The inflammatory response to extracorporeal membrane oxygenation (ECMO): a review of the pathophysiology

Extracorporeal membrane oxygenation (ECMO) is a technology capable of providing short-term mechanical support to the heart, lungs or both. Despite having first been employed clinically in the 1970s [1], the more widespread use of ECMO in critically ill adult patients is a recent phenomenon [2]. Over the last decade, the number of centres offering ECMO has grown rapidly. At the same time, the indications for ECMO in adults have expanded beyond acute severe respiratory and cardiac failure [3] to include extracorporeal cardiopulmonary resuscitation (ECPR) [4] and as a bridge to lung transplantation [5]. Due to its origins, the use of ECMO is much better established in neonatal and paediatric populations [6]. In fact, neonatal ECMO for respiratory failure accounts for 43% of the Extracorporeal Life Support Organisation (ELSO) registry, comprising 25 years of records of paediatric and adult patients undergoing ECMO [7].

Survival rates in neonates and children supported with ECMO, especially in cases of respiratory failure, are generally high [8]. For adults, it is difficult to quantify the mortality of patients undergoing ECMO. Several studies have varied in their reported mortalities based on indication and modality [8, 9], ranging from 76% in one cohort undergoing ECMO and dialysis [10] to 37% in a mixed veno-venous (VV)/veno-arterial (VA) ECMO group [11]. The ELSO registry, between 1989 and 2014, reports mortalities for respiratory and cardiac ECMO, of 57% and 41% respectively [7]. In addition, there are a number of complications that may occur, any of which is capable of inflicting serious morbidity. These can be broadly separated into those related to the ECMO device (e.g., oxygenator or pump malfunction, circuit clotting, cannula issues) and physiological complications (e.g., bleeding, haemolysis, and infection). In this review, we examine the pathophysiology of one less well-recognised complication, the inflammatory response to ECMO.

The initiation of ECMO is associated with an immediate and complex inflammatory reaction, similar to that seen in systemic inflammatory response syndrome (SIRS) [12]. At that moment when the patient’s blood first comes into contact with the foreign surface of the extracorporeal circuit, a variety of coagulative and inflammatory cascades are activated (Fig. 1). Levels of pro-inflammatory cytokines rise rapidly [13‐19], which, in association with activation of the complement and contact systems [20‐25], results in leukocyte activation [26‐28]. This innate immune response, if severe, persistent or unchecked by a compensatory anti-inflammatory response (CARS) [29], may lead to endothelial injury, disrupted microcirculation, and end-organ dysfunction [30‐35]. Despite major improvements in pump and circuit design, oxygenators and the advent of heparin-bonded surfaces, the SIRS response to ECMO remains a clinical concern. While a large volume of work has been directed towards elucidating and targeting the inflammatory response to cardiopulmonary bypass (CPB) [36‐38], a closely related form of extracorporeal circulation, much less has been devoted to studying the inflammatory response to ECMO.

ECMO and CPB are closely related and share several similarities. Both techniques generate a biomaterial-induced inflammatory response in patients. Despite this, there are important distinctions between ECMO and CPB (Table 1), of which anyone attempting to infer from one to the other must be aware.

The most obvious distinction between ECMO and CPB is in the duration of support provided. CPB is commonly employed for only minutes to hours to facilitate a surgical procedure. Conversely, ECMO, used in patients with severe organ failure, can be prolonged for weeks to months. This difference in duration requires a different approach to anticoagulation. During CPB, loading doses of unfractionated heparin between 300 and 500 U/kg may be used [39], versus 40–80 U/kg during ECMO [40]. Upon the completion of CPB, protamine sulphate is administered to reverse the effects of heparin, a practice which is avoided during and at the conclusion of ECMO. This is an important distinction, given that the formation of protamine-heparin complexes is known to exacerbate the inflammatory response (via activation of the classical and lectin complement pathways) [41, 42].

Hypothermia is mandatory during CPB, but not commonly employed during ECMO. Likewise, haemodilution, which may also be employed during CPB, is not seen to the same degree during ECMO. Large observational studies have demonstrated an association between the lowest haematocrit recorded during CPB and post-operative mortality [43]. One suggested explanation for this is that haemodilution leads to increased neutrophil activation [44].

As part of the surgical nature of conventional CPB, cardiotomy suctioning, venting of blood and venous reservoirs are incorporated into circuits. This introduces a blood-air interface. Multiple studies have detected higher levels of pro-inflammatory cytokines in cardiotomy-suctioned blood and, in some instances, lower levels of anti-inflammatory cytokines, such as IL-10 [45‐47]. The absence of an air/blood interface during ECMO may be a factor in reducing the inflammatory response.

Commonly, perfusion during CPB is non-pulsatile [48]. During VA-ECMO, depending on the residual function of the native heart, varying degrees of pulsatile flow are generated in the face of retrograde arterial perfusion by the ECMO system. This is not the case in VV-ECMO, which uses an ‘“in series configuration’, relying on the native heart for systemic perfusion [49]. Evidence from CPB suggests that pulsatile flow during extracorporeal circulation may act to reduce the inflammatory response [50]. It has been hypothesised that this may be due to the ability of pulsatile perfusion to better sustain the functional state of the microcirculation [51].

Contrary to ECMO, CPB involves inflicting an ischaemia-reperfusion injury [36]. Clamping of the aorta during surgery renders the heart, and to a large extent the lungs, ischaemic. At the completion of surgery, this clamp is removed and both organs undergo a period of reperfusion. Each phase causes a significant inflammatory reaction. In some patients this may prove significant, leading to the onset of lung ischaemia-reperfusion injury or so-called pump lung [52].

Finally, there are clear patient differences. The majority of patients undergoing CPB will have chronic disease (often severe), but are unlikely to be acutely unwell at the time of the procedure. Conversely, those undergoing ECMO are usually critically ill.

The inflammatory response caused by ECMO occurs as a reaction to the exposure of blood to the extracorporeal circulation. Both systemic and cellular factors initiate and propagate the SIRS-like cascade. These include several humoral and cellular systems, most notably the contact, intrinsic and extrinsic coagulation, and the complement systems and endothelial cells, leukocytes, platelets and cytokines. Here we will consider each in turn, although it should be borne in mind that these systems are largely interdependent and still not fully understood.

The complement system acts at the forefront of the innate immune response and, as such, is one of the body’s first host defences to be activated when ECMO is initiated [23]. An interactive cascade of over 30 complement-related proteins acts to destroy and remove foreign substances identified as “non-self”, in addition to substances released from damaged cells. Complement achieves this either by direct cell lysis, by producing a trans-membrane pore known as a membrane attack complex (MAC), or by modulating leukocytes via opsonisation or the production of pro-inflammatory anaphylatoxins [75]. The complement cascade may be activated by any or all of three pathways: the classical pathway (CP), the alternative pathway (AP), and the lectin pathway (LP).

Across the three pathways, the first common step in the cascade is the cleavage of C3 to C3a and C3b (Fig. 1). In the alternative pathway, subsequent steps result in C5 being cleaved to form C5a and C5b. The two resulting molecules have distinct effects. C5a is a potent pro-inflammatory mediator (C3a is also an anaphylatoxin, albeit not as potent) with several abilities in vivo, including increasing leukocyte recruitment, increasing vascular permeability and inciting further inflammatory mediator release [76]. On the other hand, C5b progresses through several steps to produce C5b-9 (the MAC). Given its antibody independence, it is thought that the AP is the principal means of complement activation during extracorporeal circulation (ECC), occurring as a result of blood contacting the foreign material of the circuit [77].

Although the complement response to CPB has been well-described [78‐81], the role of complement activation during ECMO is, by contrast, poorly defined. The few available studies date from the 1990s and were conducted using the much less advanced pump and circuit technology that was available to investigators at that time [19, 21‐25]. This is important, as improvements in both circuitry and pump technology have likely had a beneficial effect on the levels of complement activation seen during ECMO. Studies on CPB have shown us that the use of centrifugal pumps, now ubiquitous in modern ECMO, reduces complement activation [82, 83]. In addition, the use of contemporary heparin-bonded circuits has also been shown to reduce complement activation [84]. Studies have examined complement activation in ex-vivo models of ECMO [20, 22, 24], neonatal ECMO [20, 23, 25] and in adult patients [21]. In keeping with CPB, all of these studies revealed rapid elevations in the levels of complement factors during ECMO, with peaks occurring within 1 to 2 hours.

As previously stated, a key differentiator between the extracorporeal circulation of CPB and ECMO is their relative duration. This effect has been described by several authors. In two adult patients on ECMO for acute respiratory distress syndrome (ARDS), Vallhonrat et al., described rapid elevations in the levels of C3b and terminal complement complex (TCC) (fluid-phase MAC), which were evident within 15 minutes of commencing ECMO. These levels peaked by 60 minutes and had begun to fall by 180 minutes, reaching near-baseline levels at 2 days. Of note was the failure to demonstrate any significant elevation in C4d, a classical pathway marker [21]. In neonates, Graulich et al. detected peaks of C3a at 1 hour and of C5a and TCC at 2 hours. However, contrary to what Vallhonrat et al. observed, TCC levels did not return to baseline [20].

Ex-vivo studies of complement activation in ECMO have almost universally failed to uncover peak or falling levels of complement products. Using heparinised fresh whole human blood in an ex-vivo circuit, Moen et al. recorded levels of C3 activation products and TCC continuing to rise after 72 hours of ECMO [22]; this points towards the presence of a negative feedback mechanism in-vivo, which acts to limit inappropriate complement activation. Bergman et al. employed an ex-vivo model of ECMO, with a mixture of non-heparinised citrated fresh human blood and Ringer’s solution, to examine the influence of complement activation during ECMO on blood cell rheological properties. Close correlation between the levels of complement factors, particularly C5a and TCC, and white cell deformability were noted, leading the authors to conclude that high levels of complement activation during ECMO may play a part in the rheological deterioration of blood cells [24].

The initiation of ECMO results in the production of a variety of pro-inflammatory and anti-inflammatory cytokines. These small proteins have a variety of roles in cell signalling and are important mediators of the innate immune response. There is little evidence to support a single “master” cytokine in this process. This has implications in selecting potential therapeutic interventions. While interventions aimed at single cytokine targets are unlikely to overcome the redundancy in the system, more broad-based approaches, such as cell therapy, appear to have significant potential. We subsequently describe the most studied cytokines with respect to ECMO.

TNF-α is produced by a range of cell types in response to an inflammatory stimulus. TNF-α has pleiotropic effects on a variety of tissues; but, perhaps most importantly, it is a potent activator of neutrophils [98]. Furthermore, it induces the expression of adhesion molecules by endothelial cells, enhancing the margination of neutrophils. TNF-α can also stimulate macrophage phagocytosis, enhance the expression of prostaglandins and increase the formation of thrombin [99]. TNF-α acts as a negative inotrope at the macrocirculatory level. During ECMO, it is thought that a large proportion of early TNF-α release originates from pre-formed stores in mast cells. In a porcine ECMO study, McIlwain et al. identified gut mucosal mast cells as the key contributor to early TNF-α release. During the first 2 hours of ECMO, increased plasma TNF-α was not associated with increased tissue synthesis. This, combined with a decreased level of TNF-α found in gut mucosa, led the authors to hypothesise that mast cell degranulation, in response to complement activation, is the likely mechanism by which TNF-α concentrations are elevated [15]. Higher levels of TNF-α have been associated with non-survival in neonates undergoing ECMO [23, 27].

Interleukin-6 is a complicated cytokine, with both pro- and anti-inflammatory actions. Whilst promoting the acute phase reaction, expansion and activation of T-cells and the differentiation of B cells, IL-6 may also down-regulate the expression of other pro-inflammatory cytokines, as well as up-regulate the expression of the anti-inflammatory cytokines [100]. Studies have reported consistent elevations in IL-6 levels during ECMO [15, 18, 30, 32, 101]. IL-6 levels in the lungs have also been shown to rise after VV-ECMO in an animal model, where they were associated with parenchymal damage [30]. Interestingly, given that IL-6 has a key role initiating the acute phase response and its concentration is thus usually closely coupled to C-reactive protein (CRP) production, McIlwain et al. failed to demonstrate any link between the two during porcine ECMO [15]. In a combined study by Risnes et al., involving children and adults supported with ECMO, levels of IL-6 were inversely associated with survival, with a clear divergence in IL-6 concentrations between survivors and non-survivors after two days of ECMO support: survivors effectively normalised their IL-6 levels whilst those who died maintained a persistent elevation. This was not observed to be the case for other cytokines, such as IL1-β, IL-8 or IL-10 [102].

IL-8 is a potent neutrophil activator and a chemoattractant for neutrophils, basophils and T-lymphocytes. Studies using simulations and animal models and in neonates have demonstrated rapid elevations of IL-8 after ECMO commences [15, 18, 27]. Alterations in IL-8 concentration appear to follow a temporal trend similar to TNF-α, with increases noted within the first 15 minutes of ECMO [15, 27].

A number of other cytokines have also been implicated in the inflammatory response to ECMO and extracorporeal circulation. Of note is that low levels of the anti-inflammatory cytokine IL-10 (at the time of ECMO initiation) have been associated with poor survival [103].

In the last decade, ECMO technology has improved dramatically. This has facilitated its uptake in an increasing number of centres for an ever-growing range of indications. Ongoing research into anticoagulation, servo control and membrane lung construction will further improve the safety and efficacy of ECMO [104]. Additional research (pre-clinical and clinical) is required to assess the impact of ECMO-induced inflammation on clinically meaningful outcomes. This work will inform the requirement for novel anti-inflammatory therapies that could be used during ECMO. In future, when designing strategies to minimise the inflammatory response to ECMO, several experimental treatments appear to have potential, in particular, the use of Factor XII inhibitors and mesenchymal stromal cell (MSC) therapy.

Currently, there is substantial interest in targeting factor XII as a novel means of anticoagulation [105]. The use of a factor XII inhibitor in this setting is attractive given the potential to reduce, or eliminate, the risk of thrombosis without encountering the bleeding risk associated with current anticoagulants. A beneficial side-effect would be the inhibition of contact-mediated inflammation involving the activation of factor XII. A recombinant human antibody, 3 F7, which inhibits FXIIa, has been successfully tested in a rabbit model of VA-ECMO [55]. Any future work will need to demonstrate the safety and efficacy of factor XII inhibitors in a large animal model of ECMO, using contemporary clinical ECMO equipment, before progression to human studies.

MSC therapy is being investigated as a treatment for a number of acute inflammatory conditions. MSCs are multipotent adult stem cells found in most tissues of mesodermal origin, such as bone marrow [106]. MSCs have attracted attention due to their immunomodulatory effects. In-vitro MSCs have been shown to influence both the adaptive and innate immune response to infection and inflammation; this ability is bi-directional, that is to say, MSCs may either promote or ameliorate inflammation dependent on the contemporary milieu [107]. MSCs have been studied in a pre-clinical model of CPB as a means of reducing the harm associated with ischaemia-reperfusion injury [108]. Here, infusion of human MSCs significantly reduced the levels of inflammatory cytokines within 3 hours. Given their ability to influence the innate immune response at a range of points in the cascade and in a manner consistent with the local environment, MSCs have the potential to be efficacious in the setting of ECMO-induced inflammation.

The inflammatory response to ECMO is complex and multi-faceted. It remains unclear whether this excess inflammation is all deleterious, or if it has potential benefits to the host. In summary, it arises principally due to the contact and complement systems becoming activated as a result of blood exposure to the extracorporeal circuit. The combination of a sustained innate immune response and the pro-inflammatory aspects of coagulation result in “pan-endothelial” injury, with leukocyte activation and the production of pro-inflammatory mediators. This ultimately ends in a systemic inflammatory response and end-organ damage.

Whilst this process shares a number of similarities with the inflammatory response witnessed during CPB, there are also a number of important distinctions. Given the recent increase in the use of ECMO, there is a relative paucity of data on the inflammatory response that it induces. This is compounded by the fact that many of the available studies were conducted prior to the introduction of modern innovations in ECMO technology. Mechanistic and outcome information from adult patients is absent. Clearly all patients supported on ECMO experience some degree of inflammatory response to ECMO, but currently there is no physiological rationale to explain the extreme variation seen in the magnitude of the patients’ responses to that inflammation, nor their clinical course. Whilst it is undoubtedly difficult to isolate the relative contribution of specific components of the inflammatory response (pro-inflammatory and CARS) to outcomes after ECMO, failure to do so may lead to a missed opportunity for intervention. The first step towards addressing this deficit is to not only improve our understanding of the basic science underpinning the inflammatory response induced by modern ECMO, but to adopt minimal reporting criteria for the inflammatory response during human ECMO studies, in a manner similar to that proposed for CPB [109]. Such studies should be conducted in models that are relevant, using technology that is in contemporary clinical use. A more detailed understanding of the underlying processes will allow us to better investigate potential therapies and design future clinical studies.