JIM 2020: Weigh up early and late events in EXCEL to determine treatment for left main disease

Gregg W Stone

Gregg W Stone (Icahn School of Medicine, Mount Sinai, New York, USA) addressed the EXCEL controversy today at the Joint Interventional Meeting (JIM 2020; 13–15 February, Milan, Italy), stipulating that although percutaneous coronary intervention (PCI) has early advantages and coronary artery bypass graft (CABG) confers late advantages, in the long term there are no significant major differences in long-term survival, major adverse cardiovascular events (MACE: death, myocardial infarction [MI], or stroke), or quality of life between the two procedures.

He said that heart team discussions needed to weigh the early advantages of PCI—it is less invasive, with fewer periprocedural complications (such as, stroke, atrial fibrillation, bleeding, acute kidney injury), a lower 30-day MACE rate, and a more rapid recovery with better early quality of life and earlier angina relief—against the late advantages seen with CABG—it is more durable with fewer adverse events beyond 30 days, particularly MI and repeat revascularisation procedures—when deciding which to choose.

Looking at five-year mortality data from EXCEL—one of the main areas around which disagreement has arisen—Stone described mortality as an “underpowered exploratory endpoint” that was not specified for hypothesis testing, and said that the observed difference of 87% survival for PCI versus 90% for CABG, a difference of 3% over two years (0.6% per year) was “modest”.

Stone noted that when assessing the rates of low frequency secondary outcomes it is important to use data from all available studies. He noted there have been four randomised trials of PCI with drug-eluting stents versus CABG in patients with left main disease (EXCEL, SYNTAX, PRECOMBAT, and NOBLE) in which a total of 4,394 patients were enrolled. In a meta-analysis of these trial, there were very similar rates of five-year mortality after PCI and CABG (hazard ration [HR] 1.06, 95% confidence interval [CI] 0.83–1.35). These rates were also nearly identical at 10 years from the only trial that has reported very late outcomes, the SYNTAX trial: (HR 0.90, 95 CI 0.68–1.20).

Stone also addressed the use of periprocedural MI as an endpoint in EXCEL rather than the universal definition (UD) of MI, a further major point of contention. He explained that because the universal definition uses different criteria for MI in PCI and CABG, the same periprocedural definition of MI for both PCI and CABG was agreed upon before the trial in order to minimise ascertainment and other biases. Within the EXCEL trial, periprocedural MI according to this definition was strongly predictive of subsequent cardiovascular death whereas other definitions of MI (which would have included smaller infarctions included in the UDMI) were not.

In his presentation, Stone described a subanalysis of EXCEL (CABG n=957; PCI n=948) that identified three distinct periods of relative risk for all cause death, stroke or MI: “In the first 30 days, there was a substantial 40% reduction in major events for PCI compared to CABG [HR 0.61, 95% CI 0.42–0.88, p=0.008]. From one month to one year, there was no significant difference between the two groups [HR 1.07, 95% CI 0.68–1.7, p=0.76]. Between one year and five years, the curves clearly benefited CABG [HR is 1.61, 95% CI 1.23–2.12, p<0.001] [treatment time interaction p<0.001].”

He asked: “How do you judge the overall risk to the patient when you can have these three different periods of overall risk? If you are going to have an adverse event, would you rather have it later than early? That is one thing to take into account.”

Calculation of the mean event-free survival time in EXCEL identified the time event burden of disease, which is more in CABG than PCI: “Early on, PCI patients are more likely to be disease free than CABG patients; that peaks at about 30 to 36 months, and thereafter the lower late event risk of CABG starts to provide a benefit. At the end of five years, the average PCI patient was free of death, stroke, or MI five days longer than the average CABG patient. The 95% confidence intervals are very wide, so obviously there is no difference between these two therapies.”

“Thus there is no meaningful difference in death, stroke, or MI in terms of overall burden of disease between PCI and CABG, at least out to five years.”

He also said presence of other factors such as distal main disease, diabetes, left ventricular ejection fraction, or low, intermediate SYNTAX score led to no difference in ratios for death, MI or stroke between PCI and CABG: “The bottom line is that there was no significant interaction for any subgroup.”


Please enter your comment!
Please enter your name here