The Design of an Adaptive Clinical Trial to Evaluate the Efficacy of Extra-Corporeal Membrane Oxygenation for Out-of-Hospital Cardiac Arrest

Abstract

BACKGROUNDExtracorporeal cardiopulmonary resuscitation (ECPR) is a promising therapy for out-of-hospital cardiac arrest (OHCA) that is refractory to standard therapy, but no multicenter randomized clinical trials have been conducted to establish its efficacy. We report the design and operating characteristics of a proposed randomized Bayesian adaptive “enrichment” clinical trial designed to determine whether ECPR is effective for refractory OHCA and, if effective, to define the interval after arrest during which patients derive benefit. METHODSThrough iterative trial simulation and trial design modification, we developed a Bayesian adaptive trial of ECPR for adults who experience non-traumatic out-of-hospital cardiac arrest. Our proposed trial design addresses the threats to trial success identified during the design process, which were (1) the uncertainty surrounding the cardiac arrest (CA)-to-ECPR interval within which clinical benefit might be preserved (2) the difference in prognosis between patients with an initial rhythm that is non-shockable vs. shockable. Trial subjects will be randomized 1:1 to receive either standard care or expedited transport to a hospital for potential ECPR. The CA-to-ECPR interval will be estimated in real time based on the sum of the estimated paramedic response time (911 call to scene arrival), paramedic scene time, and transport time to hospital. A Bayesian decreasing step function will be used to estimate the efficacy of the treatment with an outcome of the 90-day utility-weighted Modified Rankin Scale (uwmRS) for each rhythm subgroup and estimated CA-to-ECPR interval at pre-specified interims. The trial will adaptively lengthen the estimated CA-to-ECPR eligibility window if the treatment appears effective at the upper limit of initial eligibility window. If ECPR appears ineffective at longer estimated CA-to-ECPR intervals, the upper limit of the window for enrollment eligibility will be shortened. The analysis will be stratified by rhythm subgroup. RESULTSWith a maximum total sample size of 400, and a cap on the maximum sample size of 300 for the non-shockable rhythm subgroup, the trial design has power ranging from 91–100% to detect a benefit from ECPR for non-shockable rhythms under the various efficacy scenarios simulated and power ranging from 69–98% for shockable rhythms under the same scenarios. The trial design also has a high probability of correctly identifying the maximum CA-to-ECPR interval within which ECPR produces a clinically significant benefit of 0.2 on the uwMRS. If ECPR is equivalent to standard CA care, the type I error is 2.5% with a 99% probability of stopping enrollment early for futility in the non-shockable subgroup and a 97% probability of stopping enrollment early for futility in the shockable subgroup. CONCLUSIONThis proposed adaptive trial design helps to ensure the population of patients who are most likely to benefit from treatment—as defined both by rhythm subgroup and estimated CA-to-ECPR interval—is enrolled. The design promotes early termination of the trial if continuation is likely to be futile.