Application of extracorporeal membrane oxygenation in patients with severe acute respiratory distress syndrome induced by avian influenza A (H7N9) viral pneumonia: national data from the Chinese multicentre collaboration

BackgroundEvidence concerning the efficacy and safety of extracorporeal membrane oxygenation (ECMO) in patients with influenza A (H7N9) has been was limited to case reports. Our study is aimed to investigate the current application, efficacy and safety of ECMO in for severe H7N9 pneumonia-associated acute respiratory distress syndrome (ARDS) in the Chinese population.MethodsA multicentre retrospective cohort study was conducted at 20 hospitals that admitted patients with avian influenza A (H7N9) viral pneumonia patients’ admission from 9 provinces in China between October 1, 2016, and March 1, 2017. Data from the National Health and Family Planning Commission of China, including general conditions, outcomes and ECMO management, were analysed. Then, successfully weaned and unsuccessfully weaned groups were compared.ResultsA total of 35 patients, aged 57 ± 1 years, were analysed; 65.7% of patients were male with 63% mortality. All patients underwent invasive positive pressure ventilation (IPPV), and rescue ventilation strategies were implemented for 23 cases (65.7%) with an average IPPV duration of 5 ± 1 d, PaO2/FiO2 of 78 ± 23 mmHg, tidal volume (VT) of 439 ± 61 ml and plateau pressure (Pplat) of 29 ± 8 cmH2O pre-ECMO. After 48 h on ECMO, PaO2 improved from 56 ± 21 mmHg to 90 ± 24 mmHg and PaCO2 declined from 52 ± 24 mmHg to 38 ± 24 mmHg. Haemorrhage, ventilator-associated pneumonia (VAP) and barotrauma occurred in 45.7%, 60% and 8.6% of patients, respectively. Compared with successfully weaned patients (n = 14), the 21 unsuccessfully weaned patients had a longer duration of IPPV pre-ECMO (6 ± 4 d vs. 2 ± 1 d, P < 0.01) as well as a higher Pplat (25 ± 5 cmH2O vs. 21 ± 3 cmH2O, P < 0.05) and VT (343 ± 96 ml vs. 246 ± 93 ml, P < 0.05) after 48 h on ECMO support. Furthermore, the unsuccessfully weaned group had a higher mortality (100% vs. 7.1%, P < 0.01) with more haemorrhage (77.3% vs. 28.6%, P < 0.01).ConclusionsECMO is effective at improving oxygenation and ventilation of patients with avian influenza A (H7N9) induced severe ARDS. Early initiation of ECMO with appropriate IPPV settings and anticoagulation strategies are necessary to reduce complications.


(Continued from previous page)
Conclusions: ECMO is effective at improving oxygenation and ventilation of patients with avian influenza A (H7N9) induced severe ARDS. Early initiation of ECMO with appropriate IPPV settings and anticoagulation strategies are necessary to reduce complications.
Keywords: Extracorporeal membrane oxygenation (ECMO), Avian influenza A (H7N9), Acute respiratory distress syndrome (ARDS), Complications, Mortality Background Avian influenza A (H7N9) viral pneumonia can manifest with varying degrees of dyspnea and is associated with a mortality of~30% [1]. In particular, 97% of patients develop rapidly progressive pneumonia and 71% progress to acute respiratory distress syndrome (ARDS). The mortality of severe ARDS is as high as 62% [2].
Timely and effective respiratory support is particularly important to treat severe ARDS caused by avian influenza A (H7N9) pneumonia. However, severe ARDS induced by avian influenza A (H7N9) pneumonia might manifest as refractory hypoxaemia even with appropriate invasive positive pressure ventilation (IPPV) support. Extracorporeal membrane oxygenation (ECMO) is the ultimate respiratory support method and directly improves the oxygenation and ventilation of patients as well as enables implementation of the "lung protective ventilation strategy" [3]. ECMO was the breakthrough treatment for the severe avian influenza A (H1N1) outbreak of 2009 and reduced mortality from this outbreak [4][5][6]. Therefore, we believe that ECMO could also be effective for other types of severe viral pneumonia.
Existing studies of ECMO treatment for avian influenza A (H7N9) pneumonia are primarily limited to case reports [7][8][9], and no study has systematically reviewed the efficacy or safety of ECMO to treat such diseases. Therefore, it is particularly important to understand the current application of ECMO for avian influenza A (H7N9) pneumonia-induced severe ARDS, investigate the application timing and management strategies of ECMO, and explore the possible reasons for treatment failure. Based on the current study, we expect to standardize the management of ECMO and provide a description of our experiences using ECMO to treat patients with avian influenza A (H7N9) pneumoniainduced severe ARDS.

Study population and data collection
Patients who had laboratory-confirmed avian influenza A (H7N9) virus-induced pneumonia were included in this study. Patients were admitted to 20 hospitals in 9 provinces of China between October 1, 2016, and March 1, 2017, and were reported to the National Health and Family Planning Commission of China.
We included patients aged >14 ys who were supported by ECMO. Patients who were lacking key detailed records of parameters during ECMO, such as ventilator or laboratory findings, were excluded.
The included patients were divided into 2 groups, namely, the "successfully weaned group" and "unsuccessfully weaned group". The former refers to a group of patients whose condition improved and were weaned from ECMO for at least 48 h; the "unsuccessfully weaned group" refers to those who died or voluntarily discontinued treatment due to lack of improvement during ECMO support.

General conditions
The general conditions included age, gender, pregnancy status, underlying disease, time from onset to antiviral drug administration, vasoactive drug administration pre-ECMO, duration of IPPV pre-ECMO, whether rescue ventilation strategies (including lung recruitment maneuvre, prone-position ventilation, and high-frequency oscillation ventilation) were implemented pre-ECMO, disease severity score, total duration of ECMO and IPPV.

Conditions during ECMO
We collected the ECMO blood flow at 24, 48, 72, and 96 h on ECMO. Improvement in circulatory and respiratory physiological indicators were considered, as well as IPPV parameters at 6 h pre-ECMO and 24, 48, and 72 h on ECMO. Furthermore, anticoagulation indicators during ECMO, including the types of anticoagulant drugs and methods of use; the maximum and minimum values of the activated coagulation time (ACT) and activated partial thromboplastin time (APTT); and the differences between the maximum and minimum ACT and APTT at 24, 48, and 72 h on ECMO were recorded. Finally, data regarding complications during ECMO therapy, including ECMO and IPPV-related complications and nosocomial infections, were collected.

Study definitions and design Outcomes
The primary outcome was in-hospital mortality. The secondary outcomes were the length of stay in the intensive care unit (ICU) and total length of hospitalization.

Definition of avian influenza A (H7N9) virus infection
Three methods were used for a laboratory diagnosis, namely, the real-time reverse transcription-polymerase chain reaction (RT-PCR), viral isolation, and serological testing for the avian influenza A (H7N9) virus using a modified haemagglutinin inhibition assay [10][11][12].

Definition of acute respiratory syndrome (ARDS)
We defined ARDS according to the Berlin definition in 2012 [13,14].

Definitions of pneumonia and severe pneumonia
Pneumonia was diagnosed as an acute illness with fever, cough, or dyspnea/tachypnea, and at least one new focal chest sign that was supported by a finding of lung shadowing on a chest radiograph and without other noninfectious causes.
The primary criteria for severe pneumonia were as follows: <1 > need for tracheal intubation and mechanical ventilation (MV) and <2 > need for vasoactive drugs after the active fluid resuscitation due to septic shock. The secondary criteria were as follows: <1 > respiratory rate ≥ 30 times/ min; <2 > PaO 2 /FiO 2 ≤ 250 mmHg; <3 > multiple lobe infiltration; <4 > disturbances of consciousness, disorientation, or both; <5 > blood urea nitrogen ≥ 7.14 mmol/L; and <6 > systolic blood pressure ≤ 90 mmHg that required active fluid resuscitation. Patients who met one primary criterion or at least three secondary criteria were diagnosed as having severe pneumonia [15].

Definitions of ventilator-associated pneumonia (VAP)
The criteria for the diagnosis of VAP are in accordance with the European Centre for Disease Prevention and Control [16] and included the following: <1 > two or more sequential chest x-rays or CT scans with a suggestive image of pneumonia for patients with underlying cardiac or pulmonary disease, or one definitive chest x-ray or CT scan in patients without underlying cardiac or pulmonary disease; <2 > a fever greater than 38°C and/ or leukocytosis greater than or equal to 12,000 WBC/ mm 3 or leukopenia less than or equal to 4000 WBC/ mm 3 ; and <3 > at least one of the following: <a > new onset of purulent sputum or change in the characteristics of the sputum; <b > cough, dyspnea, or tachycardia; <c > auscultatory findings, such as rales, bronchial breath sounds, ronchi, or wheezing; or <d > worsening gas exchange (e.g., oxygen desaturation or increased oxygen requirements or increased ventilation demand).
For all included patients, we first described the general conditions, ECMO model and parameters, IPPV parameters, the changes in circulation and respiratory physiological indicators from pre-ECMO to on ECMO status, anticoagulation on ECMO, and complications during ECMO therapy in all included patients. Then, we compared patients who were successfully or unsuccessfully weaned from ECMO with regard to above items.

Statistical analysis
All of the analyses were performed using SPSS 17.0 software. Normally distributed continuous variables are expressed as the means ± SD and were compared using the t-test or chi-square test. Non-normally distributed continuous variables are expressed as medians and quartiles and were compared using the Wilcoxon rank-sum test. Categorical variables were compared using the x 2 test. P-values < 0.05 were considered significant.

Results
A total of 473 patients were diagnosed with avian influenza A (H7N9) virus-related pneumonia. Patients were admitted to 20 hospitals in 9 provinces of China between October 1, 2016, and March 1, 2017, and were reported to the National Health and Family Planning Commission of China. The medical records of 216 patients were available, and 36 patients were reported to be supported by ECMO. One of the 36 patients lacked IPPV and ECMO parameters pre-ECMO and on ECMO and was eliminated as a participant; therefore, 35 patients were ultimately selected ( Fig. 1).

General conditions and outcomes
Data from 35 patients (65.7% males), with an average age of 57 ± 1 years, were analysed. There was no patient under the age of 16. A total of 22 patients had underlying diseases, 8 patients were treated with steroids and immunosuppressive agents within 1 month of admission to the hospital, and 1 patient was pregnant. The Sequential Organ Failure Assessment (SOFA) score was 9 ± 3 points, and the Murray score was 3.6 ± 0.4 points. The time from onset to antiviral drug administration was approximately 9 ± 5 d, and the time from onset to ECMO support was approximately 10 ± 3 d. High-dose vasoactive drugs [17] were needed to maintain blood pressure in 20 patients (57.1%). The duration of IPPV pre-ECMO was approximately 5 ± 1 d. Rescue ventilation strategies, including the recruitment manoeuvre (RM), prone-position ventilation (PP) and high frequency oscillatory ventilation (HFOV), were needed for 23 patients (65.7%) pre-ECMO. The total durations of IPPV and ECMO were approximately 20 ± 8 d and 8 d (5-13 d), respectively. Of the 35 patients, 14 (40%) were successfully weaned from ECMO, and the other 21 patients died due to an uncontrolled haemorrhage (5 patients), septic shock (6 patients due to progressive lung infection, 3 patients due to bloodstream infection), heart failure (2 patients) and discontinuation of treatment because of no improvement (5 patients). One of 14 patients showed an aggregated lung infection after weaning and eventually died due to septic shock. The inhospital mortality was 63%. The length of ICU stay was 27 ± 14 d, and the total length of hospitalization was 31 ± 14 d (Table 1).

Conditions during ECMO ECMO mode and information concerning the equipment
Of the 35 patients, 33 were treated using the veno-venous ECMO (V-V ECMO) model. A total of 2 patients with severe cardiac insufficiency and cardiogenic shock were treated using the venous-arterial ECMO (V-A ECMO) model. The ECMO equipment was mainly provided by MAQUET (Shanghai) Medical Equipment Co., Ltd. and SORIN (Shanghai) Medical Equipment Co., Ltd. The pump from SORIN was the Stockert Centrifugal Pump System (SCP/SCPC), and the oxygenator was the D905 EOS ECMO. The pump from MAQUET was the ROTA-FLOW, and the oxygenator was the QUADROX PLS.

ECMO management
Changes in IPPV parameters and physiological indicators in patients on ECMO The ventilator parameters, including FiO 2 , positive end-expiratory pressure (PEEP), P plat , and VT, were significantly decreased in patients on ECMO. The vital signs, which included the heart rate, respiratory rate, and SpO 2, and the arterial blood gas analysis (ABG), which included the pH, PaCO 2 , and PaO 2 levels, were improved in patients after ECMO support ( Table 2).
Monitoring of anticoagulation All 35 patients received a continuous infusion of unfractionated heparin for anticoagulation. However, heparin was discontinued for 2 patients with cerebral haemorrhage and 3 with active gastrointestinal haemorrhage. The ACT was 144 ± 61 to 220 ± 66 s at 24 h, 16 ± 37 to 234 ± 37 s at 48 h, and 147 ± 26 to 221 ± 40 s at 72 h on ECMO. The APTT was 58 ± 23 to 84 ± 28 s and 53 ± 15 to 75 ± 23 s at 48 and 72 h on ECMO, respectively.
Complications during ECMO therapy In this study, the rates of gastrointestinal haemorrhage, cerebral haemorrhage, brain death, renal insufficiency, disseminated intravascular coagulation (DIC), hyperglycaemia, and ECMO oxygenator thrombosis were higher compared to the relevant data from the ECLS Registry Report [18,19]. New cases of VAP developed in 21 patients during ECMO, with an incidence rate of 60%. New cases of barotrauma occurred in 3 patients, accounting for 8.6% of cases. In addition, 6 patients had a urinary infection, with an incidence rate of 17.1%, and 10 patients had a catheter-related blood stream infection (CRBSI), with an incidence rate of 28.6% (Table 3). Comparison between the patients successfully and unsuccessfully weaned from ECMO Group 1 contained 14 patients who were successfully weaned from ECMO, and group 2 included 21 patients who were unsuccessfully weaned from ECMO.

Blood flow in patients on ECMO
ECMO blood flow did not significantly differ between the two groups during the initiation of ECMO support. However, in the successfully weaned group vs. the unsuccessfully weaned group, a significant decrease in blood flow correlated with an increase in the duration of support, which was 3.65 ± 0.70 L/min vs. 4.57 ± 1.02 L/min, respectively, (P < 0.05) at 72 h on ECMO and 3.65 ± 0.86 L/ min vs. 4.62 ± 0.90 L/min, respectively, (P < 0.01) at 96 h on ECMO (Additional files 1 and 2).

Improvement in circulatory and respiratory physiological indicators on ECMO
The vital signs were improved but did not significantly differ between the two groups pre-ECMO and during ECMO support (Additional file 3). Patients who were unsuccessfully weaned from ECMO compared to patients who were successfully weaned from ECMO had severe acidosis (pH: 7.29 ± 0.14 vs. 7.40 ± 0.05, respectively, (P < 0.01), a higher PaCO 2 (57.0 ± 16.7 mmHg vs. 43.6 ± 13.0 mmHg, respectively, (P < 0.05), and a higher lactate concentration (3.6 ± 2.1 mmol/L vs. 1.7 ± 0.8 mmol/L, P < 0.05) pre-ECMO. pH and PaCO 2 did not differ significantly between the two groups during ECMO support, while patients who were eventually successfully weaned from ECMO had a gradual ascending tendency of PaO 2 at 48 and 72 h on ECMO and a sustained low level of lactate (Fig. 2, Additional file 1).

Changes in anticoagulation on ECMO
During the early stage of ECMO (24 and 48 h), the successful weaning group showed smaller differences between the ACT max and ACT min than the unsuccessful weaning group, which was 48 ± 20 s vs. 104 ± 35 s at 24 h (P < 0.05) and 40 ± 25 s vs. 98 ± 40 s at 48 h (P < 0.01). However, this trend was not found with regard to the difference between the maximum and minimum APTT.

Discussion
This study was the first to systematically and comprehensively discuss as well as elaborate on the current application of the efficacy and safety of ECMO in patients with H7N9 pneumonia-related ARDS. A few studies [17,20,21,22] have shown that the mortality of pH1N1-induced ARDS was reduced to 21-61% following ECMO treatment. Presently, no studies with large samples have investigated the mortality of H7N9-induced ARDS, while the in-hospital mortality was as high as 63% in our study. Late initiation of ECMO, inappropriate IPPV settings during ECMO, and more ECMO complications might explain the relatively high mortality. Moreover, as a multicentre collaboration study, the experiences of ECMO varied among the centres (Additional file 4), which might be another reason for the high mortality. According to the Extracorporeal Life Support Organization (ELSO) data [23,24], ECMO is indicated when death risk exceeds 80%, i.e., when PaO 2 /FiO 2 < 80 mmHg on FiO 2 > 90% and the Murray score is 3-4. Our patients met the indications for ECMO support. The duration of MV for more than 7 days pre-ECMO is an important prognostic factor for death [25]. For patients in the successfully weaned group, the duration of IPPV pre-ECMO was 5 ± 1 d; however, the duration was even longer among patients in the unsuccessfully weaned group (6 ± 4 d). Moreover, rescue ventilation strategies were implemented for most patients before ECMO, which partially delayed the timing of ECMO. In comparison, ECMO was initiated at 2 h (1-5 h) after IPPV among patients with pH1N1 in Australia and New Zealand in 2009 [5], which was significantly shorter than that in our cases. Therefore, we emphasized early implementation of ECMO in our patients.
The principle of IPPV during ECMO is the "lung rest strategy" [26]. The REVA registry study examined 123 patients with pH1N1-induced ARDS [6] and showed that the high P plat (29 cmH 2 O) on day 1 of ECMO was related to high mortality. In our study, the pre-ECMO P plat level was high (29 ± 8 cmH 2 O). High P plat can lead to overdistension of the alveoli and cause lung volutrauma. The shear force between the overdistended and collapsed alveoli further aggregates VILI [27], which ultimately increases mortality. Although the P plat values  [28] showed that a high PEEP level within 3 d of being on ECMO was related to decreased mortality. Although no difference was observed in the PEEP levels between the two groups, we speculated that the down-regulation of PEEP during ECMO might have further aggravated the occurrence of collapse-induced injury, which led to atelectasis and sputum discharge obstacles. Therefore, the IPPV parameters, including high P plat and VT levels and low PEEP settings, might have been unreasonable in our study; lung rest or the maintenance of open alveoli was not achieved. The incidence of an ECMO oxygenator thrombus, haemorrhage, and organ failure in our study was high, which suggests that some problems existed in the anticoagulation management and organ supportive treatment of ECMO. We found that the unsuccessfully weaned group had larger fluctuations in ACT (the difference between ACT max and ACT min were larger) during the early stage of anticoagulation. This effect might suggest relatively unstable anticoagulation and a higher risk of haemorrhage. Moreover, the incidence rate of VAP during ECMO was as high as 60% and was partially attributed to the long course of H7N9 pneumonia and the prolonged duration of IPPV. Therefore, intensification of airway management was extremely necessary.
Our study had limitations. The nature of the study required the collection of data at multiple consecutive time points to evaluate the efficacy of ECMO. As a retrospective study with some missing data, we were unable to successfully collect data at 6 h pre-ECMO and 24, 48, and 72 h post-ECMO. Additionally, the number of subjects was too small to perform a multiple regression analysis to explore the risk factors for unsuccessful weaning from ECMO.