Validating national registry data for randomised trials: comparison of adjudicated and registry-derived cardiovascular outcomes in the BETAMI-DANBLOCK trial

Background Registry-based randomised clinical trials are increasingly dependent on registry-derived outcomes, providing advantages in feasibility and data capture. However, registry data may introduce bias through misclassification or incomplete information. In the BETAMI-DANBLOCK trial, we compared registry-based cardiovascular events with and without adjudication, using information from national registries, electronic health records, and patient reports. Purpose To investigate whether registry-based endpoints can substitute adjudicated endpoints without materially altering trial results. Methods BETAMI-DANBLOCK enrolled 5,574 patients with myocardial infarction (MI) and a left ventricular ejection fraction ≥40% from 2018 through 2024. Patients were randomised to beta-blocker therapy or no beta-blocker therapy. The primary endpoint events (death, MI, ischemic stroke, heart failure, unplanned coronary revascularisation, and ventricular arrhythmias) were identified from the Danish and Norwegian national patient registries, self-reported questionnaires, and medical records. All registry-identified events, except death, underwent blinded adjudication. We compared events with and without adjudication by calculating incidence rates and hazard ratios (HRs) for beta-blocker treatment versus controls. Results National registries had almost complete capture of events when compared to site-reporting. In total, 75% of the primary endpoint events identified through the registries were confirmed by adjudication. The confirmation rate was lower during the first six months and varied by event type (from 92% for ischemic stroke to 45% for unplanned coronary revascularisations). The incidence rate of the primary endpoint was 6.23 per 100 person-years using registry data without adjudication and 4.43 with adjudication. Estimated treatment effects were consistent for the primary endpoint (0.88, 95% confidence interval (CI): 0.78-0.98 without adjudication and 0.85, 95% CI: 0.75-0.98 with adjudication), with no apparent differences across event types. Conclusion Registry data provide treatment effect estimates comparable to those obtained from adjudicated data in BETAMI-DANBLOCK. However, adjudication substantially reduced the total number of endpoint events and event-specific misclassification was a limitation, particularly during the first months after the index event.