Cytokine removal in human septic shock: Where are we and where are we going?

Septic shock continues to have significant mortality [1]. The underlying pathophysiology is complex with both pathogenic and host factors (PAMPs (pathogen-associated molecular patterns) and DAMPs (damage-associated molecular patterns)) playing a significant role in the development and subsequent outcome [2]. However, the heterogeneity of septic shock prevents adequate characterization of patients and may hinder subsequent clinical intervention(s). Sepsis and septic shock affect anywhere between 100 to 1000 per 100,000 person-years and 19 per 100,000 person-years depending on the cohort studied [3‐6], with reported mortality rates ranging between 20 to 50% [7‐11]. Moreover, the reported incidence is increasing, although this may be attributed to reporting bias in so-called claims-based databases, as data analysis on electronic health records cannot confirm this trend [12‐15]. Therefore, it is of no surprise that the treatment of sepsis has become a major global health issue. Indeed, in the USA during 2011 sepsis accounted for just over 5% of total hospital costs corresponding to $20 billion dollars [16].

Patients with sepsis are often treated in areas of intensive care given that close monitoring and intense therapeutic support are needed [17]. Early treatment includes the use of timely, appropriate antibiotics, intravenous fluids, oxygen therapy as well as vasopressor and inotropic support where needed. Other additional treatments including extracorporeal or so-called blood purification techniques (BPT) have also been tried [18]. These techniques include (among others): haemofiltration, haemoperfusion, intermittent or continuous high-volume haemofiltration (HVHF), plasmapheresis or adsorption. The rationale behind such an approach is to achieve “immune homeostasis” which theoretically reduces the potential damage caused by dysregulation of the host response to infection. This may be heralded by a profound rise in inflammatory mediators including cytokines which contribute to the dramatic systemic effects of sepsis, mainly in septic shock [9, 19]. The recently updated Sepsis 3.0 consensus definitions state that sepsis is an infection accompanied by life-threatening organ dysfunction caused by a dysregulated host response [20]. Given the pivotal role of cytokine production in sepsis, it follows that removal of these substances, through such BPT, may attenuate the response particularly in the early phase of sepsis [21]. Several hypotheses have been proposed as to the potential mechanisms underpinning potential benefit. These include cytotoxic theories including the peak concentration hypothesis whereby all inflammatory mediators are removed at a given rate, dependent on the BPT used and assuming they are filtered [21‐24]. Alternatively, the cytokinetic theory proposes that cytokines are removed, thereby creating a cytokine gradient between the bloodstream and tissues allowing leucocyte-enhanced trafficking [25]. In the same line, cytokine levels can also be seen as communicating messengers to talk to cells, recruit some, depress others and reduce cell metabolism for others [22]. Despite early promise, no multicentre randomized controlled studies have demonstrated a survival benefit including the use of HVHF where higher flows may lead to increased cytokine removal were tried (the cytotoxic threshold immune modulation hypothesis) [26‐28]. Other extracorporeal blood purification therapies also have failed with significant outcome data lacking with no treatment demonstrating a translatable survival benefit in any randomized controlled study [29‐31]. This somewhat too simplistic view takes into account several new concepts. Indeed, blood level of mediators (more than cytokines) implies the saturation of the interstitial and cellular compartments to be present in the blood. It is the so-called tip of the iceberg theory [31]. This theory is very similar to the “threshold immune modulation theory” [23]. In addition, plasma mediator levels depend on several factors: the intensity of production, the number of cell receptors availability, the clearance of such mediators, the affinity of the receptors for such mediators. As an example, the interleukin (IL)-6 receptor is an agonist one, which differs from tumour necrosis factor (TNF)-soluble receptors and IL-1 receptors that are inhibitory; the consequence is that a low IL-6 level with a high level of receptor induces more cellular response than high level of IL-6 with a low level of receptors [22]. This very important issue is showing that an IL-6 level alone may not be very predictive of the future response of the organism. DAMPs are also playing a key role and especially the endosomal DAMPs that are eliminated via the BPT and are capable of inducing cellular damage and apoptosis [32]. This may explain the initial “positive” observational trials with HVHF. The reported improvement in hemodynamic status associated with or not with lactate reduction might mainly be the result of the fluid replacement therapy made with large volumes of crystalloids containing high buffer concentrations. As a consequence, the induced pH increase might have changed the affinity for catecholamines to their receptors improving hypotension [28‐30].

Extracorporeal haemoperfusion with Polymyxin B (PMX-HP) has shown improvement in organ dysfunction and a survival benefit in small studies [33] including a small randomized trial [34], while larger trials failed to confirm these findings [35, 36]. Evaluating the use of PMX-HP in a randomized controlled trial of adults treated for endotoxemia and septic shock (the EUPHRATES study) included patients with persistent septic shock despite adequate fluid resuscitation and vasopressor treatment [37]. An endotoxin activity assay (EAA) was applied. No mortality difference was observed in the “per protocol population” (n =  244, multiple organ dysfunction score (MODS) > 9, EAA ≥ 0.6). However, among the patients in refractory shock with MODS of more than 9 and an EAA between 0.6 and 0.9, a significant 10.7% reduction in 28-day mortality in a post hoc analysis was recognized after receiving two sessions of PMX-B-HP [38]. Post-hoc analysis revealed a significant mortality reduction when the EAA was limited to less than 0.9, which may suggest an upper limit to a pre-treatment endotoxin burden under these treatment conditions [38]. A recent meta-analysis of 17 trials demonstrated that PMX-B-HP treatment may reduce mortality in patients with severe sepsis and septic shock. The included studies were stratified into three groups based on the mortality rates of the conventional treatment group: low-risk group (mortality rate < 0.3), intermediate-risk group (0.3–0.6) and high-risk group (> 0.6). Risk ratios with 95% confidence intervals (CI) for the mortality-stratified analysis between the PMX-HP and conventional treatment groups were calculated and presented as summary statistics. This a posteriori classification can be seen as a major caveat of this meta-analysis. Also mortality rate does not reflect cytokine concentrations and load and therefore is not a predictor of response to cytokine removal. Disease severity subgroup meta-analysis revealed a significant risk reduction in overall mortality in the intermediate-risk (mortality rate 0.3–0.6) and high-risk (mortality rate > 0.6) groups, but not in the low-risk (mortality rate < 0.3) group reinforcing the view that rigorous patient selection is crucial for treatment success [39].

Lastly, two editorials set the pace by stating that while blood purification in sepsis remains a valid approach, the use of adsorbers and their current efficacy of endotoxin/cytokine elimination cannot be recommended to reduce the mortality in absence of positive randomized controlled trials (RCTs) [40, 41].

Capillary leak represents the maladaptive and undesirable movement of fluid and electrolytes with or without protein into the interstitium that generates anasarca and end-organ oedema potentiating organ dysfunction and or failure [42]. The global increased permeability syndrome (GIPS) defined as a positive cumulative fluid balance and new onset organ dysfunction/failure is described in patients with persistent systemic inflammation resulting in continuing transcapillary albumin leakage. GIPS may represent the third phase in a continuum after the initial cytokine storm and ischaemia–reperfusion injury and could be a potential indication to start BPT [43].

The focus of this consensus meeting was to determine whether a reduction in cytokine levels is possible through the use of sorbent technology by use of the CytoSorb^® (Cytosorbents, Corporation, New Jersey, USA) device. This consists of a single-use haemoadsorption cartridge which can be used with standard blood pumps, such as those found on RRT machines or through a haemoperfusion device [44‐46]. The CytoSorb cartridge is the only currently available CE-marked device shown to consistently lower excessive cytokines in severe sepsis. The cartridge is filled with sorbent beads made from a porous polymer that adsorbs and capture cytokines as blood passes through the device. This process is concentration dependent, and so the higher the levels of cytokines in the blood, the faster the levels are reduced. Although CytoSorb therapy has been described for other indications, including intoxications, rhabdomyolysis, hyperbilirubinemia, etc., we chose to concentrate on human studies in sepsis and septic shock.

The enhanced inflammatory response seen in septic shock is associated with a high mortality [57, 58] correlated with the production of pro- and anti-inflammatory mediators [57] rather than disequilibrium between pro- and anti-inflammatory mediators [59] (Fig. 3). This has stimulated much effort towards potential attenuation of this response particularly as early studies suggested that continuous veno-venous haemofiltration (CVVH) may reduce cytokine levels [60]. However, as discussed, these early observations have not translated into clinical benefit. Meta-analysis suggests that the only potentially effective BPT for the treatment of sepsis are plasma exchange or haemoadsorption particularly with Polymyxin B [2, 39]. Although initial results using Polymyxin B haemoperfusion showed promise [31], this has not been borne out by subsequent randomized clinical trials [32]. Given the remit of sorbent technologies, we chose to focus on the use of CytoSorb as a haemoadsorption device. The adsorber has a surface of about 45,000 m² compared to a conventional hemofilter with a surface of 1–1.5 m² with a molecular cutoff of about 60 kDa removing cytokines as well as other toxins and drugs. As a consequence, CytoSorb does not adsorb endotoxin which has a molecular weight of 100 kDa. CytoSorb is saturable regarding adsorption in the clinical setting (mostly after 8 h) as evidenced by an rebound increase in the dose of vasopressors which can be tapered when changing the CytoSorb. So far, in vitro studies have not been evaluated regarding the adsorption–saturation process of adsorption. This could be an interesting area for future research. Multiple pre-clinical studies using animal models of sepsis have demonstrated reductions in various circulating cytokines and chemokines, reduced organ injury, and improved survival [61‐64].

In a feasibility study, the removal of cytokines was confirmed with cytokine extraction rates approaching 30% [65]. In this study, plasma concentrations of both IL-6 and TNF-α, but not IL-10, were significantly reduced after the first hour of therapy [65]. Overall removal was greatest for IL-6, 28% (p = 0.006), and least for tumour necrosis factor, 8.5% (p = 0.13). However, plasma concentrations for all three cytokines increased over time and were above baseline by the end of the intervention (4 h). In a more recent study, the group of Antoine Schneider [66] conducted a single-centre pilot randomized controlled trial in 30 patients undergoing elective cardiac surgery and deemed at risk of complications. Patients were randomly allocated to either standard of care (n = 15) or CytoSorb^® haemoadsorption (n = 15) during cardiopulmonary bypass (CPB). The primary outcome was the difference between the two groups in cytokines levels (IL-1a, IL-1b, IL-2, IL-4, IL-5, IL-6, IL-10, TNF-α, interferon gamma (IFN-γ), monocyte chemotactic protein-1 (MCP-1) measured at anaesthesia induction, at the end of CPB, sand 6 and 24 h post-CPB initiation. However, the intervention was associated neither with a decrease in pro- or anti-inflammatory cytokine levels nor with any improvement in relevant clinical outcomes [66]. These results are conflicting, and this is obviously a major concern regarding this new technique. So, altogether, no study using CytoSorb has shown a sustained plasma concentration reduction lasting during the whole treatment. Given the possibility that this technique may hold promise, we reviewed available data with the aim of producing current consensus statements and directing further research [67‐69].

The aim of any treatment is to select only patients who have a high probability of benefit from the intervention. To date, most of the case series included patients with a high predicted mortality in keeping with septic shock although few have included cytokine estimation [64, 71]. Where recorded, there is significant variation in IL-6 levels, and therefore, one may conclude that some patients were treated where benefit would be negligible if indeed [IL-6] levels are predictive of response to therapy. Therefore, any future studies should attempt to enrich the study population through cytokine measurement or another marker of potential therapeutic response such as PCT [70]. A threshold PCT concentration may be included as an inclusion factor in future studies. Alternatively, patient selection may concentrate on those individuals who require high doses of vasopressor agents to maintain adequate organ perfusion as a reduction in such drugs is a meaningful clinical outcome which may translate into other benefits [78, 79].

The timing of therapy remains an issue in many aspects of critical care medicine not least those involving extracorporeal therapies. Several of the case series report a less favourable outcome in those individuals who commenced treatment more than 24 h after diagnosis [70]. Data of the ACESS trial also support the concept of starting treatment within 24 h [78, 79]. It follows that any recommendation regarding timing of therapy will be dependent on the patient population and how prospective individuals are selected. Where the primary endpoint is a reduction in vasopressor, then timing will be relatively easy to define, whereas cytokine levels may be less clear especially given the differences in kinetics. Certainly, the rationale for any of the blood purification therapies would point to commencing treatment early in the disease process in order to maximize benefit. Maybe starting therapy directly in operating room during surgery for peritonitis patients with already hemodynamic impairment and catecholamine requirement should be considered for future studies. However, timing to intervene might be even more complex especially when sepsis-induced immunodepression is taken in account. This depression is essential to limit the consequences on the host tissues of the host response related to infection (i.e. maintain the inflammation–repair–healing cycle). Characterization of this immunodepression might be one of the best indicators to decide when to use adsorber technique to reduce inflammation, namely when the downregulation is modest or absent [22].

The optimum length of treatment is still undecided. Moreover, the length of time of initial therapy is also unclear. Schadler et al. treated patients for 6 h a day for 7 days [73, 74], whereas other reported studies either did not describe the treatment period or used the cartridge for 24 h [70, 71]. No evidence to date exists for the appropriate duration of adsorption therapy or indeed the initial treatment period. Again, this may be directed in future studies by the use of IL-6 or rather PCT as a surrogate for total cytokine removal and reduction in systemic levels coupled with clinical improvement may be a logical place to start regarding treatment times. Also, it is unknown whether the initial therapy should be for 12 or 24 h. Again, a pragmatic approach may be directed by cytokine estimation. Again, the study by Schadler [73, 74] may provide some answers. Although the 6-h treatment did not result in an overall reduction in IL-6 compared to controls over a 24-h period, there was increased IL-6 elimination with the use of haemoadsorption [73, 74]. It follows that if the aim of treatment is to reduce cytokine levels, then a prolonged treatment period should be associated with a reduction in systemic levels. However, this removal in cytokines may be attenuated by cytokine shift from the interstitium into the blood compartment thereby negating any overall effect [22, 68, 69]. As a consequence, pharmacokinetic studies on cytokine clearance would be extremely useful to conduct in the future to better delineate the therapy duration of hemadsorption in relation to benefits for the patient with a cytokine storm.

Future studies should aim to include as homogeneous cohorts of septic shock patients, especially those with very high vasopressor needs. Another interesting population to study may be patients with acute respiratory distress syndrome (ARDS). Mortality rate of ARDS still remains high between 30 and 50% [80]. ARDS is a heterogeneous syndrome, characterized by increased pulmonary capillary permeability [81]. It is the accumulation of protein-rich fluid inside the alveoli that triggers damage to the capillary endothelium and alveolar epithelium leading to release of cytokines, producing diffuse alveolar damage [82]. Extracorporeal membrane oxygenation (ECMO) may serve as a salvage therapy with the CESAR trial that have led to an enhanced use of this therapy [83]. Several case reports mention the positive effect of incorporation of CysoSorb therapy for respiratory failure due to ARDS, using ECMO and CytoSorb therapy. The use of CytoSorb appeared to result in rapid resolution of neutropenia, reversal of toxic shock and rapid weaning of the high-dose vasopressor infusions and a significant reduction in the levels of circulating inflammatory mediators [84‐88].

The intervention is more likely to benefit patients with high severity of illness (e.g. APACHE-II greater than 25) [75]. Refractory shock indicated by high doses of vasopressor support and early multiple organ failure with some evidence of a cytokine storm (i.e.: high PCT) should be preferred [70, 71, 79].

It is uncertain whether cytokine concentrations can identify the ideal patient for treatment. Detecting persistent ‘cytokine storm’—despite adequate resuscitation and source control—could be an alarming signal that cytokine removal may be beneficial. PCT as one of the most studied inflammatory biomarker in sepsis could be added as a biomarker to determine in which patient to start therapy [87, 88]. It is difficult to determine the absolute cutoff values above which treatment may be indicated, but there is some evidence, that PCT kinetics, i.e. failure of PCT to decline or rapidly increasing levels may help to identify persisting cytokine storm [89]. It is indeed the reality that the molecular weight of PCT is about 13 kDa and PCT might be removed by the adsorbent. Therefore, PCT is a good biomarker to decide when to start but not to follow the response to therapy nor to decide when to stop. IL-6 may also be a promising biomarker, but at present, little is known about IL-6 kinetics, and its response to appropriate or inappropriate treatment, which needs further research. Nevertheless, the biomarkers to characterize inflammatory status cannot be summarized by limited to IL-6 associated and PCT: we also do need to obtain information on cellular function (immune and tissular) [22].

Based on the above, future studies must define the population in terms of physiological parameters (APACHE-II) or SOFA (Sequential Organ Failure Assessment) scores [90], vasopressor requirement, etc.), and there should be biomarker enrichment of patient selection probably using markers such as IL-6 or PCT levels. Secondly, clinically relevant endpoints should be selected. In the first instance, this may well be vasopressor dose under strict protocol guidance and adherence. In all cases, invasive hemodynamic monitoring including cardiac output and derived variables must be performed and data recorded, in order to have a better understanding on the pathophysiological effects of the therapy. With regard to treatment length, this should be guided by response in terms of other therapies as well as biomarker reduction if possible.

Regarding study endpoints, the consensus panel agreed that other endpoints rather than mortality should be chosen in the future to test the actual effect of this therapy. Monitoring organ dysfunction with measures like the SOFA score may be a better option, in order to understand the physiological effects of CytoSorb treatment. Furthermore, a hemodynamic primary endpoint such as the change in vasopressor need could also be applied, with mortality, ICU length of stay, length of mechanical ventilation, dialysis dependence, etc., as secondary endpoints. Finally, potential side effects should be also be evaluated like the removal of antibiotics, and especially lipophilic antibiotics ones [91].