Comparing effect estimates in randomized trials and observational studies from the same population: An application to percutaneous coronary intervention

Background The ability for real world data to deliver similar results as a trial that asks the same question about the risks or benefits of a clinical intervention can be restricted not only by lack of randomization, but also limited information on eligibility criteria and outcomes. To understand when results from observational studies and randomized trials are comparable, we carried out an observational emulation of a target trial designed to ask similar questions as the VALIDATE randomized trial. VALIDATE compared the effect of bivalirudin and heparin during percutaneous coronary intervention on the risk of death, myocardial infarction, and bleeding across Sweden. Methods We specified the protocol of a target trial similar to the VALIDATE trial protocol, then emulated the target trial in the period before the trial took place using data from the SWEDEHEART registry; the same registry in which the trial was undertaken. Results The target trial emulation and the VALIDATE trial both estimated no difference in the effect of bivalirudin and heparin on the risk of death or myocardial infarction by 180 days: emulation risk ratio for death 1.21 (0.88, 1.54); VALIDATE hazard ratio for death 1.05 (0.78, 1.41). The observational data, however, could not capture less severe cases of bleeding, resulting in an inability to define a bleeding outcome like the trial, and could not account for intractable confounding early in follow-up (risk ratio for death by 14 days 1.85 (0.95, 3.63)). Conclusion Using real world data to emulate a target trial can deliver accurate long-term effect estimates. Yet, even with rich observational data, it is not always possible to estimate the short-term effect of interventions, or the effect on outcomes for which data are not routinely collected. If registries included information on reasons for treatment decisions, researchers may be better positioned to identify important confounders.