Stents vs. Bypass Surgery for Three-Vessel Heart Disease: What the FAME 3 Study Found

Table of Contents

• Key Points
• Why This Research Matters
• How the Study Was Designed
• Who Took Part in the Study
• What the Two Treatments Involved
• What the Researchers Measured
• How the Numbers Were Analyzed
• Main Results: The Primary End Point
• Individual Outcomes and Repeat Procedures
• Safety: Where Bypass Surgery Caused More Problems
• What These Findings Mean for Patients
• Limitations of the Study
• Practical Takeaways
• Frequently Asked Questions
• Source Information

Key Points

• In a 1,500-patient trial, FFR-guided stenting did not match bypass surgery for preventing death, heart attack, stroke, or repeat procedures at one year (10.6% vs 6.9%).
• Bypass surgery caused more short-term problems: major bleeding, kidney injury, irregular heart rhythms, longer hospital stays, and more 30-day readmissions.
• Stenting offered a faster, less invasive recovery, with a median hospital stay of 3 days versus 11 days for bypass surgery.
• The trial included mostly men (81–83%) and mostly White patients (92–94%), so results may not fully apply to women or more diverse groups.
• Results cover one year only; longer follow-up is needed to see whether the gap between stenting and bypass narrows, widens, or stays the same.

Why This Research Matters

Large randomized trials have already shown that open-heart bypass surgery produces better outcomes than percutaneous coronary intervention (PCI). Open-heart bypass surgery is also called coronary-artery bypass grafting, or CABG. Percutaneous coronary intervention (PCI) is the older approach of opening blockages with balloons and stents. These trials studied patients with three-vessel coronary artery disease. But those earlier trials had two important gaps.

First, they rarely used the newer, second-generation drug-eluting stents. A drug-eluting stent is a tiny wire mesh tube coated with medication that slowly releases into the artery wall to stop the blockage from growing back. Second-generation versions have improved both early and late results. They cause less stent thrombosis, which is a blood clot forming inside the stent. They cause fewer heart attacks during and after procedures. They cause less restenosis, which is the artery narrowing again. They cause fewer deaths than the first-generation stents did.

Second, the older trials did not routinely measure a value called fractional flow reserve to guide the stenting. FFR is measured with a thin coronary pressure wire threaded into the artery. It gives a more accurate picture of how much a blockage actually restricts blood flow than an angiogram (an X-ray dye picture of the arteries) does on its own. A low FFR means the blockage is truly limiting blood flow; a high FFR means it is not.

PCI guided by FFR leads to better short-term and long-term outcomes than PCI guided by angiography alone, or than treating with medication only. The FAME 3 investigators therefore set out to test whether FFR-guided PCI using current-generation drug-eluting stents could hold its own against bypass surgery in patients with three-vessel disease.

How the Study Was Designed

FAME 3 was an investigator-initiated, multicenter, international, randomized, controlled trial run at 48 sites. It was designed as a "noninferiority" trial. That means the researchers were not trying to prove stenting was better than surgery — they were testing whether stenting was no worse than surgery by more than a pre-agreed amount.

The trial was supported by research grants to Stanford University from Medtronic and Abbott Vascular. Neither company had any role in designing the trial, running it, or preparing the manuscript. Stanford University oversaw the trial.

Patients were assigned in a 1:1 ratio — half to CABG, half to FFR-guided PCI. Assignment was done through a web-based system and was stratified by diabetes status and by trial site, so that each group would be balanced on those two factors.

The main requirement to join the study was three-vessel coronary artery disease. Three-vessel coronary artery disease was defined as at least 50% narrowing (diameter stenosis) in each of the three major heart vessels or their major side branches. This was judged by visual estimation. The left main coronary artery could not be involved. The blockages also had to be treatable by either PCI or CABG, as decided by the heart team at each site.

People were excluded if they had a recent ST-segment elevation myocardial infarction (STEMI), the most severe type of heart attack. People were excluded if they had cardiogenic shock, the heart failing so badly it cannot pump enough blood for the body. People were excluded if they had a left ventricular ejection fraction below 30%. Ejection fraction is the percentage of blood in the heart's main pumping chamber that is pushed out with each beat; below 30% indicates a seriously weakened heart.

The trial protocol was approved by the relevant institutional review boards or ethics committees, and every participant gave written informed consent.

Who Took Part in the Study

A total of 1,500 patients were enrolled: 757 assigned to PCI and 743 to CABG. The two groups were well matched at the start.

The average age was about 65 years (65.2 years in the PCI group and 65.1 in the CABG group). Most patients were men: 81.4% in the PCI group and 83.3% in the CABG group. About 94% of patients in the PCI group and 92% in the CABG group were White. Average body-mass index was 28.6 in the PCI group and 28.7 in the CABG group.

Other baseline characteristics were closely similar between the groups:

• Diabetes: 28.3% of the PCI group and 28.8% of the CABG group. About 7–8% of patients in each group needed insulin (7.3% vs 8.2%).
• High blood pressure (hypertension): 71.2% vs 75.0%.
• Abnormal cholesterol levels (dyslipidemia): 68.9% vs 71.7%.
• Current tobacco use: 19.2% vs 18.4%; former smokers made up 39.2% vs 39.9%.
• Family history of coronary artery disease: 32.5% vs 28.8%.
• Previous heart attack: 33.3% vs 33.5%.
• Previous PCI: 13.0% vs 14.0%.
• Prior transient ischemic attack (TIA) or stroke (cerebrovascular accident): 6.5% vs 7.6%.
• Kidney disease (defined as an estimated glomerular filtration rate below 60 ml per minute per 1.73 m² of body-surface area): 4.9% vs 5.9%.
• Reduced left ventricular ejection fraction (50% or less): 18.2% vs 17.6%.
• Hospitalized with a non–ST-segment elevation acute coronary syndrome: 39.7% vs 38.7%.

On the angiograms, patients had a mean of 4.3 blockages (lesions). Twenty-two percent had at least one completely blocked vessel (chronic total occlusion), and 68% had at least one bifurcation lesion — a blockage at a point where one artery splits into two. The mean SYNTAX score, which grades how complex the coronary disease is, was 26. Lower SYNTAX scores indicate simpler disease and predict a better outcome with stenting.

What the Two Treatments Involved

Bypass surgery was performed according to the standard practice at each participating center. Complete arterial revascularization was strongly recommended. This means surgeons were encouraged to use arteries rather than veins for the grafts. Measuring FFR to guide the surgery was not required, but if it had already been done during the diagnostic angiogram, the surgeon could use that information.

Everyone assigned to PCI first had FFR measured with a coronary pressure wire (Abbott Vascular) using adenosine given intravenously or directly into the coronary artery. Only blockages with an FFR of 0.80 or lower were treated with a stent. Stenting used durable-polymer zotarolimus-eluting stents (Resolute Integrity or Resolute Onyx, Medtronic). Measuring FFR again after stenting was encouraged, and intravascular imaging was used when the treating physician felt it was needed.

All patients in both groups were to receive aspirin, a high-dose statin, and guideline-directed medical therapy. Patients who had PCI also took a second antiplatelet medication for at least 6 months afterward. Follow-up visits took place at hospital discharge and at 1, 6, and 12 months.

How the procedures compared in practice:

• Time to the procedure: a median of 4 days after randomization for PCI (interquartile range 1 to 13 days) versus 13 days for CABG (interquartile range 6 to 26 days).
• Procedure duration: a median of 87 minutes for PCI (67 to 113 minutes) versus 197 minutes for CABG (155 to 239 minutes).
• Hospital stay: a median of 3 days for PCI (1 to 7) versus 11 days for CABG (7 to 16).
• Stents: an average of 3.7 stents per patient (plus or minus 1.9), with a median total stented length of 80 mm (52 to 116 mm). Twenty-two percent of PCI patients needed a staged procedure — a second session on another day.
• Bypass grafts: an average of 3.4 distal anastomoses (graft connections) per patient; 97% received a left internal thoracic artery graft, and 24.5% received multiple arterial grafts. About 24% had off-pump surgery, meaning the heart was not stopped and no heart-lung machine was used.
• Intravascular imaging during PCI: used in about 12% of cases.

FFR was actually measured in 82% of the blockages in the PCI group. The most common reason for skipping it was that the vessel was subtotally or completely blocked. The mean FFR was 0.70. Of the lesions that had been intended for treatment, 24% turned out to have an FFR greater than 0.80. This means they were not significantly limiting blood flow and were left alone. FFR was measured again after stenting in 60% of the treated blockages, with a mean value of 0.88, indicating that blood flow had been restored. In the CABG group, FFR was measured before surgery in 10% of patients.

What the Researchers Measured

The primary end point was the occurrence within 1 year of a major adverse cardiac or cerebrovascular event. This was defined as any one of the following: death from any cause, myocardial infarction (heart attack), stroke, or repeat revascularization (another procedure to reopen or bypass arteries).

Heart attacks were classified as either procedural or spontaneous. In both groups, a procedural heart attack was defined by a rise in cardiac troponin within 72 hours after the procedure. Cardiac troponin is a protein released when heart muscle is damaged. The rise had to be to more than 10 times the 99th percentile of the upper reference limit. This applied to patients whose troponin was normal beforehand. For patients whose troponin was already elevated, the threshold was an increase of more than 20%. In addition, at least one of these had to be present. One was new pathologic Q waves on the electrocardiogram or a new left bundle-branch block. One was angiographic proof of a new graft or major native coronary blockage. One was imaging showing new loss of viable heart muscle or a new regional wall-motion abnormality. This definition matches the type 5 category (post-bypass procedural heart attack) in the Third and Fourth Universal Definitions of Myocardial Infarction.

Spontaneous heart attack was defined as a rise or fall in troponin with at least one value above the 99th percentile of the upper reference limit. There also had to be evidence of reduced blood supply to the heart. That evidence had to be shown by at least one of the following. One was symptoms of ischemia, such as chest pain or its equivalents. One was electrocardiographic changes indicating new ischemia. One was development of pathologic Q waves. One was imaging evidence of new loss of viable heart muscle or a new regional wall-motion abnormality.

Secondary end points included the composite of death, heart attack, or stroke, along with safety outcomes. An independent clinical events committee reviewed and adjudicated all events without knowing which treatment each patient had received.

How the Numbers Were Analyzed

The main analysis followed the intention-to-treat principle, meaning patients were analyzed in the group they were originally assigned to, whether or not they received that treatment.

The researchers initially assumed that 12% of patients assigned to CABG would have a primary end-point event within 1 year. They also assumed that PCI patients would not be at higher risk than CABG patients. Under those assumptions, they used a noninferiority boundary of less than 1.45 for the upper end of the 95% confidence interval of the hazard ratio. They calculated that 712 patients per group would be needed. That is 1,424 patients in total. This would give 90% statistical power to claim noninferiority.

During enrollment, and without knowing what the actual event rates were, the trial steering committee decided to relax the noninferiority margin to less than 1.65. They judged this more appropriate based on newly published trials comparing CABG with PCI. Those trials reported major adverse cardiac or cerebrovascular events in no more than 10% of patients assigned to bypass surgery. Those trials used noninferiority margins similar to a hazard ratio of 1.65. With the wider margin, only 645 patients per group (1,290 total) were needed for 90% power. Even so, the committee completed the scheduled enrollment of 1,500 patients.

Between-group differences were shown with cumulative-incidence curves and estimated with a Cox proportional hazards model. That model adjusted for the stratification factors by including baseline diabetes status and by allowing the baseline risk to vary from site to site. Each component of the primary end point was compared separately. Safety end points were compared using chi-square or Fisher's exact tests as appropriate. The proportional hazards assumption was checked with a two-sided score test of the scaled Schoenfeld residuals at the 0.05 level. Noninferiority was assessed with a Wald test at the 0.025 significance level.

Subgroup analyses used the same Cox proportional hazards framework and were displayed as a forest plot for these prespecified characteristics: age (under 65 vs. 65 or older), sex, presence or absence of diabetes, presence or absence of acute coronary syndrome, left ventricular ejection fraction (above 50% vs. 50% or below), previous PCI, and core laboratory–assessed SYNTAX score. A post hoc sensitivity analysis used an alternative definition of procedure-related clinically relevant heart attack proposed by the Society for Cardiovascular Angiography and Interventions. Analyses were independently reproduced within the statistical team using SAS version 9.4 and R version 4.0.15.

Main Results: The Primary End Point

Follow-up at 1 year was completed in 99.7% of patients — an exceptionally high rate that makes the results highly reliable.

FFR-guided PCI did not meet the trial's definition of noninferiority. At 1 year, the primary end point had occurred in 80 of 757 PCI patients (10.6%, about 11 in 100) and in 51 of 743 CABG patients (6.9%, about 7 in 100). The hazard ratio was 1.5, with a 95% confidence interval of 1.1 to 2.2. The P value for noninferiority was 0.35.

In plain terms, that P value means the result failed the statistical test required to declare stenting "no worse" than surgery. The confidence interval is also informative. Even the most optimistic end of the range (1.1) is above 1.0. The entire range sits well above the pre-specified noninferiority boundary of 1.65 only at its lowest end. It does not. When a confidence interval does not fall entirely below the noninferiority limit, noninferiority cannot be claimed.

Individual Outcomes and Repeat Procedures

When the researchers looked at each piece of the primary end point separately, there was no clear evidence of a difference between the groups. The numbers were as follows:

• Death from any cause: 1.6% with PCI (about 2 in 100) vs. 0.9% with CABG (about 1 in 100); hazard ratio 1.7 (95% CI, 0.7 to 4.3).
• Death from cardiac causes: 0.8% vs. 0.5%.
• Heart attack (myocardial infarction): 5.2% vs. 3.5%; hazard ratio 1.5 (95% CI, 0.9 to 2.5). Of these, spontaneous heart attacks occurred in 3.3% vs. 2.3%, and procedural heart attacks in 1.7% vs. 1.2%.
• Stroke: 0.9% with PCI vs. 1.1% with CABG; hazard ratio 0.9 (95% CI, 0.3 to 2.4).
• Combined death, heart attack, or stroke: 7.3% with PCI (about 7 in 100) vs. 5.2% with CABG (about 5 in 100); hazard ratio 1.4 (95% CI, 0.9 to 2.1).
• Repeat revascularization: 5.9% with PCI vs. 3.9% with CABG; hazard ratio 1.5 (95% CI, 0.9 to 2.3). Of these repeat procedures, 5.2% vs. 3.5% were repeat PCIs, and 0.8% vs. 0.4% were repeat bypass operations.

The authors note an important caution here. The confidence intervals for these secondary end points were not adjusted for multiple comparisons. So they should not be interpreted as definitive proof of a treatment effect on any single component.

There were no obvious differences between the groups in the medications patients were taking at 1 year. The one exception was that more patients in the FFR-guided PCI group were still on dual antiplatelet therapy and nitrate therapy.

Safety: Where Bypass Surgery Caused More Problems

The safety picture ran in the opposite direction from the effectiveness picture. Patients randomly assigned to CABG had longer hospital stays and higher rates of several complications. Patients lost to follow-up before the end of the first year were excluded from these comparisons.

• Major bleeding (Bleeding Academic Research Consortium type 3 to 5, meaning severe bleeding): 1.6% with PCI (about 2 in 100) vs. 3.8% with CABG (about 4 in 100); P = 0.009.
• Acute kidney injury: 0.1% with PCI (about 1 in 1,000) vs. 0.9% with CABG (about 1 in 100); P = 0.04. Acute kidney injury was defined as a rise in serum creatinine of at least 0.3 mg per deciliter (26.5 micromoles per liter) within 48 hours. Acute kidney injury was also defined as a rise to at least 1.5 times a known or presumed baseline within the previous 7 days. Acute kidney injury was also defined as urine output below 0.5 ml per kilogram of body weight per hour for 6 hours.
• Atrial fibrillation or clinically significant arrhythmia (irregular heart rhythm): 2.4% with PCI (about 2 in 100) vs. 14.1% with CABG (about 14 in 100); P < 0.001. This was the largest safety difference in the trial.
• Definite stent thrombosis: 0.8% in the PCI group (6 patients); not applicable to the CABG group.
• Definite symptomatic graft occlusion: 1.3% in the CABG group (10 patients); not applicable to the PCI group.
• Rehospitalization within 30 days: 5.5% with PCI (about 6 in 100) vs. 10.2% with CABG (about 10 in 100); P < 0.001.

What These Findings Mean for Patients

The headline conclusion is straightforward. In patients with three-vessel coronary artery disease, FFR-guided PCI was not found to be noninferior to CABG. This applied to preventing the combined outcome of death, heart attack, stroke, or repeat revascularization at 1 year.

The authors point out that these results align with earlier trials comparing CABG with PCI. But there is one important difference. In this trial, FFR was measured routinely to guide the stenting, with the expectation that this would lead to more judicious use of stents. The idea was that an FFR-guided strategy would treat only the blockages that genuinely restricted blood flow. Those blockages are known to cause more adverse events when treated with medication alone. The strategy would also avoid stenting blockages that were not limiting flow. Those blockages would respond just as well to medication.

Even with that more selective approach and with modern drug-eluting stents, stenting still fell short of surgery on the trial's primary measure. That makes the finding especially meaningful. The advantage of bypass surgery in this patient group does not appear to be explained away by older stent technology. The advantage also does not appear to be explained away by stenting blockages that did not need stents.

At the same time, the two treatments carry very different risk profiles, and patients and their care teams must weigh both sides. Bypass surgery involved a longer wait to treatment (median 13 days vs. 4), a longer operation (median 197 minutes vs. 87), a longer hospital stay (median 11 days vs. 3), more severe bleeding (3.8% vs. 1.6%), far more irregular heart rhythms (14.1% vs. 2.4%), more kidney injury (0.9% vs. 0.1%), and more readmissions within 30 days (10.2% vs. 5.5%). Stenting, on the other hand, carried a higher rate of the combined serious-event outcome over the first year.

This means the decision is genuinely a trade-off. For many patients with three-vessel disease, surgery offers better protection against future heart attacks, strokes, repeat procedures, and death. But surgery comes at the cost of a harder recovery and more short-term complications. Stenting offers a faster, less invasive recovery, but with a higher risk of needing another procedure or having a serious event in the first year.

Limitations of the Study

Several aspects of this trial deserve careful thought before drawing broad conclusions.

1. The results cover one year only. Longer-term follow-up will be needed to see whether the gap between the two treatments narrows, widens, or stays the same over 5 or 10 years.
2. The noninferiority margin was changed mid-trial. The steering committee widened the acceptable margin from a hazard ratio below 1.45 to below 1.65 while enrollment was ongoing. Even with the wider margin, the trial still failed to show noninferiority — which strengthens rather than weakens the conclusion, but the change should be noted.
3. The secondary end points were not adjusted for multiple comparisons. The authors explicitly warn that these confidence intervals should not be used to make definitive claims about individual outcomes such as death or stroke.
4. Certain patients were excluded. People with recent STEMI heart attacks, cardiogenic shock, or an ejection fraction below 30% were not included, so the findings may not apply to the sickest patients.
5. The population was mostly male (about 81–83%) and mostly White (about 92–94%). Results may not generalize fully to women or to more diverse populations.
6. FFR was measured before CABG in only 10% of patients, so the bypass procedures were largely not guided by flow measurement, unlike the PCI procedures.

Practical Takeaways

If you or someone you care for has blockages in all three main heart arteries, here is what this trial suggests.

1. Ask whether your case has been reviewed by a heart team. In this trial, both PCI and CABG had to be technically feasible, as judged by a multidisciplinary team. That same discussion is the right setting for your decision.
2. Understand that both options are legitimate, but they carry different risks. Bypass surgery gave better protection against the combined outcome of death, heart attack, stroke, or repeat procedures at 1 year. Stenting gave a far easier short-term course, with fewer rhythm problems, less bleeding, less kidney injury, and shorter hospital stays.
3. Weigh the recovery period honestly. In this trial, bypass patients stayed in the hospital a median of 11 days, compared with 3 days for stenting. They were also more than twice as likely to be readmitted within 30 days (10.2% vs. 5.5%).
4. Follow the medication plan closely, whichever procedure you have. Everyone in this trial was expected to take aspirin, a high-dose statin, and guideline-directed medical therapy. PCI patients also took a second antiplatelet drug for at least 6 months.
5. Keep your follow-up appointments. The trial's high follow-up rate (99.7% at 1 year) is part of why its results are trustworthy. Regular visits at discharge and at 1, 6, and 12 months help catch problems early.
6. Talk with your cardiologist and cardiac surgeon about your personal risk factors — diabetes, kidney function, ejection fraction, and how complex your blockages are (the SYNTAX score) — since these influence which approach is likely to serve you best.

Frequently Asked Questions

What is FFR-guided stenting?

FFR (fractional flow reserve) is measured with a thin pressure wire in the artery to show how much a blockage limits blood flow. In FFR-guided stenting, only blockages with an FFR of 0.80 or lower are treated with a stent. A low FFR means the blockage truly restricts flow; a high FFR means it does not.

Who was included in the FAME 3 trial?

The trial enrolled 1,500 patients with three-vessel coronary artery disease. This means at least 50% narrowing in each of the three main heart arteries or their major branches, with the left main artery not involved. Blockages had to be treatable by either stenting or bypass. People with recent severe heart attack, cardiogenic shock, or ejection fraction below 30% were excluded.

What were the main results at one year?

At one year, 10.6% of the stenting group (about 11 in 100) had died, had a heart attack, had a stroke, or needed another procedure, compared with 6.9% of the bypass group (about 7 in 100). The hazard ratio was 1.5. Stenting did not meet the trial's definition of noninferiority to bypass surgery.

What does a hazard ratio of 1.5 mean?

A hazard ratio of 1.5 means that, over the same period, the risk of the combined outcome was about 50% higher in the stenting group than in the bypass group. The 95% confidence interval was 1.1 to 2.2, meaning the data are consistent with a risk increase somewhere between 10% and 120%.

How did recovery compare between stenting and bypass?

In this trial, stenting was faster: median time to procedure was 4 days versus 13 for bypass, procedure duration was 87 minutes versus 197, and hospital stay was 3 days versus 11. Bypass patients were more than twice as likely to be readmitted within 30 days (10.2% versus 5.5%).

What should patients discuss with their care team?

Ask whether your case has been reviewed by a heart team, since both stenting and bypass had to be technically feasible in this trial. Discuss your personal risk factors, including diabetes, kidney function, ejection fraction, and how complex your blockages are (the SYNTAX score), because these influence which approach may serve you better.

When should a patient with three-vessel coronary artery disease seek a second opinion about stenting versus bypass surgery?

When you have blockages of at least 50% in all three main heart arteries and both stenting and bypass are technically possible, a second opinion can help you weigh the trade-off. FFR-guided stenting did not match bypass surgery at one year for death, heart attack, stroke, or repeat procedures (10.6% versus 6.9%). Bypass caused more bleeding, irregular rhythms, kidney injury, and longer hospital stays. A second review can check whether your case was assessed by a heart team and how your diabetes, kidney function, ejection fraction, and SYNTAX score fit. Diagnostic Detectives Network provides independent expert second opinions.

Source Information

Original article title: Fractional Flow Reserve–Guided PCI as Compared with Coronary Bypass Surgery FAME3

Authors: W.F. Fearon, F.M. Zimmermann, B. De Bruyne, Z. Piroth, A.H.M. van Straten, L. Szekely, G. Davidavičius, G. Kalinauskas, S. Mansour, R. Kharbanda, N. Östlund‑Papadogeorgos, A. Aminian, K.G. Oldroyd, N. Al‑Attar, N. Jagic, J.‑H.E. Dambrink, P. Kala, O. Angerås, P. MacCarthy, O. Wendler, F. Casselman, N. Witt, K. Mavromatis, S.E.S. Miner, J. Sarma, T. Engstrøm, E.H. Christiansen, P.A.L. Tonino, M.J. Reardon, D. Lu, V.Y. Ding, Y. Kobayashi, M.A. Hlatky, K.W. Mahaffey, M. Desai, Y.J. Woo, A.C. Yeung, and N.H.J. Pijls, for the FAME 3 Investigators

Publication: The New England Journal of Medicine, 2022; volume 386, pages 128–137. Published online November 4, 2021; updated January 13, 2022. DOI: 10.1056/NEJMoa2112299. Copyright © 2021 Massachusetts Medical Society.

Funding: Medtronic and Abbott Vascular provided research grants to Stanford University. The funders had no role in the design or conduct of the trial or in the preparation of the manuscript.

Trial registration: FAME 3, ClinicalTrials.gov number NCT02100722.

This patient-friendly article is based on peer-reviewed research. It is intended for educational purposes and does not replace personalized medical advice from your own care team.