Prone positioning in severe ARDS requiring extracorporeal membrane oxygenation

All patients suffered from severe ARDS. VV ECMO support was initiated in cases of severe hypoxic respiratory failure or CO₂ retention despite of mechanical ventilation as suggested by the ELSO guidelines. Patients receiving PP during ECMO support were allocated to the prone group, whereas the remaining patients formed the supine group. PP before initiation of ECMO support did not influence the allocation of patients in one or the other group. Primary endpoints were successful ECMO weaning, and ICU and hospital survival. Successful ECMO weaning was defined as being free from ECMO and alive for at least 48 h after decannulation. Unsuccessful weaning was defined as the inability to explant the ECMO device because of persistent respiratory failure or death during ECMO support and the need for re-cannulation within 48 h. Moreover, ventilator settings of the first 10 days after ECMO initiation were analysed.

To compare the patients’ disease severity, the RESP [15], SOFA [16], and APACHE II scores [17] as well as the Horowitz index (PaO₂/FiO₂) were analysed.

Our institution features a 24/7 ECMO centre localised within a tertiary hospital with a 30-bed medical intensive care unit. Cannulations in our ECMO centre are performed by two experienced intensivists and a perfusionist in Seldinger’s technique without primary surgical cut down. All member of the ECMO team can be gathered within 30 min. Typical numbers for veno-arterial and veno-venous cannulations are 65 and 35 per year, respectively. There is a 24 h/7 days outreach team. For this research, only in-house cases were considered. As ECMO system, either SCPC (Sorin Centrifugal Pump Console, LivaNova, London, UK) or Cardiohelp (Maquet Getlinge Group, Rastatt, Germany) was used. Cannulation was predominantly performed with dual-lumen cannula (Avalon, Maquet, Rastatt, Germany). For patients without life-threatening bleeding, anticoagulation was provided by intravenous unfractionated heparin aiming at a partial thromboplastin time 1.5 times upper normal limit. The management of vasopressors and fluid therapy was driven by clinical judgement of the ECMO experienced intensivist in charge and has been reported earlier [18].

Treatment algorithms and standard operating procedures were subject to optimizations during the observational period, reflecting current state-of-the-art recommendations and scientific knowledge.

Controlled MV mode used at our institution mostly was biphasic positive airway pressure (BIPAP). In few patients, airway pressure release ventilation (APRV) was used, when considered beneficial. VV ECMO support was implemented in case of severe but potentially reversible respiratory failure, when lung-protective MV resulted in hypoxemia or hypercapnia. Lung-protective MV was defined as positive end expiratory pressure (PEEP) ≤ 15 cmH₂O, plateau pressure ≤ 30 cmH₂O, driving pressure ≤ 15 cmH₂O, and FiO₂ ≤ 50%. Cannulation was performed predominately jugulary using a dual-lumen cannula.

After initiation of the VV ECMO support, invasivity of MV was reduced and ECMO flow was adjusted aiming for a peripheral oxygen saturation of 85–90% and partial pressure arterial oxygen of approximately 50 mmHg, respectively. Typical ventilator settings were as follows: PEEP 15 cmH₂O, plateau pressure 25 cmH₂O, FiO₂ 50%, and respiratory rate 10/min.

ARDS treatment was carried out according to the currently valid guidelines [19]. The decision on whether to perform PP in the individual case lays with the treating medical team’s judgement.

Prone positioning was done face down. Sedation for PP patients at our institution was titrated to preserve spontaneous breathing if possible. Neuromuscular blockade was not given on a routine basis for executing PP. However, in individual cases, especially in cases of strong respiratory drive and concerns about a self-inflicted lung injury [20], neuromuscular blocking agents were used.

Summary results for categorical variables are presented as frequency and percentage. Results for numeric variables are presented as median with interquartile range (IQR). Fisher’s exact test and Pearson’s chi-squared test were used for analysing nominal variables. In dependence of normal distribution, Student’s t test or Mann-Whitney U test was performed for continuous variables.

Propensity score matching was performed using SPSS with a nearest neighbour matching algorithm using a calliper of 0.01. Matching was performed for age, sex, SOFA score, the duration of MV before ECMO, and performance of prior PP before ECMO. Cumulative incidences of 60-day mortality were calculated using competing risk regression (Fine and Gray method) with discharge alive considered a competing event [21]. Statistical calculations were performed using IBM SPSS statistics 25.0 (Armonk, NY: IBM Corp, 2017).

A total of 158 patients with complete medical data could be analysed (age 54.5 (41.8–64.0) years, 67% male). The collective showed a relatively high rate of comorbidities, and this was especially true for immunosuppression (36%, Table 1).

Thirty-eight patients (24.1%) received PP during ECMO therapy. No relevant complications (e.g. decannulation) occurred during the positioning procedures. Patients with PP during ECMO support had a higher rate of pre-existing chronic renal failure and pneumonia-induced ARDS. Patients in the prone group displayed a different pulmonary pathogen spectrum (more viral and fungal infections, especially Pneumocystis jirovecii, Table 1). Survival prediction scores (SOFA, APACHE II, and RESP) did not differ between both groups. PP before ECMO initiation was performed in 16.5% of the patients in both groups.

On average, the first PP during ECMO support was performed after 1.7 (0.5–5.0) days on ECMO support, with 2.0 (1.0–3.0) PP manoeuvres performed per patient. Average PP duration was 19.5 (16.8–20.8) h (Additional file 1, table E1).

Patients with PP during ECMO support showed higher PEEP levels from day 4 and higher plateau pressures from day 4 to 8 (Additional file 1, figure E1). There was no difference in driving pressures as well as in tidal volumes. However, patients with PP during ECMO support showed less spontaneous breathing on day 5 and day 8 to 10.

There were no differences in ECMO weaning rate (47.4% vs. 46.7%, p = 0.940), and ICU or hospital survival (36.8% vs. 36.7%, respectively, p = 0.984) between the prone and the supine groups (Table 2). Cumulative incidences of 60-day in-hospital death were 55% and 64% for the prone and supine groups, respectively (p = 0.207, Fig. 1).

Thirty-eight propensity score matched pairs (76 patients) with similar baseline characteristics could be analysed (Fig. 2, see also Additional file 1, table E2). Successful ECMO weaning rate was 47.4% vs. 44.7% (p = 0.818) in patients with and without PP during ECMO support, respectively. Furthermore, there was no difference in survival between both groups (36.8% vs. 36.8%, p = 1.0). Cumulative incidences of 60-day in-hospital death were 58% and 65% for the prone and supine groups, respectively (p = 0.482, Additional file 1, figure E1).

Underlying lung fibrosis, status of immunosuppression, and aspiration were associated with death, whereas proof of bacterial infections was associated with survival (Table 3). Moreover, a high proportion of spontaneous breathing in the first 10 days was strongly associated with survival. In multivariate analysis, only underlying lung fibrosis (odds ratio 0.15 [95% CI 0.0–0.7]) and a high proportion of spontaneous breathing in the first 10 days (odds ratio 20.0 [95% CI 5.4–73.5]) were independent predictors for death and survival, respectively.

In patients with PP, higher age, acute renal failure, and underlying pulmonary disease were associated with death. Proof of pulmonary bacterial infection and timing of the first PP after ECMO initiation were associated with survival in a univariate analysis (Additional file 1, table E4). In a multivariate analysis, only early initiation of PP (< 17 h) was associated with survival (odds ratio 20.6 [95% CI 1.4–312.9], Fig. 3).

Patients treated with early PP during ECMO (n = 11) showed a superior survival to patients treated with late PP or without PP during ECMO support (81.8% vs. 33.3%). Cumulative incidences of 60-day in-hospital death were 18% for the early PP group and 65% for the late and no PP group, respectively (p = 0.027, Fig. 4). Also, in a separate comparison of patients with late PP as well as patients without PP, early PP showed superior survival rates (81.8% vs. 18.5% and 36.7%, p < 0.001 and p = 0.003, respectively).

Patients in the early PP group were younger than patients with late or without PP during ECMO support (40.0 vs. 56.0 years, p = 0.004). The groups did not differ concerning vasoactive support or in SOFA and APACHE II scores at the time of ECMO implantation. Moreover, there was no difference in the SOFA score between both groups in the first 3 days (Additional file 1, table E8). The RESP score of the patients with early PP was higher (2.0 (1.0–6.0) vs. 0 (− 2.0–2.0), p = 0.025, Additional file 1, table E5). The RESP score without including age was 3 (2.0–6.0) vs. 2 (0–4.0), p = 0.096).

This is a retrospective observational study and therefore contains the risk of selection and reporting bias. Another limitation is the small sample size of only 38 patients with PP and 76 patients in the matched pair analysis, respectively. Moreover, this is a single-centre report and specific processes may influence the presented results. The same internal standard operating procedures applied to the entire treating physician team. However, the indication for performing PP during ECMO support was on basis of the treating ECMO physician and therefore was not standardised. Despite using propensity score matching for outcome analysis, this among other factors might be remaining confounders that we did not control for. Together, due to these limitations, our findings should be considered as hypothesis generating and should not prompt clinical decision-making.