Acute respiratory distress syndrome in patients with cancer: the YELENNA prospective multinational observational cohort study

Abstract

Purpose

Acute respiratory failure is the leading reason for intensive care unit (ICU) admission among critically ill patients with cancer. We aimed to describe the clinical characteristics, risk factors, and outcomes of patients with cancer and acute respiratory distress syndrome (ARDS) and to evaluate associations of venovenous extracorporeal membrane oxygenation (ECMO) with outcomes in the subgroup with severe ARDS.

Methods

We conducted a multinational, prospective, observational cohort study of patients with cancer and ARDS in 13 countries in Europe and North America. The primary endpoint was 90-day mortality.

Results

Among 715 included patients, 73.4% had hematologic malignancies and 26.6% solid tumors; 31.2% had undergone hematopoietic stem-cell transplantation (168 allogeneic). ICU, hospital, and 90-day mortality rates were 55.3%, 70.9%, and 73.2%, respectively. By multivariate analysis, independent predictors of higher 90-day mortality were older age, peripheral vascular disease, severe ARDS at inclusion, acute kidney injury, and ICU admission as a time-limited trial (vs. full code). Conversely, lymphoma was associated with lower 90-day mortality. Among the 322 patients (45.7%) with severe ARDS at inclusion, 90-day mortality was 82.2%; with no difference between patients who received ECMO (n = 58, 18%) and those who did not (82.6% vs. 80.7%, P = 0.89). This finding remained unchanged in a double-adjusted overlap- and propensity-weighted Cox mixed-effects model (adjusted hazard ratio, 1.12; 95% confidence interval 0.65–1.94; P = 0.69).

Conclusion

Patients with cancer and ARDS, particularly severe forms, experience high 90-day mortality, irrespective of ECMO use. These findings suggest a need for nuanced ICU goals-of-care discussions and raise concerns about the generalizability of ECMO guidelines to this population.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00134-025-08113-7.

Take-home message:

Patients with malignancies and ARDS, notably severe ARDS, have a high 90-day mortality rate. Using ECMO in patients with severe ARDS is not associated with lower mortality.

Introduction

Acute respiratory distress syndrome (ARDS) in patients with hematologic or solid malignancies is a life-threatening complication, the prognosis of which was historically considered dismal [1]. Despite recent improvements in both cancer treatments and critical care, the prognosis of ARDS remains substantially worse in patients with cancer [2]. This persistent survival gap raises challenges in determining the appropriate level of therapeutic aggressiveness in these highly vulnerable individuals. Many intensive care unit (ICU) teams are still reluctant to provide full-code care in patients with a higher severity of illness, arguing that the risk of treatment-related toxicity, poor long-term outcomes, and non-beneficial treatments [3] support early palliative care [4, 5].

However, over the past two decades, cancer care has changed profoundly, with the introduction of treatments that are both more effective and better tolerated. The introduction of targeted agents, biotherapies, immune-based treatments, and adoptive cell therapies (including chimeric antigen receptor [CAR] T-cells and checkpoint inhibitors) has markedly improved the outcomes of patients with refractory or relapsed malignancies [6, 7]. In hematology, high-dose chemotherapy followed by autologous or allogeneic hematopoietic stem-cell transplantation (HSCT) has become the standard of care in various settings [8]. These advances have led not only to increased survival rates but also to longer-lasting remissions [9].

ICU physicians must, therefore, reappraise the goals of care in the light of the improved prognosis of cancer [10]. Additionally, there is considerable progress made in critical care interventions. Innovations such as in-ICU chemotherapy [11], optimized infectious-risk control, and the development of structured diagnostic strategies for complex complications have improved survival among critically ill patients with cancer [12]. The expansion of personalized medicine, driven by new insights into disease trajectories and tailored risk–benefit evaluations, has led to substantial changes, notably regarding organ support strategies for patients with ARDS [13].

Nonetheless, ARDS still carries higher mortality rates in patients with cancer [2]. The reasons include baseline immunosuppression, multiple sources of lung injury (including infections, drug toxicities, and disease infiltration), and frequent diagnostic delays [14]. Also, while supportive strategies for ARDS such as lung-protective ventilation and prone positioning are broadly accepted [15], the use of extracorporeal membrane oxygenation (ECMO) remains controversial [16–18]. Although ECMO offers full extracorporeal gas-exchange support for patients with severe, refractory hypoxemia, its risk–benefit ratio in immunocompromised patients with complex cancer-related disease is still unclear, notably due to uncertainty about survival benefits, complications, and long-term quality of life [19].

The primary objective of this multinational prospective observational cohort study was to report 90-day mortality and related risk factors in patients with hematological or solid malignancies who required ICU admission for ARDS. We also analyzed the role of venovenous ECMO (ECMO) therapy on outcome in this cohort.

Patients and methods

Study design and participants

YELENNA was a multinational, prospective observational cohort study performed by the Nine-I (Caring for critically ill immunocompromised patients) study group. This group includes 19 ICUs in 13 countries with extensive experience in the management of critically ill patients with cancer. Institutional review board approval was obtained by each participating ICU in accordance with local ethics regulations. Written informed consent was obtained from each patient, or from a relative if the patient was too ill to provide consent, before study inclusion. The enrolment period was January 1, 2017, to June 30, 2023.

Inclusion criteria were age ≥ 18 years, malignancy (active or diagnosed within the past 5 years), ARDS of any severity (PaO₂/FiO₂ ≤ 300) with intubation [20], and ICU status without treatment limitations (either full code or ICU trial without limitations) [21]. Exclusion criteria were recent intracranial bleeding, post-surgical admission, and do-not-resuscitate or do-not-use-ECMO order.

Data collection

At each participating ICU, the study data were collected by trained study personnel. A standardized electronic case-report form was used for data collection after testing in 24 patients followed by feedback-based corrections (Castor EDC^®, Amsterdam, The Netherlands). Due to local regulations, however, three ICUs collected the data on paper-based forms, and these data were then included in the database. Data that are shown in tables and figures were collected prospectively. SOFA score at admission was calculated as previously reported [22].

Treatment of the study patients

All decisions regarding the management of ARDS including invasive mechanical ventilation, prone positioning, ECMO, and general intensive-care measures (e.g., management of other organ dysfunctions and/or infections, sedation, and cancer-directed treatments) were at the discretion of the physicians in charge according to standard practice in each ICU.

Diagnostic tests to identify the cause of ARDS were chosen based on previous work by the Nine-I study group [1, 15]. ARDS etiologies were determined based on pre-defined criteria in each participating ICU [23]. All diagnoses were then validated by two study investigators (PS and MD).

Study outcomes

The primary outcome was 90-day cumulative survival, with censoring at the last known follow-up. Secondary outcomes included mortality stratified by cancer characteristics and ARDS etiology, the presence of non-respiratory organ dysfunctions, the use of life-support interventions such as ECMO, and identification of risk factors associated with 90-day mortality.

Statistical analysis

We planned to recruit a total of 600 patients with mild-to-moderate ARDS, which may progress to severe ARDS likely in approximately 200 patients based on previous data [1]. Given that an estimated 20 of the later would receive ECMO at the discretion of the treating physician, a propensity score-based approach was considered feasible and used to limit bias of between-group comparison to assess the impact of ECMO when compared to non-ECMO treatment.

No imputation of missing data was performed. Overall, six patients had missing data regarding day-90 mortality; however, data regarding outcome at hospital discharge being available, these patients were included in the final survival analysis. The overall rate of missing data across all variables of patients included in the final analysis was 3.5%.

Data are reported as absolute values with percentages for categorical variables or medians with interquartile intervals for quantitative variables. Comparisons of proportions between groups are made using the χ² or Fisher tests according to data. Comparisons of continuous variables between groups were made using the Wilcoxon rank-sum test.

Cox model was performed to assess factors associated with day-90 mortality. This was built using conditional backward stepwise variable selection process based upon variable influence in the univariate analysis. Critical entry and exit P values were 0.2 and 0.1 respectively. Correlation and interaction were checked within the final model as well as the proportional hazard assumptions.

To explore the association of ECMO on the outcome of patients with severe ARDS at inclusion, a double adjustment strategy was performed. First, to avoid analysis conditioned on future events, it was preplanned to restrict the analysis to patients with > 72 h ICU survival and patients in whom ECMO was initiated within 72 h. A single patient with ECMO performed on day 16 was excluded.

Then a first adjustment of factors associated with ECMO was performed using propensity score weighting analysis. Briefly, overlap weighting was performed. This strategy allows weighting patients from each treatment group with probability to be assigned to the other treatment group [24]. This assigns higher weight to patients with intermediate risk and lower weight to outliers in both treatment groups. The analysis emphasizes the proportion of the population where the most treatment equipoise exists in clinical practice [25]. This model has demonstrated high stability and the ability to obtain precise adjustment in various situations [26]. The propensity score was built using logistic regression according to variables associated with ECMO initiation and likelihood of use of this treatment. These variables were selected using directed acyclic graph (Fig. S4). Covariates included in the model were age, experience of the center with regard to ECMO (tercile of ECMO use), underlying malignancy, and severity of ARDS as defined by PaCO₂ and PaO₂/FiO₂ ratio. Quality of matching was assessed using propensity score distribution before and after weighting and variables distribution after weighting. Influence of ECMO was then assessed by weighted Cox model, adjusted for predefined variables associated with outcome (namely SOFA score, platelet count and preexisting cardio-vascular comorbidity) and for center effect.

Seven sensitivity analyses were performed. First, a preplanned analysis comparing the benefit of ECMO in the subgroup of patients with severe ARDS at inclusion and the theoretical indication of ECMO (severe ARDS (PaO₂/FiO₂ ≤ 100) and Murray score ≥ 3.0 or uncompensated hypercapnia (pH < 7.2) and prone positioning for at least 6 h) was performed using double adjustment described above. In addition, six post-hoc sensitivity analyses were performed forcing variables which were believed to influence outcome, namely performance status (ECOG and Frailty score), hematopoietic stem cell transplantation recipients, unknown diagnosis as etiology of acute respiratory failure (ARF), neutropenia at the onset of ARDS, uncontrolled malignancy, and lactate levels at ARDS onset. No sensitivity analyses were performed for hematologic malignancy or solid tumor. These variables had already been adjusted for in the covariate balancing (Fig. S3) and, hence, assessing this specific point in the doubly adjusted model would mean adjusting twice for the same variable.

Statistical significance was considered using two-sided tests with a critical alpha risk of 0.05.

Statistical analyses were performed using R version 3.6.2 (R Foundation for Statistical Computing), including ‘survival’, ‘survey’, and ‘WeightIt’ packages.

Results

Patients

A total of 776 patients with malignancies and ARDS were enrolled across 13 countries over the study period (Fig. S1). After exclusion of 61 patients due to missing 90-day mortality data, 715 patients were included in the descriptive analysis (Fig. 1). Table 1 reports the baseline clinical features of the whole population analysis cohort as well as mortality after exclusion of six patients missing time to outcome. The median age was 61 [53–69] years and 60.2% of patients were male. Out of the 715 patients, 525 (73.4%) had hematological malignancies including 168 (168/715, 23.5%) with allogeneic HSCT and 55 (55/715, 7.7%) with autologous HSCT. The remaining 190 (26.6%) patients had solid malignancies. At least one comorbidity was present in 49.8% of patients. Chemotherapy was administered to 15.4% of patients during their ICU stay.

Fig. 1.

Patient flowchart. ARDS acute respiratory distress syndrome, ECMO extracorporeal membrane oxygenation. Descriptive analyses are performed in 715 patients after exclusion of 61 patients due to missing 90-day survival status; the whole population cohort analysis is performed in 709 patients after exclusion of a further six patients due to missing time to event status

Table 1.

Baseline characteristics of all patients according to 90-day mortality

	All patients	90-day survivors	90-day non-survivors	P value
n =	709	190	519
Male gender, n (%)	441 (62.2)	110 (57.9)	331 (63.9)	0.17
Age, median [IQR]	61 [53–69]	59 [51–66]	62 [53–69]	0.02
Hematologic malignancy, n (%)				0.01
Acute myeloid leukemia	191 (26.9)	41 (21.6)	150 (28.9)
Acute lymphoblastic leukemia	48 (6.8)	7 (3.7)	41 (7.9)
Chronic lymphocytic leukemia	28 (3.9)	9 (4.7)	19 (3.7)
Chronic myeloid leukemia	19 (2.7)	4 (2.1)	15 (2.9)
Non-hodgkin lymphoma	134 (18.9)	44 (23.2)	90 (17.3)
Myeloma	51 (7.2)	18 (9.5)	33 (6.4)
Hodgkin's lymphoma	9 (1.3)	1 (0.5)	8 (1.5)
Other	92 (13.0)	18 (9.5)	74 (14.3)
Solid tumor, n (%)	190 (26.8)	64 (33.7)	126 (24.3)	0.02
Hematopoietic stem cell transplantation, n (%)				0.01
Allogeneic	168 (23.7)	30 (15.8)	138 (26.6)
Autologous	54 (7.6)	16 (8.4)	38 (7.3)
Acute graft-versus-host disease, n (%)				0.006
Active	17 (2.4)	0 (0.0)	17 (3.3)
Controlled	54 (7.6)	9 (4.7)	46 (8.9)
Chronic graft-versus-host disease, n (%)				0.34
Active	15 (2.1)	3 (1.6)	12 (2.3)
Controlled	20 (2.8)	8 (4.2)	12 (2.3)
Progressive disease, n (%)	47 (6.6)	6 (3.2)	41 (7.9)	0.04
ECOG score, median [IQR])	2 [1–2]	1 [1–2]	2 [1–3]	0.04
Clinical frailty scale, median [IQR]	3 [2–6]	3 [2–5]	4 [2–6]	0.004
No comorbidity, n (%)	255 (36)	98 (51.6)	257 (49.5)	0.69
SOFA score, median [IQR]	11 [8–14]	10 [7–12]	12 [9–14]	< 0.001
ARDS severity, n (%)				< 0.001
Mild	105 (14.8)	37 (20.0)	68 (13.3)
Moderate	272 (38.4)	91 (49.2)	181 (35.3)
Severe	321 (45.3)	57 (30.8)	264 (51.5)
ARDS etiology, n (%)				0.15
Non-bacterial pneumonia, n (%)	252 (35.5)	55 (28.9)	197 (38.0)
Bacterial pneumonia	130 (18.3)	37 (19.5)	93 (17.9)
Extra-pulmonary sepsis	60 (8.5)	14 (7.4)	46 (8.9)
Specific infiltration	21 (3.0)	4 (2.1)	17 (3.3)
Aspiration	20 (2.8)	7 (3.7)	13 (2.5)
Toxicity	5 (0.7)	3 (1.6)	2 (0.4)
Other	35 (4.9)	12 (6.3)	23 (4.4)
Unknown	186 (26.2)	58 (30.5)	128 (24.7)
Vasopressors, n (%)	587 (82.8)	151 (79.5)	436 (84.0)	0.66
Renal replacement therapy, n (%)	249 (35.1)	41 (21.6)	208 (40.1)	< 0.001
Duration of IMV (days), median [IQR]	7.0 [3.0–13.0]	7.5 [3.0–14.0]	7.0 [3.0–13.0]	0.15
ICU LOS (days), median [IQR]	9.0 [4.0–18.0]	13 [7–25]	8 (3.5–15)	< 0.001
ICU mortality, n (%)	392 (55.3)	0 (0.0)	392 (75.5)	< 0.001
Hospital mortality, n (%)	523 (73.8)	20 (10.5)	503 (96.9)	< 0.001

Descriptive analyses were performed in 715 patients after exclusion of 61 patients due to missing 90-day survival status; the whole population cohort analysis was performed in 709 patients after exclusion of further six patients due to missing time to event status

ARDS acute respiratory distress syndrome, IQR interquartile range, ECOG Eastern Cooperative Oncology Group, ICU intensive care unit, LOS length of stay, SOFA Sequential Organ Failure Assessment

At study inclusion, ARDS severity was classified as mild in 15.5% of patients, moderate in 38.8%, and severe in 45.7%. Notably, 55.0% of patients met the criteria for severe ARDS at some point during their ICU stay. The main causes of ARDS were pneumonia (54.0%), extrapulmonary sepsis (8.4%), and malignant infiltrates (2.9%). No cause was identified in 26.3% of patients. During the ICU stay, vasopressors were used in 82.8%, and renal replacement therapy in 35.1%.

The most common complications were bleeding of any severity (74.7%), septic shock (45.3%) [27], ventilator-associated pneumonia (29.1%), and invasive fungal infections (29.1%). The median lengths of mechanical ventilation and ICU stay were 7 [3–13] and 9 [4–18] days, respectively. ICU, hospital, and 90-day mortality rates were 55.3%, 70.9%, and 73.2%, respectively. Further characteristics and unadjusted comparisons between survivors and non-survivors are given in Table 1 and Fig. S2.

Factors associated with 90-day mortality

The mixed effects Cox model for 90-day mortality identified the following factors independently associated with higher mortality: older age tercile (56–66 years, adjusted hazard ratio [aHR]: 1.30; 95% confidence interval [95% CI] 1.04–1.62; P = 0.019; and 66–87 years, aHR: 1.63; 95% CI 1.3–2.06; P < 0.001 vs. 19–56 years as the reference); peripheral vascular disease (aHR: 1.83; 95% CI 1.21–2.77; P = 0.004); severe ARDS at inclusion (aHR: 1.76; 95% CI 1.33–2.32; P < 0.001); acute kidney injury (aHR, 1.38; 95% CI 1.13–1.69; P = 0.002); and admission for an ICU trial (aHR, 1.41; 95% CI 1.07–1.85; P = 0.014 vs. full code status). Lymphoma was independently associated with lower mortality (aHR, 0.75; 95% CI 0.57–0.97; P = 0.029; vs. acute leukemia as reference; Table 2).

Table 2.

Cox model assessing day-90 mortality in the overall population

	Adjusted hazard ratio (95% CI)	Adjusted P value
Age tercile: ref. = [19–56] years
[56–66]	1.3 (1.04–1.62)	0.02
[66–87]	1.63 (1.3–2.06)	< 0.001
Hematologic malignancy: ref. = leukemia
Lymphoma	0.75 (0.57–0.97)	0.03
Myeloma	0.7 (0.47–1.02)	0.06
Other	0.87 (0.7–1.07)	0.18
Peripheral vascular comorbidity	1.83 (1.21–2.77)	0.004
Goals of care: ICU trial (ref. = full code)	1.41 (1.07–1.85)	0.01
ARDS severity (ref. = mild)
Moderate	1.14 (0.86–1.51)	0.36
Severe	1.76 (1.33–2.32)	< 0.001
Acute kidney injury (ref. = no AKI)	1.38 (1.13–1.69)	0.002

ARDS acute respiratory distress syndrome, CI confidence interval, ref. reference

In the sensitivity analysis, when forced in the final model, higher clinical frailty score (CFS) quartiles were associated with mortality (3–6: aHR, 1.6; 95% CI 1.24–2.07; P < 0.001; 6–9: aHR, 1.56; 95% CI 1.11–2.18; P = 0.01]) but did not change the final model. When unknown etiology of ARDS or HSCT were forced into the model, they were neither associated with outcome nor did they change the final model.

Subgroup of patients with severe acute respiratory distress syndrome

We analyzed the subgroup of 322 patients with severe ARDS at inclusion. They had a median age of 59 [50.0–67.38] years and 65.4% were male. Hematological malignancies predominated (84.2%). Allogeneic and autologous HSCT recipients contributed 24.5% and 6.8% of patients with severe ARDS. The comorbidity burden was similar to the overall population of patients with non-severe ARDS, as was the severity of the acute illness (median SOFA score, 11 [8–14]). The median PaO₂/FiO₂ ratio was 81 [63.5–93] and the median PaCO₂ was 47 [40–59] mmHg.

The treatment followed recent evidence-based recommendations with protective ventilation (Fig. S5). Prone positioning was used in 73% of patients with severe ARDS.

Venovenous ECMO was used in 58/322 (18%) patients. Patients treated with ECMO had a higher median platelet count when compared to patients who did not receive ECMO (95 [41–232] vs. 45 [20–127] × 10⁹/L; P = 0.001). The 90-day mortality rate was not different between patients with and without ECMO (82.6% vs. 80.7%; P = 0.89). This finding remained unchanged with the doubly-adjusted mixed-effects model for time to death within 90 days (aHR with ECMO 1.12; 95% CI 0.65–1.94; P = 0.69]; Table 3, Fig. 2 and Figure S3).

Table 3.

Doubly adjusted Cox model in patients with severe acute respiratory distress syndrome

	Adjusted hazard ratio (95% CI)	Adjusted P value
SOFA score quartile: reference = [2–8]
[8–11]	0.83 (0.42–1.64)	0.59
[11–14]	0.77 (0.36–1.65)	0.51
[14–21]	0.95 (0.39–2.35)	0.92
Peripheral vascular comorbidity: yes (ref. = no)	2.8 (0.78–10.02)	0.11
Platelets quartile: reference = [0–21] 109/L
[21–45]	0.96 (0.44–2.12)	0.93
[54–142]	1.29 (0.55–3.03)	0.55
[142–609]	0.63 (0.27–1.51)	0.30
ECMO (ref. = no ECMO)	1.12 (0.65–1.94)	0.69

SOFA Sequential Organ Failure Assessment, ARDS acute respiratory distress syndrome, CI confidence interval, ECMO extracorporeal membrane oxygenation

Fig. 2.

Cumulative adjusted and weighted survival according to ECMO status in the 322 patients with severe ARDS. ECMO extracorporeal membrane oxygenation

In sensitivity analyses, the following variables were forced in the final model individually: HSCT status, unknown etiology of ARDS, CFS score, ECOG performance status, vascular comorbidities, progressive status of the underlying malignancy, neutropenia, lactate level at baseline, and time of ICU admission (Table S3). The addition of these variables was not associated with outcome, did not change the impact of ECMO, and did not change the final model.

Finally, as preplanned, the same analyses were conducted in the subgroup of 272 out of 322 patients (84.5%) with severe ARDS at inclusion who met the predefined theoretical ECMO eligibility criteria. The results were consistent with the overall findings, showing no significant unadjusted association between ECMO use and 90-day mortality (Table S1). After applying a double adjustment, ECMO remained unassociated with 90-day mortality in this subgroup (adjusted hazard ratio [aHR] 1.09; 95% CI 0.60–1.97; Table S2).

Discussion

In this large multinational cohort of critically ill patients with active or recently treated cancer and ARDS, the majority had hematological malignancies, and severe ARDS was common already at inclusion or developed during the ICU stay. Most cases were triggered by pneumonia, and patients frequently required organ support, including vasopressors and renal replacement therapy. Complications such as bleeding, septic shock, and invasive fungal infections were frequent. Overall mortality was high, with nearly three-quarters of patients dying within 90 days. Factors independently associated with increased mortality included older age, peripheral vascular disease, severe ARDS at inclusion, acute kidney injury, and ICU admission under a trial status. In contrast, lymphoma was associated with better outcomes compared to acute leukemia. Among patients with severe ARDS, around one in five received ECMO; however, ECMO use was not associated with improved survival, even after rigorous adjustment and in those meeting theoretical eligibility criteria.

Acute respiratory failure is the most common reason for medical ICU admission in patients with malignancies. Non-invasive ventilation and high-flow nasal oxygen therapy were not superior to standard oxygen in randomized controlled trials, and intubation is often necessary [28–30]. Nevertheless, survival rates have improved substantially in recent decades [31, 32]. Possible reasons include better selection whether and when to admit to the ICU, avoidance of delayed intubation, improvements in infection diagnosis and antimicrobial therapy, as well as advances in critical care [21, 33–35].

Mortality was very high in our cohort, particularly across those with severe ARDS. The mortality rate in our subgroup with severe ARDS is consistent with a large retrospective study of patients treated in 1990–2011 in France and Belgium [1]. A secondary analysis of data from the EFRAIM trial in immunocompromised patients with any severity of ARDS showed a 56.3% hospital mortality, the risk factors being no identified cause of ARDS, use of vasopressors, and renal replacement therapy but not, interestingly, the severity of ARDS [15].

Several factors may explain the high mortality in our cohort. First, the proportion of patients with severe ARDS was markedly higher than in previous investigations [1]. Second, among the patients with severe ARDS, over 80% met pre-defined theoretical ECMO criteria, indicating extremely severe illness. Third, failure to identify the cause of ARDS, a known factor of poor prognosis, occurred in a quarter of the patients [14, 34]. Fourth, the median CFS score and ECOG performance score suggest a liberal ICU admission policy in patients with functional impairment. In the sensitivity analysis, as expected, worse CFS score values were associated with higher mortality. Finally, nearly a fourth of our patients had received allogeneic HSCT, a known poor-prognosis factor, although not associated with mortality in our analysis.

Our findings have several clinical implications. On the one hand, the optimal management for patients with cancer and severe ARDS clearly requires further attention to evaluate strategies to improve outcome. On the other hand, outcomes were better in some subgroups, notably in patients with mild or moderate ARDS. In these patients, the combination of non-HSCT status, absence of acute kidney injury, and age below 70 years was associated with a 90-day mortality of about 50%. These patient characteristics, together with frailty, comorbidities, and ICU-trial status, should be carefully considered in each individual patient when determining the appropriate level of aggressiveness of critical care.

ECMO was not associated with lower mortality in patients with severe ARDS. Based on the meta-analyses of two randomized controlled trials [36, 37], ECMO is currently recommended in patients with severe ARDS. Importantly, the trials were done in experienced high-volume ECMO centers that adhered to protocol-defined inclusion criteria and management strategies with a low representation of immunocompromised patients. The few cohort studies of patients with cancer and severe ARDS treated with venovenous ECMO produced conflicting results regarding outcomes [38–40]. Outcomes were very poor in patients within one year of allogeneic HSCT [18, 41]. In a retrospective multicenter observational study of 203 immunosuppressed patients, six-month survival of patients with hematologic or solid malignancies treated with ECMO was only 24% and 20.5%, respectively [16]. Similarly, a multicenter retrospective study of 297 patients with severe ARDS and malignancies (including 54% with solid tumors) who were treated with ECMO showed a 60-day survival rate of 26.8%, which was similar to that in a propensity-score-matched group of patients who did not receive ECMO [17]. In the ECMO group, risk factors for mortality were low platelet count, high lactate level, and progressive or newly diagnosed malignancy.

We did not identify subgroups of patients likely to benefit from ECMO therapy. The sensitivity analyses also showed no mortality difference supporting the robustness of our findings. Bleeding was common in the overall cohort, but neither bleeding nor infections were more common with vs. without ECMO. Of note, among patients with severe ARDS, platelet counts were lower in the subgroup that did not receive ECMO. This finding may be due to potentially higher extrapulmonary severity or characteristics of the malignant disease and could constitute an unmeasurable selection bias. Furthermore, we might not have been able to adjust for additional unmeasured confounders in our analysis. In all, these factors may have influenced clinicians’ a priori beliefs, ECMO indications and, consequently, our results cannot be taken as firm evidence that ECMO is not effective. An international consensus statement on ECMO indications was issued in 2023 but was confined to HSCT recipients [18]. Older age, comorbidities, functional impairments, uncontrolled malignancy, multiple non-pulmonary organ failures, unidentified ARDS etiology, and non-full-code ICU status argue against the use of ECMO. Nonetheless, the overall body of evidence available to date does not allow advising for or against the use of ECMO in patients with malignancies and severe ARDS. Our findings raise doubts on whether the inclusion of ECMO into routine clinical care would be appropriate at large scale. ECMO decisions in these patients should remain highly individualized. The risk of prolonging death should be discussed among the healthcare providers and with the patient and family. In this context, the prospect of impaired long-term survival, functional disability and impairment in emotional, physical, and general well-being has to be considered [42–44].

The main limitation of our study is the observational design and reason to initiate ECMO. Although the propensity-score analyses can expect to minimize bias, especially with overlap weighting which accounts for extreme weights, unadjusted confounders are to be expected and additional studies are needed to confirm our findings. Also, a potential lack of statistical power to detect an effect and an imprecision of the effect estimate must be acknowledged. The prospective design resulted in a very low proportion of missing data. We also only considered patients with available data on ECMO use and 90-day survival. While the clinical management was not standardized across ICUs in different countries, which practices may vary, the multinational study design supports the generalizability of our findings. Even though diagnostic tests to identify the cause of ARDS were chosen based on previous work by our group [1, 15], we found a two-fold increase in the rate of unknown etiology of ARDS when compared to previous studies in similar populations. This could be explained in part by generally recognized limitations in immunocompromised patients, such as neutropenia or prior antibiotic use, and should be acknowledged when interpreting our data. Besides, we did not report information on single pathogens. Furthermore, the participating ICUs followed recommendations for ARDS management, as shown by the rates of use of protective ventilation and prone positioning. Nonetheless, we cannot rule out that the lack of standardization of ECMO use may have masked outcome differences in specific patient subgroups. Finally, we included only patients receiving invasive mechanical ventilation and excluded patients with do-not-use-ECMO orders.

Conclusions

Patients with malignancies and ARDS, notably severe ARDS, had a high 90-day mortality rate. Using ECMO in patients with severe ARDS was not associated with lower mortality, although this finding should be interpreted in the light of the observational design, without standardization of ECMO indications or other components of management. These findings invite caution when considering ECMO in patients with malignancies and severe ARDS. General recommendations about ECMO indications in ARDS may not apply to this specific population. Further research is needed to determine whether specific subpopulations of patients with malignancies and severe ARDS can benefit from ECMO.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P.S., E.A., T.S., and M.D. designed the study; M.D., P.S., E.A., and N.B. analyzed the data; P.S., E.A., and M.D. drafted the manuscript. All other authors contributed significantly to the acquisition and interpretation of data, revised the manuscript for important intellectual content, and approved the final version.

Funding

Open access funding provided by Medical University of Vienna. This study was supported by a research grant of the European Society of Intensive Care Medicine (ESICM).

Data availability

Data will be shared upon reasonable request tot he corresponding author.

Declarations

Conflicts of interest

PS reports speaker fees from Fresenius Medical and Getinge; MD reports payments to his institution made by Gilead; TS reports speaker fees from Baxter, CSL Behring, Fresenius Medical, Getinge, and Mitsubishi Pharma; VL reports being the treasurer of a research group (GrrrOh), which received payments from Alexion, Fisher Paykel, Gilead, Sanofi, and Shionogi; all other authors did not report any conflicts of interest. MD and VM are Section Editors for Intensive Care Medicine. They have not taken part in the review or selection process of this article.

Ethics approval

Institutional review board approval was obtained by each participating ICU in accordance with local ethics regulations. Written informed consent was obtained from each patient, or from a relative if the patient was too ill to provide consent, before study inclusion.