Unreported exclusion and sampling bias in interpretation of randomized controlled trials in patients with STEMI

To assess the impact of sampling bias due to reported as well as unreported exclusion of the target population in a multi-center randomized controlled trial (RCT) of ST-elevation myocardial infarction (STEMI).

We compared clinical characteristics and mortality between participants in the DANAMI-3 trial to contemporary non-participants with STEMI using unselected registries. A total of 179 DANAMI-3 participants (8%) and 617 contemporary non-participants (22%) had died (Log-Rank: P < 0.001) after a median follow-up of 1333 days (range: 1-2021 days). In an unadjusted Cox regression model all groups of non-participants had a higher hazard ratio to predict mortality compared to participants: eligible excluded (n = 144) (hazard ratio: 3.41 (95% CI: (2.69-4.32)), ineligible excluded (n = 472) (hazard ratio: 3.42 (95% CI: (2.44-4.80), eligible non-screened (n = 154) (hazard ratio: 3.37 (95% CI: (2.36-4.82)), ineligible non-screened (n = 154) (hazard ratio: 6.48 (95% CI: (4.77-8.80).

Sampling bias had occurred due to both reported and unreported exclusion of eligible patients and the difference in mortality between participants and non-participants could not be explained only by the trial exclusion criteria. Thus, screening logs may not be suited to address the risks of sampling bias.