Extracorporeal membrane oxygenation as a rescue therapy for leukaemic children with pulmonary failure


Bernhard Meister, MD, Department of Paediatrics, Innsbruck Medical University, Anichstrasse 35, 6020 Innsbruck, Austria. E-mail: bernhard.meister@i-med.ac.at


In patients with leukaemia, acute respiratory distress syndrome (ARDS) secondary to intensified chemotherapy-induced immunosuppression is a devastating disorder resulting in high morbidity and mortality. Compared to standard indications for extracorporeal membrane oxygenation (ECMO), cytopenia further increases the risks of infection and bleeding. We describe the use of ECMO in four children with ARDS and leukaemia. Two patients (50%) survived, pulmonary function recovered and they are in prolonged first remission. The two other patients died from ARDS and pulmonary leukaemic infiltration. Although ECMO support is a high-risk setup for nosocomial infection we observed no additional septic episodes. All patients had a highly increased demand for packed platelet and red blood cell transfusions. This increased demand and unmanageable chronic bleeding into both lungs in one patient were probably caused by a combination of coagulopathy from the primary illness, the use of anticoagulants, chemotherapy-induced cytopenia, and a reduced survival rate of platelets and red cells due to permanent contact to foreign surface. We concluded that ECMO is a supportive tool to reduce the incidence of early death, treatment-related mortality and, ultimately, to improve overall survival in childhood leukaemia.

The lung is one of the organs most severely affected by complications during the course of haematological disorders (Poletti et al, 2000). Therefore patients with leukaemia may develop acute respiratory distress syndrome (ARDS), a devastating disorder of overwhelming pulmonary inflammation and hypoxia, resulting in high morbidity and mortality. As more than 80% of children with lymphoblastic leukaemia (ALL) are long-term survivors, infections and their complications remain a major cause of morbidity and mortality after steroid and intensive combination-chemotherapy. Chemotherapy and steroids often induce prolonged periods of marrow aplasia, leading to neutropenia and defective immunity, rendering patients more susceptible to bacterial (Pound et al, 2008), viral and fungal infections (Grigull et al, 2003). In a population-based analysis of the Austrian Berlin-Frankfurt-Münster (BFM) study group, we found a progressive reduction in death rates of children with ALL that may be explained by treatment in experienced, specialized haemato-oncological centres with improved supportive and intensive care (Prucker et al, 2009). The estimated 10-year cumulative incidence of death significantly decreased from 6 ± 1% to 2 ± 1%. Sixty-eight percent of the patients died from infectious complications. In contrast, such progress lags behind in acute myeloid leukaemia (AML) with cure rates of approximately 40–50%. Despite some advances in treatment and supportive care of this disease during the last years, infections remain a major and substantial cause of therapy-associated morbidity and death (Slats et al, 2005; Brown et al, 2006; Laws et al, 2007). For instance, 104 patients (11·5%) enrolled into the clinical trials AML-BFM 93 and AML-BFM 98 died shortly after diagnosis or as a result of treatment-related complications (Creutzig et al, 2004). In juvenile myelomonocytic leukaemia (JMML), pulmonary leukaemic cell infiltration and subsequent pulmonary failure also remains an unresolved problem and a major cause of fatal casualties in this disease.

To reduce the incidence of early death and treatment-related mortality and, ultimately, to improve overall survival in children with leukaemia, regimens for better management of complications have to be assessed. Extracorporeal membrane oxygenation (ECMO) could be such a technique to improve survival in children with leukaemia and pulmonary failure. ECMO is an established treatment in children and adults for severe haemodynamic or respiratory failure (Brown et al, 2006), refractory to standard medical treatment but not in children with a malignant disease. As ECMO technology and expertise has substantially increased over the last decade, with less morbidity and mortality, criteria for the application of ECMO may also be extended to ARDS-patients with leukaemia. In the present study, we report on our experience of veno-venous and arterio-venous ECMO in leukaemic children with respiratory failure. We also discuss the dilemmas that arise in considering this innovative therapy for critically ill children when there is little data to support its use.


Patient population

Four children with leukaemia received ECMO treatment at the Paediatric Intensive Care Unit, Medical University of Innsbruck, Austria. All of the cases were reviewed at diagnosis of leukaemia by the national study centre in Vienna and patients were treated according to the ALL-BFM 2000, the AML-BFM 2004 and the European Working Group on myelodysplastic syndrome in childhood (EWOG-MDS) 2006 protocols after informed consent from the patient’s parents. Studies were conducted in accordance with the declaration of Helsinki and approval was delivered by the relevant ethic committee. Table I shows the type of malignancy, clinical characteristics, and the paediatric logistic organ dysfunction (PELOD) score (Leteurtre et al, 2003) for each patient. This score measures the severity of organ dysfunction syndrome in paediatric care units, using a method of assessing physiological instability for paediatric intensive care unit patients, Hospitality Course and Mortality rate.

Table I.   Clinical characteristics, organ dysfunction, clinical course, complications and outcome of extracorporeal membrane oxygenation.
Patient no.1234
  1. c-ALL, common B-cell precursor acute lymphoblastic leukaemia; APL, acute promyelocytic leukaemia; T-ALL, T cell acute lymphoblastic leukaemia; JMML, juvenile myelomonocytic leukaemia; ECMO, extracorporeal membrane oxygenation; PELOD, paediatric logistic organ dysfunction; PICU, paediatric intensive care unit; G-CSF, granulocyte colony-stimulating factor; n.d., not done.

Type of malignancyc-ALLAPLT-ALLJMML
Age (years)1513150·5
Administration of chemotherapy
 Within the past 4 weeksNoYesYesNo
 During ECMO supportNoNoNoYes
PELOD on day 1 of ECMO support11132
PELOD on day 7 of ECMO support12412
PELOD on discharge from PICU11n.d.n.d.
Length of stay in PICU (days)45391729
Duration of ECMO support (days)9151517
G-CSF stimulation during ECMO support (days)0950
Ventilation days40341629
Number of vasopressors used during ECMO support1321
Transfused Packed red blood cells during ECMO support14143110
Transfused apherisis platelet concentrates during ECMO support3275417
Infection before ECMOYesYesYesYes
Additional infection during ECMONoNoNoNo
Bleeding complicationsNoNoYesNo
Survival at PICU dischargeAliveAliveDeceasedDeceased
Survival at 2 years post PICUAliveAliveDeceasedDeceased


Patient 1

A 15-year-old female with newly diagnosed common B-cell precursor ALL was treated according to the ALL-BFM 2000 protocol. Following minimal residual disease criteria, she was assigned to the medium risk group. Four weeks after starting standard reinduction therapy with dexamethasone (10 mg/m2) on days 1–21, weekly vincristine (1·5 mg/m2) on days 8, 15, 22 and 29, weekly doxorubicin (30 mg/m2) on days 8, 15, 22 and 29 and Escherichia coli asparaginase medac (CASP; Kyoma, Hakko, Kyogo, Japan, 1000 U/m2) on days 8, 11, 15 and 18 she developed multiple nodular pulmonary lesions diagnosed by computed tomography (CT). Concomitant cytopenia required empiric therapy with Voriconazole (Vfend®; Pfizer Ltd, Sandwich, Kent, UK). Following disease progression, a CT-guided biopsy revealed zygomycosis. This aggressive and rapidly progressive opportunistic fungal infection in immuno-compromised patients is caused by Rhizopus species. The mortality of zygomycosis is very high, especially for disseminated disease. Antifungal therapy was changed to liposomal amphotericin but further disease progression required partial resection of the lower lobe of the right lung and wedge resection of the upper lobe of the right lung. Postoperatively, she required high levels of conventional positive pressure ventilation, and high frequency oscillation. Subsequently, ARDS was diagnosed based on physiological and radiological criteria. Thirteen days postoperatively, veno-venous ECMO for pulmonary support had to be initiated. At initiation of ECMO this patient did not suffer from neutropenia or thrombocytopenia, but was immuno-compromised after long-term chemotherapy and steroid medication. However, after the initiation of ECMO, platelet counts (Fig 1) and leucocyte counts (Fig 2) decreased and the patient needed 3 packed platelet and 13 red blood cell transfusions. After 10 d, pulmonary gas exchange improved; the patient was decanulated but needed conventional ventilation for an additional 20 d. Thereafter, she recovered and is still in first complete remission five years after diagnosis of ALL. No major bleeding complication or additional infection was observed during ECMO support.

Figure 1.

 Course of platelet counts in days after initiation of ECMO in Patients 1–4.

Figure 2.

 Course of leucocyte counts in days after initiation of ECMO in Patients 1–4.

Patient 2

A 13-year-old male patient with newly diagnosed acute promyelocytic leukaemia (APL) was treated according to the AML-BFM 2004 protocol. Ten days after the fourth course of chemotherapy (six doses of high-dose Cytarabine 3 g/m2, four doses of Etoposide 125 mg/m2 per day and Cytarabine 40 mg i.t.) and the fourth course of all-trans retinoic acid treatment, the patient developed a septic shock syndrome caused by Streptococcus oralis and consequently ARDS. At the initiation of veno-venous ECMO, the patient was severely neutropenic (leucocytes <0·1 × 109/l), thrombocytopenic and anaemic. The course of platelet counts is shown in Fig 1, the course of leucocyte counts in Fig 2. Nine doses of granulocyte colony-stimulating factor (G-CSF, Neupogen®, Amgen Europe B.V., Breda, The Netherlands, 5 μg/kg per day) were necessary to stimulate recovery from severe neutropenia 7 d after initiation of ECMO. During the 15-d treatment with ECMO, the patient received 5 fresh frozen plasma concentrates, 14 packed red blood cells, 27 apheresis platelet concentrates, and 11 cryoprecipitates. Despite prolonged thrombocytopenia and prolonged severe neutropenia whilst on ECMO he developed no severe bleeding or additional infection. Ventilation time after discontinuation of ECMO was 19 d. This patient is still in first complete remission 4 years after diagnosis of APL.

Patient 3

This 15-year-old male patient had T cell (T-III) ALL. Due to poor response to prednisone on day 8 (22·6 × 109 blasts/l peripheral blood) he was treated according to the high-risk arm of the ALL-BFM 2000 protocol with induction therapy and six high-risk intensified consolidation blocks as previously published (Moricke et al, 2008). Following this therapy, histologically proven pulmonary mucormycosis was diagnosed and further chemotherapy had to be postponed for three months. After 2 months of combined antimycotic therapy with liposomal amphotericin (Abelcet®, Cephalon GmbH, Martinsried, Germany) and posaconazole (Noxafil®, SP Europe, Bruxelle, Belgium), one residual nodular lesion was resected thoracoscopically. Three weeks after starting reinduction therapy with dexamethasone (10 mg/m2) on days 1–21, weekly vincristine (1·5 mg/m2) on days 8, 15 and 22, doxorubicin (30 mg/m2) on days 8 and 15, and pegylated Asparaginase (1000 U/m2) on day 8, this patient developed E. coli sepsis, bilateral pneumonia and multiple cerebral lesions. After he developed ARDS, which failed to respond to conventional ventilation and oscillation, veno-venous ECMO was commenced. Five doses (5 μg/kg per day) of G-CSF (Neupogen®) were administered for treatment of neutropenia. During the 15-d ECMO support this patient required massive blood component transfusions including one fresh frozen plasma, 31 packed red blood cells, 54 apheresis platelet concentrates and 4 cryoprecipitates. Nevertheless his clinical course was complicated by apparent episodes of pulmonary bleeding and no improvement of pulmonary gas exchange. He died 15 d after initiation of ECMO treatment because of unmanageable chronic bleeding into both lungs. Postmortem examination revealed severe haemorrhagic ARDS of both lungs with haemorrhagic destruction of lung tissue (Fig 3A, B).

Figure 3.

 (A) Lung tissue with severe haemorrhage in Patient 3. Chromotrope aniline blue (CAB) stain, 100×. (B) Complete haemorrhagic infarction with dimly visible lung tissue in Patient 3. Haematoxylin and eosin (H&E), 40×.

Patient 4

A 6-month-old boy was admitted to the intensive care unit with dry cough, dyspnoea, hepatosplenomegaly, leucocytosis (24·3 × 109/l), monocytosis (23·5%, absolute monocyte count 5·7 × 109/l), thrombocytopenia (51 × 109/l) and 7·8% HbF. Due to respiratory failure he was immediately intubated and placed on mechanical ventilation. Bone marrow aspiration showed 9·5% blasts and clonal monosomy 7. The boy developed ARDS, triggered by a respiratory syncytial virus infection and possibly by pulmonary leukaemic infiltration. As conventional ventilation and oscillation failed, he was placed on arterio-venous ECMO. When the diagnosis of JMML was established by detection of monosomy 7, he was treated with azacitidine (100 mg/m2 intravenously over an hour intravenously on five consecutive days) as recently published (Furlan et al, 2009) and additional cytosine-arabinoside (75 mg/m2 intravenously for four consecutive days). Despite good haematological response and resolution of peripheral blood monocytosis, pulmonary gas exchange did not improve and the boy died. At postmortem examination both lungs showed solid consistency caused by severe diffuse alveolar damage (Fig 4A) and focal residual infiltrates of the JMML (Fig 4B) on light microscopy.

Figure 4.

 (A) Patient 4: severe diffuse alveolar damage with hyaline membranes and clots. H&E, 100×. (B) Residual disease of juvenile myelomonocytic leukaemia in Patient 4. H&E, 200×.


Compared with standard indications for ECMO, the treatment of patients with leukaemia, chemotherapy-induced cytopenia and respiratory failure poses additional medical and ethical challenges. Firstly, in immuno-compromised severely neutropenic paediatric patients the risk of ECMO-induced septic episodes is further increased. Even in immuno-competent patients, ECMO support is a high-risk procedure for nosocomial infection and septic episodes (Brown et al, 2006). Brown et al reported that over the entire study period, 39 out of 215 children had 47 septic episodes, with a rate of 24·9 per 1000 ECMO d. Three of our patients similarly developed severe bacterial and fungal infections at the end of scheduled intensive chemotherapy, during a period when patients are seriously immuno-compromised. Therefore, we had great reservations at the initiation of ECMO. Patient 2 was severely neutropenic at initiation of ECMO, but recovered 6 d after stimulation with G-CSF. In Patient 4, we started therapy with azacitidine and cytosine-arabinoside 1 d after initiation of ECMO support, when clonal monosomy 7 emerged. Despite our concerns about additional infectious complications in these immuno-compromised patients, none were observed during ECMO support.

Secondly, there is a high risk of bleeding and a highly increased demand for transfusions during EMCO. Chemotherapy reduces myelopoiesis on one hand and on the other hand consumption of blood cells is increased because of permanent contact to foreign surfaces, coagulopathy and simultaneous need of anticoagulation. In ARDS, the pulmonary endothelium is a main target of circulating cells and humoral mediators under injury, resulting in deregulated coagulation and fibrinolysis (Maniatis et al, 2008). Furthermore, the need for anticoagulation leads to haemorrhage with increased requirement for blood transfusions. During this highly specialized assistance, children are anticoagulated with heparin and require regular monitoring of coagulation parameters and platelet count (Gibson et al, 2004). In fact, ECMO is associated with severe bleeding, even in patients with initially normal platelet counts. The combination of coagulopathy from the primary illness, the haemostatic defects associated with ECMO and haemodilution contribute to a high risk of intracranial haemorrhage (Gibson et al, 2004). Transfusion guidelines for neonates and older children suggest that platelet counts should be maintained above 100 × 109/l (Gibson et al, 2004). After initiation of ECMO, all four patients needed the transfusion of multiple blood products as no alternative to these repeated transfusions exists. Even with massive transfusions this goal is hardly achievable in leukaemic patients during and after chemotherapy. In fact, in Patient 3, the combination of anticoagulation treatment, low platelet counts and ARDS resulted in severe bleeding into both lungs and consequently this patient died because of haemorrhagic ARDS. This high demand of blood products potentially increases the risk of transfusion-related acute lung injury (TRALI) (Silliman & Kelher, 2005), a life-threatening adverse effect of transfusion, which may induce further lung damage. The threshold model of TRALI proposed by Bux and Sachs (2007) suggests that a certain threshold must be overcome to induce a TRALI reaction. In an individual ‘at risk’, such as a patient with a comorbidity, related activation of the pulmonary endothelium, the transfusion of mediators with a relatively low-neutrophil-priming activity will be sufficient to overcome the threshold. Active infection, recent surgery and massive transfusions also predispose the patient for the development of acute lung injury and TRALI.

Thirdly, the decision to initiate ECMO in patients with leukaemia and pulmonary failure is an ethical challenge aggravated by serious prognosis. Two of the four patients suffered from zygomycosis, a disease associated with very high mortality. Invasive fungal infections are known to be responsible for one-fifth of the infectious deaths in children with ALL (Grigull et al, 2003). Furthermore, it was shown that children who require ECMO for cardiopulmonary support after haematopoietic stem cell transplantation have a poor prognosis (Gow et al, 2006). Overall, 15 (79%) died during ECMO. Initiation of ECMO was for pulmonary support (n = 17), cardiac support (n = 1), or cardiopulmonary resuscitation (n = 1). The median duration of ECMO support was 5·1 d (range, 30 h to 42 d). Furthermore, of those who survived their ECMO run, only one patient survived to discharge from the hospital (Gow et al, 2006). The authors concluded that clinicians must be cautious in presenting this option to parents and discuss the appropriate expectations in this high-risk population.

Our experience indicates that sepsis, neutropenia and thrombocytopenia in children with leukaemia should not be regarded as contraindications for ECMO treatment. As successful treatment is associated with early initiation, we suggest that ECMO should be commenced in the presence of a severe haemodynamic or respiratory failure refractory to standard medical support. Our cases illustrate the ethical imperative to consider therapies for critically ill children even without solid evidence predictive of success. More experience needs to be published to allow identification of risk factors, establish initiation criteria and treatment protocols for such therapies to advance the standard of care. ECMO by itself does not cure pulmonary failure but gives the lungs time to rest and the body or the treatment time to cure the underlying disease.


We are indebted to W. Högler, W. Streif and U. Schweigmann for valuable discussion and reading the manuscript.