The H1N1 epidemic demonstrated that extracorporeal membrane oxygenation (ECMO) use in the most critically ill patients with acute respiratory distress syndrome (ARDS) resulted in a substantial decrease in mortality (Noah MA, et al. JAMA. 2011;306(15):1659-1668). The CESAR trial was a landmark randomized controlled trial (RCT) that showed improved survival and increased quality of life with the use of ECMO (Peek GJ, et al. Lancet. 2009;374:1351-1363). A major critique of the trial was the failure to define the proper use of ECMO in ARDS, with many critics believing the demonstrated survival benefit was related to referral to a specialty care center as opposed to the use of ECMO.
The EOLIA trial (Combes A, et al. N Engl J Med. 2018;378:1965-1975) was a multicenter, international RCT that randomized patients with severe ARDS as defined by one of three criteria: P/F ratio < 50 mm Hg for > 3 hours, P/F ratio < 80 mm Hg for > 6 hours, or pH < 7.25 coupled with Paco2 ≥ 60 mm Hg for > 6 hours (with respiratory rate < 35 beats/min and plateau pressure ≤ 32 cm H2O). Physicians were encouraged, but not required, to use paralysis and prone positioning before randomization. Exclusion criteria were vast and, most importantly, limited the study to salvageable, non-obese adult patients on mechanical ventilation for < 7 days with no contraindications for heparin use or vascular access. Randomization was stratified by the center, with the primary end point being 60-day mortality. A key secondary end point was treatment failure, which included crossover from conventional treatment to ECMO and overall death rates.
In order to properly power the study to detect a 20% mortality difference, a maximum of 331 patients were expected under an intention-to-treat protocol. However, after recruitment of 240 patients, an interim analysis failed to demonstrate a significant difference in 60-day mortality and led to early trial termination. A total of 1,015 patients were screened, with 124 assigned to receive ECMO and 125 assigned to the control group. Patient characteristics were similar between the two groups, with a majority of patients having bacterial (45%) or viral (18%) pneumonia. The key development for this study was that patients in both groups had access to similar rescue techniques and received low-pressure, low-volume ventilation according to current standard of care. Of significance, 35 patients in the control group (28%) crossed over to the ECMO group due to refractory hypoxemia at 6.5 ± 9.7 days.
At 60 days, 44 patients in the ECMO group (35%) and 57 in the control group (46%) had died (RR, 0.76; CI, 0.55–1.04; p = 0.09) with a hazard ratio in the ECMO group of 0.70 at 60 days (CI, 0.47–1.04; p = 0.07). The study also showed that the relative risk of treatment failure (death at day 60 on ECMO, crossover to ECMO, or death in the control group) was 0.62 (CI, 0.47–0.82; p = 0.001). At 60 days, patients on ECMO required less prone positioning, experienced less ischemic stroke, and had less need for renal replacement therapy. However patents receiving ECMO had a higher rate of bleeding requiring red blood cell transfusion (46% vs. 28%) and severe thrombocytopenia (27% vs. 16%).
When compared to previous RCTs, the EOLIA trial corrected many previous deficits: The authors standardized the mechanical ventilation settings for all subjects, all patients were offered similar rescue modalities (neuromuscular blockade, inhaled vasodilators, and prone positioning), subjects randomized to ECMO were nearly universally cannulated (98%, with 1 patient improving before ECMO use and 2 dying immediately after randomization), and finally, all patients randomized to ECMO were started on ECMO before transport to a specialty center. Additionally, the authors were methodical about creating succinct definitions of respiratory failure that met criteria for ECMO use.
Interpretation of results is difficult because many patients in the control group were transitioned to ECMO (28%) due to refractory hypoxemia but were analyzed as control group subjects due to the intention-to-treat principle. If solely examining treatment failure (as defined previously), there was a statistically significant benefit in favor of ECMO use. The authors admit that this exhibits bias toward the control group but note that the decision to transition to ECMO therapy was made only after every other rescue modality had been tried. Furthermore, the study did demonstrate a trend toward mortality benefit with the use of ECMO, although this was not statistically significant.
The most important criticism to surface on final assessment is the underpowering of the study to detect an absolute between-group mortality difference of 20%. This severely limited the primary analysis and led to early termination.
Coauthors of this installment of Concise Critical Appraisal:
Kevin M. Jones, MD, MPH, FACEP, is an assistant professor of emergency medicine with the Program for Trauma at the R. Adams Cowley Shock Trauma Center at the University of Maryland Medical Center, Baltimore, Maryland, USA.
James Lantry III, MD, is an assistant professor of emergency and critical care medicine and the associate program director of the Critical Care Fellowship at the University of Maryland Medical Center in Baltimore, Maryland, USA. Dr. Lantry is an editor of Concise Critical Appraisal.