The 2009 randomized CESAR trial in Lancet concluded that in severe ARDS in the U.K., referral to an ECMO center saved lives. However, patients in the control (non-ECMO) group didn't consistently get low-tidal ventilation, and many patients randomized to ECMO never received it, creating skepticism of the findings. A case series from Australia/New Zealand (ANZ ECMO) in JAMA showed a 70% survival in people with ARDS due to H1N1 treated with ECMO, but baked-in selection bias and likely unmeasured confounders precluded any strong conclusion of ECMO's independent benefit.
In the U.K., regional referral biases exist in the rates of transfer of patients with severe ARDS to receive ECMO at one of the regional centers (they had 1, and 3 were added during the pandemic). Noah et al used this to their advantage by analyzing the outcomes of the 80 patients referred for ECMO for severe ARDS during the 2009-10 H1N1 pandemic (mean age 36), and comparing them to a propensity-matched cohort of 80 non-ECMO-referred patients (mean age 37-38).
The results were impressive: 23.7% of the ECMO-treated patients died in-hospital, compared to ~50% of the matched conventionally treated patients. This, despite the ECMO patients being very, very sick: 80% were on FiO2 of 1.0; 50% had failed low-tidal ventilation and were being vented with alternate strategies (which was likely why they were referred for ECMO). 86% of patients referred for ECMO actually got ECMO here (in CESAR it was 75%).
Although propensity-matching should always inspire a healthy skepticism, their two cohorts looked well-matched and their methods robust. They used three models that all agreed; their findings held up when various exclusion criteria were entered and the models re-run (most notably, deaths within 48 hours after ECMO transfer). They had a big group from which to draw their control cohort. Their tables are clear and easy to understand. (It's not easy to get into JAMA.)
At the same time, authors acknowledge the vulnerability of unmeasured confounders to influence the findings. Even with good propensity matching, those deemed "too sick for ECMO" and therefore not referred may have been more likely to die, and biased the control cohort toward a higher mortality. (What confounders might be hidden inside these existing regional referral patterns?) There was no way to tell if the control cohort patients were managed properly (remember that in CESAR, there was no standardization of treatment and many patients got insufficient rates of low tidal volume ventilation).
ECMO was no walk in the park, and serious complications were common. Of 80 patients: 8 patients had intracranial hemorrhage; 1 a fatal pulmonary hemorrhage; 4 had hemothorax, and 2 had retroperitoneal bleeds. (Of course, that makes their survival statistics that much more impressive.)
As usual with critical care research, not knowing the long-term outcomes is a serious flaw. Surviving to discharge isn't always a wonderful thing. But because the surviving patients were mostly in their 30s, they likely did well post-discharge.
Technology and manufacturing advances have made ECMO simpler and safer to implement. It no longer requires a dedicated team at the bedside; a nurse with special training can run and monitor the circuit, analagous to continuous renal replacement therapy. Of course, an experienced team of physicians must be continually on-call, and for this reason I would expect ECMO's expansion to be slow and (in the U.S.) limited to academic centers competing to be seen as on the leading-edge.
Noah MA et al. Referral to an Extracorporeal Membrane Oxygenation Center and Mortality Among Patients With Severe 2009 Influenza A(H1N1). JAMA 2011; ePub October 5, 2011.