CABG versus PCI in high-risk coronary revascularisation patients shows mixed outcomes over 5 years

Aim This study aimed to investigate whether treatment effects (TE) for coronary artery bypass grafting (CABG) versus percutaneous coronary intervention (PCI) were consistent for people requiring coronary revascularisation with high-risk characteristics. Methods We used target trial emulation (TTE) as study design and observational data from Hospital Episode Statistics (HES). The target population was people requiring multivessel coronary revascularisation with at least one of seven high-risk characteristics: age >75 years, female, heart failure, chronic kidney disease, peripheral vascular disease, intermediate frailty risk, or presentation with acute coronary syndrome. The intervention was CABG and the comparison was PCI. Outcomes included all-cause and cardiovascular (CV) mortality, CV hospitalisation, and major adverse cardiovascular events within 5 years of the index procedure. This study included four research stages: (1) latent class analysis (LCA) to identify mutually exclusive patient clusters representing different clinical phenotypes, (2) instrumental variable analysis (IVA) to estimate the TE for the whole population and each patient class; (3) repeating IVA in an emulated trial population obtained by matching a previous cardiac surgery trial enriched for high-risk characteristics to the HES population; (4) co-designed a pragmatic randomised controlled trials (RCT) with multiple stakeholders to address uncertainty identified from the analyses above. Results Of 103,670 patients in the target population, 62,048 (59.9%) received CABG and 41,622 (40.1%) received PCI. Seven patient classes were identified as the best solution from LCA. The emulated trial consisted of 3124 patients, 1,588 (50.8%) in CABG arm and 1,536 (49.2%) in PCI arm. Patients aged >75 years without CKD (Class 1) who received CABG showed consistent benefits in all primary and secondary outcomes. For the other six patient classes, heterogeneity in benefits and harms between CABG and PCI existed both between and within classes in clinical outcomes. An RCT to resolve the remaining uncertainty using Bayesian approach would require 3000 participants to detect a hazard ratio of 0.7 with a family-wise type 1 error rate <5% and 90% power across all seven classes. Conclusion TTE of coronary revascularisation options in people with high-risk characteristics demonstrated mixed benefits and harms both between and within disease phenotypes. A trial to address uncertainty as to the indications for CABG versus PCI in this target population is feasible.