Registry-based randomized clinical trials are increasingly used because they can improve feasibility and simplify data capture. However, reliance on administrative or national registries introduces the possibility of bias through misclassification or missing information. The BETAMI‑DANBLOCK trial was designed to examine how registry-derived cardiovascular events compare with events subjected to blinded adjudication in a large randomized MI population.
The BETAMI‑DANBLOCK study randomized 5,574 patients who had experienced a myocardial infarction and did not have heart failure to receive either beta-blocker therapy or no beta-blocker. The trial evaluated a composite primary endpoint consisting of all-cause mortality, recurrent myocardial infarction, ischemic stroke, heart failure, unplanned coronary revascularization, and malignant ventricular arrhythmias.
Primary endpoint events were ascertained from three complementary sources: the Danish and Norwegian national patient registries, participant self-reported questionnaires, and review of medical records. All events identified via the national registries were submitted to a blinded adjudication process except for all-cause mortality, which typically can be captured reliably in national death registries and therefore was not adjudicated in this workflow.
The authors compared non‑adjudicated registry events with adjudicated events using incidence rates (events per 100 person‑years) and hazard ratios (HRs) for the effect of beta‑blocker therapy versus control. These metrics allowed assessment of both event capture completeness and the potential influence of misclassification on estimated treatment effects.
National registries captured 99.7% of primary endpoint events that were also registered by self-report and medical records, demonstrating very high completeness of event capture in the registries used.
However, validation by blinded adjudication showed that only 75% of registry-identified primary endpoint events were confirmed. This reduced the observed incidence from 6.23 to 4.43 events per 100 person‑years after adjudication, indicating that a substantial proportion of registry-coded events did not meet adjudication criteria.
Confirmation rates varied by event type. The highest confirmation was observed for ischemic stroke (92%), while much lower confirmation rates were reported for unplanned coronary revascularizations and heart failure (both around 45%). Additionally, the confirmation rate tended to be lower during the first 6 months after the index myocardial infarction, suggesting time‑dependent differences in coding accuracy or clinical classification in the early post‑MI period.
Despite the reduction in event counts after adjudication, estimated treatment effects for the composite primary endpoint were consistent across non‑adjudicated registry data and adjudicated events. The HR for beta‑blocker therapy versus control using registry-derived, non‑adjudicated events was 0.88 (95% confidence interval [CI] 0.78–0.98). Using adjudicated events, the HR was 0.85 (95% CI 0.75–0.98). These overlapping estimates indicate that, for the primary composite endpoint, registry-derived effect estimates were comparable to those obtained after adjudication.
The authors noted, however, that differences between non‑adjudicated and adjudicated results were apparent for specific component endpoints, particularly heart failure and revascularization, where misclassification appeared more pronounced.
BETAMI‑DANBLOCK demonstrates that national registries can provide highly complete event capture and yield treatment effect estimates for a composite cardiovascular endpoint that are broadly similar to adjudicated outcomes. At the same time, blinded adjudication removed a substantial number of events and revealed event‑specific misclassification, especially early after the index MI and for heart failure and unplanned coronary revascularization.
These findings imply a trade-off: registry-only ascertainment may be sufficient to estimate treatment effects for some composite outcomes, while adjudication can improve specificity, reduce event counts, and identify misclassification in certain event types or time windows. The study supports the pragmatic value of registries but also highlights situations where adjudication remains important to refine endpoint accuracy.
The trial registration numbers reported were NCT03646357 and NCT03778554. Keywords reported by the authors include beta-blocker, myocardial infarction, randomized controlled trial, registries, and validation.
All statements and results in this summary are drawn from the source abstract of the BETAMI‑DANBLOCK publication. Detailed methods of adjudication, specific adjudication criteria, per‑event counts, subgroup analyses, and other granular data were not reported in the abstract and therefore are not included here.