Extracorporeal Membrane Oxygenation in Patients with Burns, Hero or Futile Medical Care? A Systematic Review and Meta-Analysis

Severely burned patients, particularly when compounded with inhalation injuries, are at high risk for cardiopulmonary failure. Recently, promising studies have stimulated interest in using extracorporeal membrane oxygenation (ECMO) as a potential therapy for burn patients with refractory cardiac and/or respiratory failure. Several observational studies have been reported, but the ndings in these vary. In this study, we conducted a systematic review and meta-analysis using standardized mortality ratios (SMRs) to elucidate the benets associated with the use of ECMO in patients with burn and/or inhalation injuries. A literature search using PubMed, EMBASE, MEDLINE, and the Cochrane Library were performed from inception to October 20, 2020. Clinical outcomes in the selected studies were compared.


Introduction
Severely burned patients, particularly when compounded with inhalation injuries, are at high risk for cardiopulmonary failure [1]. Despite advances in burn care, the morbidity and mortality for these patients remain extremely high [2,3]. Severe acute respiratory distress syndrome (ARDS) with refractory respiratory failure is one of the most dominant causes of death in patients with burns [2,4]. ARDS results from smoke inhalation injuries, pneumonia, and an overwhelming cascade of airway in ammation, extraordinarily elevating the mortality rates in burn patients [5,6]. Mechanical ventilation is the primary therapy to treat ARDS, which uses a lung-protective strategy to avoid superimposing additional damage on the alreadyinjured pulmonary alveoli in order to let the "lung rest". However, such ventilation is unable to provide lifesaving respiratory support when a critical volume of the alveolar unit has failed [7]. Extracorporeal membrane oxygenation (ECMO) is considered as an alternative treatment to solve this problem without overstretching the injured lungs, and provides cardiac support, for extended periods, from hours to several weeks [8].
The two most common forms of ECMO are veno-arterial (VA) ECMO and veno-venous (VV) ECMO. VA-ECMO support is required for cardiac and/or respiratory failure; VV-ECMO provides adequate oxygenation and carbon dioxide removal in isolated refractory respiratory failure [7]. In early studies, the high incidences of bleeding and thrombotic complications were attributed to practitioners' inexperience, resulting in unfavourable outcomes in ECMO-treated groups [9]. In recent years, ECMO has become more reliable with improvements in equipment, and increased practitioners' experience has led ECMO to become an alternative tool to treat patients with severe cardiac and pulmonary dysfunctions [10,11]. These promising studies have stimulated interest in using ECMO as a potential therapy for burn patients with refractory cardiac and/or respiratory failure.
However, burn patients with cardiac and/or respiratory failure are rare, making it di cult to perform randomizsd trials. Should a disaster occur, there may be many patients with major burns accompanied by cardiac and/or respiratory failure. However, a massive in ux of burn patients would shock the workforce of hospitals in the surrounding area, making it di cult to conduct clinical studies at that moment. Medical ethics is another concerning issue in this regard.
Standardized mortality ratios (SMRs) indicate the mortality in a cohort relative to the mortality in a reference population. A meta-analysis of SMRs investigated the all-cause and cause-speci c SMRs, eliminating the effect of differing patient characteristics in the two compared populations, and thus provides a better picture of the changes in survival [33,34]. Because most of the available literature on burn patients being treated with ECMO are observational studies, and there is a lack of systematic studies evaluating cause-speci c mortality, we conducted a systematic review of the literature and performed a meta-analysis on the available clinical data using SMRs. This was performed to elucidate the bene ts associated with the use of ECMO in the patients with burn and/or inhalation injuries.

Methods
This study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.

Search strategy
A systemic literature search was carried out in PubMed, EMBASE, MEDLINE, and the Cochrane Library databases on October 20, 2020 using the following search terms: "burn," "burns," "ARDS," "adult respiratory distress syndrome", "Extracorporeal Membrane Oxygenation," "ECMO," "inhalation injury," "smoke," and "respiratory failure." All published articles were limited to human studies without language restrictions. All identi ed articles were screened for cross references.

Study selection
Review articles, observational controlled studies, letters, and case reports were included in the study. The titles and abstracts of all of the identi ed articles were screened and selected according to the following inclusion criteria: (1) participants were children or adults with a diagnosis of a thermal burn and/or smoke inhalation requiring ECMO as determined by a physician; (2) an identi ed group of patients who received ECMO as part of their therapeutic regimen; and (3) refer to disease severity in patients treated with ECMO using the revised Baux system, or with details provided for further calculation. For multiple studies using the same cohort, studies with the longest follow-up durations and that met the study inclusion criteria were selected. Studies meeting one of the following criteria were excluded from our analysis: (1) studies that were duplicate publications and (2) studies with appropriate data that could not be extracted based on the published results.
Two reviewers (Y.A. K. and Y.J. C.) independently examined the titles and abstracts of the articles independently. A subsequent full-text review was performed manually when the abstracts did not provide su cient information. Any disagreements were discussed with a third reviewer (Y.J. H.) and resolved by consensus.

Page 5/21
The outcomes evaluated included patient mortality rates and SMRs. The revised Baux scoring system described by Osler et al. has been widely adopted using age, total body surface area burned, and inhalation injuries as predictors to produce outcome estimates on a continuous scale [35]. Revised Baux scores were calculated as age(years) + TBSA (%) + (17 * inhalation injury). Predicted mortality was calculated using a logistic regression model = .
For each study, the expected mortality was calculated by multiplying the number of cases by the revised Baux score predicted mortality rate.

Data extraction
Two reviewers (Y. A. K. and Y. J. C.) extracted the following data separately from all of the studies that met the inclusion criteria independently. The extracted data from the studies included the: study types, sample sizes, inclusion dates, treatment regimens, ages, sexes, countries, burn types, TBSA, presence of inhalation injuries or ARDS, ECMO settings, mortality status, mortality rates, revised Baux scores, and revised Baux score-based SMR with 95% CI. All the extracted data were crosschecked to rule out any discrepancies.

Data synthesis
The meta-analysis was performed using MetaXL version 5.2 following the PRISMA guidelines. We calculated the pooled crude mortality rate of patients receiving ECMO. The results were expressed as overall odds ratios (ORs) with associated 95% con dence interval (CI). For all studies that provided the revised Baux scores of patients, logistic regression calculations between the revised Baux scores and predicted mortality rates were performed. SMR was de ned as the ratio of observed mortality to expected mortality, and the accompanying 95% CI was based on the methods used by Ury and Wiggins [34]. We produced a pooled SMR for ECMO treatment, with the results expressed as overall SMRs and associated 95% CIs. Subgroup analyses of different ECMO settings and paediatric patients were also performed.
Heterogeneity across studies was evaluated using the X 2 test, P values, and I 2 statistics. A random effects model was used for all analyses because of the large heterogeneity of the sample. Funnel plots were used to identify the present of publication bias [36]. When the mortality rate was zero, we added 0.5 to both the observed deaths and expected deaths and used the adjusted SMRs in our analysis.

Study selection
The abstraction process is detailed in Fig. 1 The meta-analysis found that the observed mortality was signi cantly higher than the predicted mortality in patients receiving ECMO, with a pooled SMR of 2.07 (95% CI: 1.04-4.14) as shown in Fig. 2A. The funnel plot did not indicate any publication biases (Fig. 3). However, in the burn patients with inhalation injuries subgroup, all patients receiving V-V ECMO had a lower mortality than their predicted mortality, with a pooled SMR of 0.95 (95% CI: 0.52-1.73) as shown in Fig. 4C. Other subgroup analyses, including an adult group, paediatric group, V-V setting group, and V-A setting group, did not report any bene ts from ECMO; however, these results were not statistically signi cant (Figs. 2 and 4). Interestingly, the pooled SMRs decreased as patients' revised Baux scores increased, with a high correlation (R= -0.92), as shown in Fig. 5. The pooled SMRs were less than one when the selected patients' revised Baux scores exceeded approximately 90, indicating that the potential bene ts from ECMO treatment increased as the severity of patients with burns increased, especially when the patients' revised Baux scores exceeded 90.

Assessment of bias
Funnel plots revealed no evidence of publication bias, as shown in Fig. 3. A random effects model was used for all analyses due to the large heterogeneity of the sample. According to the GRADE classi cation, we judged the quality of evidence of included studies. Subcategories of bias (such as indication, selection, allocation, performance, attrition or reporting bias) were not assessed.

Discussion
To the best of our knowledge, this study is the rst review and meta-analysis of burn patients receiving ECMO therapy that is based on SMRs. The pooled all-cause mortality of burn patients receiving ECMO was 48%. The pooled overall SMR of 2.07 (95% CI: 1.04-4.14) suggested that ECMO recipients have signi cantly higher mortality rates compared to their predicted mortality rates calculated using their revised Baux scores.
The use of ECMO may rather increase mortality in unsuitable patients. Moreover, our subgroup analysis showed no bene ts in terms of patient survival when using ECMO in different settings or depending on different age populations. However, in the subgroup of burn patients with inhalation injuries who received V-V ECMO and those with major burn injuries with revised Baux scores exceeding 90, the observed mortality rates were lower than the predicted mortality rates, with a pooled SMR of 0.95 (95% CI: 0.52-1.73) and 0.90 (95% CI: 0.42-1.93).
SMRs based on generic mortality prediction models have been widely applied to predict deaths in the general population [37]. Various mathematical models have been developed and widely used to predict mortality as an outcome of burn injuries. They are associated with several factors, including age, TBSA, inhalation injuries, and so on [38]. Lots of prediction models such as the revised Baux score [39], Abbreviated Burn  [38]. Several studies have reported that the revised Baux score system is more accurate for predicting survival not only in adult patients but also in paediatric patients [38,39,[44][45][46][47][48][49]. Moreover, this model is simple to calculate and has good calibration and discriminatory power.
As a result, our SMR calculations were based on the revised Baux score system when conducting the analyses in this study.
In recent decades, ECMO has become an essential tool in the care of patients with severe cardiac and pulmonary dysfunctions that are refractory to conventional management [10,11,50] There were some limitations to this analysis. First, all included studies were case series or retrospective studies with a limited sample size. However, burn patients with cardiac and/or respiratory failure are nearly impossible to perform randomised trials due to ethical considerations and the rarity of the injuries with ECMO therapy. Second, SMRs that are based on prediction scoring systems such as in our study may have biases, other comorbidities and complications during hospitalisation were not evaluated as well. Last, since ECMO therapy is a rapidly evolving technology, older studies may follow different protocols or indications, causing different outcomes and selective biases.

Conclusions
This study revealed that burn patients receiving ECMO treatment were at high risk of death. As a rescue treatment, it should not be routinely used in all burn patients. However, select patients including those with inhalation injuries and patients with a revised Baux score exceeding 90, may bene t from ECMO treatment.
The potential bene ts from ECMO treatment increase as the severity of the patients with burns increases. We suggest that these patients consider early ECMO intervention and should consider being transferred to an ECMO centre.

Competing interests
None.

Funding
This study was supported by a grant from the Taipei Veterans General hospital, Taipei, Taiwan (V106A-008 and V110B-038).
Authors' contributions YJC and YAK contributed to the study design, study selection, data extraction, quality assessment and writing of the manuscript. YCH and YAK analyzed the data and veri ed the analytical methods. TWC and HM revised the manuscript for important intellectual content. YAK and HM contributed to the nal version of the manuscript and supervised the project. All authors read and approved the nal manuscript.   For all studies that provided the revised Baux scores of patients, logistic regression calculations between the revised Baux scores and predicted mortality rates were performed. A pooled SMR for ECMO treatment, with the results expressed as overall SMRs and associated 95% CIs. The observed mortality was signi cantly higher than the predicted mortality in patients receiving ECMO, with a pooled SMR of 2.07 (95% CI: 1.04-4.14) (A); adult group and paediatric group did not report bene ts from ECMO (B) and (C). CI: con dence interval Figure 2 For all studies that provided the revised Baux scores of patients, logistic regression calculations between the revised Baux scores and predicted mortality rates were performed. A pooled SMR for ECMO treatment, with the results expressed as overall SMRs and associated 95% CIs. The observed mortality was signi cantly higher than the predicted mortality in patients receiving ECMO, with a pooled SMR of 2.07 (95% CI: 1.04-4.14) (A); adult group and paediatric group did not report bene ts from ECMO (B) and (C). CI: con dence interval   In V-A setting group, and V-V setting group, did not report bene ts from ECMO (A) and (B); in the burn patients with inhalation injuries subgroup, all patients receiving V-V ECMO had a lower mortality than their predicted mortality, with a pooled SMR of 0.95 (95% CI: 0.52-1.73) (C). CI: con dence interval In V-A setting group, and V-V setting group, did not report bene ts from ECMO (A) and (B); in the burn patients with inhalation injuries subgroup, all patients receiving V-V ECMO had a lower mortality than their predicted mortality, with a pooled SMR of 0.95 (95% CI: 0.52-1.73) (C). CI: con dence interval The pooled SMR decreased as the patients' revised Baux scores increased, with a high correlation (R= -0.92).
The pooled SMR would cross over 1 when the patient's revised Baux exceeded approximately 90, indicating that the potential bene ts from ECMO treatment increased as the severity of patients with burns increased, especially when the patients' revised Baux scores exceeded 90.