OUP user menu

★ fast track Editor's choice ★

Coronary artery bypass grafting vs. percutaneous coronary intervention for patients with three-vessel disease: final five-year follow-up of the SYNTAX trial

Stuart J. Head, Piroze M. Davierwala, Patrick W. Serruys, Simon R. Redwood, Antonio Colombo, Michael J. Mack, Marie-Claude Morice, David R. Holmes Jr, Ted E. Feldman, Elisabeth Ståhle, Paul Underwood, Keith D. Dawkins, A. Pieter Kappetein, Friedrich W. Mohr
DOI: http://dx.doi.org/10.1093/eurheartj/ehu213 2821-2830 First published online: 21 May 2014


Aims Coronary artery bypass grafting (CABG) has been considered the standard of care for patients with three-vessel disease (3VD), but long-term comparative results from randomized trials of CABG vs. percutaneous coronary intervention (PCI) using drug-eluting stents (DES) remain limited.

Methods and results Patients with de novo 3VD or left main disease were randomly assigned to PCI with the paclitaxel-eluting first-generation stent or CABG in the SYNTAX trial. This pre-specified analysis presents the 5-year outcomes of patients with 3VD (n = 1095). The rate of major adverse cardiac and cerebrovascular events (MACCE) was significantly higher in patients with PCI compared with CABG (37.5 vs. 24.2%, respectively; P < 0.001). Percutaneous coronary intervention as opposed to CABG resulted in significantly higher rates of the composite of death/stroke/myocardial infarction (MI) (22.0 vs. 14.0%, respectively; P < 0.001), all-cause death (14.6 vs. 9.2%, respectively; P = 0.006), MI (9.2 vs. 4.0%, respectively; P = 0.001), and repeat revascularization (25.4 vs. 12.6%, respectively; P < 0.001); however, stroke was similar between groups at 5 years (3.0 vs. 3.5%, respectively; P = 0.66). Results were dependent on lesion complexity (P for interaction = 0.12); in patients with a low (0–22) SYNTAX score, PCI vs. CABG resulted in similar rates of MACCE (33.3% vs. 26.8%, respectively; P = 0.21) but significantly more repeat revascularization (25.4% vs. 12.6%, respectively; P = 0.038), while in intermediate (23–32) or high (≥33) SYNTAX score terciles, CABG demonstrated clear superiority in terms of MACCE, death, MI, and repeat revascularization. Differences in MACCE between PCI and CABG were larger in diabetics [hazard ratio (HR) = 2.30] than non-diabetics (HR = 1.51), although the P for interaction failed to reach significance for MACCE (P for interaction = 0.095) or any of the other endpoints.

Conclusion Five-year results of patients with 3VD treated with CABG or PCI using the first-generation paclitaxel-eluting DES suggest that CABG should remain the standard of care as it resulted in significantly lower rates of death, MI, and repeat revascularization, while stroke rates were similar. For patients with low SYNTAX scores, PCI is an acceptable revascularization strategy, although at a price of significantly higher rates of repeat revascularization.

Clinical trial registration NCT00114972.

  • Percutaneous coronary intervention
  • Coronary artery bypass grafting
  • Three-vessel disease
  • Randomized trial
  • Incomplete revascularization
  • Diabetes

See page 2788 for the editorial comment on this article (doi:10.1093/eurheartj/ehu253)


Because of the rapid progress in percutaneous coronary intervention (PCI) technology from balloon angioplasty to bare-metal stents (BMS) and subsequently drug-eluting stents (DES), several randomized clinical trials have been conducted over the last two decades to compare the outcomes of PCI with coronary artery bypass grafting (CABG), which has been considered the ‘gold standard’ for treatment of multivessel stable coronary artery disease (CAD).1,2 Consistent improvements in outcomes of PCI led to a wider spectrum of patients being treated with PCI, including those with more complex CAD.

The SYNTAX trial, which is one of the most recent and largest randomized controlled trials comparing PCI using the paclitaxel-eluting first-generation stent with CABG, aimed at providing the best possible evidence to determine the most appropriate treatment option for patients encountered by surgeons and interventional cardiologists in their ‘real-world’ daily practice. The trial failed to establish non-inferiority of PCI to CABG for the treatment of left main (LM) and/or three-vessel disease (3VD) at 1 year, because of significantly higher rates of major adverse cardiac and cerebrovascular events (MACCE) after PCI.3 However, the data demonstrated considerable variation in the treatment effect of PCI, especially with reference to the pre-specified subgroups of LM or 3VD, and complexity of CAD as determined by the SYNTAX score.3 Based on these 1 year results, both European and North-American guidelines recommend PCI as a valuable treatment option for patients with LM disease and an alternative to CABG in patients with less complex 3VD (SYNTAX score < 23).4,5

For judicious decision-making, it is essential to consider the risk/benefit ratios of PCI and CABG for 3VD; weighing procedural invasiveness and the associated short-term complications against long-term event rates of death, myocardial infarction (MI), repeat revascularization, and improvements in health-related quality of life.6 The present report therefore presents a comprehensive analysis of the 5-year outcomes of the predefined 3VD subgroup of patients receiving PCI or CABG in the SYNTAX trial.


Study design

The study design, methods as well as mid-term outcomes of the SYNTAX trial have been published previously.3,7,8 Briefly, SYNTAX was a prospective, multinational, randomized, clinical study in which 1800 patients with de novo LM and/or 3VD, in whom clinical equipoise was deemed with revascularization by either treatment, were randomly assigned to undergo PCI with TAXUS Express paclitaxel-eluting stents (Boston Scientific, Natick, MA, USA) or CABG. Patients in whom PCI or CABG was considered the treatment of choice by a multidisciplinary Heart Team were enrolled in a CABG-ineligible PCI registry (n = 198) or PCI-ineligible CABG registry (n = 1077), respectively.9 A yearly follow-up of all patients was performed by clinic visits or telephone contact up to 5 years.

Randomization was stratified by clinical site, the presence of LM disease, and medically treated diabetes mellitus. A separate analysis of patients with 3VD was pre-specified in the trial protocol, provided non-inferiority was met for the primary endpoint. However, since non-inferiority was not met, the current analysis should be interpreted as ‘hypothesis generating’. The trial is registered at ClinicalTrials.gov with number NCT00114972.


The primary endpoint of the SYNTAX trial was the rate of MACCE at 12 months after allocation and is a composite of all-cause death, stroke, MI, and repeat revascularization. Secondary endpoints consisted of (i) a composite safety endpoint of death/stroke/MI, (ii) individual endpoints of all-cause death, cardiac death, stroke, MI, and repeat revascularization, and (iii) symptomatic graft occlusion and stent thrombosis (ST). These events were considered at 1 month post-procedure and yearly post-allocation. Definitions of these events have been reported previously.7 An independent clinical events committee adjudicated all primary clinical events.

The vessels (diameter ≥1.5 mm and significant stenosis >50%) deemed important to revascularize were determined by the Heart Team prior to randomization. The group of 3VD consists of patients with significant stenosis in vessels supplying all three major epicardial territories in the absence of LM disease, as determined by the site investigators. Incomplete revascularization (IR) was assessed by correlating this pre-operative documentation to the actual revascularization performed during the procedure and was recorded by the investigator.

Statistical methods

Continuous variables are expressed as mean ± SD and compared using the Student t-test. Discrete data are presented as frequencies and compared with a χ2 or Fisher's exact tests, where appropriate. Analyses were performed according to intention to treat. Short-term outcomes were considered to be within 30 days post-procedure and were calculated as binary rates. Long-term event rates were estimated using Kaplan–Meier curves with a statistical comparison made by log-rank test. Patients lost to follow-up or who withdrew consent were non-assessable and were censored at the last available follow-up time, assumed to be event free. A sensitivity analysis was performed in which non-assessable patients were assumed to have had an event.

Subgroup analyses were performed according to SYNTAX score tercile, diabetic status, and IR, with P-values for interaction calculated using χ2 testing. Multivariable proportional hazard models with a P-value of 0.10 as entry and exit criteria were constructed to identify whether PCI treatment vs. CABG was an independent predictor of MACCE, the composite safety endpoint, and all-cause death during follow-up. Variables included in the model were believed to be clinically relevant, and PCI vs. CABG was forced in each model (see Supplementary material online, Appendix). A two-sided P-value of <0.05 was considered to indicate statistical significance. All statistical analyses were performed using the SPSS software, version 20 (IBM Corporation, Armonk, NY, USA).


Baseline characteristics

Of the 1800 patients randomized in the SYNTAX trial, 1095 patients had 3VD. Of these, 549 were randomly assigned to CABG and 546 to PCI. There were no differences between treatment groups in terms of age, gender, presence of cardiac risk factors and associated comorbidities, or expected surgical mortality as assessed by the logistic EuroSCORE (Table 1). The extent and complexity of CAD were also evenly distributed between the two groups. The majority of patients underwent elective revascularization with no differences between groups.

View this table:
Table 1

Baseline characteristics of patients with three-vessel disease

CABG (n = 549)PCI (n = 546)P-value
Age64.5 ± 9.6 (549)65.1 ± 9.6 (546)0.32
Male81.1% (445/549)79.3% (433/546)0.47
Body mass index28.0 ± 4.2 (549)28.1 ± 4.7 (545)0.88
Current smoker20.8% (114/548)18.9% (103/546)0.42
Diabetes30.4% (167/549)31.1% (170/546)0.80
 Medically treated diabetes26.0% (143/549)28.0% (153/546)0.46
 Insulin-requiring diabetes10.7% (59/549)11.4% (62/546)0.75
Hyperlipidaemia78.3% (426/544)77.2% (416/539)0.66
Blood pressure ≥130/85 mmHg65.0% (357/549)70.1% (383/546)0.07
Previous MI39.1% (213/545)34.2% (184/538)0.10
Congestive heart failure5.7% (31/540)4.2% (23/543)0.26
Carotid artery disease7.5% (41/549)7.7% (42/546)0.89
Previous stroke5.3% (29/546)3.9% (21/544)0.25
Previous transient ischaemic attack5.7% (31/545)4.2% (23/544)0.27
Peripheral vascular disease10.0% (55/549)8.2% (45/546)0.31
Creatinine >200 µmol/L1.5% (8/549)0.9% (5/546)0.41
Chronic obstructive pulmonary disease9.5% (52/549)7.7% (42/546)0.29
Unstable angina27.3% (150/549)27.8% (152/546)0.85
Ejection fraction <30%a3.1% (17/549)1.3% (7/546)0.040
Elective revascularization94.1% (497/528)95.2% (518/544)0.43
Logistic EuroSCORE3.7 ± 3.7 (549)3.5 ± 3.5 (546)0.56
Parsonnet score8.0 ± 6.5 (549)8.3 ± 6.3 (546)0.50
Coronary complexityb
 Number of lesions4.9 ± 1.6 (545)4.9 ± 1.5 (543)0.39
 Any total occlusion6.2% (167/2686)6.5% (171/2632)0.68
 Any bifurcation25.9% (697/2686)25.3% (667/2632)0.61
 Diffuse disease or small vessels11.6% (63/545)13.4% (73/543)0.35
 SYNTAX score28.4 ± 10.4 (545)27.6 ± 9.8 (543)0.17
  • CABG, coronary artery bypass grafting; MI, myocardial infarction; PCI, percutaneous coronary intervention.

  • Values are shown as mean ± SD (n) or % (n/N).

  • aOr indicated by clinical site as ‘poor’ if exact value was not available.

  • bCore laboratory reported.

Procedural characteristics

Patients randomized to CABG had significantly longer waiting times for the procedure as opposed to PCI (Table 2). Procedure times and post-procedural hospital stay for patients who underwent CABG were significantly longer than those who underwent PCI.

View this table:
Table 2

Procedural characteristics of patients with three-vessel disease

CABG (n=549)PCI (n = 546)P-value
Allocation to procedure (days)19.2 ± 28.1 (528)7.6 ± 14.1 (544)<0.001
Procedure time (h)3.5 ± 1.0 (522)1.8 ± 1.0 (544)<0.001
Bypass time (min)89.2 ± 33.6 (436)
Cross clamp time (min)57.3 ± 23.6 (402)
Off-pump surgery13.9% (72/517)
Complete arterial17.0% (88/517)
Bilateral internal mammary artery use30.6% (158/517)
Number of grafts, n2.9 ± 0.7 (517)
 Arterial grafts, n1.5 ± 0.7 (517)
 Venous grafts, n1.4 ± 0.9 (517)
Number of distal anastomoses, n3.4 ± 0.8 (517)
Number of stents implanted, n5.3 ± 2.1 (530)
Total stent length implanted (mm)99.4 ± 45.5 (530)
Long stenting (>100 mm)41.5% (220/530)
Incomplete revascularization42.8% (226/528)48.3% (263/544)0.069
Post-procedural hospital stay (days)9.5 ± 8.1 (528)3.6 ± 5.1 (544)<0.001
30-day post-procedural outcomes
 MACCE4.8% (25/524)7.0% (38/543)0.12
 Death/stroke/MI4.0% (21/524)5.5% (30/543)0.25
 All-cause death1.3% (7/524)2.2% (12/543)0.28
 Stroke1.3% (7/524)0.4% (2/543)0.10
 MI1.7% (9/524)4.2% (23/543)0.016
 Repeat revascularization1.1% (6/524)3.9% (21/543)0.005
  PCI0.4% (2/424)2.6% (14/543)0.003
  CABG0.8% (4/524)1.3% (7/543)0.40
  • CABG, coronary artery bypass grafting; MACCE, major adverse cardiac and cerebrovascular events; MI, myocardial infarction; PCI, percutaneous coronary intervention.

  • Values are shown as mean ± SD (n) or % (n/N).

Amongst patients randomized to CABG, off-pump surgery was performed in only 13.9% of patients. Bilateral internal mammary arteries were used in 30.6% of patients, with total arterial revascularization being performed in 17.0% of patients. The mean number of grafts and distal anastomoses per patient were 2.9 ± 0.7 and 3.4 ± 0.8, respectively. Patients undergoing PCI received a mean of 5.3 ± 2.1 stents with a mean stent length of 99.4 ± 45.5 mm per patient. A total stent length of >100 mm was implanted in >40% of patients. Patients who underwent CABG less commonly had IR; however, it did not reach statistical significance.

Patients who underwent PCI generally received more secondary preventive medication during follow-up than those who underwent CABG. Antiplatelet therapy was given more frequently in the PCI group at all time points, and at 5 years was still significantly higher following PCI than CABG for administration of a thienopyridine (36.7 vs. 11.0%, respectively; P < 0.001) and dual antiplatelet therapy (20.8 vs. 6.4%, respectively; P < 0.001). The use of other medications was exclusively higher after PCI during the first year of follow-up, but at 5 years rates were comparable between PCI and CABG: beta-blockers (75.7 vs. 76.1%, respectively; P = 0.90), ACE-inhibitors (55.2 vs. 52.9%, respectively; P = 0.49), calcium-channel blockers (25.4 vs. 22.3%, respectively; P = 0.28), angiotensin II-receptor antagonists (20.3 vs. 20.9%, respectively; P = 0.82), and statins (83.2 vs. 85.3%, respectively; P = 0.40) (Supplementary material online, Table S1).

Pre-procedural MACCE was 0.7% (n = 4) in the CABG group and 0.2% (n = 1) in the PCI group (P = 0.38). A total of 19 patients died within 30 days after revascularization: 12 after PCI and 7 after CABG (P = 0.28) (Table 2). Although the stroke rate was non-significantly higher after CABG (P = 0.10), MI and repeat revascularization occurred significantly more frequently after PCI. The rate of MACCE at 30 days was comparable between PCI and CABG (7.0 vs. 4.8%, respectively; P = 0.064).

Five-year outcomes

At 5 years follow-up, 88% of CABG-randomized and 96% of PCI-randomized 3VD patients were available for analysis (see Supplementary material online, Figure S1). Overall, the occurrence of MACCE at 5 years was significantly higher in patients who underwent PCI when compared with those who underwent CABG [37.5 (n = 123) vs. 24.2% (n = 201), respectively; P < 0.001] (Figure 1A). Similarly, the composite safety endpoint of death/stroke/MI was more commonly observed after PCI than after CABG [22.0 (n = 118) vs. 14.0% (n = 71), respectively; P < 0.001] (Figure 1B), with the difference driven by significantly higher rates of all-cause death [14.6 (n = 78) vs. 9.2% (n = 46), respectively; P = 0.006] and MI [10.6 (n = 55) vs. 3.3% (n = 17), respectively; P < 0.001] (Figures 1C and E). Furthermore, patients had significantly higher rates of cardiac death after PCI vs. CABG [9.2 (n = 48) vs. 4.0% (n = 20), respectively; hazard ratio (HR) 2.34, 95% CI 1.39–3.95; P < 0.001]. Patients required repeat revascularization more frequently after PCI than after CABG [25.4 (n = 130) vs. 12.6% (n = 61), respectively; P < 0.001] (Figure 1F), driven by significant differences in both repeat PCI [21.5 (n = 110) vs. 12.3% (n = 59), respectively; HR 1.96, 95% CI 1.41–2.65; P < 0.001] and repeat CABG [5.6 (n = 28) vs. 1.0% (n = 5), respectively; HR 5.56, 95% CI 2.15–14.41; P < 0.001]. Graft occlusion after surgery occurred in 18 patients (3.7%), resulting in an MI in 3 patients, repeat revascularization in 12 patients, and 3 patients required no treatment; there were no deaths. Thirty PCI patients (5.8%) experienced ST; of these, 9 patients died, 11 experienced an MI, and 10 underwent repeat revascularization. Occurrence of angina was also significantly higher at all time points after PCI than after CABG (Figure 2).

Figure 1

Five-year estimates of adverse events. Kaplan–Meier curves for (A) MACCE, (B) all-cause death/stroke/MI, (C) all-cause death, (D) stroke, (E) MI, and (F) repeat revascularization. Bars represent ±1.5 standard error. P-values from log-rank test. CABG, coronary artery bypass grafting; CI, confidence interval; HR, hazard ratio; PCI, percutaneous coronary intervention; MACCE, major adverse cardiac and cerebrovascular events; MI, myocardial infarction.

Figure 2

Presence of angina during follow-up. CABG, coronary artery bypass grafting; PCI, percutaneous coronary intervention.

In the sensitivity analysis, assuming all patients who were lost to follow-up or withdrew consent had an event, MACCE remained significantly higher after PCI than after CABG (40.7 vs. 34.6%, respectively; HR 1.23, 95% CI 1.01–1.49; P = 0.040), although there was no longer a difference in the composite of all-cause death/stroke/MI (26.0 vs. 26.6%, respectively; HR 0.96, 95% CI 0.76–1.21; P = 0.72).

Predictors of 5-year outcomes

The multivariable proportional hazard model revealed treatment with PCI to be an independent predictor of MACCE over 5-year follow-up (HR 1.66, 95% CI 1.32–2.09; P < 0.001), in addition to several pre-operative patient characteristics and IR (Table 3). Furthermore, treatment with PCI not only remained a significant independent predictor of the composite safety endpoint (HR 1.81, 95% CI 1.33–2.46; P < 0.001) but also of all-cause death (HR 1.81, 95% CI 1.24–2.67; P = 0.002).

View this table:
Table 3

Multivariable predictors of adverse events

HR (95% CI)P-value
 PCI treatment vs. CABG1.66 (1.32–2.09)<0.001
 Age1.01 (1.00–1.03)0.024
 Poor LV function2.08 (1.03–4.21)0.042
 Hypertension1.31 (0.98–1.75)0.067
 Peripheral vascular disease1.76 (1.27–2.43)0.001
 Incomplete revascularization1.37 (1.10–1.72)0.006
 Enrolment in the United States1.47 (1.06–2.06)0.023
Composite safety endpoint of death/stroke/MI
 PCI treatment vs. CABG1.81 (1.33–2.46)<0.001
 Age1.04 (1.02–1.06)<0.001
 Poor LV function3.23 (1.50–6.96)0.003
 Moderate LV function1.43 (1.02–2.02)0.040
 Hypertension1.47 (0.98–2.20)0.061
 Peripheral vascular disease1.90 (1.28–2.83)0.002
 History of TIA or stroke1.50 (0.97–2.30)0.066
 Chronic obstructive pulmonary disease1.87 (1.23–2.84)0.003
All-cause death
 PCI treatment vs. CABG1.81 (1.24–2.67)0.002
 Age1.07 (1.05–1.10)<0.001
 Previous MI1.38 (0.94–2.03)0.098
 Poor LV function3.17 (1.25–8.08)0.015
 Moderate LV function1.58 (1.03–2.41)0.035
 Peripheral vascular disease3.13 (2.00–4.89)<0.001
 Chronic obstructive pulmonary disease2.25 (1.38–3.68)0.001
 Creatinine ≥200 µmol/L2.46 (0.87–6.94)0.090
  • CABG, coronary artery bypass grafting; HR, hazard ratio; LV, left ventricular; MACCE, major adverse cardiac and cerebrovascular events; MI, myocardial infarction; PCI, percutaneous coronary intervention; TIA, transient ischaemic attack.

Subgroup analyses of 5-year outcomes

Medically treated diabetes

Overall, 296 patients (CABG, n = 143 and PCI, n = 153) with 3VD had medically treated diabetes (Figure 3A). The MACCE rates in diabetic patients with 3VD were significantly higher in the PCI group than the CABG group (HR 2.30, 95% CI 1.50–3.55; P < 0.001). Significantly higher event rates were also seen after PCI vs. CABG in the composite safety endpoint of death/stroke/MI, all-cause death, MI, and repeat revascularization. Stroke rates were comparable. Although the differences in event rates among non-diabetic patients (CABG, n = 406 and PCI, n = 393) were smaller between groups, PCI in comparison with CABG still had significantly higher rates of MACCE (HR 1.51, 95% CI 1.15–1.96; P = 0.002), the composite safety endpoint, MI, and repeat revascularization, but with similar rates of all-cause death and stroke (Figure 3B). There were no significant treatment-by-diabetes interactions.

Figure 3

Five-year estimates of adverse events in patients with diabetes (A) and without diabetes (B). Treatment-by-diabetes interactions failed to reach statistical significance for MACCE (P=0.095), the composite safety endpoint (P=0.44), all-cause death (P = 0.37), stroke (P = 0.63), MI (P = 0.88), and repeat revascularization (P = 0.15). Values are Kaplan–Meier event rates with P-values from log-rank test. CABG, coronary artery bypass grafting; composite, composite safety endpoint of all-cause death/stroke/MI; PCI, percutaneous coronary intervention; MACCE, major adverse cardiac and cerebrovascular events; MI, myocardial infarction.

SYNTAX terciles

Differences in event rates between PCI and CABG showed a step-wise increase with increasing SYNTAX scores for all events, except stroke, for which no differences were observed through all SYNTAX score terciles (Figure 4). For patients with a low SYNTAX score (0–22), no significant differences were noted between PCI and CABG treatment groups in rates of MACCE, the composite safety endpoint of death/stroke/MI, all-cause death, and MI. Patients who underwent PCI did have a significantly higher rate of repeat revascularization than those who underwent CABG. Patients with intermediate (23–33) or high SYNTAX scores (≥33), patients who underwent CABG had significantly lower rates of MACCE, the composite safety endpoint of death/stroke/MI, as well as individual components all-cause death, MI and repeat revascularization (Figure 4). There was no evidence of significant treatment-by-SYNTAX score tercile interaction.

Figure 4

Five-year estimates of adverse events according to the SYNTAX score. (A) MACCE, (B) the composite safety endpoint of all-cause death/stroke/MI, (C) all-cause death, (D) stroke, (E) MI, and (F) repeat revascularization. Treatment-by-SYNTAX score tercile interactions failed to reach statistical significance for MACCE (P = 0.12), the composite safety endpoint (P = 0.085), all-cause death (P = 0.15), stroke (P = 0.12), MI (P = 0.16), and repeat revascularization (P = 0.32). Values are Kaplan–Meier event rates with P-values from log-rank test. CABG, coronary artery bypass grafting; PCI, percutaneous coronary intervention; MACCE, major adverse cardiac and cerebrovascular events; MI, myocardial infarction.

Incomplete revascularization

In the CABG cohort, outcomes were similar in patients with complete and IR (Table 4). Conversely, in the PCI cohort, patients who underwent incomplete as opposed to complete revascularization had significantly higher rates of MACCE (42.6 vs. 32.7%, respectively; P = 0.010) and repeat revascularization (30.6 vs. 20.4%, respectively; P = 0.003). However, occurrence of the composite of death/stroke/MI and its individual endpoints were similar (Table 4). There were no significant interactions between completeness of revascularization and treatment.

View this table:
Table 4

Five-year outcomes according to completeness of revascularization

Clinical outcomeCABG (n = 528)PCI (n = 544)Interaction P-value
Incomplete (n = 226)Complete (n = 302)P-valueIncomplete (n = 263)Complete (n = 281)P-value
MACCE26.9% (59)22.2% (63)0.1742.6% (111)32.7% (90)0.0100.62
Death/stroke/MI14.6% (32)13.3% (38)0.6424.2% (63)20.0% (55)0.230.72
All-cause death10.1% (22)8.3% (23)0.4116.2% (42)13.1% (36)0.310.97
Stroke4.7% (10)2.4% (7)0.183.8% (9)2.3% (6)0.360.81
MI3.2% (7)3.4% (10)0.9010.0% (25)11.3% (30)0.690.95
Repeat revascularization15.3% (32)10.6% (29)0.09130.6% (77)20.4% (53)0.0030.75
 PCI14.9% (31)10.3% (28)0.08825.3% (63)18.0% (63)0.026
 CABG0.9% (2)1.0% (3)0.918.1% (20)3.2% (20)0.012
  • CABG, coronary artery bypass grafting; MACCE, major adverse cardiac and cerebrovascular events; MI, myocardial infarction; PCI, percutaneous coronary intervention. Values are Kaplan–Meier event rates with P-values from log-rank test.


The 5-year analysis of patients with 3VD randomized in the SYNTAX trial demonstrated that treatment with PCI resulted in significantly higher rates of MACCE, which were driven not only by increased rates of repeat revascularization but also by significantly higher all-cause death and MI. Stroke rates were, however, comparable between CABG and PCI. The 5-year outcomes show a persistent divergence in all event rates except stroke. These results were further validated by the fact that treatment with PCI was found to be an independent predictor, not only of 5-year MACCE but also of all-cause death and composite safety endpoint of death/stroke/MI.

The SYNTAX trial has been crucial for establishing the optimal revascularization strategies for patients with LM disease and 3VD. Remarkably, the treatment effect of PCI vs. CABG differed significantly between patients with LM disease and 3VD. Left main PCI has become more common since the results from the SYNTAX trial established similar safety and efficacy in comparison with CABG.10 However, for multivessel disease, unlike previous trials using simple balloon angioplasty or BMS,1,2 this study for the first time provides robust evidence that CABG is superior to PCI using first-generation DES in reducing long-term hard clinical endpoints of death and MI. The present study included more complex patients than previous trials, in which inclusion of patients with two-vessel disease probably mitigated the advantages of better survival and freedom from MI after CABG. Until this analysis, only large registries have demonstrated a long-term survival benefit in patients with 3VD disease treated with CABG. Registries always encompass a selection bias for which no degree of adjustment can be applied, thus potentially having unknown cofounders. Nevertheless, Hannan and coauthors reported significantly better adjusted HRs for death after CABG vs. PCI with BMS, and subsequently found similar results when PCI was performed with first-generation DES.11,12 The ASCERT study, which is the largest propensity-adjusted analysis of registry data involving nearly 200,000 patients, also reported a significantly lower 4-year mortality after CABG than after PCI (16.4 vs. 20.8%, respectively; risk ratio 0.79, 95% CI 0.76–0.82).13

Although occurrence of MI was significantly higher even at 30 days after PCI than after CABG, 30-day mortality and the composite of death/stroke/MI were comparable between the two treatment groups. The 30-day stroke rate, though higher, was not significantly increased after CABG (Table 2). This non-significant difference in stroke rate was evident during the first year (1.9 vs. 0.8%, P = 0.09), but declined over extended follow-up.14 The early occurrence of stroke after CABG could potentially have been reduced by use of off-pump surgery with minimal or no aortic manipulation and the use of epiaortic scanning, especially in patients with aortic calcification.15 A reduction in post-operative MI and long-term improvement in graft patency may have been achieved by use of intraoperative graft flow measurements.16 Although not standard of care, greater use of multiple arterial grafts, particularly bilateral internal mammary arteries (30.6%) and total arterial grafting (17.0%) may have also further improved the long-term results of CABG. Numerous such technical advancements in CABG continue to enhance post-operative short- and long-term outcomes of CABG.15,17 A recent analysis of ∼1 500 000 patients in the Society of Thoracic Surgeons database who underwent isolated CABG between 2000 and 2009 demonstrated a significant reduction in in-hospital mortality and morbidity, despite an increasing operative risk.18 Furthermore, in concordance with previous studies, secondary preventive medication remains suboptimal following CABG in comparison to PCI, in particular within the first year of follow-up. Why this continuous to be an issue remains unclear, but it appears that country-wide intensive education and quality improvement programs required to significantly increase compliance following revascularization could potentially further enhance long-term results following surgical revascularization.

Likewise, several newer-generation DES have been developed since the use of the paclitaxel-eluting stent in the SYNTAX trial and have replaced the latter in current clinical practice due to significant reductions in MI, ST, and repeat revascularization.19 Recent analyses even revealed lower rates of all-cause death with newer stents/scaffolds.20,21 Biodegradable polymer scaffolds may further improve outcomes following PCI.22 Despite these technological improvements in stents, the basic method of implantation within the diseased segment of the coronary artery offers no safeguard against the advancement of the pre-existing or evolution of new disease, as the native coronary artery continues to remain the source of inflow. Contrarily, bypass conduits, which are the new sources of inflow after CABG, are most often anastomosed to coronary arteries beyond the diseased segments, thus rendering the type, length, degree, and complexity of stenoses irrelevant. While newer-generation DES might diminish the difference between PCI and CABG, especially in repeat revascularization and MI due to ST, the principle difference in the method of revascularization is the major contributing factor to the long-term benefit seen with CABG.

Additionally, IR may be another factor responsible for differences in 5-year outcomes between PCI and CABG in the current study, as it negatively impacted only PCI and not CABG results. This could be explained by either a small sample size, because IR with CABG was associated with increased adverse events in an analysis of the combined SYNTAX trial and registry,23,24 or it was caused by dissimilarities in the type and, hence, consequences of IR between the two therapies. IR occurs more frequently in patients with highly complex CAD, in particular chronic total occlusions, undergoing PCI;24,25 resulting in higher post-operative residual SYNTAX scores with large areas of non-revascularized/ischaemic myocardium that could potentially increase the risk of death during follow-up.26,27 Conversely, IR in CABG usually encompasses either small vessels with less ischaemic myocardium at risk or diffusely diseased vessels which are often well collateralized. It must also be considered that lesions of 50% angiographic severity are counted in the SYNTAX score and are not routinely bypassed, but contribute to IR by definition. For these reasons, IR with CABG is often considered to be more acceptable than with PCI.28 The increased incidence of IR especially with higher SYNTAX scores correlated well with the differing treatment effect of PCI vs. CABG according to SYNTAX score terciles; although no significant IR-by-treatment interaction (P = 0.23) was observed. CABG was superior to PCI in patients with intermediate or high SYNTAX scores. Although PCI and CABG had equivalent safety profiles in patients with low-SYNTAX scores, PCI resulted in significantly higher rates of repeat revascularization.

However, accumulating data suggest that complete anatomical revascularization is not always necessary but decisions depend on the functional assessment of a lesion. Revascularization of non-significant stenosis, as measured by fractional flow reserve (FFR), increases the risk of death and MI during follow-up.29 The FAME trial demonstrated that exclusive functional and not pure anatomical complete revascularization by FFR- instead of angiography-guided PCI resulted in a significant reduction in use of stents and consequently reduced the 1-year rate of a composite of death, MI, and repeat revascularization.30 This has been confirmed recently in a ‘real-world’ scenario.31 Similarly, a recent study revealed that FFR- vs. angiography-guided CABG was associated with a lower number of bypass grafts, but comparable rates of major adverse cardiac events at 3-year follow-up (12 vs. 11%, respectively; P = 0.91). However, rates of recurrent angina significantly declined (31 vs. 47%, respectively; P < 0.001).32 The impact of FFR-guided revascularization on the PCI vs. CABG treatment effect will be determined in the FAME III trial.

In accordance with the data on the SYNTAX score and completeness of revascularization, the current European guidelines provide a IIa B recommendation for PCI in patients with 3VD and SYNTAX scores ≤22, provided full functional revascularization is feasible.4 However, it should be noted that this SYNTAX score cut-off value can always be overruled when important factors other than anatomical complexity of the lesion have to be taken into consideration. Moreover, the current SYNTAX score cut-offs of 23 and 32 were derived from the SYNTAX trial and still require confirmation from other large randomized trials comparing PCI with CABG.9 The choice of treatment strategy should therefore not solely depend on lesion complexity, but the Heart Team should also consider clinical parameters, comorbidities, treatment preferences, and operator skills, particularly in patients with low SYNTAX scores for whom both PCI and CABG are excellent options.33 A much considered factor is diabetes, the presence of which has been shown to be a strong indicator for CABG, since these patients are at an increased risk of adverse events during follow-up and have a higher long-term mortality with PCI than CABG.3436 The current analysis also confirmed that PCI was associated with HRs of 2.30 in diabetic and 1.51 in non-diabetic patients. Additionally, the difference in MACCE and composite safety endpoint of death/stroke/MI remained significant in favour of CABG amongst both diabetic and non-diabetic patients. The non-significant interaction term (P = 0.095) suggests that the treatment effect of PCI vs. CABG for 3VD remains consistent irrespective of diabetic status.

In order to incorporate clinical parameters into decision-making and to improve patient selection, the SYNTAX II score was developed based on anatomical complexity and six clinical factors found to be independent predictors of 4-year mortality in the randomized SYNTAX trial. Further validation will be forthcoming from the SYNTAX II trial (NCT02015832).37 Using SYNTAX II scores for patients with 3VD to predict 4-year mortality, CABG was favoured in 84.2% of patients in the SYNTAX trial. In the current analysis of 5-year outcomes, the Kaplan–Meier curves continued to diverge beyond this point with all-cause death rates between 4 and 5 years of 2.4% for CABG and 2.8% for PCI (P = 0.74) (Figure 1).

Study limitations

The SYNTAX trial, which involved a hierarchical statistical testing plan, was powered only for the overall 3VD/LM population and subsequently for the 3VD subgroup only if non-inferiority for the primary endpoint for the whole population was met; however, given that the latter pre-requisite was not satisfied, the analyses presented here must be considered observational and ‘hypothesis generating’ only. In addition, the number of patients in each subgroup according to SYNTAX score terciles was small and not adequately powered. Groups of patients who underwent off-pump and/or total arterial grafting were too small with limited statistical power to evaluate the impact of these techniques on the treatment effect of PCI vs. CABG.


The 5-year outcomes in patients with 3VD in the randomized arm of the SYNTAX trial suggest that treatment with CABG as opposed to PCI with the first-generation paclitaxel-eluting DES resulted in significantly lower rates of death and MI, and should therefore remain the standard of care for 3VD. Patients with complex CAD (intermediate or high SYNTAX scores) particularly benefit from CABG, whereas PCI is an acceptable treatment option in patients with less complex disease (low SYNTAX scores), though at a price of significantly higher rates of repeat revascularization. Randomized trials are needed to reassess revascularization options in the era of newer-generation DES.


This study was supported by funds from Boston Scientific Corporation.

Conflict of interest: T.E.F. has received consulting and lecture fees and research support from Boston Scientific Corporation and Abbott. P.U. and K.D.D. own stock in and are full-time employees of Boston Scientific Corporation. All other authors declare that they have no conflicts of interest.


The authors thank Kristine Roy, PhD and Vicki Houle, PhD (Boston Scientific Corporation) for assistance in manuscript preparation and Jian Huang, MD MS (Boston Scientific Corporation) for statistical analysis and assistance.


  • These authors contributed equally to the manuscript.


View Abstract