CABG Still Superior to Stents Despite FAME 3 Endpoint Swap

Medscape

06-05-2025

What happened with the 5-year FAME 3 results is not right.
The reporting of this trial comparing stenting vs coronary artery bypass graft surgery (CABG) in patients with multivessel coronary artery disease (CAD) defies proper scientific principles.
The FAME 3 Trial of Stents vs Surgery
FAME 3 tested the best strategy to revascularize patients with multivessel CAD. Three previous trials comparing percutaneous coronary intervention (PCI) vs CABG (SYNTAX, FREEDOM, and BEST) had all found surgery to be superior.
But both technologies had advanced, perhaps PCI more than CABG. PCI is now done via the radial artery, dramatically reducing bleeding complications. What's more, the use of fractional flow reserve (FFR) to assess stenosis severity should ensure that drug-eluting stents are placed only in hemodynamically significant lesions.
FAME 3 was conducted because it was now possible that PCI could prove equal to CABG.
It was designed and powered as a 1-year noninferiority trial with follow-up at 3 and 5 years, 'if funding allows.'
Original 1-Year Results
The 1-year results were clear. The primary endpoint of all-cause death, myocardial infarction (MI), stroke, and repeat revascularization occurred in 10.6% of patients in the PCI group and 6.9% of those in the CABG group. The 50% higher rate of events with stenting had an upper bound of 2.2, which was well over the prespecified noninferiority margin of 1.45. (HR = 1.5; 95% CI, 1.1-2.2). Three components of the composite endpoint — death, MI, and repeat revascularization — were numerically higher in the PCI arm.
The authors concluded that contemporary FFR-guided PCI was not noninferior to CABG at 1 year. The title of the accompanying editorial was "CABG versus PCI — End of the Debate?" The writer answered his own question in the affirmative.
The FAME 3 trial bolsters the role of CABG as the benchmark for patients with multivessel coronary disease.
"Bolster" is the correct verb because FAME 3 clearly corroborated the three previous trials.
3-Year Results and an Endpoint Change
Then something happened. The FAME 3 investigators used a different endpoint (death, MI, and stroke) in reporting the 3-year results. Repeat revascularization was dropped. The result of this change was that the 30% higher rate of events in the PCI arm did not reach statistical significance (12% vs 9.2%; HR = 1.3; 95% CI, 0.98-1.83; P =.07).
But if they had stuck to their original primary endpoint, the 3-year result would have been similar to the 1-year result. In fact, the accrual of more events from year 1 to 3, particularly in the stent arm (18.6% vs 12.5% for surgery), led to more confidence in this endpoint, as evidenced by tighter confidence intervals (HR = 1.5; 95% CI, 1.2-2.00; P =.002).
The change in endpoint prompted the authors to now conclude there was 'no difference' in major adverse events between FFR-guided PCI and CABG.
5-Year Results
The authors continued with their new endpoint for the 5-year results. Now, 16% of patients in the PCI arm experienced either death, MI, or stroke vs 14.1% in the CABG arm. The higher rate of events in the PCI arm did not reach statistical significance (HR = 1.16; 95% CI, 0.89-1.52; P =.27).
If they had used the original primary endpoint that includes repeat revascularization, CABG would have been superior (25% vs 18% for CABG; HR = 1.44; 95% CI, 1.15-1.81). More events meant even tighter confidence intervals. The rates of MI and repeat revascularization were also higher after PCI.
The authors once again ignored the trial's original endpoint and concluded that there was no difference in the death/MI/stroke endpoint. They also noted that a landmark analysis looking at years 2-5 found no accrual of benefit in the surgery arm.
The results were presented at the 2025 American College of Cardiology scientific sessions where the lead sentence of the press release noted that 'in contrast to previous studies patients with severe triple-vessel heart disease fared equally well whether they underwent CABG or PCI.' Headlines in news coverage implied that stents had closed the gap and were equivalent to surgery.
Comments
To conclude that the two therapies were similar required a different endpoint that omitted repeat revascularization. Had the authors kept the original endpoint, FAME 3 at 5 years would have drawn a similar conclusion to the 1-year results and the three previous trials: CABG is superior to PCI for patients with multivessel CAD.
I will make three arguments why this endpoint change and revised conclusion were inappropriate:
First, repeat revascularization is a relevant clinical outcome. The authors obviously believed so because they chose to include it in the composite primary endpoint for the 1-year outcome. What's more, two observational studies have associated repeat revascularization with higher mortality. This makes sense because the rate of MI in the PCI arm of FAME 3 at 5 years was 60% higher than for surgery, so many of these repeat procedures probably were done for acute coronary syndrome.
Second, changing the endpoint is improper science. The authors may argue that the change was prespecified, but I see no clinical reason why the PCI-vs-CABG question requires different endpoints at different follow-up times.
The third reason that use of fewer events in the revised endpoint is problematic relates to statistical rules. The authors list as their first limitation that FAME 3 was not powered for secondary endpoints. Indeed the confidence intervals for the revised endpoint were wide and included the chance that PCI was more than 50% worse. What's more, the secondary endpoint of death, MI, and stroke was not adjusted for multiple comparisons, further introducing uncertainty.
Conclusion
My take-home is that while PCI and CABG have improved over time, CABG remains the superior strategy in patients with multivessel disease. The findings for FAME 3 at 5 years are no different from the 1-year outcomes. The only change has been in how the authors presented the results.