A Bayesian Perspective - Extracorporeal CPR for Refractory Out-of-Hospital Cardiac Arrest

BackgroundWhether extracorporeal CPR (eCPR) has survival benefits over conventional CPR (cCPR) in patients with refractory out-of-hospital cardiac arrest is an unresolved clinical question. Performing trials in this environment is exceedingly challenging and inferences need careful examination. ObjectiveDetermine if a Bayesian perspective provides additional inferential insights. MethodsThe INCEPTION trial of patients with refractory out-of-hospital cardiac arrest reported eCPR and cCPR had similar effects on the primary outcome, 30 day survival with a favorable neurologic outcome. Herein the probability of eCPR superiority, equivalence or inferiority to cCPR is re-evaluated with a Bayesian analysis using both vague and informative priors (from previously completed randomized clinical trials (RCTs)). ResultsDepending on the chosen prior, the Bayesian reanalysis of the INCEPTION intention-to-treat (ITT) data suggests an equivalence probability < 10% (defined as an absolute risk difference (RD) < 1%) but a clinical superiority probability of 66 - 99 % (defined as RD > 1.0). An INCEPTION per protocol (PP) analysis with a vague prior suggested a 1% probability of clinical benefit but this posterior probability increased to 86% when informative PP data from previous RCTs were considered. ConclusionBayesian INCEPTION trial re-analyses provide additional quantative insights. The totality of the ITT evidence reveals a high probability for a clinically meaningful eCPR benefit over cCPR at 30 days. A PP analysis shows a less definitive probability of benefit. (Abstract word count 197, Manuscript word count 1477)