The study was approved by the ethics committee of the Regensburg University Hospital (reference 07/012), and the ethics committees of all involved study centers. Before enrolment, written informed consent was obtained from the legal proxy of each patient. Patients were enrolled from September 2007 through December 2010 at eight intensive care units (ICU) in Germany and two ICU’s in Austria.

Entry criteria were: (1) presence of ARDS according to the American-European Consensus Conference [¹²] defined by bilateral infiltrates on chest X-ray, and a PaO₂/FIO₂ <200 present for at least 2 h. At the time of screening patients could not have any evidence of left ventricular failure; (2) age ≥18 years; (3) history of mechanical ventilation <7 days); (4) plateau pressure >25 cmH₂O at defined ventilator settings (PEEP/FIO₂-table + V_T = 6 ml/kg); (5) absence of severe hemodynamic instability with high demand for vasopressors (mean arterial pressure ≥70 mmHg with continuous norepinephrine infusion ≤0.4 μg/kg/min). Exclusion criteria were: decompensated heart insufficiency, acute coronary syndrome, severe chronic obstructive pulmonary disease, advanced malignancy with life expectancy <6 months, chronic dialysis treatment, lung transplant patients, proven heparin-induced thrombocytopenia (HIT), morbid obesity (body mass index >40 kg/m²), cirrhosis of the liver Child Class ≥B (Child–Pugh scores ≥7), or acute fulminant hepatic failure, severe peripheral arterial occlusive disease, absence of limb doppler pulse, and acute brain injury (Glasgow Coma Scale ≤9).

To identify patients presenting with established ARDS and to help further rule out patients with acute cardiogenic pulmonary edema, the screening was followed by a stabilization period for 24 h, characterized by lung protective mechanical ventilation with high PEEP (≥12 cmH₂O), the use of supportive measures, and hemodynamic evaluation (echocardiography). Patients who had ARDS criteria (PaO₂/FIO₂ <200) after 24 h despite optimal supportive treatment were identified as those with established ARDS [¹³], and were randomized through phone hot line by a random number table generated by the involved statistician with respect to the stratum pulmonary/non-pulmonary ARDS (Fig. 1). The principal investigator (TB) performed visits aimed at information and supervision of the study protocol in the participating centers. An independent Data Safety Monitoring Board monitored the study.

The intensive care management in both groups followed the ‘best clinical evidence’ recommended by the European Society of Intensive Care Medicine [¹⁴]. The daily monitoring included the assessment of awakeness/sedation (Richmond Agitation Sedation Scale [RASS]) and the daily evaluation of the Sequential Organ Failure Assessment (SOFA) and the Simplified Acute Physiology Score (SAPS-II).

After randomization to the treatment group, percutaneous cannulation and initiation of pumpless extracorporeal lung assist (iLA AV, Novalung, Heilbronn, Germany) was performed. Unlike “classic” pump-driven extracorporeal carbon dioxide removal, iLA does not require a blood pump, because the extremely low resistance of this circuit allows flows of about 1–2 l/min with normal arterial pressures (for more detailed description: [¹⁰, ¹⁵]). Evaluation, cannulation and clinical monitoring for avECCO₂-R can be found in electronic supplementary material (ESM).

After initiation of avECCO₂-R, adaptation of the ventilation strategy according to study protocol was performed as follows: a rapid titration down to V_T 3 ml/kg/PBW, PEEP following ARDSNet “high-PEEP/FIO₂” table [¹⁶], respiratory rate 10–25/min with an inspiratory/expiratory ratio of 1:1. Termination of the avECCO₂-R therapy and decannulation was performed according to a defined algorithm (see ESM).

The ventilatory management followed the algorithm of the study group except for the use of a V_T = 6 ml/kg/PBW. The target blood gases for both groups were: PaO₂ ≥60 mmHg and arterial pH ≥7.2. The use of buffering (tris-(hydroxymethyl) aminomethane [TRIS]) was permitted if the patient had hypercapnia and respiratory acidosis (pH <7.2).

In both groups daily screening was performed as a precondition for carrying out a spontaneous breathing trial (SBT): ability to breathe spontaneously (FIO₂ ≤0.4, SaO₂ ≥90 %, PEEP ≤8 cmH₂O, hemodynamic stability, RASS score ≥−1, body temperature <38.5°).

SBT was carried out over 1 h with augmented spontaneous breathing (ASB) 5 cmH₂O with an active humidifier or, 10 cmH₂O with HME filter (PEEP ≤8 cmH₂O, flow trigger <3 l/min). Patients were extubated when no deterioration was observed over a 1 h period.

At the time of screening (24 h prior to randomization) physiologic data were recorded and relevant laboratory, radiographic and clinical findings were collected. Throughout the complete study period, data on ventilator settings, laboratory, physiologic, radiographic and interventional data were recorded. Relevant data of the ventilation management (V_T, PEEP, P_plat, minute ventilation, proportion of spontaneous ventilation on minute ventilation, arterial oxygen saturation) as well as results of the blood gas analyses (PaO₂, PaCO₂, pH) were recorded three times daily (8 a.m., 2 p.m., 10 p.m.) and the mean values of repeated measures were used for further analysis. SOFA score, SAPS-II and RASS scores were calculated daily (8 a.m.).

In addition, in a subgroup of study patients of the Regensburg University Hospital, depending on ‘in-the-daytime’-availability of requested lab values, the levels of proinflammatory cytokines (tumour necrosis factor [TNF], interleukin 6 [IL-6] and interleukin 8 [IL-8]) were assessed from serum probes of avECCO₂-R-patients (n = 20) and control patients (n = 15) at various time points during the study, since the measurement of these parameters is part of the routine laboratory program in this hospital. Furthermore, in all participating centers the daily cumulative doses of vasopressors (norepinephrine, adrenaline, dobutamine), sedative and analgesic agents (sufentanil, propofol, midazolam) and net fluid balances were calculated in each patient and recorded.

The primary outcome parameter was the proportion of days without assisted ventilation in a 28-day period (“ventilator-free” days within 28 days [28-VFD]) and in a 60 days period (“ventilator-free” days within 60 days [60-VFD]) [¹⁷].

Secondary outcomes were: inspiratory plateau pressure levels (P_plat), the proportion of spontaneous breathing as a percentage of the minute ventilation (automatically calculated by the ventilator’s software), the RASS score, hemodynamic changes, the incidence of complications or adverse reactions, the frequency and duration of other adjunctive therapeutic measures, transfusion requirements [packed red blood cell transfusions (units), fresh frozen plasma units, platelet transfusion], the daily cumulative doses of analgesic and sedative agents, cumulative catecholamine requirements/24 h throughout the study period, frequency and duration of renal replacement therapy, the number of failing organs, the “organ-failure-free days” within 28 days after randomisation, and “in-hospital” mortality.

Assuming an increase in 28-VFD from 6.0 ± 10 (control group) to 11.0 ± 8 (study group [¹³]), we estimated that 53 patients would be needed per group for a power of 0.8 and an alpha of 0.05 (calculation by Charitè Centrum für Therapieforschung, Berlin, Germany). We assumed a “drop-out-rate” of 10 % and we calculated that 120 patients would need to be enrolled. However, after an interim analysis with 56 patients by the Data Safety Monitoring Board, it was decided to limit the study period to 3 years, since a statistical significant difference was not expected in a longer study period.

We used a Student’s t-test, the Wilcoxon test, the chi-square test or Fisher’s exact test to assess differences between the groups, as appropriate. A Kaplan–Meier curve was calculated and plotted to assess the days until successful weaning from the ventilator. Furthermore, a nonparametric analysis of longitudinal data in factorial experiments was performed. Changes in interesting clinical outcomes with respect to time were analyzed using multivariate nonparametric analysis of longitudinal data in a two-factorial design (1st [independent] factor: groups, 2nd [dependent] factor: repetitions in time). Therefore, all the time points were simultaneously compared on the corresponding response curves (analysis according to Brunner [Brunner-analysis ¹⁸]). Additionally, a post hoc analysis was performed in patients with greater hypoxemia (PaO₂/FIO₂ ≤150) regarding primary outcome parameters. The reported P values are two-sided and significance was set at P < 0.05.

We found no statistical differences in VFD-28 or VFD-60 between groups, (Table 3), but a post hoc analysis demonstrated that surviving patients with greater hypoxemia (PaO₂/FIO₂ ≤150 at randomization) treated with avECCO₂-R had a significantly shorter period of ventilation (VFD-60 = 40.9 ± 12.8) compared to control (28.2 ± 16.4, p = 0.033) Fig. 2). Non-pulmonary organ failure free days within 60 days, intensive care and hospital days were not statistically different between groups.

After insertion of pumpless extracorporeal lung assist, a V_T of 3 ml/kg PBW was achieved in most patients which increased moderately (as spontaneous breathing increased) to day 6 after randomization (Fig. 3), while in control patients V_T increased from 6 ml to 8 ml/kg/PBW (p < 0.001 between groups). Minute ventilation was significantly decreased, and the difference between plateau pressure and PEEP (P_plat − PEEP) was significantly reduced in study patients compared to control.

The fractional amount of augmented spontaneous breathing as a percent of total minute ventilation was significantly increased from day 3 in avECCO₂-R compared to control (Fig. 4).

In line with previous data, extracorporeal CO₂ removal required higher FIO₂ which is thought to be due to the lower partial pressure of alveolar oxygen (PAO₂) secondary to a decreased lung respiratory quotient [¹⁹], and due to a reduction in functional residual capacity with a V_T of 3 ml/kg compared to 6 ml/kg [²⁰]. PaCO₂-values, arterial pH, and mean respiratory rates/min were similar between groups, but mean PEEP levels tended to be elevated in the study group. During the study period, patients with avECCO₂-R had a modest reduction in MAP without increased demand for continuous norepinephrine infusion.

The cumulative amount of opioid (sufentanil) and benzodiazepine infusion (midazolam) was significantly lower in the study group, but mean RASS-scores were not different between groups during the study period (Fig. 5, ESM).

The number of units of red blood cells transfused was significantly higher in the avECCO₂-R -group in the period between randomization and day 10 (3.7 ± 2.4 units RBC) compared with control (1.5 ± 1.3, p < 0.05). The incidence of avECCO₂-R -related adverse events was low (3 patients = 7.5 %): in one patient, transient ischemia of the lower limb was observed, while two patients developed a ‘false’ aneurysm as a result of arterial cannulation. The overall hospital mortality was low (16.5 %) and we found no significant difference between avECCO₂-R-group (17.5 %) and control (15.4 %).

TNF-α, IL-6, and IL-8 levels are presented in Table 4. Serum TNF-α levels remained unchanged during the early study period in both groups, while IL-6 levels decreased significantly in the avECCO₂-R-group, but not in the control group within 24 h after study initiation. IL-8 levels increased in the first few study days in the control group, but not in study patients.

T. Bein and A. Slutsky are consults for Novalung and received honoraria. A. Slutsky is also a consultant to Maquet Medical. The other authors declare no conflicts of interest.

The data are presented as mean ± standard deviation or absolute numbers

The data are presented as mean ± standard deviation

^$p < 0.05 in comparison with before

^$$p < 0.01 in comparison with before