Document resource
Introduction The comparative long-term survival of venous graft coronary artery bypass grafting (CABG) versus percutaneous coronary intervention with drug-eluting stents (DES-PCI) in multivessel coronary artery disease remains controversial, with conflicting evidence from observational studies complicated by severe confounding by indication. We aimed to compare 20-year all-cause mortality between DES-PCI and venous graft CABG in patients with multivessel coronary artery disease.Methods We conducted a retrospective observational cohort study of consecutive adults undergoing first coronary revascularization for multivessel disease (≥2 vessels with significant stenosis) at a single tertiary cardiac center between May 2003 and December 2018. Patients received either DES-PCI exclusively or isolated CABG with venous graft configuration. To address severe confounding by indication, we employed inverse probability of treatment weighting (IPTW) using stabilized weights. Weights were trimmed at the 99th percentile to minimize influence of extreme observations. Covariate balance was assessed using standardized mean differences (SMD <0.10 threshold for adequate balance). Survival was compared using weighted Kaplan-Meier methods and weighted Cox proportional hazards regression with robust variance estimation. Proportional hazards assumptions were verified using scaled Schoenfeld residuals. Sensitivity analyses included multivariable Cox regression in the full cohort, doubly robust estimation combining IPTW with covariate adjustment, and prespecified subgroup analyses stratified by procedural urgency. Secondary analyses identified independent predictors of mortality within each treatment group using multivariable Cox regression.Results Among 4,901 patients undergoing first revascularization (venous graft CABG n=207, 4.2%; DES-PCI n=4,694, 95.8%), baseline characteristics revealed substantially greater disease severity in CABG patients: 73.4% had three-vessel disease versus 13.0% in DES-PCI (standardized mean difference 1.54), with additional imbalances in hypercholesterolemia (73.4% vs 55.3%, SMD 0.39), ex-smoking status (55.1% vs 30.7%, SMD 0.52), respiratory disease (16.4% vs 3.6%, SMD 0.44), and left main stenosis (28.0% vs 13.5%, SMD 0.36) ( table 1). After inverse probability weighting, balance was successfully achieved for 18 of 21 baseline covariates including the critical covariate of coronary disease extent (post-weighting SMD 0.02); residual modest imbalance persisted only in age (SMD -0.18), left main stenosis (SMD -0.17), and never-smoker category (SMD 0.10). The effective sample size after weighting was 52 for CABG and 4,598 for DES-PCI, reflecting the extreme baseline differences requiring substantial reweighting to achieve covariate balance. Median follow-up in the full cohort was 10.6 years (interquartile range 6.1–15.2) for CABG and 6.4 years (3.0–11.3) for DES-PCI; deaths occurred in 43/207 CABG patients (20.8%) versus 840/4,694 DES-PCI patients (17.9%) (table 2). Kaplan-Meier analysis in the IPTW-weighted cohort demonstrated comparable survival throughout follow-up (figure 1). Weighted Cox proportional hazards regression revealed no statistically significant difference in mortality risk between treatment groups (hazard ratio for CABG vs DES-PCI 1.22, 95% CI 0.76–1.97, P=0.41). This finding remained directionally consistent across sensitivity analyses: multivariable-adjusted model in the full cohort demonstrated HR 1.65 (95% CI 1.19–2.30, P=0.003), likely reflecting residual confounding by indication, while doubly robust estimation yielded HR 1.50 (95% CI 0.96–2.36, P=0.078). Proportional hazards assumptions were satisfied across all models (Schoenfeld residual tests: unadjusted P=0.77, IPTW P=0.99, multivariable P=0.91). Separate multivariable Cox regression models within each treatment arm revealed age as the dominant mortality predictor in DES-PCI (HR 1.02 per year, 95% CI 1.01–1.04, P=0.005) alongside three-vessel disease (HR 2.26, 95% CI 1.51–3.38, P<0.001), cerebrovascular disease (HR 3.30, 95% CI 1.02–10.7, P=0.047), and elective procedure status (HR 1.50, 95% CI 1.05–2.15, P=0.026), whereas CABG mortality was predominantly associated with severe left ventricular dysfunction (LVEF <30% HR 11.56, 95% CI 1.69–78.87, P=0.012) and cerebrovascular disease (HR 16.71, 95% CI 2.23–120.4, P=0.006), though these point estimates reflect the small CABG sample size with correspondingly wide confidence intervals.Conclusion In this real-world cohort with up to 20-year follow-up, we found no statistically significant difference in long-term all-cause mortality between revascularization strategies. These findings support clinical equipoise regarding optimal revascularization approach for multivessel disease and emphasize the necessity of individualized shared decision-making incorporating anatomical complexity, left ventricular function, comorbid conditions, procedural risks, and patient preferences.Abstract 538 Table 1Baseline characteristicsAbstract 538 Table 2Survival probability for unadjusted and IPTW adjusted cohortsAbstract 538 Figure 1IPTW-adjusted Kaplan Meier survival curve