Introduction: Although randomized controlled trials (RCT) limit bias and confounding, their restrictive selection criteria have led to concerns about their representativeness of real world practice and their applicability outside of RCT populations.
Hypothesis: A minority of patients in an observational dataset will be eligible for RCTs; their characteristics will differ from those ineligible; the 5-year survival of RCT-eligible will differ from ineligible patients and from published RCT results.
Methods: Based on selection criteria for ARTS, Courage, ERACI II, MASS II, RITA II and SoS, we identified Duke Cardiovascular Databank patients meeting each RCT’s inclusion and exclusion criteria from 7/92-12/12 and compared RCT-eligible vs. ineligible patients using Chi-square and t-tests, and Kaplan-Meier 5-year survival by intervention.
Results: Out of 35,010 Duke patients, 15% met ARTS eligibility criteria, 21% Courage, 8.8% ERACI II, 2.9% MASS II, 36% RITA II and 22% SoS. For RCTs comparing 2 interventions, 17%-24% of each trial’s Duke-eligible patients did not receive either of those interventions. RCT-eligible Duke patients differed significantly from ineligible patients (e.g., age, gender, EF and # diseased vessels). The 5-year survival of RCT-eligible minus ineligible patients ranged from 0.6-10% for MED, -2.5-2.6% for PCI, and 4.3-8.3% for CABG. The trial reported 5-year survival minus that of RCT-eligible Duke patients ranged from 11-33% for MED, 11-12% for PCI (except -0.7% for MASS II), and 5.5-14% for CABG (except -0.2% for MASS II). The 5-year survival of PCI minus MED ranged from 0.1-0.7% in trials versus 12-21% in RCT-eligible patients, and the 5-year survival of CABG minus PCI ranged from -4.4-3.4% in trials versus -1.9-5.8% in RCT-eligible patients.
Conclusions: Our results show that RCT-eligible patients represent a minority of patients in a large academic real world dataset. Characteristics of RCT-eligible patients differed significantly from ineligible ones, and for most RCTs, RCT-eligible patients had higher 5-year survivals than ineligible patients. Lastly, nearly all 5-year survivals using RCT-eligible patients underestimate those reported by each RCT, and intervention survival differences vary from RCT estimates.