Does extracorporeal membrane oxygenation attenuate hypoxic pulmonary vasoconstriction in a porcine model of global alveolar hypoxia?

During severe respiratory failure, hypoxic pulmonary vasoconstriction (HPV) is partly suppressed, but may still play a role in increasing pulmonary vascular resistance (PVR). Experimental studies suggest that the degree of HPV during severe respiratory failure is dependent on pulmonary oxygen tension (PvO2). Therefore, it has been suggested that increasing PvO2 by veno‐venous extracorporeal membrane oxygenation (V‐V ECMO) would adequately reduce PVR in V‐V ECMO patients.

the precapillary sensing site of the pulmonary circulation can be expressed as a function of P A O 2 and PvO 2 : 9 P A O 2 cannot be measured directly and has therefore to be estimated by the alveolar gas Equation 12 : The specific pathways for hypoxia sensing are not fully understood. 13,14 Nevertheless, low oxygen tensions activate pulmonary arterial smooth muscle contraction and thus increase PVR.
When conventional treatment in ARDS fails, extracorporeal membrane oxygenation (ECMO) is considered as a rescue procedure. 15,16 ECMO is established either by the veno-venous (V-V ECMO) or the veno-arterial (V-A ECMO) approach. 17 V-V ECMO is the method of choice for respiratory failure as long as cardiac function is preserved.
In an experimental study of atelectatic lungs it has been found that PvO 2 >100 mm Hg (>13.3 kPa) effectively reduced PVR. 18 Moreover, case series and reviews propose that V-V ECMO might adequately reduce PVR by increasing PvO 2 when HPV is present during severe respiratory failure, and thus, reduce the right ventricular afterload. 2,[19][20][21][22] The rationale is that the increased PvO 2 would attenuate HPV and PVR by oxygen sensing and signal transduction mechanisms located in the precapillary region independently from P A O 2 . 8 In contrast to these reports, in our experience the increased PvO 2 by V-V ECMO does usually not counteract HPV in the most severe forms of ARDS. 17,23 We hypothesized that this disagreement could be due to a difference in the degree of severity of the ARDS, that is, the amount shunting through lung units with low P A O 2. Therefore, the aim of this study was to carefully dissect the specific influence of PvO 2 and P A O 2 during V-V ECMO on HPV in a porcine model of global alveolar hypoxia.

| Ethics
The study was approved by the local ethics committee for animal research in Uppsala, Sweden, Dnr. 5.8.18-14077/2017. The animals were taken care off according to the ethical guidelines and the European Union's directive. Twelve landrace pigs from a local breeder were used in the study, which was performed at the Hedenstierna Laboratory, Uppsala University, Sweden.

| Anesthesia and mechanical ventilation
Anesthesia was performed as previously reported. 24 After arrival at the laboratory the pigs received an intramuscular injection of tiletamine/zolazepam (6 mg kg −1 , Zoletil ® ; Virbac) and xylazine (2.2 mg kg −1 , Rompun ® , Bayer) into the neck or thigh for induction of anesthesia. After a sufficient degree of anesthesia was established, a peripheral intravascular catheter was placed in a left or right auricular vein. The animals were placed in supine position, and fentanyl (Leptanal ® , Janssen-Cilag AB, Sweden, 2-4.5 µk kg −1 ) was administered before a tracheotomy was performed. Anesthesia was maintained by continuous infusion of midazolam (midazolam Actavis, Actavis Group, Hafnersfjordur, by the ventilator's oxygen sensor during the experiments after its (1)

Editorial Comment
In the setting of advanced global lung injury and respiratory failure, some degree of hypoxic pulmonary vasoconstriction reactivity is presumed to remain present. In this large animal study with controlled global alveolar hypoxia as the model, the effect of veno-venous ECMO for varying blood oxygen values and the effects of this pulmonary vascular resistance was studied. The findings included that the degree of alveolar hypoxia remained important for resulting pulmonary vascular resistance in this model. accuracy was confirmed with a Deltatrac ® (Datex-Ohmeda) in two pilot animals, and continuously recorded during the experiments by a custom-build data acquisition system. During the experiment, preand post-oxygenator blood samples were taken to evaluate membrane lung performance and oxygen transport by ECMO. Blood samples were immediately analyzed on an ABL 500 and an OSM3 blood gas analyzer (Radiometer).

| ECMO
After stabilization and unchanged circulatory and ventilatory parameters for 15 minutes, the animals were cannulated for V-V ECMO. ECMO cannulas were inserted into the right or left femoral vein and the right or left external jugular vein, by either an open surgical or a semi-percutaneous approach. Heparin was administered intravenously before cannulation (50 IE kg −1 , LEO Pharma).
We cannulated three animals for femoro-atrial V-V ECMO with a 21F/38cm cannula (Maquet) for drainage of desaturated blood and a 19F/23cm cannula for reinfusion and six animals for atriofemoral V-V ECMO (21F/38cm or 25F/55cm for drainage and 17, 19, or 21F/23 cm for reinfusion) in a non-randomized manner. For the semi-percutaneous approach the diameter of the vessel was studied with ultrasound before cannulation, the open surgical approach did not require this examination. The largest possible cannula size was chosen to optimize the flow. We planned to treat hemodynamic instability during initiation of ECMO by continuous infusion of norepinephrine and/or dobutamine (mean arterial pressure below 60 mm Hg or cardiac output below 3.0 L min −1 ).
Because alpha and beta agonists may change PVR and thereby bias the results, the protocol did not allow any changes of these drugs during the experiment regardless the hemodynamic situation. Cannula position was examined by X-ray. The position of the cannulas was assumed being correct when the tip of the atrial cannula was located at the beginning of the right atrium and the femoral cannula could not be detected in the same examination.  Novalung, X-lung set 3/8 " (Xenios AG) (n = 7).

| Experimental design
The experiment was designed as a longitudinal study with repeatedmeasures within subjects where every animal was exposed to different experimental conditions. We studied the effect of increasing  15 (12)(13)(14)(15)(16)(17) Note: Weight in kg; PIP: peak inspiratory pressure (cmH After the last measurements, the animals, still in deep anesthesia, were killed by an intravenous injection of 20 mL (100 mmoL) KCl.

| Calculations
PsO 2 (mm Hg) and P A O 2 (mm Hg) were calculated according to

| Sample size and power calculation
We calculated the sample size before study start based on the results of published case series and one experimental study. 19,20,26 We expected a change of approximately 20%-25% of MPAP between two different conditions during the experiment. To reach a power of 80% (1-β = 0.8) with a moderate effect size and a type I error of 5% the number of animals to be studied was calculated as n = 32. After analysis of data of 9 of 12 consecutive included animals the study had reached a power of 99% for MPAP and PVR. Therefore, the study was stopped. F I G U R E 2 Rank differences plots of (A) pulmonary vascular resistance (PVR) and (B) mean pulmonary artery pressure (MPAP). Nine comparisons were performed (see Figure 1). Individual values and median with interquartile range for rank differences are displayed for every matched pair comparison. ***P < .001; **P < .01; *P < .05    pulmonary artery oxygen tension (kPa); PsO A P-value of <.05 (two sided tests) was considered significant.

| Statistical calculation
We report the multiplicity-adjusted P-value of the rank sum differences for each comparison.

| RE SULTS
The data from nine animals (six females and three males; weight 44.7-58.  Table 1. Detailed data of the atrio-femoral and femoro-atrial groups are presented in a digital online supplement.

| ENTIRE S TUDY P OPUL ATI ON
Rank sums differed significantly in seven of nine comparisons for PVR and in eight for MPAP (Figure 2A

| D ISCUSS I ON
This is, to the best of our knowledge, the first controlled experimental study that has investigated the effect of increasing PvO 2 by In a small case series, Reis Miranda and co-workers found that initiating V-V ECMO in patients with severe ARDS resulted in a statistical significant, but minor decrease in MPAP. 19 However, these patients had respiratory acidosis before initiation of ECMO, which may had contributed to the pulmonary hypertension. In a case report by Mongodi and colleagues, initiation of V-V ECMO was associated with improved right ventricular function and stabilized circulation. 20 However, first, the results in case and case series reports are prone to chance, and second, it is not possible to know whether the reduction in right ventricular afterload in the reports were due to an increase in PvO 2 , a combination of improved oxygenation in PvO 2 and increased pH by decarboxylation, by the corrected respiratory acidosis alone, or by change in ventilator settings . 28,29 In our study mechanical ventilation and pH were kept constant throughout the entire experiment and did therefore not influence the results. Bishop Figure 3A-C).
There is no consensus about ventilator settings during V-V ECMO, but usually F I O 2 will be reduced significantly to avoid or reduce toxic effects of high oxygen concentrations. 32 This and reducing inspiratory pressure and PEEP may lead to decreased alveolar oxygen concentrations in the ventilated areas of the lung. As a consequence the increase in PVR will result in increased right ventricular afterload. Therefore, in our clinical experience, the progression of right ventricular failure during V-V ECMO is usually depended on lung function and not an "ominous sign, pointing to lack of recovery" as described by Krishnan and Schmidt. 33 Attempts to increase PvO 2 by increasing the extracorporeal flow usually fail to reduce PVR. Hence, the results from this study support our clinical experience, and we are convinced that the circuit should preferentially be changed to V-A ECMO in these situations to stabilize the circulation. 17,23

| LI M ITATI O N S
Our study has several limitations: (1) this is an experimental study in an animal model with all its inherent limitations, and although the HPV response to hypoxia is similar in most mammals, the results can only cautiously be transferred to human conditions. (2) Three animals could not be studied due to complications during pre-experimental preparations. However, the set-up of this study was elaborate, and all complications were procedural. According to the 3-R principle, we did not include the a priori calculated 32 animals because the results were clear-cut with a calculated power of 99% after nine animals.
(3) We studied a model of global alveolar hypoxia in normal lungs and not of ARDS. This was done to estimate P A O 2 , with the assumption that the results from an ARDS model would give similar results when HPV is present. However, this notion has not been examined in this study. (4) The estimated P A O 2 when V-V ECMO was started after inducing pulmonary hypertension with an F I O 2 0.15 could be somewhat higher than the real value, due to CO 2 removal by ECMO, reducing the predicted RQ of 0.8. 34 (5) ECMO may theoretically interfere with thermodilution and continuous cardiac output measurements with a pulmonary artery catheter, but, to the best of our knowledge, this has not been studied yet. However, due to these reasons, our findings regarding cardiac output should be interpreted cautiously. Moreover, the long updating time during continuous cardiac output measurement could make sudden changes in cardiac output undetectable. 35 Since there was no relationship between MPAP and CO, the increase in MPAP could not be due to increased flow.
Moreover, MPAP and PVR were strongly correlated indicating that changes in MPAP adequately represented the changes in PVR.

| CON CLUS ION
In this porcine model of alveolar hypoxia, V-V ECMO, despite it increased PvO 2, did not decrease MPAP, and PVR, or attenuated systemic hypoxemia at F I O 2 below 0.15. Our results could contribute to explain the fact that V-V ECMO sometimes in our experience fails to stabilize hemodynamics in patients with the most severe forms of ARDS. In these patients conversion from V-V to V-A ECMO is usually needed until the lungs have recovered.

ACK N OWLED G M ENTS
The authors thank Kerstin Ahlgren, Mariette Andersson, Liselotte Pihl, Agneta Roneus, and Maria Swälas from the Hedenstierna laboratory for their invaluable help with performing the experiments.

CO N FLI C T O F I NTE R E S T
BH has received lecture fees from Xenios. HK, SE, and AL do not report any conflict of interest.