Validation of RESP and PRESERVE score for ARDS patients with pumpless extracorporeal lung assist (pECLA)

Background RESP score and PRESERVE score have been validated for veno-venous Extracorporeal Membrane Oxygenation in severe ARDS to assume individual mortality risk. ARDS patients with low-flow Extracorporeal Carbon Dioxide Removal, especially pumpless Extracorporeal Lung Assist, have also a high mortality rate, but there are no validated specific or general outcome scores. This retrospective study tested whether these established specific risk scores can be validated for pumpless Extracorporeal Lung Assist in ARDS patients in comparison to a general organ dysfunction score, the SOFA score. Methods In a retrospective single center cohort study we calculated and evaluated RESP, PRESERVE, and SOFA score for 73 ARDS patients with pumpless Extracorporeal Lung Assist treated between 2002 and 2016 using the XENIOS iLA Membrane Ventilator. Six patients had a mild, 40 a moderate and 27 a severe ARDS according to the Berlin criteria. Demographic data and hospital mortality as well as ventilator settings, hemodynamic parameters, and blood gas measurement before and during extracorporeal therapy were recorded. Results Pumpless Extracorporeal Lung Assist of mechanical ventilated ARDS patients resulted in an optimized lung protective ventilation, significant reduction of PaCO2, and compensation of acidosis. Scoring showed a mean score of alive versus deceased patients of 3 ± 1 versus − 1 ± 1 for RESP (p < 0.01), 3 ± 0 versus 6 ± 0 for PRESERVE (p < 0.05) and 8 ± 1 versus 10 ± 1 for SOFA (p < 0.05). Using receiver operating characteristic curves, area under the curve (AUC) was 0.78 (95% confidence interval (CI) 0.67–0.89, p < 0.01) for RESP score, 0.80 (95% CI 0.70–0.90, p < 0.0001) for PRESERVE score and 0.66 (95% CI 0.53–0.79, p < 0.05) for SOFA score. Conclusions RESP and PRESERVE scores were superior to SOFA, as non-specific critical care score. Although scores were developed for veno-venous ECMO, we could validate RESP and PRESERVE score for pumpless Extracorporeal Lung Assist. In conclusion, RESP and PRESERVE score are suitable to estimate mortality risk of ARDS patients with an arterio-venous pumpless Extracorporeal Carbon Dioxide Removal.


Background
Specific mortality risk scores, especially the Respiratory ECMO Survival Prediction (RESP) score [1] and the PRedicting dEath for SEvere ARDS on VV-ECMO (PRE-SERVE) score [2], were developed and validated for ARDS patients with veno-venous high-flow Extracorporeal Membrane Oxygenation (ECMO). ARDS with severe hypercapnia without life-threatening hypoxemia can be treated with Extracorporeal Carbon Dioxide Removal (ECCO 2 R), especially pumpless Extracorporeal Lung Assist (pECLA). Despite a high mortality rate validated risk scores are lacking for these devices.
During the past decade, ECMO was frequently used for patients suffering severe hypoxemic ARDS, indicated by a Horowitz index PaO 2 /FiO 2 below 50-80 mmHg despite lung protective ventilation, to maintain gas exchange and facilitate lung protection [3]. In ARDS patients with severe hypercapnia and respiratory acidosis without life-threatening hypoxemia ECCO 2 R was propagated to achieve lung protective ventilation [4]. Arteriovenous pECLA represents a specific subgroup of ECCO 2 R using a simplified extracorporeal lung assist technique for patients with hypercapnia and respiratory acidosis without cardiac failure. It demonstrated efficient extracorporeal carbon dioxide elimination resulting in lung protective ventilation without respiratory acidosis [5] and reducing the risk of ventilator induced lung injury (VILI) [5][6][7]. pECLA therapy is limited by a low oxygen transfer with only moderate increase of oxygenation.
High mortality rates of ECMO and allocation of limited ECMO resources were leading to the development of mortality prediction scores for veno-venous ECMO in severe ARDS. Especially the RESP score [1] and the PRESERVE score [2] have been used to identify risk factors for death of ECMO patients (additional files 1 and 2). Additionally, non-ARDS-specific scores have been used in critical care. The Sequential Organ Failure Assessment (SOFA) score, published in 1996, evaluates morbidity by scoring the organ failure of lung, coagulation, liver, cardiovascular system, brain, and kidney (additional file 3) [8]. In the prospective observational LUNG SAFE study SOFA score was associated with outcome of ARDS [9]. RESP and/or PRESERVE scores have been compared and evaluated in several studies for ECMO therapy [10][11][12][13][14][15][16], but both scores as well as SOFA score have not been validated for ARDS patients treated with a primary extracorporeal CO 2 removal, like pECLA.
In this retrospective study we tested the hypothesis that RESP and PRESERVE score are suitable to assume the mortality risk of pECLA therapy in case of ARDS and are superior to the SOFA score, which is not specific for Extracorporeal Lung Support and ARDS.

Methods
We conducted a retrospective single center cohort study of ARDS patients undergoing pECLA therapy between 2002 and 2016 at RWTH Aachen University Hospital to validate RESP, PRESERVE and SOFA score. General ethical approval was received by the RWTH Aachen University regional research ethics committee for retrospective studies and confirmed for this retrospective study (AF 047/16). Inclusion criteria were ARDS according to the Berlin criteria [17] with pECLA therapy and exclusion criteria missing data necessary for calculation of scores.
Standard therapy included a lung protective ventilation strategy with a pressure controlled ventilation mode, usually Biphasic Positive Airway Pressure ventilation: Additionally prone position was initiated in moderate to severe ARDS and inhaled nitric oxide was used as rescue therapy in hypoxemia according to the local standard [18]. In our institution, indication for pECLA and ECMO is confirmed multidisciplinary by physicians of all involved medical faculties. In case of severe hypoxemia due to ARDS indicated by persistent PaO 2 /FiO 2 < 60 mmHg despite optimized conservative therapy, patients were treated with veno-venous ECMO as rescue therapy. An indication for pECLA was a severe hypercapnia especially in case of concomitant respiratory acidosis (pHa > 7.2 and/or PaCO 2 > 60 mmHg) as well as achievement of lung protective ventilation, especially when plateau pressure was more than 30 mbar despite optimization of conservative ARDS therapy. The pECLA consisted of a polymethylpentene oxygenator with heparin coating and a membrane surface area of 1.3 m 2 (iLA Membrane Lung®, Xenios AG, Heilbronn, Germany). Filling volume was 250 ml. The cannulas were inserted in the femoral artery (13 or 15 Fr) and in the femoral vein (15 or 17 Fr). pECLA initiation and therapy was performed according to the manufacturer's instructions of use and local standards.
The collected data contained origin of ARDS at ICU admission, demographic parameters such as age, sex, height, weight, diseases, hours of ventilation before pECLA initiation, and SOFA score before pECLA. Furthermore, subjects were retrospectively classified in PRE-SERVE and RESP scores according to the work of Schmidt et al. [1,2]. We recorded ventilator settings with airway pressures (peak/plateau inspiratory pressure, PEEP, driving pressure) and tidal volume. As all patients were ventilated in a pressure controlled mode peak inspiratory pressure and plateau pressure were equal. Registered hemodynamic parameters were mean arterial pressure (MAP), central venous pressure, heart rate, and norepinephrine dose per minute, and additionally, blood gas measurement with Horowitz index (P aO2 /F IO2 ), P aCO2 , pH, and S aO2 . All parameters were registered straight before pECLA initiation, as well as 2 and 24 h after pECLA initiation. Calculating the scores required specific additional information, such as laboratory values, organ function, comorbidity, medication, and specific interventions before pECLA initiation. Hospital mortality rate was recorded according to the development of RESP Score by Schmidt et al. [1].
For statistical analysis, data are presented as mean and standard deviation (mean ± SD). After confirmation of normal distribution with the Kolmogorov-Smirnov test, significance was tested within groups with repeatedmeasures ANOVA with post-test and between groups with unpaired t-test (InStat version 3.06, GraphPad, San Diego, CA, USA). A value of p < 0.05 was considered statistically significant. A multivariable regression analysis including a variable selection assessed the correlation with mortality. With GraphPad Prism 7 (GraphPad, San Diego, CA, USA) receiver operating characteristic (ROC) curves of the scores were calculated and an optimum threshold was defined by calculating the maximum Youden index (J = Sensitivity + Specifity -1).

Results
Between 2002 and 2016 79 ARDS patients were treated with pECLA at RWTH Aachen University Hospital. After retrospective screening six patients were excluded due to missing data and 73 subjects were included in the study. Table 1 presents demographic data including severity and origin of ARDS as well as morbidity before pECLA in detail. Thirteen subjects had an immunocompromised status with a significantly higher mortality rate of 85%, defined as hematologic malignancies, solid tumor, solid organ transplantation, human immunodeficiency virus, or liver cirrhosis. All subjects fulfilled the ARDS criteria including a PEEP of at least 5 cm H 2 O according to the Berlin definition [17]. Most patients had a moderate ARDS (Table 1). Fifty-two patients had a severe hypercapnia with a P aCO2 ≥ 60 mmHg and 28 a severe acidosis with a pH < 7.2. All subjects were sedated and invasive mechanically ventilated in a pressure controlled mode with a shorter duration before pECLA in the survivor group. During pECLA all patients received invasive mechanical ventilation.
Overall hospital mortality rate was 49%, but demonstrated significant age-related differences. Subjects who died in hospital were significantly older and SOFA score was higher before initiation of pECLA. Main Causes of death were septic shock with multi organ failure (44%), non-infectious multi organ failure (17%) and persistent respiratory failure (28%). 11% died due to infaust neurologic prognosis (3 severe head injury after trauma and 1 intracranial bleeding under anticoagulation).
Ventilation, oxygenation, acid-base status, and hemodynamics are presented before initiation of pECLA, after 2 and after 24 h in Table 2. After starting pECLA therapy a significant reduction of inspiratory pressure and driving pressure was observed in all subjects. After 2 and 24 h P aCO2 was significantly reduced and pre-pECLA acidosis was compensated in all subjects. A significant increase of oxygenation index was achieved after 2 h, but remained significantly increased after 24 h only for the surviving cohort. Overall pECLA therapy achieved a stabilization of cardiovascular parameters such as heart ratio, mean arterial pressure, and central venous pressure ( Table 2).
The results of the multivariable regression analysis are presented in Table 3 demonstrating the correlation between parameters before pECLA and mortality.

Discussion
With this retrospective study we could demonstrate that RESP and Preserve score are correlating with the mortality of ARDS patients with pECLA. For the first time two specific risk scores were validated for an ECCO 2 R device and were superior to a general organ dysfunction score, the SOFA score. In the past RESP and PRESERVE score were developed and multiple validated for veno-venous ECMO in hypoxemic ARDS.
In the ELSO registry, used for the RESP score definition, only 21% of the subjects had a bacterial pneumonia, and major diagnostic groups were other acute respiratory diagnosis with 28% as well as unspecified with 30%. This origin of ARDS also contributes to the calculated RESP score [1]. Nevertheless in the recently published EOLIA ECMO trial 45% of ARDS subjects suffered from a bacterial pneumonia and 18% from viral pneumonia [19]. In our study, bacterial pneumonia was also the most frequent origin of ARDS with 40% and viral pneumonia was observed in 14%, demonstrating a typical collective of ARDS patients. RESP and PRESERVE score development and validation showed, that age, immunocompromised status, duration of mechanical ventilation, and SOFA score are relevant risk factors for outcome of ECMO [1,2]. We observed also a significantly younger age, less immunocompromised status, shorter pre-pECLA duration of mechanical ventilation and lower SOFA score in the survivor group (Table 1). There was no direct impact of ARDS etiology to survival rate. Pre-and post-pECLA salvage therapy was not different between survivors and non-survivors. The multivariate analysis of our data revealed also age, SOFA score, immunocompromised status and P aCO2 before pECLA as relevant factors for mortality ( Table 3). As in former pECLA studies extracorporeal CO 2 removal allowed an enhanced lung protective ventilation.
The PRESERVE score used a database of 140 ARDS subjects with ECMO to identify risk factors and to generate this score [2]. Subjects presented with a median P aO2 /    [1]. Blood gas analysis revealed similar values before ECMO initiation with a median P aO2 /F IO2 of 59 mmHg (interquartile range 48-75 mmHg), median P aCO2 of 56 mmHg (44-73 mmHg) and a median pH of 7.25 (7.15-7.35). In our study, subjects presented with a better oxygenation, indicated by a Horowitz index of 126 ± 59 mmHg, but with a severe respiratory acidosis (P aCO2 79.4 ± 30.6 mmHg and pH 7.23 ± 0.14). Patients with a severe disturbed oxygenation comparable to the PRESERE and RESP validation studies were not suitable for pECLA due to the limited oxygen uptake. These patients were primary connected to veno-venous ECMO. Nine pECLA patients were switched to veno-venous ECMO after further deteriorating oxygenation.
Nevertheless, oxygenation and acid base status were more compromised than in the prospective randomized controlled Xtravent study, which evaluated pECLA in combination with an ultraprotective ventilation strategy compared to lung protective ventilation in severe ARDS [20]. ECCO 2 R therapy as arterio-venous pECLA or low-flow veno-venous device seems a promising option to ensure optimized lung protection avoiding further ventilator induced lung injury (VILI) [21] and clinical trials are ongoing [22]. Although there was no leading severe hypoxemia, hospital mortality was 49% in our study compared to 43% in the RESP score study by Schmid et al. [1]. Therefore, in case of extracorporeal carbon dioxide removal a specific risk score seems also useful to identify high-risk patients.
In the PRESERVE and RESP score validation study most of the included patients suffered from severe hypoxemic ARDS [1,2], whereas only 33% of our subjects had a severe ARDS before pECLA start. In the Berlin definition of ARDS, severity of disturbed oxygenation defines the grade and correlates with mortality [9,17]. On the other hand severe hypercapnia is independently associated with mortality of ARDS [23]. Therefore, a direct transfer of the RESP and PRESERVE score from ECMO to ECCO 2 R seems not suitable, because patients have different ARDS characteristics with leading hypercapnia and concomitant acidosis but without lifethreatening hypoxemia. After positive validation for ARDS patients with leading hypercapnia and ECCO 2 R therapy the established RESP and PRESERVE scores could be used for hypoxic as well as hypercapnic ARDS patients intended for extracorporeal lung support.
Validation of pECLA in our study demonstrated comparable results to other studies analyzing PRESERVE and RESP score for veno-venous ECMO (Table 5). We additionally tested, if a non-specific SOFA score could be an alternative tool to assess the risk profile, but AUC as indicator for accuracy was lower. Nevertheless a SOFA score > 12 represents a risk factor in the PRE-SERVE score but not in the RESP score. Overall, only the specific scores demonstrated a good diagnostic accuracy for pECLA. Comparing both scores, the  PRESERVE score requires less items and as a result seems easier to handle than the RESP score. In conclusion both scores seem suitable for pECLA as ECCO 2 R device.
As mentioned above several studies evaluated RESP and PRESERVE scores for other ECMO populations with differing accuracy and without superiority of one score (Table 4). Survival in the different predefined risk classes demonstrated some inconsistent results but with a generally increasing mortality for a higher risk score (Table 5). Compared to these studies the performance of PRESERVE and RESP was non-inferior for pECLA in our study. Limitations of our study are the retrospective small validation cohort from one ARDS center without additional data from other centers to verify our results, the missing long-term survival data and the restriction to one specific low-flow device for ECCO 2 R. A prospective registry of ECCO 2 R could be able to generate more detailed as well as long-term data. With our retrospective study, PRESERVE and RESP score could be sufficiently validated to identify a high-risk profile before starting an extracorporeal carbon dioxide elimination. Nevertheless, ARDS therapy and especially time of initiation and decision for conventional therapy versus ECCO 2 R or ECMO require clinical assessment and could not be replaced by a simple scoring.
In our study we focused on pumpless ECLA as ECCO 2 R device, but other veno-venous low-flow ECLA systems are also used for hypercapnic ARDS. For venovenous devices, there is an ongoing transition from leading decarboxylation to decarboxylation plus oxygenation with increasing blood flow. As RESP and PRESERVE were primary validated for classical high-flow ECMO and now were additionally validated for pECLA as decarboxylation device by our study, we hypothesize that these scoring systems are also suitable for other lowflow ECLA systems. Further investigations of low-flow veno-venous ECCO 2 R could be used to confirm this assumption.

Conclusions
Performance of RESP and PRESERVE score was at least as good for pECLA as for veno-venous ECMO, the primary validation cohort and this is the first study expanding the scope from high-flow ECMO to an ECCO 2 R therapy. We demonstrated that these risk scores are suitable for ARDS with leading hypercapnia and pECLA additional to severe hypoxemic ARDS with high-flow ECMO.
Both scores, RESP and PRESERVE, but not SOFA score seem suitable to point out the risk profile of ARDS