Long-term lung function recovery after ECMO versus non-ECMO management in acute respiratory failure: a systematic review and meta-analysis

Abstract

Background

Extracorporeal membrane oxygenation (ECMO) is increasingly employed to support lung function in patients with acute respiratory failure (ARF). However, the long-term outcomes of the approach have not been encouraging when compared to those of conventional mechanical ventilation. Further, the long-term effects of ECMO on lung function and recovery are unclear. For this review, we examined the long-term lung function outcomes of patients with ARF treated with and without ECMO.

Methods

We searched the Embase, CENTRAL, Web of Science, and PubMed sites for studies comparing long-term (≥ 6 months) pulmonary function test results in patients with ARF treated with and without ECMO published until January 2024. We conducted a meta-analysis for percentage predicted values.

Results

We included five studies. Our meta-analysis showed similar values of forced vital capacity (FVC%) (MD, 0.47; 95% CI, -3.56–4.50) and forced expiratory flow in the first second % (MD, 1.79; 95% CI, -2.17–5.75) in patients with ARF treated with or without ECMO. The FEV1/FVC % values were slightly higher in patients treated with ECMO than in those without ECMO (MD, 2.03; 95% CI, 0.01–4.04; p-value = 0.05). According to the meta-analysis, the values for total lung capacity % (MD, -3.20; 95% CI, -8.83–2.44) and carbon monoxide diffusion capacity % (MD, -0.72; 95% CI, -3.83–2.39) were also similar between patients undergoing ECMO and those without it.

Conclusion

The meta-analysis of a small number of studies with significant selection bias indicates that patients with ARF treated with ECMO may have comparable long-term pulmonary function recovery to those treated with conventional strategies. Further investigations including a larger number of patients and focusing on the long-term impact of ECMO are needed to supplement the current evidence.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-024-03321-1.

Introduction

Acute respiratory failure (ARF) is a common cause of hospitalizations and intensive care unit (ICU) admissions globally [1, 2]. The condition is characterized by rapid and progressive hypoxemia developing due to any underlying respiratory, cardiovascular, or systemic illness. The pathogenesis involves hypoventilation, diffusion impairment, shunting, ventilation-perfusion mismatch, or a combination of these mechanisms [3]. Approximately 10% of all cases are diagnosed as having a severe form of ARF called acute respiratory distress syndrome (ARDS), characterized by lung infiltrations secondary to numerous underlying diseases, or injuries [4]. Mortality after ARF and ARDS is high and ranges from 40–54% [5, 6] and survivors present significant disease-related sequelae like reduced exercise capacity, decreased quality of life, cognitive impairment, and psychiatric disorders [7, 8].

ARF management is primarily supportive to maintain adequate oxygenation and ventilation, with pharmacotherapy for treating the underlying condition [4]. Extracorporeal membrane oxygenation (ECMO) has gained rapid acceptance worldwide for the management of ARF. ECMO is a mechanical device that temporarily supports a failing heart or lung by providing gas exchange [9]. In ARF, ECMO can provide lung-protective ventilation by allowing reduced respiratory rates and tidal volumes as well as lower driving and plateau pressures of the mechanical ventilator [10]. Despite these theoretical advantages, the inconsistent results in the literature need to be addressed to clarify its benefit to improve outcomes [11]. Some evidence supports the use of ECMO for life-threatening ARF, but the risk-to-benefit ratio needs to be accurately calculated [10]. Moreover, ECMO is a high-risk, expensive, and complex treatment modality and its effects on long-term outcomes are unknown. Since ARF can lead to significant long-term pulmonary sequelae [7, 8], the impact of ECMO on long-term lung function must be clarified for clinicians to be able to make optimal informed decisions on its allocation and use. Thus, we conducted this systematic review to compare the long-term functions of ARF survivors treated with or without ECMO.

Materials and methods

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement reporting guidelines during the study and to present our findings [12]. We pre-registered the study on the International Register of Systematic Reviews, PROSPERO (CRD42024506957).

Literature retrieving

We conducted an online search on the databases of Embase, CENTRAL, Web of Science, and PubMed. Two reviewers independently screened all the studies published until 31st January 2024 and included all articles without language restrictions. The search was restricted to human studies published as full-texts.

The search terms used were: “extracorporeal membrane oxygenation”; “ECMO”, “Extracorporeal lung assist”, “acute respiratory failure”; “acute respiratory distress syndrome”; “spirometry”, “lung function”; and “pulmonary function”. We generated search strings by combining these keywords with Boolean operators and used them across all databases (Supplementary Table 1). In addition, we examined Google Scholar as a source for grey literature and hand-searched references of included original articles and pertinent reviews.

Eligibility criteria and selection of studies

We screened the searched articles against the following PICOS inclusion criteria:

Population (P): Adult patients with ARF who survived after treatment.

Exposure (E): Management with ECMO.

Comparison (C): Standard care (without ECMO).

Outcomes (O): Pulmonary function tests (PFTs) or any component of PFTs. The duration of follow-ups had to be at least 6 months.

Study type (S): randomized controlled trials (RCT), prospective or retrospective cohort studies.

We excluded single-arm studies, studies not exclusively on ARF, not examining relevant outcomes, editorials, reviews, and non-peer-reviewed articles. For publications reporting results on the same data, we included only the article with the largest sample size.

We initially combined the search results of all databases in a single reference manager software (EndNote) and eliminated duplicates. The two reviewers then carefully read the title and abstract of each study for screening, and they read full-texts of the selected studies to determine their suitability for our review according to the inclusion criteria. Any differences between reviewers were solved by dialogue.

Risk of bias and data management

We assessed the quality of studies via the Newcastle Ottawa Scale (NOS) [13]. Two reviewers judged the studies according to criteria for participant selection, comparability of groups, and validity of results adjudicating a number of stars (ranging from 1 to 9) to each study (a higher star number indicated better study quality).

Two investigators extracted data from each study including study information, patient demographic factors, COVID-19 disease history, body mass index, sequential organ function assessment (SOFA) score, tidal volume, partial pressure of oxygen/ fraction of inspiratory oxygen concentration ratio (PaO₂/FiO₂), use of tracheostomy, proning history, indication and duration of ECMO, length of ICU stay, outcomes, and follow-up results. We identified all available PFT components and pooled their percentage predicted values. Finally, we conducted a meta-analysis for forced vital capacity (FVC%), forced expiratory flow in the first second (FEV1%), FEV1/FVC%, total lung capacity (TLC%), and diffusing capacity of the lungs for carbon monoxide (DLCO%).

Statistical analysis

We used “Review Manager” (RevMan, version 5.3) to calculate the pooled mean differences (MDs) and 95% confidence intervals (CIs) of PFTs between ECMO and non-ECMO groups. We combined data in an inverse variance random-effect model. Due to the low number of suitable studies, we did not assess publication bias via funnel plots and omitted a sensitivity analysis. We assessed inter-study heterogeneity by calculating chi-square-based Q and I² statistics. We considered a p-value < 0.10 for Q statistic and > 50% for I² as indicative of substantial heterogeneity. In addition, we considered all P values < 0.05 as reflecting statistical significance.

Results

The combined search results obtained from all databases included 6368 articles. Of those, 1546 were unique articles. The reviewers found 28 articles relevant to the review after screening based on the titles and abstracts. Finally, only five articles met the inclusion criteria [14–18] after the full-text analysis (Fig. 1). There was high agreement between the reviewers for the study selection process (Kappa = 0.89).

Fig. 1.

Chronology of literature search

Table 1 shows the details of data collected from the studies. The five studies were from the USA, UK, Italy, France, and China with 178 patients receiving ECMO and 258 managed without ECMO. Indications for ECMO varied as per the local hospital protocols. Table 2 lists the indications reported by individual studies. All patients received veno-venous ECMO. One study was an RCT [14], whereas all others were retrospective cohort studies. The RCT included all cases of ARF and the retrospective studies were on ARDS. One study [18] also included patients with coronavirus disease-19 (COVID-19). Most ECMO-treated patients exhibited higher SOFA scores and worse PaO₂/FiO₂ ratios than the non-ECMO-treated patients. Tidal volumes at the beginning of treatment were not routinely reported. The mean/median durations of ECMO ranged from 6 to 10 days. The lengths of ICU stay were longer in the ECMO group than in the non-ECMO group. Two studies [14, 18] reported a follow-up of 6 months, whereas three studies reported data after 1 year of treatment. None of the observational studies accounted for comparability between the ECMO and non-ECMO groups, so we did not awarded any stars for this domain in the NOS. All observational studies scored 6, while the RCT scored 8.

Table 1.

Details of included studies

Study	Location	Groups	Sample size	Age, years	Males (%)	BMI, kg/m2	SOFA score	P/F ratio, mmHg	TS (%)	Proning (%)	Tidal volume, mL	Duration of ECMO, days	Length of ICU stay, days	Follow-up	NOS score
Snyder 2023[18]	USA	ECMO Non-ECMO	34 76	35 51	56 57	33[26–40] 34[28–39]	7[5–9] 4[3–6]	104[81–158] 150[103–210]	NR	82 74	386 [285–463] 397 [340–427]	NR	29[19,43]	6 months	6
Grasselli 2019[16]	Italy	ECMO Non-ECMO	18 19	NR	NR	NR	NR	NR	66 47	77 47	NR	10[6–13]	29[16–36] 13[5–29]	1 year	6
Wang 2017[15]	China	ECMO Non-ECMO	24 48	38 44.3	75 69	26.5 ± 6.1 26.9 ± 6.2	10.8 ± 3.5 7.9 ± 3.1	68.3 ± 16.1 84.8 ± 16.5	17 21	NR	3.4 [3.1–4.2]* 6.3[5.8–6.9]	6 ± 2.3	13[9.8–22.3] 11[8–18]	1 year	6
Luyt 2012[17]	France	ECMO Non-ECMO	12 25	35.5 42	42 52	30.3[23.7–31.7] 25.3[22.5–28.6]	8[6-9.5] 7[4–8]	73.5[67–121] 90[71–139]	67 8	NR	NR	9.5[7.5–17]	37.5[19–67] 19[12-27.5]	1 year	6
Peek 2009[14]	UK	ECMO Non-ECMO	90 90	NR	57 59	NR	NR	75·9 ± 29·5 75 ± 35.7	NR	40 42	NR	9[6–16]	24[13-40.5] 13[11–16]	6 months	8

TS, Tracheostomy; DM, diabetes mellitus; P/F, PaO₂ /FiO₂; SOFA, Sequential Organ Function Assessment; BMI, body mass index; ECMO, extracorporeal membrane oxygenation; NR, not reported

*mL/kg

Data in bold indicates statistical significant difference between ECMO and non-ECMO groups

Table 2.

Criteria for ECMO

Study	Criteria
Snyder 2023[18]	(1) Hypercapnia (PaCO2 > 60 mmHg with pH < 7.25 or inability to ventilate the patient with plateau pressure < 30 cmH2O) or (2) severe hypoxemia (P/F ratio] < 50 mmHg with FiO2 > 80% FiO2 for > 3 h or P/F ratio < 80 mmHg on 80% FiO2 for > 6 h despite optimization of mechanical ventilation). Relative contraindications to ECMO were: (1) Age > 60 years old; (2) BMI > 40 kg/m2; 3) > 10 days mechanically ventilated; 4) home oxygen requirement; 5) severe neurological injury/insult; 6) terminal dis- ease with low 1-year survival; 7) severe underlying liver disease; 8) acute hepatic failure; 9) Jehovah’s Witness (unwilling to receive blood); 10) Acquired Immune Deficiency Syndrome; 11) WBC < 1000 cells/mL3 of blood; 12) poor baseline functional status.
Grasselli 2019[16]	(1) Oxygenation index > 30; (2) PaO2/FiO2 < 70 mmHg with PEEP > 15 cmH2O (in patients admitted to an ECMO center) or PaO2/FiO2 < 100 mmHg with PEEP > 10cmH2O (in patients still to be transferred); (3) pH < 7.25 for at least 2 h. In general, ECMO support is started after failure of rescue therapies such as pulmonary recruitment maneuvers, inhaled nitric oxide and prone positioning. Absolute contraindication to ECMO were: (1) non-reversible cause of respiratory failure; (2) intracranial bleeding or major contraindication to anticoagulation; (3) active malignancy; (4) previous severe disability. Age and comorbidities did not constitute absolute contraindications to ECMO. A relative contraindication to ECMO was use of invasive mechanical ventilation for > 7 days prior to ECMO consideration.
Wang 2017[15]	Severe respiratory failure (PaO2/FiO2 < 100 mmHg on FiO2 of 1, a Murray lung injury score of 3 or higher, decompensated hypercapnia with a pH of < 7.2) despite optimal conventional treatment, age < 65 years, and less invasive mechanical ventilation time (< 7 days) prior to ECMO
Luyt 2012[17]	Persistent refractory hypoxemia, defined as a PaO2, 50 mm Hg despite an FiO2 = 1, prone positioning, and/or nitric oxide administration during at least 3 h, or persistently high plateau pressure (≥ 35 cm H2O) despite a low tidal volume (but ≥ 4 mL/kg of ideal body weight) with pH ≤ 7.15 and a positive end-expiratory pressure of at least 5.
Peek 2009[14]	If patients were haemodynamically stable, a standard acute respiratory distress syndrome treatment protocol was used, which comprised pressure-restricted mechanical ventilation at 30 cm H2O, positive end- expiratory pressure titrated to optimum SaO2, FiO2 titrated to maintain SaO2 at more than 90%, diuresis to dry weight, target packed cell volume of 40%, prone positioning, and full nutrition. If the patient did not respond to this protocol within 12 h (FiO2 > 90% needed to maintain SaO2 > 90%, respiratory or metabolic acidosis < 7·2) or was haemodynamically unstable, they received cannulation and ECMO.

PaO2, partial pressure of oxygen in arterial blood; FiO2, fraction of inspiratory oxygen concentration; SaO2, arterial oxygen saturation level; ECMO, extracorporeal membrane oxygenation; NR, not reported; P/F, PaO₂ /FiO₂; WBC, white blood cells

Figures 2 and 3 show the forest plots of pooled data available from all five studies. The meta-analysis showed similar values of FVC% (MD, 0.47; 95% CI, -3.56–4.50) and FEV1% (MD, 1.79; 95% CI, -2.17–5.75) in both groups of patients with ARF (Fig. 2). There was no inter-study heterogeneity in either meta-analysis (I² = 0%). However, we found the FEV1/FVC % to be slightly higher in patients treated with ECMO than in the others (MD, 2.03; 95% CI, 0.01–4.04; p-value = 0.05). The inter-study heterogeneity was nil (I² = 0%). The meta-analysis results for TLC% (MD, -3.20; 95% CI, -8.83–2.44) and DLCO% (MD, -0.72; 95% CI, -3.83–2.39) were similar for both ECMO and non-ECMO groups with low interstudy heterogeneity (I² values of 0% and 3%, respectively for each variable).

Fig. 2.

Meta-analysis of FVC% and FEV1% between ECMO and non-ECMO groups

Fig. 3.

Meta-analysis of FEV1/FVC%, TLC1%, DLCO% between ECMO and non-ECMO groups

Discussion

The benefit of ECMO as a treatment modality for ARF has generated significant controversy. The CESAR RCT published in 2009 demonstrated a significant improvement in survival without disability in patients transferred to an ECMO center as compared to the outcomes of patients receiving conventional mechanical ventilation [14]. In 2018, another international RCT (EOLIA) showed that ECMO did not significantly reduce the 60-day mortality of patients with severe ARDS when compared to the 60-day mortality of patients who had undergone conventional mechanical ventilation combined with other rescue strategies [19]. However, both these trials had limitations. In the CESAR trial, only a subset of patients assigned to receive ECMO actually received the treatment, and the EOLIA trial was stopped early due to ECMO futility. However, an individual patient-data meta-analysis [20] of pooled analyses with other observational studies [21] revealed that compared to conventional mechanical ventilation, ECMO is associated with a significant reduction in 60- and 90-day mortality. Further, the experience of ECMO in the HINI and the COVID-19 pandemic data have strengthened the idea that ECMO provides a benefit to patients with ARF [22, 23].

ARF management with conventional ventilation entails the use of large tidal volumes (10–15 mL/kg) to maintain appropriate oxygenation and ventilation. High tidal volumes and inspiratory pressures can induce ventilator-associated lung injury by over-distending the alveoli, disrupting the alveolar–capillary membrane, and increasing inflammation [24]. Research has shown that ARF survivors are left with impaired DLCO, airway obstruction, chest restriction, and combined obstructive-restrictive patterns on long-term follow-ups [24, 25]. Lung-protective ventilation or low tidal volumes have been suggested to protect the lung function. Such a strategy can reduce the risk of mortality when compared to conventional ventilation with high tidal volumes in ARDS, but the long-term data on lung function remains limited [26].

The benefits of ECMO are primarily backed by its lung-protective effects and the reduction in the risk of ventilator-associated lung injury by allowing reduced tidal volumes, respiratory rates, plateau and driving pressures, and mechanical powers [10]. However, whether the “protective effect” of ECMO translates into long-term preservation of lung function when compared to conventional treatment remains to be seen. To the best of our knowledge, our meta-analysis presents the first evidence for long-term pulmonary function in ARF survivors treated with and without ECMO. However, the evidence is derived primarily from observational studies and not from RCTs. We found baseline differences between the ECMO and non-ECMO cohorts in all studies but in the RCT [14]. None of the observational studies ensured the comparability of ECMO and non-ECMO groups and their low NOS scores highlight their high risk of bias. In addition, treatment protocols were decided on a case-by-case basis in the observational studies, which were therefore prone to selection bias. After pooling the data, we found similar percentage predicted values of FVC, FEV1, FEV1/FVC, TLC, and DLCO in both groups after follow-ups of 6 months to 1 year. These results were consistent across all studies with none demonstrating any beneficial effects of ECMO on the long-term lung function. Our results also confirm the findings of a previous meta-analysis by Wilcox et al. [27] who pooled data from just two studies to demonstrate similar long-term FVC% and FEV1% values in ECMO- and non-ECMO– treated patients with ARF.

The criteria for selecting patients for ECMO varied across studies, but most studies deployed ECMO when proven conventional management techniques like mechanical ventilation (including lung-protective) and prone positioning had failed. Combes et al. [28] have suggested that ECMO should be implemented in settings where the patients meet the EOLIA trial [19] inclusion criteria, which are: PaO₂/FiO₂ < 50 mmHg for > 3 h; PaO₂/FiO₂ < 80 mmHg for > 6 h; pH < 7.25 with a PaCO₂ ≥ 60 mmHg for > 6 h, and after all proven conventional management techniques have failed. ECMO is still an evolving technology and its application is highly dependent on the availability of equipment, skill of the intensivist, and patient characteristics. Given these limitations, complete standardization of patients receiving ECMO is difficult and clinicians have to rely on heterogenous data to formulate guidelines. The studies included in our meta-analyses all had different management protocols for the non-ECMO groups: There were variations in the type of invasive ventilation, use of proning, corticosteroids, inhaled vasodilators, and muscle relaxants based on individual patient characteristics and the choice of the clinician. This limitation was noted during the CESAR trial wherein no standard protocol was mandated for the non-ECMO group as the participating centers could not reach a consensus on what the best ARF treatment was [14]. The variable ARF manifestations and the conflicting outcomes of various treatment modalities have led to the formulation of different guidelines without a gold-standard [4].

Other meta-analysis limitations include the scarcity of relevant publications. Despite our comprehensive search, we found only five studies with relatively small sample sizes for the meta-analysis. Most of the studies had high risks of bias as suggested by their low NOS scores. Moreover, we were unable to examine outcomes like the six-minute walk test, arterial blood gas measurements, or others, which could have provided insights into long-term function. The baseline differences in ECMO and non-ECMO cohorts point to selection bias wherein younger and non-comorbid patients were prioritized. Also, in most studies, the ECMO patients were sicker with higher SOFA scores, worse PaO₂/FiO₂ ratios, and longer lengths of ICU stay, factors which can impair outcomes. None of the included studies used propensity-score matching to account for baseline differences. Moreover, there were variations in the included populations amongst the studies. One study [18] included patients with COVID-19 who probably presented different outcome risks. According to a study, patients with COVID-19 treated with ECMO had longer ECMO durations, higher complication rates, and increased in-hospital mortality than patients with other ARF diseases [29]. Given the limited number of COVID-19 patients, we could not report separate outcomes for COVID-19 and non-COVID-19 survivors. Another source of bias in our study had to do with the fact that research was conducted on ARF survivors willing to participate in long-term follow-ups. This may have led to the exclusion of the most disabled and most fully recovered patients. The high loss of participants during the follow-up in most studies is testimony to the selection bias in the available evidence. OF note, ECMO protocols and ventilator settings may have a long-term impact on lung function. The number of patients in the studies included who received lung-protective ventilation and how it impacted their long-term outcomes is unknown. Also, ECMO outcomes are significantly dependent on the clinical expertise and volume of patients at a healthcare center, and high-volume centers have reported better ECMO outcomes than other centers [30]. Due to absence of relevant data, we were unable to assess the influence of such variables on patient outcomes. Lastly, spirometry itself has certain limitations and may not be entirely reflective of lung function. Improper maneuvers and ICU-acquired weakness may have impacted the test results.

To the best of our knowledge, our review is the first one to collate data on the impact of ECMO on long-term lung function. Our analysis represents the largest-ever direct comparison of lung function with and without ECMO. Our results have important clinical implications and suggest that long-term pulmonary function in an ECMO survivor may be similar to that of a non-ECMO survivor. Indeed, long-term impaired pulmonary function in a patient with ARF would not be a criterium for selection of ECMO versus other conventional strategies. However, gaining insights into future long-term problems with a new therapeutic modality like ECMO is important. We believe that our findings will help intensivists make informed decisions on using ECMO and its long-term implications. Additionally, only long-term high-quality RCTs can provide better evidence on the matter. However, given the logistical and ethical hurdles of enrolling patients for ECMO trials, it is likely that intensivists will need to rely on biased observational data for a while.

Conclusions

The results of our meta-analysis of a small number of studies with significant selection bias indicates that patients with ARF treated with ECMO may have similar long-term pulmonary function recovery to patients treated with conventional strategies. Further investigations on a larger number of patients focusing on the long-term impact of ECMO are needed to supplement the current evidence.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

BW contributed to the study conception and design. XY and BW performed searches, data collection, and analyses. BW wrote the first draft of the manuscript. All authors read and approved the final manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.