It is a general rule among intensivists, that our job satisfaction frequently varies inversely in proportion to our distance from the nearest ECMO circuit. It is therefore all the more remarkable that ECMO has seen so little attention from the CICM examiners, by whom all the major ECMO centres in Australia are extensively colonised. The only question ever to involve any discussion of the evidence was Question 11 from the second paper of 2010. Later, Question 23 from the first paper of 2014 asked about the indications for ECMO and the relative merits of VV and VA ECMO, but only a very limited discussion of the literature was required. Given the known 18-month lag between publication of something interesting and its apppearance in the exam, one might expect something about ECMO to appear in the Part II papers in the second part of 2019, approximately 18 months after the recently published EOLIA trial (Combes et al, 2018).
The LITFL ECMO literature summaries page is an excellent resource for the time-poor candidate who is in looking for a quick update. A quick update is all that would be required, given the limited number of high-quality papers and the neglect of this topic by Part II SAQ writers.
- VV ECMO in adults with ARDS: methodologically flawed trials (CESAR, 2009; EOLIA, 2018). CESAR found a mortality benefit where it shouldn't have, and EOLIA didn't find a mortality benefit where it should have.
- There is no evidence that the routine use of ECMO improves mortality in ARDS, as compared to maximal conventional care (including prone position and opn-lung, lung protective ventilation)
- Use of ECMO should be limited to patients with the highest severity of hypoxia
- Prone ventilation should be considered first, wherever it is not contraindicated
- VA ECMO in adults with cardiogenic shock: prospective observational studies only; ECMO appears to be superior to IABP, but only where IABP is clearly not working (i.e. pt. remains shocked in spite of it).
- VA ECMO in adults difficult to separate from bypass: mainly retrospective oberservational studies; ECMO appears to be associated with a very high mortality in the over-70 age group (75%), and usually requires approximately 7 days of ECMO. Only abut 50% of patients can be weaned.
- VA ECMO in neonates: surprisingly, mortality and neurological outcomes are usually quite good.
This has been going on since before 1986. Bartlett et al reported that after the experience, 63% of his patients were "normal or near normal". A more recent (2006) review of VA ECMO used as rescue therapy in 27 paediatric and neonatal patients also reported good mortality results, in spite of the fact that many of these little patients had profoundly (congenitally) abnormal myocardia.
The early data (eg. Suchyta et al, 1991) was not exactly favourable towards ECMO. The survival in ARDS during this time is obviously discoloured by the fact that these data come from a historical dark age, pre-ALVEOLI, pre-ARDSNET; these people were ventilated with profoudly stupid settings and therefore probably damaged irreversibly by their exposure to "conventional" ventilation. ECMO was instituted late in the game after the organs have already been trashed by biotrauma. Suchyta et al compared their late 1980s experience to the data from the late 1970s, and celebrated an improvement in mortality from 89% to merely 55%. Later, Lewandowski et al (1997) incorporated ECMO into their package of rescue therapies and pushed survival to 75%, which is actually admirable even by today's standards.
The 2009 CESAR trial from the UK is quoted in the college answer to Question 11 from the second paper of 2010. This trial randomised 180 patients to ECMO or conventional treatment. The study found a 63% survival at 6 months in the ECMO group, as compared to 47% in the conventional group. Critics of the study pointed out the fact that the groups were treated differently, and received significantly different levels of care (i.e. more lung-protective ventilation and sterids in the ECMO group). Furthermore, almost 25% of the patients who were being considered for ECMO didn't actually receive any - which might have, you know, killed them. These significant confounders make it difficult to make recommendations on the basis of CESAR alone. In summary, most agree that the mortality difference observed by the authors probably should not be attributed to ECMO.
The 2009 Australasian H1N1 experience with ECMO found a 21% survival rate at the end of the 3-month observation period, with many patients still in ICU at the end of the study. Mortality in the non-ECMO group was only 13%, which only goes to show that ECMO was being used as a last-ditch effort to rescue the sickest patients. Furthermore, the ECMO mortality in this ANZ cohort was lower than the worldwide reported mortality (which was around 30%) for ECMO-treated ARDS of a comparable severity.
In the time ensuing, the lack of high quality data has not stopped people from enthusiastically embracing extracorporeal methods, to the point where sane people (Kallet, April 2018) complained that everybody was using an expensive and unproven therapy instead of one which was proven and cheap. Li et al (2018) found that 69% of patients with ARDS went straight on ECMO without any trial of prone position ventilation, demonstrating a certain bias for the technique. In fact, "the proportion of all venovenous ECMO patients in whom prone positioning was used before ECMO was lower in studies published after 2013", after the publication of the PROSEVA trial. The authors' conclusions were uncharacteristically cynical; "...clinicians assume that a more complicated treatment is superior..." they alleged, and follow to accuse the ECMO centres of profiteering: " ... recently published rates suggest that a busy ECMO program could profit institutions by millions of dollars per year... In contrast, prone position is not reimbursed as an additional procedure".
The more recent EOLIA trial (Combes et al, May 2018) found no mortality difference in ARDS patients randomised to either ECMO or conventional ventilation. Of the controls, the vast majority (90%) were prone-ventilated, making this something of a "prone vs. ECMO" trial. Unfortunately 28% of these controls (ones which were sicker at the time of randomisation) migrated to the ECMO arm anyway, which makes the result difficult to interpret. Or easier - if they stayed in the conventional ventilation arm, presumably all would have died. The mortality in the cross-overs was much higher than in any of the others, and all were much sicker at the time of randomisation. Overall, the editorial from NEJM (Hardin & Hibbert, 2018) suggests that this trial supports the use of ECMO in patients who have clearly exceeded the capacity of conventional ventilation and prone position, rather than routinely using it for everybody with a P/F ratio worse than 80. Most would agree that the lack of mortality difference is surprising and that the dilution of results by crossover patients probably diminished the real contribution of ECMO, making it look less effective.
For these people, ECMO does not appear to be a bridge to anything. In general, in these older studies nobody over the age of 75 survived to discharge. This does not seem to have improved with time. A 2015 article on the use of VV or VA ECMO in cardiogenic shock still reports 70% mortality in both VV and VA groups. There's basically nothing here except for prospective observational studies (at best) or retrospective audits of registries.
One could go on, but these two studies are representative of the five articles (n = 233) which met the inclusion criteria in the biggest meta-analysis of this matter to day, by Ouweneel et al (2016). Overall, VA-ECMO showed a 33% higher 30-day survival compared to IABP alone, and was about the same as the Impella device. According to an abstract by Arora et al (2018), poor survival rates have not stopped people from using this technique more and more over the 2008-2018 period, particularly in privately insured Caucasian patients.
Basically, this group of ECMO candidates are a subset of the cardiogenic shock specimens. They were clearly fit for heart surgery (if elective) or it was urgent and unavoidable (in which case they might already have been in cardiogenic shock). In most of the studies, IABP is viewed as the less extreme invasive measure, and most of the patients end up on both IABP and ECMO after they fail on IABP alone.
Magovern et al (1994) published some of the earliest data on tis group. "The major problem limiting recovery was left ventricular distention secondary to inadequate left ventricular decompression", they complained. All the patients who had pre-op cardiac arrest died. All patients requiring ECMO after mitral valve surgery also died. Only the post-CABG patients survived, with a mortality rate of 20%. The indication for ECMO was failure to wean from bypass in spite of maximal inotropes and IABP.
Doll et al (2004), in another retrospective case series, found that only 37 out of 219 patients survived to 5 year follow-up. The mean duration of ECMO was about 2.8 days, which is much shorter than any of the other case series (most of the time it's one week, minimum). Again, in 144 of these, IABP was a co-intervention with ECMO.
Hsu et al (2010), in a retrospective analysis of 5 years of VA ECMO experience, found that about 53% of these patients could be weaned successfully off ECMO, i.e. there was some light at the end of the tunnel. In-hospital mortality overall was 67%.
Welp et al (2017) were able to demonstrate that the elderly who fail to separate from bypass still do poorly on ECMO in comparison to other populations. Their cohort of over-70s had a mortality rate of 75%. Again, the mean duration of ECMO was around 7-8 days.
ECMO as CPR has initially failed to yield satisfying results. In one study from 2003, the survival rate for patients in whom VA ECMO was started during CPR was 31%. Multi-organ system failure was the culprit there. However, a more recent observational study has produced some encouraging data.These days, it is beginning to look more and more promising.It may even improve outcomes: a 2016 review by Patel et al was able to find several propensity-matched cohort studies to support this assertion. Survival with good neurological outcome appeared to improve substantially; effect size was 28.8% vs 12.3% in one study (in-hospital) and 23.5% vs 5.9% in another (2-year). The obvious disadvantage is that you need an ED which has a circuit cycling 24/7, as well as dedicated night-time intensivist cover (unless one trusts their senior trainees to insert the cannulae).
Recently, exactly this scenario was played at a local centre of excellence, with encouraging results. As a part of a larger package (involving early reperfusion and therapeutic hypothermia) the CHEER trial demonstrated feasibility of urgent ECMO in the out-of-hospital arrest population, albeit in the highly unrealistic sci-fi utopia of a major ECMO centre. The investigators reported good (54%) survival with good neurological outcome.
The 2CHEER study (2020), which was a prospective observational cohort study trying to replicate the CHEER protocol, only reported 44% survival to hospital discharge with good neurological outcome (in 11 of 25 patients). Only twenty five patients were enrolled over 2.5 years among two major inner city hospitals in metropolitan Sydney, mainly because very few of them met criteria for ECPR. Interestingly, somehow the survival was poorer then in CHEER, even though 2CHEER enrolled patients with in-hospital cardiac arrest (who typically have better survival). Still, 44% was much better than the non-ECMO outcomes (for example, PARAMEDIC 2 had 30-day survival of 3.2% and 2.4% for their unselected out-of-hospital cardiac arrest group)
The ARREST trial (2020) also gave a similar number (43% survival to hospital discharge), and unlike 2CHEER was an actual trial with a control group (in whom the survival rate was 7%). So impressed were the Data Safety Monitoring Board that the trial was terminated after enrolling only 29 patients (14 and 15 in each arm), as it was viewed as unethical to continue killing patients with conventional resuscitation techniques. Unfortunately, the result of this move was a fragility index of 1, i.e. the trial findings would have shifted into the region of statistical insignificance if even one additional control group patient managed to survive.
Each has advantages and disadvantages.