Emulated Trials, Latent Classes, and Causal Inference in High-Risk Coronary Revascularization

RCTs in revascularization have long sidestepped the most vulnerable—old age, frailty, and comorbidity—leaving a shaky evidence base for the sickest patients. This methods paper outlines a next-gen hybrid: blend national registry data, latent class clustering, and instrumental variable inference, then stress-test these ideas in a simulated trial to plan a future adaptive RCT. The payoff is a trial design reflecting real-case mixes, not idealized, cherry-picked populations. The risk is that stacking complex modern statistics atop wobbly routine data will cloud, not clarify, trial interpretation, especially for clinicians on the receiving end of a protocol concocted far from the bedside.