Study design for an emulated trial of a 2 arm, parallel, stratified, adaptive, RCT of CABG versus PCI in people requiring myocardial revascularization at high risk (High-Risk REVASC).

AIM: This study aims to use routinely collected health data and trial emulation methodology to inform the design of a pragmatic randomized controlled trial (RCT) in people requiring multivessel coronary revascularization with severe symptomatic multivessel disease and high-risk characteristics, typically underrepresented in previous RCTs.

METHODS: Hospital episode statistics (HES) linked to Office for National Statistics will be the main data source. The study population is patients who require multivessel myocardial revascularization with at least one of the following high-risk characteristics: age >75 years, female, diagnosed with acute coronary syndrome, heart failure, chronic kidney disease, peripheral vascular disease, or intermediate frailty risk. The intervention procedure is coronary artery bypass grafting (CABG) and the control (reference) is percutaneous coronary intervention (PCI). Outcomes include all-cause and cardiovascular (CV) death, CV hospitalization, major adverse cardiovascular events, and major vascular complications or bleeding within 5 years of the index procedure. This study includes 3 stages of statistical analyses: (1) latent class analysis (LCA) to identify mutually exclusive patient clusters (latent classes) representing different clinical phenotypes, (2) instrumental variable analysis (IVA) to estimate the average treatment effect (ATE) in the whole population and each patient cluster; and (3) repeating stage 2 in an emulated trial population obtained by matching the HES population with individual participant data from an RCT. We will then co-design the protocol for a definitive clinical trial in partnership with patients, public, and stakeholders.

DISCUSSION: This study introduces a novel, stepwise data science framework that integrates machine learning (unsupervised learning through LCA), causal inference, and trial emulation methods applied in big data, to design a future stratified and adaptive RCT of CABG versus PCI in high-risk patients. Our proposed approach fosters new collaborations among data scientists, trial methodologists, clinicians, and patient and public representatives in complex trial designs for diverse, high-risk populations. This study represents a new framework for co-production in trials of cardiovascular interventions, which offers a scalable model and has the potential to transfer to other disease areas.

CLINICAL TRIAL REGISTRATION: URL: https://www.

CLINICALTRIALS: gov/study/NCT05853536. Unique identifier: NCT05853536.