Introduction

A prolonged course of acute lung injury (ALI) or acute respiratory distress syndrome (ARDS) with no recovery during the first week has been termed as late-phase ARDS [1]. It has been characterized by organization and collagen remodelling and pronounced reductions in pulmonary compliance [2]. In this subset of patients even low tidal volume ventilation may cause lung overdistension and attendant detrimental high inspiratory airway plateau pressures [3]. Furthermore, hypercapnia results in increased inspiratory flows and tidal volumes, and decreased breath-to-breath variability in tidal volumes complicating patient–ventilator interaction and prolonging weaning from mechanical ventilation [4, 5]. In fact, severe hypercapnia, tachypnoea and possible patient distress have been highlighted as reasons why low tidal volume ventilation strategies are not universally implemented in ALI/ARDS patients [6, 7].

To optimize plateau pressures and patient–ventilator interaction long-term ventilated patients with high CO2-levels often receive high doses of analgesics and sedatives or muscle relaxants [8, 9]. All these factors may compromise weaning and sedation protocols [10] and negate the benefits of early spontaneous breathing [11, 12].

We hypothesized that extracorporeal CO2 removal using a pumpless extracorporeal lung assist system (p-ECLA, NovaLung®) facilitates lung protective ventilation and the management of sedation in late-phase ALI/ARDS patients with severe hypercapnia.

Methods

We retrospectively analysed a consecutive series of postpneumonic late-phase ALI/ARDS patients [1, 2, 13] with pronounced impairment of pulmonary compliance and severe hypercapnia who were treated in a university-affiliated tertiary referral centre for severe ALI/ARDS in a 18 months period. We assessed the impact of extracorporeal CO2 removal [14, 15] on airway pressures, tidal volumes, PaCO2 levels, breathing pattern and use of sedatives.

Upon transfer from outside hospitals all patients were ventilated in a pressure-controlled mode (biphasic positive airway pressure ventilation allowing assisted spontaneous breathing, BIPAP/ASB) to be aimed at a tidal volume of 6 ml/kg predicted body weight (PBW). Patients received standardized treatment including permissive hypercapnia, PEEP adjusted to oxygenation, intermittent prone positioning and reduction of extravascular lung water [16–18]. Sedation levels were assessed three times daily using the Ramsay Score with a target of 2–3 [19]. The depth of sedation was increased when hypercapnia was associated with tachypnoe > 40/min, high inspiratory flows, tidal volumes > 6 ml/kg PBW or poor patient–ventilator interaction. Triggers for putting patients on the p-ECLA device were inspiratory plateau pressures > 35 cm H2O, a pCO2 > 80 mmHg and/or respiratory acidosis with a pH < 7.2 [16]. We performed p-ECLA as described elsewhere [16, 20].

During the period of review we treated a total of 12 patients with p-ECLA. Two patients died due to refractory septic shock within 24 h of beginning p-ECLA. We did not include these two patients in the analysis as no data regarding the primary outcome measures were available.

Measurements

Ventilator settings, arterial blood gases, doses of analgesics and sedatives before (baseline) and after (1 h, 12 h, 2–4 days) initiation of treatment with p-ECLA were retrieved from an electronic patient data management system (PDMS). Ventilator settings were electronically sent to the PDMS every 10 min. The inspiratory plateau pressures (Pplat) were automatically computed from the ventilator on mandatory mechanical breaths. The ASB pressures were set to equal the Pplat minus PEEP (ΔPaw). After beginning p-ECLA the primary goal was to reduce inspiratory plateau pressures below 32 cmH2O even if tidal volumes resulted in less than 6 ml/kg PBW [16]. PEEP values were adjusted according to FiO2. The mandatory rate was decreased when the spontaneous rate increased after reduction of sedation.

Statistics

For repeated measurements Friedman’s two-way analysis of variance was used. Post hoc testing of values prior to p-ECLA and values at day 4 after p-ECLA implementation was performed by Wilcoxon-testing for paired samples; P < 0.05 (two-sided) was considered statistically significant. Data for continuous variables are given as median and 25th percentiles. Statistical analyses were performed using SPSS, Version 14, SPSS, Inc., Chicago, Illinois, USA.

Results

Ten patients with ALI/ARDS caused by pneumonia were included (Table 1). In three patients there was evidence of a pre-existing pulmonary fibrosis and in one patient a bronchiolitis obliterans and organizing pneumonia (BOOP) was demonstrated histologically. Nine of the ten patients had already been treated in another hospital for a median of 7 (2.25/13.25) days prior to transfer to our hospital. Three of the patients (patients 6, 7 and 8 in Table 1) had ARDS due to a second infective hit [21] following drug-induced lung injury, immuno-suppression related aspergillosis [22] and pulmonary reinfection, respectively. These patients had also required significant periods of non-invasive ventilation and intensive respiratory therapy prior to intubation and mechanical ventilation. Median duration of ventilation prior to implementation of p-ECLA therapy was 11.5 (2.5/18.75) days. Duration of p-ECLA treatment ranged from 7 to 77—18.5 (8.8/60)—days. In the four patients who survived, the median duration of p-ECLA therapy was 38 (7.25/75) days.

Table 1 Clinical and demographic characteristics of patients
Full size table

Prior to p-ECLA, median semi-static lung compliance was 14.5 (11.6/19.7) ml/cmH2O, median pCO2 was 120 (81.9/154.5) mmHg and median pH was 7.06 (6.9/7.3). In seven patients p-ECLA was established during the first 24 h of ICU admission and in the three remaining patients between 7 and 12 days after admission. For extracorporeal gas exchange we placed a 5 mm (15F) cannula in the femoral artery and a 5.7 mm (17F) cannula in the femoral vein. In one patient dissection and temporary occlusion of the left deep femoral artery occurred during cannulation. In this patient cannulation of the right femoral artery was subsequently performed under angiographic control. Recanalization of the left deep femoral profound artery was in this case achieved through anticoagulation with heparin. No further intervention was necessary.

Detailed data concerning gas exchange and respiratory mechanics are provided in Table 2. Median gas flow through the membrane (sweep flow) was 7.8 (6.6/9) l/min and median p-ECLA blood flow was 1.2 (1.0/1.3) l/min within the first 4 days of the study. The PaO2/FiO2 ratios did not change significantly after initiation of p-ECLA; however, the FiO2 could be reduced during the observation period. Median PaCO2 was reduced by 50% within 12 h of starting p-ECLA. We observed a significant reduction of Pplat and ΔPaw by reducing tidal volumes below 4 ml/kg PBW during p-ECLA therapy. The pulmonary compliance remained unchanged. The median pre-intervention infusion rates of fentanyl [8.9 (7.6/10.5) μg/kg per h] (Fig. 1) and midazolam [0.28 (0.17/0.51) mg/kg per h] (Fig. 2) were decreased significantly to 4.4 (2.7/7.1) and 0.19 (0.1/0.2) mg/kg per h, respectively within 4 days of p-ECLA treatment. But there was only a minor change in the depth of sedation: Ramsay score 6 (5/6) at baseline and 5 (4/6) by day 4 (P > 0.05). However, a corresponding and statistically significant increase in the proportion of spontaneous breathing (RRspon) was observed. Norepinephrine administration was significantly reduced from 0.2 (0.06/0.7) μg/kg per min at baseline to 0.02 (0/0.13) at day 4 of p-ECLA treatment.

Table 2 Gas exchange and respiratory parameters before and after p-ECLA insertion
Full size table
Fig. 1
figure 1

Cumulative daily dosage of Midazolam (mg/kg per h) before and during p-ECLA (pumpless extracorporeal lung assist); # Friedman’s two-way analysis of variance was used for repeated measurements, P < 0.05; * post hoc testing by Wilcoxon-test for paired observations, cumulative dosage of midazolam prior p-ECLA and on day 4 after p-ECLA implementation were compared, P < 0.01

Full size image
Fig. 2
figure 2

Cumulative daily dosage of Fentanyl (μg/kg per h) before and during p-ECLA (pumpless extracorporeal lung assist); # Friedman’s two-way analysis of variance was used for repeated measurements P < 0.05; * post hoc testing by Wilcoxon-test for paired observations, cumulative dosage of fentanyl prior p-ECLA and on day 4 after p-ECLA implementation were compared, P < 0.01

Full size image

Discussion

We reported that extracorporeal CO2 removal with a novel pumpless extracorporeal lung assist device facilitated normocapnic ventilation with tidal volumes of less than 4 ml/kg PBW. Reduction of tidal volumes resulted in a significant reduction of inspiratory plateau pressures and ∆Paw reflecting improved lung protection in patients with late-phase ALI/ARDS [1]. Following initiation of p-ECLA we were able to significantly reduce doses of sedatives and analgesics and observed a parallel increase in the proportion of assisted spontaneous breathing.

It has been shown that low stretch ventilation with low tidal volumes (4–6 ml/kg, PBW) reduces ARDS mortality [23, 24]. However, in patients with severely impaired lung compliance low tidal volume ventilation with 4–6 ml/kg PBW may still represent overexpansion of the lung. Extracorporeal CO2 elimination may have the potential to reduce inspiratory plateau pressures, ∆Paw, and ultimately ventilator-induced lung injury [25, 26] by enabling ventilation with tidal volumes of less than 4 ml/kg PBW.

We noted a non-significant trend towards a reduced PaO2/FiO2 ratio after institution of p-ECLA, while FiO2 could be reduced. It has been discussed that tidal volumes below 4 ml/kg PBW may result in alveolar derecruitment if insufficient PEEP levels are applied [27]. On the other hand this might have been caused by a decrease in a CO2 dependent cardiac output.

Depth of sedation is adversely related to duration of ventilation, length of ICU stay and outcome [10, 28]. Our patients with hypercapnic respiratory drive had high cumulative doses of analgesics and sedatives (Figs. 1, 2) and deep sedation [Ramsay score 6 (5/6)] prior to p-ECLA. Although p-ECLA treatment resulted in a reduction of sedation requirements, we could not observe significant changes in the depth of sedation. This may indicate that dosages prior to p-ECLA were associated with accumulation and reflects a common problem in late-phase ALI/ARDS patients with hypercapnia—namely that high doses of sedation are required in order to achieve an acceptable level of patient–ventilator interaction. However, we observed that reduction of sedatives and analgesics resulted in an increase of spontaneous breathing in the context of a normocapnic respiratory drive.

Our study has a number of limitations. Data concerning breath-to-breath variability in tidal volumes in order to better estimate to what extent extracorporeal CO2 elimination effects this parameter were not available.

Experiences with the treatment of patients with multiple organ failure and coagulation disorders with a p-ECLA device are limited. The indication of p-ECLA treatment in our cohort was an individual decision with respect to the protracted deleterious mechanical ventilation.

Two patients who died during p-ECLA therapy within the first 24 h were in profound septic shock with massive instable hemodynamics at the time of starting p-ECLA therapy. In both cases p-ECLA was being used as an attempted rescue therapy of profound respiratory acidosis and catecholamine insensitivity. However, there are no data available to substantiate the value of p-ECLA therapy in such cases.

Conclusions

In this study p-ECLA treatment allowed ventilation with tidal volumes below 4 ml/kg PBW in patients with significantly impaired pulmonary compliance. Pulmonary shear stress was reduced and normal pCO2 and pH could be maintained. We also reported a reduction in sedation requirements and a corresponding increase in assisted spontaneous breathing without the synchronization problems frequently seen in hypercapnic patients. Prospective randomized trials are required to determine whether these observations hold true in larger patient groups and whether there is a beneficial impact on outcome.