important consideration when deciding between PCI and CABG. The lower events with CABG compared to PCI after 30 days is also noteworthy. In the 5-year follow-up study, the outcome curves diverge in favor of CABG (19.2% vs 22.0%), and the upper bound of the 95% CI for the primary outcome was 6.5%, suggesting that PCI did not meet non-inferiority at 5 years, using the original non-inferiority threshold which was 4.2%. Additionally, mortality was higher in the PCI group at 5-years (13.0% vs. 9.9%, mean difference: 3.1%, 95% CI: 0.2 - 6.1). Despite this, the authors still reported this in their conclusion: “there was no significant difference between PCI and CABG with respect to the rate of the composite outcome of death, stroke, or myocardial infarction at 5 years”.

Another controversial aspect of this trial is the myocardial infarction definition used in the final publication. The authors used the SCAI definition for trial reporting, deviating from the Universal definition specified in the protocol. In a correspondence, they reported the incidence of 5-year myocardial infarction according to both definitions: 9.0% with CABG vs 10.2% with PCI using the publication’s definition, and 4.7% with CABG vs 9.6% with PCI using the Universal definition. This substantial discrepancy, largely due to differences in procedural myocardial infarction classification, raises concerns.

The "neutral" reporting of the 5-year results, along with discrepancies between the protocol and the final publication in the definition of myocardial infarction, created controversy among physicians and researchers, and led a lead surgical investigator, Dr. David P. Taggart, to withdrew his name from the final publication.

In summary, this trial does not provide conclusive evidence on the best treatment strategy for patients with left main coronary artery disease.