2011 Oct-9 Extracorporeal Membrane Oxygenation as a First-Line Treatment Strategy for ARDS Is the Evidence Sufficiently Strong?
William Checkley, Published online October 5, 2011. doi: 10.1001/jama.2011.1504
The 2009 influenza A(H1N1) pandemic was associated with a high attributable mortality among critically ill patients who developed acute respiratory distress syndrome (ARDS) and required mechanical ventilation. In this issue of JAMA, Noah and colleagues1 present evidence in support of extracorporeal membrane oxygenation (ECMO) in combination with lung protective ventilation as a treatment strategy early in the course of ARDS related to H1N1 infection. The authors found that among 80 patients with severe suspected or confirmed H1N1 and ARDS who were transferred to 4 UK specialized centers for treatment with ECMO, 22 died (27.5%) before hospital discharge.
1 This mortality rate was lower than that among matched critically ill patients with equally severe (suspected or confirmed) H1N1 and ARDS who were not transferred for treatment with ECMO. This study adds to a series of recent investigations that favor the use of ECMO for severe respiratory failure in adults.
1,2,3 In all of these studies, ECMO was initiated in the first 7 days of mechanical ventilation. Average duration of ECMO use was 9 to 10 days, and reported mortalities ranged from 21% to 37%.
1,2,3
Use of ECMO in severe respiratory failure, particularly in the treatment of ARDS, is occurring more commonly. Extracorporeal gas exchange may allow the use of low tidal volumes and lower levels of inspired oxygen, and use of higher positive end-expiratory pressure if desired. A treatment strategy that capitalizes on volume and pressure limitation may decrease the risk of regional overdistention of heterogeneously compromised lungs and prevent further injury. The study by Noah et al
1 further suggests that a UK-wide system for referral and use of ECMO can be established quickly and effectively during an influenza epidemic. Furthermore, the Conventional Ventilatory Support Versus Extracorporeal Membrane Oxygenation for Severe Adult Respiratory Failure (CESAR) trial reported by Peek et al
2 suggested safety and improved outcomes for patients with severe respiratory failure who were referred to a specialized center for ECMO-based management vs those managed with conventional ventilation.
However, some limitations may diminish enthusiasm for wider use of ECMO. First, ECMO can be accomplished effectively only in highly specialized centers; it is invasive, requires anticoagulation for the duration of therapy, and can be associated with serious complications. Second, the design of both the study by Noah et al1 and the CESAR trial
2 required transfer of patients to highly specialized centers for management of ECMO, making it unclear whether the mortality benefit associated with ECMO was attributable to management of severe respiratory failure in a specialized center or to the use of ECMO. Evidence for the former is that patients who receive mechanical ventilation in tertiary care centers with high case volumes have a lower mortality than patients treated in centers with lower case volumes. In a study by Kahn et al4 of approximately 20 000 patients who received mechanical ventilation registered in the Acute Physiology and Chronic Health Evaluation clinical information database across 37 centers, hospital mortality was 34% in low-volume sites compared with 26% in the high-volume sites. Nonetheless, a sensitivity analysis conducted by Noah et al
1 revealed that the findings of a survival advantage favoring the group of ECMO-referred patients were unchanged when the group of non–ECMO-referred patients was restricted to the subgroup of patients recruited at high-volume centers.
Other studies highlight how difficult it is to disentangle the effect of treatment with ECMO from the experience and care in a highly specialized center. Hospital mortality for patients with ARDS from all causes enrolled in the US National Institutes of Health's Acute Respiratory Distress Syndrome Network trials and managed with lung protective ventilation between 1996 and 2005
5,6,7 was similar to that of patients with H1N1-related ARDS managed with ECMO under the treatment strategy presented by Noah et al.
1 Specifically, hospital mortality at 90 days was 29% among 1715 patients with a ratio of PaO2 to fraction of inspired oxygen of less than 200 mm Hg prior to enrollment in ARDS Network trials. Considering the lower age range (28-46 years) of patients who received ECMO in the study by Noah et al,
1 hospital mortality at 90 days among patients enrolled in ARDS Network trials in this age range was 21%. The Canadian Critical Trials Group H1N1 Collaborative reported a 90-day mortality of 21% in 118 adult critically ill patients with confirmed or suspected H1N1 infection, yet less than 5% received treatment with ECMO.
8 The implications of international comparisons, however, are limited because differences in admission criteria and discharge practices may result in different mortality rates.
Therefore, the question remains as to whether there was a true difference in mortality between patients managed at the same highly specialized centers with conventional lung protective ventilation compared with a strategy that incorporates ECMO. In the report by Noah et al,
1 the authors explored whether differences in the quality of care could have influenced their results. While the investigators did not have access to detailed information regarding use of lung protective ventilation strategies, they did restrict the analysis to selection of non–ECMO-referred patients from centers with better than expected severity-adjusted outcomes and found that the results were unchanged.
Third, the study by Noah et al1 was an observational, prospective study of patients with ARDS from H1N1 infection who received ECMO and lung protective ventilation at 4 referral hospitals (64% of patients were referred to Glenfield Hospital in Leicester, England) compared with patients of similar disease severity who were treated at UK hospitals outside the 4 referral centers. Both ECMO-referred patients and non–ECMO-referred patients were identified from hospitals participating in the Swine Flu Triage study.9 Even though the study was not randomized, the authors used novel analytic techniques based on propensity score matching to achieve the best possible balance between individual risk factors across the 2 study groups. A critical concept of randomization is that it achieves balance between observed and unobserved factors to prevent systematic differences between treatment groups, therefore allowing for a causal interpretation in findings. While such balance is difficult to achieve in observational studies, propensity score analysis aims to select matched patients from each of the 2 study groups to reduce bias. As such, the authors have provided careful analyses of their data and used several appropriate matching methods. Propensity score matching offers a complementary approach to the analysis of observational studies but does not replace a randomized clinical trial.
An additional complicating factor in the analysis of an observational study involves having data from multiple centers. Because the main thesis of the current study was that specialized centers that incorporate ECMO make a difference in survival from ARDS related to H1N1 infection, hospitals and not individuals are the unit of analysis. Thus, a better study design and analysis than that conducted by Noah et al1 may be to compare the averages of hospital mortalities across several centers with or without ECMO-based management. Additional propensity score matching at the hospital level would help reduce bias from differences between centers in case volume, severity of illness, and other important determinants. However, such a design would require many more patients and hospital centers than were available in the current study. It is therefore possible that the current study may have been underpowered to determine if ECMO was associated with a survival advantage when using hospitals as the unit of analysis.
In summary, the study by Noah et al1 involving critically ill patients with H1N1 joins other recent investigations
2,3 that have revitalized interest in the use of ECMO as a treatment strategy for ARDS. While underlying risk factors may be different, severe respiratory failure from H1N1 infection presents a clinical challenge similar to that involving ARDS from other causes. Despite several decades of investigation into potential treatment strategies, use of low tidal volumes remains the only proven therapy to decrease mortality in ARDS.
5 In light of the large observed differences in mortality with and without ECMO, large consortia of trialists may be enticed to consider ECMO as a potential target for a randomized controlled trial early in the course of severe ARDS from all causes.
Weblink
here
