Optimal Strategy for Multivessel Coronary Revascularization

Graphical Abstract: Schematic illustrating factors including individual patient risk profile, life expectancy, and coronary disease characteristics that influence choice of optimal revascularization strategy for patients with multivessel disease. CTO, chronic total occlusion; CABG, coronary artery bypass grafting.

The optimal revascularization strategy for multivessel coronary disease depends on the individual patient risk profile, life expectancy, and disease burden. Large, randomized trials have consistently shown superior survival or freedom from major adverse cardiac events at 5 years with surgical revascularization, compared with percutaneous coronary intervention (PCI) using drug-eluting stents, in patients with multivessel or left main stem disease, and in patients with diabetes.^[1–4] Specifically, the Nordic–Baltic–British Left Main Revascularization study (NOBLE) randomized 1201 patients with left main stem disease, reporting higher rates of death, myocardial infarction, and repeat revascularization after PCI at 5 years [28% vs. 18%, hazard ratio (HR) 1.5; 95% confidence interval (CI) 1.1–2.0] compared with coronary artery bypass graft (CABG).^[1] The Evaluation of XIENCE versus Coronary Artery Bypass Surgery for Effectiveness of Left Main Revascularization (EXCEL) trial randomized 1905 patients with left main stem disease and showed significantly worse survival at 5 years with PCI (13% vs. 9.9%, odds ratio 1.4, 95% CI 1.03–1.85) compared with CABG.^[2] The Future Revascularization Evaluation in Patients with Diabetes Mellitus: Optimal Management of Multivessel Disease (FREEDOM) trial randomized 1900 diabetic patients with multivessel disease in whom 5-year mortality was significantly higher after PCI (16% vs. 11%, P = 0.049) compared with CABG.^[3] The Synergy between Percutaneous Coronary Intervention (PCI) with Taxus and Cardiac Surgery (SYNTAX) trial reported, among 1800 patients with three-vessel disease randomized to coronary artery bypass or PCI, that mortality at 5 years was higher with PCI (11% vs. 19%, P = 0.005) compared with CABG in patients with complex coronary disease defined by a SYNTAX score ≥33.^[4]

In this issue of the European Heart Journal, Davierwala and colleagues compare 10-year survival after PCI with either multiarterial or single arterial multivessel CABG.^[5] The benefits of left thoracic internal artery compared with vein grafts and stents are firmly established, and attributable to superior patency exceeding 95% at 5 years.^[6,7] It has been hypothesized that using two or more arterial conduits provides better clinical outcomes than a single artery with vein grafts, and by extension better outcomes than multivessel PCI (Graphical Abstract).^[8] To test this hypothesis, Davierwala and colleagues perform a post-hoc analysis of the SYNTAX study, one of >40 already published, with follow-up extended by >10 years beyond the primary endpoint of the trial.^[5] As in other randomized trials of PCI and CABG, the choice of conduit in SYNTAX was left to individual surgeons: 310 (37%) coronary bypass operations involved a second arterial conduit in addition to the left internal thoracic artery bypass.^[4] Davierwala and colleagues compare long-term survival of these 310 patients with that of the 502 patients who received a single arterial conduit and 901 who underwent PCI in the trial, and perform further subgroup analyses in patients with left main disease and diabetes. The authors report that multiple arterial grafting was associated with significantly lower all-cause mortality at 12 years compared with PCI (adjusted HR 0.66, 95% CI 0.49–0.89), but single arterial grafting was not (HR 0.83, 95% CI 0.67–1.03). In subgroup analyses of patients with left main stem disease and in diabetic patients, the authors observed no significant differences in long-term survival between the three revascularization strategies. The authors' finding that long-term survival after multiarterial revascularization is superior to that after single revascularization is consistent with >20 years of observational data spanning early single institutional reports to more recent population-level registry studies, the as-treated analysis of 3102 patients in the Arterial Revascularization Trial (ART), and a patient-level meta-analysis of six randomized trials.^[8–12] However, several notes of caution must temper any conclusions based on this and the authors' other findings. These cautions arise from the major methodological limitations of their study, many of which are shared by another extended post-hoc analysis of SYNTAX, and were described by Boden and colleagues in a recent accompanying Editorial in this journal.^[13,14]

First, there is the substantial problem of selection bias and presence of unmeasured confounders. Although this is acknowledged by the authors, the fact that this alone may explain their study findings is problematic, particularly as no attempt is made to quantify the impact of unmeasured confounders on the study results. The primary value of well-designed randomized trials such as the five landmark studies described above is the ability to eliminate selection bias. In that context, contradictory findings from observational data are unlikely to represent an advance in knowledge. A good example of this is ART, in which 3102 patients were randomized to single vs. bilateral internal thoracic artery multivessel coronary bypass surgery, but 14% of patients assigned to bilateral internal thoracic artery only received one, and 22% of patients assigned to single internal thoracic artery received a second arterial conduit.^[8] The characteristics of patients in the as-treated study arms illustrates how, even within the confines of a randomized trial, surgeons preferentially and systematically perform multiarterial revascularization in the patients they adjudicate to have the longest life expectancy based on characteristics that are easily measured such as younger age and less comorbidity, as well as factors less easily quantified such as frailty, non-compliance, and lack of access to healthcare. Consequently, the as-treated analysis of ART showed a survival benefit with multiarterial revascularization, but the intention-to-treat analysis did not. It is therefore reasonable to believe that the SYNTAX surgeons (who were allowed to choose their surgery revascularization strategy) systematically selected patients with the best life expectancy for multiarterial revascularization, leaving a sicker, frailer group for single revascularization. When Davierwala and colleagues compare the latter patients with the overall PCI cohort, they are not comparing like with like despite propensity matching, because important unmeasured confounders such as frailty and non-compliance are not included in their propensity model.

Second, the follow-up in this extended post-hoc analysis is limited to mortality. Consequently, there is no information about medication compliance, repeat revascularization, crossover to surgery, or any other adverse cardiovascular events after the original 5-year study endpoint. This is relevant because 54 patients randomized to PCI in SYNTAX underwent repeat revascularization with CABG within the 5-year follow-up time frame.^[15] However, the authors of this extended post-hoc analysis have no data on such crossover beyond the 5-year follow-up point, so are only able to speculate about causes for late convergence in survival curves they observe. Additionally, their supplementary tables 9–11 indicate that patients who underwent CABG were significantly less likely to be discharged and remain on optimal medical therapy, including aspirin and statins, than patients who underwent PCI, but there are no similar data available beyond 5 years. This early treatment bias favours PCI, and raises questions about how longer term treatment bias and any similar differences in the management of diabetes between the study arms may have negated a late survival benefit with CABG in diabetic patients.

Third, the subgroup analyses comparing outcomes of these three revascularization strategies in different patient populations are underpowered for survival. Only 452 of 1800 patients in SYNTAX had diabetes treated with insulin or oral hypoglycaemics, so the trial was 'woefully underpowered to ascertain any differential treatment benefit of revascularization, particularly for mortality'.^[9] Consequently, the absence of a significant difference in survival between PCI and CABG in patients with left main disease, or diabetes, provides no additional insights into the relative merits of each revascularization strategy in these important patient populations beyond the compelling data supporting surgical revascularization provided by the NOBLE, EXCEL, and FREEDOM trials, which were specifically designed and powered to evaluate PCI and CABG in these patient subgroups.^[1–4]

In summary, major methodological limitations largely explain the discordance between the authors' findings and those of four landmark randomized trials that consistently show significant and substantial event-free survival benefits with CABG compared with PCI in patients with multivessel disease, left main stem disease, and diabetes. A large, randomized trial incorporating newer generation drug-eluting stents and antiplatelet therapy, advances in intravascular imaging, and more complete revascularization could answer the very valid question concerning the long-term superiority of multiarterial revascularization over PCI that this post-hoc analysis of SYNTAX cannot adequately address. However, the SYNTAX, FREEDOM, NOBLE, and EXCEL landmark randomized trials already provide compelling evidence supporting coronary artery bypass over PCI for optimal survival in most patients with multivessel coronary disease, and particularly patients with complex multivessel disease, left main stem disease, or diabetes.