Background: A high proportion of patients requiring cardiopulmonary resuscitation (CPR) before randomisation has been proposed as a key reason for the lack of a mortality benefit of veno-arterial extracorporeal membrane oxygenation (VA-ECMO) in patients with infarct-related cardiogenic shock in the ECLS-SHOCK trial.
Purpose: To determine whether experimentally varying the proportion of resuscitated patients meaningfully alters the estimated treatment effect of VA-ECMO on mortality.
Methods: Subgroup-specific odds ratios (ORs) for 30-day and 1-year mortality were obtained for patients with and without preceding CPR. Two complementary methods were used to assess the potential impact of CPR prevalence on overall treatment effect: (1) analytic mixture modelling combining subgroup effects across hypothetical CPR proportions (0-100%) and (2) bootstrap resampling generating synthetic trial cohorts (n=420 each) with varying CPR proportions, preserving randomisation. For each scenario, ORs and empirical 95% confidence intervals (CI) were estimated.
Results: Thirty-day mortality ORs were 1.0 (95% CI 0.65-1.55) in CPR patients and 0.81 (95% CI 0.36-1.82) in non-CPR patients in the total study cohort (Figure). Across both analytic and resampling methods, estimated treatment effects remained close to unity across CPR proportions. At minimal CPR prevalence, OR estimates shifted modestly toward potential benefit of VA-ECMO (approximate OR 0.7-0.8) but CIs consistently crossed 1.0. One-year outcomes showed similar neutrality, without a scenario yielding a statistically robust mortality reduction.
Conclusion: Simulations across the full spectrum of CPR prevalence demonstrated that changing the proportion of resuscitated patients does not materially influence mortality outcomes with VA-ECMO. The neutral effect observed in the randomised trial may not be solely attributed to case-mix differences in CPR. These findings suggest that other determinants beyond resuscitation status may also be considered to explain divergent outcomes across mechanical support studies.
