Principal Findings:

Of all the patients who were screened, a team consisting of a surgeon and an interventionalist decided if the anatomy was suitable for both CABG and PCI, CABG only, or PCI only. The patients who were amenable to both options constituted the study population for this study. They were further stratified by the presence of LM disease and diabetes mellitus. The investigators also calculated the SYNTAX score for each patient, based on the coronary lesion complexity.

A total of 1,800 patients were randomized, 897 to CABG and 903 to DES-PCI. About 25.1% were diabetic, with 33% having evidence of prior myocardial infarction (MI). Unstable angina was present in about 28.5% of the patients, and the additive mean EuroSCORE was 3.8. The total SYNTAX score was similar between the CABG and PCI arms (29.1 vs. 28.4, p = 0.19). The time to procedure was significantly shorter in the PCI arm (17.4 vs. 6.9 days, p < 0.0001). Medical management at discharge was better in the PCI arm.

The mean number of lesions was about 4.4, with about 66% having 3-VD only (without LM involvement), and 3.4% having pure LM disease. LM + 1-, 2-, and 3-VD was present in about 5.3%, 11.8%, and 13.7% of the patients, respectively. The number of bifurcation and trifurcation lesions was similar between the two groups (73.3% vs. 72.4%, 10.6% vs. 10.7%).

Of the patients undergoing PCI, the mean number of stents implanted was 4.6, with an average of 3.6 lesions being treated per patient. About 14.1% had a staged procedure. About 15% of the patients in the CABG arm underwent off-pump CABG; at least one arterial graft was used in 97.3% of the patients (95.6% with an arterial graft to the left anterior descending artery). The mean number of grafts per patient was 2.8.

The incidence of the primary endpoint of major adverse cardiac and cerebrovascular events (MACCE) at 12 months was lower in the CABG arm compared with PCI (12.4% vs. 17.8%, p = 0.002), and did not meet the prespecified noninferiority threshold for PCI. This was driven predominantly by a significant reduction in the incidence of repeat revascularization in the CABG arm compared with PCI (5.9% vs. 13.5%, p < 0.001). There was no difference in the incidence of death (3.5% vs. 4.4%, p = 0.37) or MI between the two arms (3.3% vs. 4.8, p = 0.11). The incidence of cerebrovascular accident was significantly higher in the CABG arm (2.2% vs. 0.6%, p = 0.003), whereas the incidence of symptomatic graft occlusion and stent thrombosis was similar between the two arms (3.4% vs. 3.3%, p = 0.89).

On subgroup analysis of the LM patients, the overall 12-month MACCE event rate was lower with CABG (13.7% vs. 15.8%), although patients with LM only (8.5% vs. 7.1%) and LM + 1-VD (13.2% vs. 7.5%) seemed to do slightly better with PCI. Patients with LM + 2-VD (14.4% vs. 19.8%), LM + 3-VD (15.4% vs. 19.4%), or 3-VD alone (11.5% vs. 19.2%) seemed to do better with CABG than PCI. When stratified by diabetes status, patients with diabetes had lower 12-month MACCE event rates with CABG than with PCI (14.2% vs. 26.0%, p = 0.0025), whereas nondiabetic patients showed a trend toward a benefit with CABG, but it was not statistically significant (11.8% vs. 15.1%, p = 0.08).

Three-year follow-up: The incidence of the primary endpoint of MACCE was still lower in the CABG arm as compared with the PCI arm (20.2% vs. 28.0%, p < 0.001), once again driven by a significant reduction in repeat revascularization with CABG (10.7% vs. 19.7%, p < 0.0001). Interestingly, the rates of MI were also lower in the CABG arm at 3 years (3.6% vs. 7.1%, p = 0.0002); strokes were numerically higher with CABG, but no longer statistically significant (3.4% vs. 2.0%, p = 0.07). On subgroup analysis, no difference in the primary endpoint was noted in the LM subset (22.3% vs. 26.8%, p = 0.20), but was statistically lower with CABG in the 3-VD subset (18.8% vs. 28.8%, p < 0.0001). When stratified by SYNTAX score, all patients with low (0-22) and some patients with intermediate (23-32) SYNTAX scores did equally well with PCI as compared with CABG.

Five-year follow-up: The incidence of the primary endpoint of MACCE was still lower in the CABG arm as compared with the PCI arm (26.9% vs. 37.3%, p < 0.0001). Reductions were also noted in the incidence of repeat revascularization (13.7% vs. 25.9%, p < 0.0001) and MI (3.8% vs. 9.7%, p < 0.0001); strokes remained numerically higher with CABG, but not statistically significant (3.7% vs. 2.4%, p = 0.09). The incidence of stent thrombosis or graft occlusion was similar (4.0% vs. 5.5%). All-cause mortality was similar between CABG and PCI (11.4% vs. 13.9%, p = 0.1). However, CABG was superior to PCI for CV (5.8% vs. 9.6%, p = 0.008) and cardiac (5.3% vs. 9.0%, p = 0.003) mortality. MI-related mortality was particularly low with CABG (0.4% vs. 4.1%, p < 0.0001).

On subgroup analysis, no difference in the primary endpoint was noted in the LM subset (31.0% vs. 36.9%, p = 0.12), but was statistically lower with CABG in the 3-VD subset (24.2% vs. 37.5%, p < 0.0001). When stratified by SYNTAX score, all patients with low (0-22) and the LM subset of patients with intermediate (23-32) SYNTAX scores did equally well with PCI as compared with CABG.

In the overall LM subset, MACCE rates at 5 years were similar between PCI and CABG (36.9% vs. 31.0%, p = 0.12), with higher rates of stroke with CABG (1.5% vs. 4.3%, p = 0.03) and higher rates of repeat revascularization with PCI (26.7% vs. 15.5%, p < 0.01).

In the overall 3-VD subset, MACCE rates at 5 years were higher in the PCI arm compared with CABG (37.5% vs. 24.2%, p < 0.001), including a higher risk of all-cause mortality (14.6% vs. 9.2%, p = 0.006), MI (9.2% vs. 4.0%, p = 0.001), and repeat revascularization (25.4% vs. 12.6%, p < 0.001), with a similar risk of stroke (3.0% vs. 3.5%). When stratified by SYNTAX score, patients with low (0-22) SYNTAX scores had comparable MACCE rates with PCI or CABG (33.3% vs. 26.8%, p = 0.21), with a higher rate of repeat revascularization (25.4% vs. 12.6%, p = 0.038) with PCI.

Ten-year follow-up (n = 1,689):

• Mortality: 27% with PCI vs. 24% with CABG (p = 0.092)
• Mortality among those with 3-VD: 28% with PCI vs. 21% with CABG
• Mortality among those with LM disease: 26% with PCI vs. 28% with CABG (p for interaction = 0.019)
• There was no treatment interaction according to diabetes status (p for interaction = 0.66)

Cost-effectiveness: Total procedure costs were lower in the CABG arm compared with PCI ($8,504 vs. $11,919) due to a large number of stents required (mean 4.6). However, overall hospitalization costs were higher in the CABG arm ($33,190 vs. $23,154; Δ = $10,036; p < 0.001). Over 5 years of follow-up, the cumulative cost difference narrowed to $5,619. The lifetime analysis demonstrated that CABG was likely to be highly cost-effective (incremental cost-effectiveness ratio = $16,537 per quality-adjusted life-year gained). On subgroup analysis, PCI was the dominant strategy in patients with LM disease and those with a low SYNTAX score; CABG was cost-effective in patients with 3-VD and intermediate and high SYNTAX scores.

Impact of revascularization on long-term outcomes: Repeat revascularization was higher with initial PCI vs. CABG (25.9% vs. 13.7%, p < 0.0001). Multiple repeat revascularizations were also higher with PCI compared with CABG (9.0% vs. 2.8%, p = 0.022). At 5-year follow-up, patients who underwent repeat revascularization vs. patients not undergoing repeat revascularization had significantly higher rates of the composite safety endpoint of death, stroke, and myocardial infarction (MI) after initial PCI (33.8 vs. 16.6%, respectively, p < 0.001), and a trend was found after initial CABG (22.4 vs. 15.8%, respectively, p = 0.07). These differences were statistically significant after multivariable adjustment.

Quality-of-life assessments: Compared with baseline, both the DES-PCI and CABG groups demonstrated significant improvement in cardiac-specific health status, as assessed by the Seattle Angina Questionnaire (SAQ) by 1-month follow-up with further improvement at 6 months and 1 year; these changes were largely sustained at 3 and 5 years. Compared with CABG, patients assigned to DES-PCI had more rapid health status improvement with higher scores on the SAQ-PL, SAQ-QOL, and SAQ-TS scales at 1 month. These differences were not sustained at 1 year. The proportion of patients who were angina-free was significantly greater with CABG than with DES-PCI at both 1-year (76.8% vs. 71.5%; p = 0.01) and 5-year (78.8% vs. 74.1%; p = 0.03) follow-up. Patients with high SYNTAX scores were more frequently angina-free at 5 years with CABG than DES-PCI (82.2% vs. 73.3%); however, there were no significant differences in the proportion of angina-free patients with low (72.3% vs. 72.5%; p = 0.94) or intermediate (81.6% vs. 75.6%; p = 0.21) SYNTAX scores.