ECMO in severe acute respiratory distress syndrome

As summarised last year in a Comment in The Lancet Respiratory Medicine,¹ four randomised trials2, 3, 4, 5 of extracorporeal membrane oxygenation (ECMO) have been done in patients with severe acute respiratory distress syndrome (ARDS) in the past 40 years. The 1979 and 1994 trials2, 3 did not include a lung-protective ventilation strategy for patients in their control groups because they were done before publication of the US National Heart, Lung, and Blood Institute's ARDS Network-sponsored trial,⁶ which was published in 2000 and showed a major reduction in mortality with a low tidal volume and plateau pressure limited ventilation strategy. Since then, CESAR, the venovenous ECMO trial⁴ that was published in 2009, showed that patients who were transferred to a centre that could institute ECMO had significantly lower mortality and severe disability at 6 months than those who were not transferred to an ECMO centre. However, the result was not definitive because only some of the patients referred to the ECMO centre received ECMO, and there was no evidence that patients in the control group received lung-protective ventilation. In the most recent trial, EOLIA,⁵ which was published in 2018, 60-day mortality was numerically lower in the venovenous ECMO group (35%) than in the control group (46%), in which patients received lung-protective ventilation, but this difference was not significant (p=0·09). For several secondary endpoints, however, ECMO was significantly better than the control group, including more days alive and free of the need for vasopressor therapy, renal replacement therapy, or prone positioning to treat their respiratory failure.⁵ The decision to stop the EOLIA trial for futility after enrolment of 249 patients, before the planned enrolment of 331 patients, was based on methods that can decrease power for rejection of the null hypothesis, and might have led to a type 2 error.⁷ The trial was powered for a 20% absolute reduction in mortality, which was probably an unrealistic goal.

In this context of uncertainty about interpretation of the results of the EOLIA trial⁵ and the increasing use of ECMO, Laveena Munshi and colleagues report the results of a systematic review and meta-analysis⁸ of the use of ECMO in adult patients with severe ARDS in this issue of The Lancet Respiratory Medicine. The authors included three observational studies and two randomised controlled trials, CESAR and EOLIA. For the three observational studies, matching methods were used to compare ECMO-treated patients with those who did not receive ECMO. Two of the observational trials were done primarily in patients with H1N1 influenza.

The primary analysis, which was restricted to the two randomised controlled trials (429 patients overall), showed that, compared with conventional mechanical ventilation, ECMO was associated with a reduced risk 60-day mortality (risk ratio [RR] 0·73 [95% CI 0·58–0·92]). ECMO was also associated with a reduced risk of treatment failure (0·58 [0·39–0·84]), which was defined as death in the ECMO group and death or crossover to ECMO in the control group. In a meta-analysis of all five studies, which included 773 patients overall, ECMO was associated with a reduced risk of 30-day mortality (RR 0·69 [95% CI 0·50–0·95). However, this meta-analysis was prone to bias because of the inclusion of the observational studies, which did not include randomisation for ECMO treatment. Evidence also suggested that ECMO is associated with a risk of major haemorrhage. Munshi and colleagues concluded that the evidence favouring ECMO is of moderate-to-high quality according to the Grading of Recommendations Assessment, Development and Evaluation criteria. The availability of only two randomised controlled trials for this meta-analysis is an important limitation. It is unusual for a meta-analysis to be based on only two randomised controlled trials, but the authors recognise this limitation and tried to adjust for heterogeneity between the trials. Despite this limitation, this meta-analysis is an important contribution because clinicians need guidance on how to interpret the evidence from CESAR⁴ and EOLIA.⁵ Although another trial of ECMO versus non-ECMO treatment for severe acute respiratory syndrome has been called for,¹ such a trial is not likely to be organised or funded, and if a trial were done it would not be completed for many years.

In addition to Munshi and colleagues' meta-analysis,⁸ Goligher and colleagues have done a Bayesian post-hoc analysis⁹ of the EOLIA trial. The Bayesian approach defines the probability of a treatment effect rather than ruling out the absence of a treatment effect as in conventional trial design. Goligher and colleagues used several estimates of posterior probabilities of enthusiasm or scepticism for efficacy of ECMO in patients with severe ARDS. The Bayesian analysis showed that, across a range of assumptions about the probability of benefit from early ECMO, the posterior probability of any mortality benefit with early ECMO in EOLIA was high, ranging from 88% to 99%. A 2018 clinical case analysis¹⁰ also provided a timely example of a patient with severe ARDS. The authors of that analysis addressed the question of whether or not to use ECMO in this clinical setting and provided both pro and con viewpoints.

In view of the results of the EOLIA trial,⁵ Munshi and colleagues' meta-analysis,⁸ and the Bayesian analysis of EOLIA,⁹ what should clinicians conclude about the use of ECMO in patients with severe ARDS? It is important for clinicians to be certain they have instituted other therapies before considering ECMO, including optimal lung-protective ventilation, diuresis, and neuromuscular blockade with deep sedation, prone positioning, and possibly inhaled nitric oxide, recruitment manoeuvres, and renal replacement therapy.¹¹ If these therapies do not stabilise the patient, and there are no other exclusion criteria (as in the EOLIA trial), I believe that the balance of evidence favours use of ECMO in severe ARDS if available from a medical centre experienced in provision of ECMO.

© 2019 Dr Barry Slaven/SPL