A comprehensive search strategy was developed to investigate acute pancreatitis in adult humans, using MeSH terms for a number of immunological mechanisms (see Supplementary Data A). A search was conducted through Medline (PubMed), Web of Science, Cochrane Registry, and the National Institutes of Health Clinical Trials Registry. The time frame for the search was limited to January 1980 to May 2021 given the technical and historical limitations in investigating immunological mechanisms prior to this. For practical purposes only studies published in English were considered for inclusion.

Studies were included if they investigated an immunological mechanism, intervention or outcome relating to acute pancreatitis in humans. Animal studies, paediatric studies and studies of chronic pancreatitis were excluded. We also excluded studies investigating auto-immune pancreatitis due to there being a distinct immunological mechanism and treatment for that syndrome. Review articles, meta-analyses and case reports were excluded however their reference lists were reviewed to ensure relevant studies were not omitted.

Studies were imported into Covidence^© (Melbourne, Australia) review management software and duplicates were removed. Title and abstract screening was performed by two authors independently. Full text screening was subsequently performed by the two authors and reference lists for relevant publications were also screened for study inclusion. Any uncertainty or conflict regarding study inclusion was then deferred to a third author for consensus. Data was extracted and collected in Covidence^© (see data extraction form in Supplementary Data B) using a pre-specified data dictionary.

Simple quantitative analysis was performed to measure trial demographics and patient characteristics, including median values, percentages and interquartile ranges. We also quantified disease factors including symptom duration prior to hospital presentation, method of classifying disease severity and aetiology of AP. Studies were analysed to quantify timing and duration of outcome measurement and qualitative analysis was performed on immunological assays and outcomes.

A total of 2831 studies were identified after the initial search. After removal of 137 duplicates, 2489 studies were removed during abstract screening and full text review, with 205 studies remaining for data analysis (see PRISMA flow diagram – Figure 1).

Of the 205 retrieved studies (see Table 2), 17 (8.3%) were randomised controlled trials (RCTs), 9 (4.4%) pseudo-randomised controlled trials, 9 (4.4%) randomised non-controlled intervention study, 1 (0.5%) retrospective review of RCT data, 99 (48.3%) case-control studies, 68 (33.2%) case series, and 2 (1%) cohort studies. There were a total of 23,206 participants (including controls) across the 205 studies. The time frame for retrieved studies was 1984 to 2020. The majority of studies, 190 (93%), were conducted in a single centre. The median number of participants per trial was 62 (IQR 40-102). 129 (63%) of the studies had a control cohort. Of the 205 studies, 77 (38%) did not specify a time frame for symptom onset for inclusion.

During the full text screening process, six recurring themes emerged and provided a structure for the review. These themes were: aberrant innate immune response, cytokine profile, cellular responses, genetic predisposition, alterations to intestinal barrier function, and intervention studies. The main immunopathological findings are summarised in a flow chart in Figure 2.

The studies included were conducted across multiple countries and regions – of note 24% occurred in China, 9% in Japan, and 8% in Finland (Table 3). While these countries do have relatively higher incidences of AP, these figures do not reflect the global distribution of the disease burden (1).

Pancreatitis severity was classified using the Atlanta 2012 Consensus criteria in 110 (54%) of the studies). Most studies included a spectrum of disease: 97 (47%) mild and severe, 38 (18.5%) mild, moderate and severe, 6 (3%) moderate and severe, while 50 (24%) included severe disease only. With regards to aetiology, 33 studies (16%) did not specify an identified cause of AP in their patient population. Of the remaining studies that did specify, 166 studies (81%) specified gallstone/biliary aetiology, 164 (80%) alcohol, 59 (29%) hypertriglyceridaemia, 16 (8%) drug-induced, and 149 (70%) ‘other’ (which may have included less common pathologies for that particular population e.g. hypertriglyceridaemia or drugs that were specified in other study designs).

One of the evolved responses of the innate immune system is to recognise not only foreign pathogens but also to detect and respond to damage signals that may be indicative of host tissue injury. In AP, the initial tissue injury triggers the release of DAMPs into the blood, which typically occurs early in the acute phase of the illness. DAMPs directly interact with pattern-recognition receptors (PRRs), which are predominantly expressed by innate immune cells. These interactions trigger leukocyte activation and cytokine release, amplifying the inflammatory response to tissue injury but incidentally may also cause further tissue and organ dysfunction (213). Early DAMP release into the blood and their association with disease severity suggests that they may be useful biomarkers for prediction of severity on admission.

High-mobility group box-1 protein (HMGB-1) is a highly-conserved intranuclear protein normally involved in gene transcription but is also released into the circulation following cellular necrosis. Elevated HMGB-1 levels have been demonstrated in several inflammatory syndromes including sepsis, acute myocardial infarction and rheumatoid arthritis, and in this context is postulated to function as a pro-inflammatory mediator (152). In a case series of 15 patients with AP, HMGB-1 levels were significantly higher in patients with SAP compared to MAP, as well as in non-survivors compared with survivors (10). These elevations were observed from the time of admission and in the SAP cohort were sustained up to day 7 of illness. Similar findings were observed in a case-control study comparing admission levels of HMGB-1 in patients with MAP or SAP with a sepsis group and healthy controls (19). Both MAP and SAP had significantly higher HMGB-1 levels than healthy controls. In the SAP subgroup, HMGB-1 levels were elevated compared to the MAP patients, and were higher than levels expressed in the sepsis cohort. The early elevation of HMGB-1 and its correlation with disease severity suggests that it could be a useful biomarker for prediction of disease severity, and this requires further validation in larger studies.

Histones are intranuclear proteins involved in DNA packaging and gene regulation. In health, plasma histone levels are present at low levels, however elevated levels are seen in cellular injury syndromes such as trauma and sepsis (214). Elevated circulating histone levels have been implicated in early SAP. In a case-control study of 236 AP patients with 47 controls, circulating histone levels were comparable between volunteers, and patients with MAP or moderately severe AP (MSAP), however histone levels were significantly raised in the SAP cohort (25). In patients who presented with early organ failure (within 24 hours), histone levels were elevated compared to those with delayed organ failure. Similar findings were observed in another series where admission circulating histone levels were higher in SAP compared to MSAP and MAP (37). In both studies, receiver operating characteristics (ROC) area under the curve (AUC) for the accuracy of histones in predicting AP severity on admission were high (0.96 and 0.87 respectively). This suggests that plasma histones may be a useful biomarker of disease severity in AP, warranting further investigation.

Neutrophil extracellular traps (NETs) are expressed on neutrophil surfaces in response to stimulation by cytokine, lipopolysaccharide or intrinsic DAMPs. NETs are composed of decondensed DNA complexes and histones; their intrinsic purpose is an antimicrobial effect via the release of enzymes and proteins (a process known as NETosis). There is increasing evidence for the role of NETs in autoimmune and inflammatory disorders via direct tissue injury as well their potentiation of microvascular thrombosis and organ ischaemia (215). In AP, DAMP-induced NET activation appears to play a key role in both pancreatic and extra-pancreatic organ dysfunction. There is limited human data investigating NETs in AP, however one study assessed serum levels of free DNA and MPO-DNA complexes (markers of NET activity) in patients with SAP, MAP and healthy controls (216). The SAP cohort had significantly higher levels of both markers, which supports the implication of NETs in AP.

Other intranuclear products include free DNA, RNA products and nucleosomes, all of which can act as damage signals and trigger the inflammatory response. Elevated free DNA levels were observed in patients with SAP and sepsis on admission, however patients with milder disease had levels comparable to healthy controls (19). Plasma DNA levels were directly correlated with worsening severity of the CT score (radiological score of pancreatic injury). Micro RNAs (miRNAs) levels have been shown to increase in AP, significantly more so in SAP and with strong correlation with clinical scores of pancreatitis severity (41). Cold-inducible RNA-binding protein (CIRP) is an intranuclear protein that is activated by cellular stress and induces a proinflammatory response through a damage-signalling mechanism. In a prospective observational study of patients with SAP, MAP and healthy controls, admission serum CIRP levels were elevated in SAP patients, with these levels associated with increased risk of organ dysfunction, local pancreatic complications and mortality (14). Conversely, CIRP levels in MAP patients were comparable to those of healthy controls. A similar trend is observed with plasma nucleosome levels, with higher admission nucleosome levels in more severe disease, and therefore potential utility for nucleosomes as biomarkers of severity and prediction of organ dysfunction (217).

Heat shock proteins are intracellular proteins also activated in response to cellular stress, however these have protective effects on cells by inhibiting inflammatory mediators such as nuclear factor kappa-beta (NF-KB). In a small study, converse to the other DAMPs, Heat shock protein 70 (Hsp70) levels were lower on admission in SAP compared to MAP and MSAP, with sustained reduction in Hsp70 levels in the SAP cohort over the 7-day study period (10). Hsp70 levels were similarly depressed in non-survivors compared with survivors. A similar trend was observed for heat shock factor 1 (HSF1) levels in a case-control study, where SAP admission HSF1 levels were reduced compared to MAP and healthy controls (84). The elevated heat shock molecule levels in milder disease is postulated to exert a cytoprotective effect and inhibit the dysregulated inflammatory response, whereas the early disease progression in SAP may in part reflect failure of the heat shock systems.

PRRs expressed by immune cells serve as the receptors for the above damage signals. These receptors include toll-like receptors (TLRs), receptors for advanced glycosylation end products (RAGEs), nucleotide oligomerisation domain-like receptors (NLRs) and the absent-in melanoma-2 (AIM2) protein. Interaction between DAMPs and these receptors activates inflammatory responses and cytokine release, which subsequently triggers leukocyte recruitment to the site of tissue injury.

Higher TLR4 expression on peripheral blood mononuclear cells (PBMCs) in patients with AP has been reported, as compared to healthy controls (22). There were no differences in TLR4 expression between MAP and SAP, however these findings may have been confounded by the use of empiric glucocorticoids in the SAP cohort. These findings are supported by studies of mRNA derived from PBMCs that found high levels of TLR4 and TLR2 transcripts in AP as compared to healthy controls (49, 181). In summary, all three studies suggest higher TLR4 activity in AP compared to controls, however they do not determine whether TLR4 expression differs between mild and severe disease.

RAGEs act as receptors for HMGB-1, augmenting its inflammatory action. Interestingly, in a prospective case control study, plasma soluble RAGE (sRAGE) levels in SAP were significantly lower than healthy controls (19). In the same study, sRAGE levels in MAP patients and those with severe sepsis were significantly higher than the SAP cohort and were elevated compared to the healthy controls. In patients with severe sepsis, there was a direct correlation between HMGB-1 and sRAGE levels, whereas the relationship was inverse in the AP patients. The mechanism for this relationship is unclear and requires further delineation.

On activation by damaged molecules, AIM2 forms an intracellular inflammasome. Inflammasomes are cytosolic complexes that regulate proteolytic caspase-1 and the release of pro-inflammatory cytokines. On admission, patients with AP had higher levels of AIM2 compared to controls, with augmented expression in more severe disease (135). ROC AUC for AIM2 predicting SAP was high (0.945) suggesting it may be a useful admission biomarker for severity. NLRP3 is another inflammasome complex that that likely contributes to AP disease pathogenesis, contributing to both pro and anti-inflammatory cytokine pathways processes (218). The NLRP3 inflammasome has been extensively studied in animal models, but investigations in human AP are comparatively scarce. In one observational study, serum ASC levels, a molecule closely linked to the activity of the NLRP3 inflammasome and a surrogate marker of its activity, were measured in healthy controls and patients of varying degrees of AP severity (219). ASC levels were found to correlate with disease severity and IL-18 levels (see below), implicating NLRP3 in the pathogenesis of human AP, and warranting further investigation of this protein complex in human studies.

Increased expression of markers of leukocyte adhesion and activation, hallmarks of inflammation, have been reported in AP. With regards to the granulocytic response, a peripheral blood neutrophilia and monocytosis predominates in the acute phase, with higher counts observed in more severe disease (142). Adhesion molecules facilitate cellular migration from the blood vessels into the tissue to exert their inflammatory effects. A number of adhesion molecules have been identified as being involved early in the disease course of AP, including intercellular adhesion molecule-1 (ICAM-1 or CD54), vascular cell adhesion molecule-1 (VCAM-1), E-selectin, Integrin alpha L (CD11a) and Integrin alpha M (CD11b). These molecules are expressed both on peripheral blood monocytes (142), as well as by mononuclear cells isolated from acute pancreatic tissue (136). Expression of these molecules is increased in correlation with disease severity, indicating increased leukocyte activation in more severe AP, which likely perpetuates tissue injury and further drives the inflammatory response (12, 27, 136, 142, 191). The early elevation of these molecules in AP suggests their potential utility as biomarkers of severity.

AP is associated with high levels of circulating free fatty acids (FFAs), which is likely precipitated by elevated lipolytic enzymes at both the local pancreatic and systemic level (220). In a case control study comparing FFA levels in patients with and without necrotizing pancreatitis, FFA levels were elevated in both groups on admission, but normalized by day 7 in the group without necrotizing pancreatitis (221). In the necrotising pancreatitis cohort, FFA levels remained elevated, possibly indicating ongoing pancreatic and extrapancreatic fat necrosis. There is evidence that FFAs, including unsaturated fatty acids (UFAs) and non-esterified fatty acids (NEFAs) contribute to the inflammatory response and tissue injury by triggering pro-inflammatory cytokine activity as well as direct deleterious effects on mitochondrial function (222, 223). Investigating the pathophysiology of fatty acid lipotoxicity in humans patients with AP is comparatively limited and warrants further investigation.

Definitively determining the cytokine pathways upregulated in AP that contribute to the pathogenesis is of significant value in the current era of monoclonal antibodies and other targeted therapies. For example, many of the implicated cytokines are currently targeted in other disease states, including anti-TNFα and IL-1 (i.e. via infliximab and anakinra, respectively) in inflammatory arthritis and autoinflammatory disorders, IFNγ (via JAK inhibitor tofacitinib) in rheumatoid arthritis and inflammatory bowel disease as well as anti-IL-6 antibody tocilizumab in rheumatoid arthritis, juvenile idiopathic arthritis and severe COVID-19 (225). While the cytokine profile in AP has been extensively reported, there are several limitations in evaluating the data. Of the 137 studies that were selected for inclusion, 103 (75%) employed enzyme immunoassay techniques for cytokine measurement, however the specific assay type and/or manufacturer was not often reported, which can lead to inter-assay variability. Interpreting cytokine levels and trends is also made difficult by the inconsistency of timing of patient recruitment to the studies. However, further advancements in these areas have the potential to rationally advance anti-cytokine therapies into clinical trials in AP.

As an MHC-II molecule, human leukocyte antigen-DR (HLA-DR) is a surface marker of cellular activation and plays a key role in T cell antigen presentation and regulation of pro-inflammatory cytokine release. Reduced HLA-DR expression is a reliable biomarker of immunodeficiency and is associated with a higher risk of infection (228). In patients with mild and moderate AP, there is an observed reduction in HLA-DR expression within the first few days of disease onset, however this returns to near-normal within the first week (75). In contrast, patients with SAP – particularly those with multi-organ failure – downregulated HLA-DR expression persists beyond the first week of illness and is associated with increased rates of secondary infectious complications (40, 42). In a study comparing HLA-DR expression in AP patients with and without infectious complications, the cohort that developed secondary infections exhibited significantly lower HLA-DR expression (28). A similar inverse correlation has been observed between HLA-DR expression and mortality (16). Given the association between HLA-DR expression and disease severity and complications, it has also been proposed as an early biomarker of disease severity.

Further evidence of immunosuppression in AP is presented by the heightened expression of the immune checkpoint molecules programmed cell death 1 (PD1) and programmed cell death ligand 1 (PD-L1). PD1/PD-L1 is a well-known immune inhibitory pathway involved in prohibiting T cell activation and proliferation as well as the induction of T cell apoptosis. Heightened PD1/PD-L1 expression can be a means of immune system evasion in metastatic malignancy, but is also a marker of T cell exhaustion, such as in septic shock where there is an associated higher risk of nosocomial infectious complications (229). Recent data implicates heightened activity of the PD1/PD-L1 system in severe pancreatitis, with greater expression in patients who develop secondary infectious complications (28). Similar activity has been observed with the soluble isoform of PD-L1 (sPD-L1), demonstrating higher sPD-L1 levels in more severe AP and in those with infectious complications (11). Immune checkpoint inhibitors targeting the PD-1 are currently in clinical trials in sepsis, to determine their efficacy in reversing T cell anergy and exhaustion (230) and findings here may have some relevance to synonymous phenomena in AP.

In addition to quantitative deficiencies, there is increasing evidence to suggest functional impairment of both lymphocyte and monocyte populations is present in severe disease. A state of cellular anergy in SAP would certainly increase the risk of subsequent infectious complications. One study performed in vitro assessments of lymphocyte NF-KB signalling in response to incubation with microbial pathogens, comparing the lymphocytes of patients with SAP (with HLA-DR expression <80%) with those of healthy volunteers (146). T and B cells from the SAP cohort had significantly reduced responses, suggesting a degree of functional exhaustion. Cells were subsequently treated with TNFα and NF-KB activity measured; the lymphocytes from SAP patients had significantly diminished responses compared to controls, indicating reduced number of TNFα-responding cells. A similar study assessed in vitro monocyte phagocytosis in peripheral blood on admission in mild, moderate and severe cohorts of AP, and compared this to healthy controls. In all cohorts, phagocytic activity was increased, with activity positively correlating with disease severity (153). However, on subgroup analysis, patients who developed infectious complications had significantly reduced monocyte phagocytic activity. Phagocytic activity was then further stimulated by incubating lymphocytes with protein kinase C activator (PMA – a protein that enhances macrophage oxidative metabolism and phagocytic function). Monocytes from the mild, moderate and severe (without infectious complications) all had significantly higher oxidative activity, whereas monocytes from patients who later developed infectious complications had no response to PMA, indicating cellular metabolic dysfunction.

Several pharmacological interventions have been trialled in AP with limited to no efficacy in reducing mortality despite improving biochemical, clinical and immunological markers.

Given the profound inflammatory response associated with SAP, glucocorticoids have been trialled albeit with no high quality RCTs testing their efficacy in this disease. We identified three studies that trialled glucocorticoids; one retrospective propensity-matched cohort study, one case control study and one case series (233–235). In the propensity matched cohort study, it is unclear what the indication for glucocorticoid therapy was, and patients were treated with varying formulations at differing doses and durations. In the glucocorticoid arm, compared to the propensity matched arm, there were improvements in mortality, ICU length of stay (LOS) and treatment costs. In the case control study, patients were treated with with 1mg/kg dexamethasone three times daily compared to standard care; the steroid arm had improved rates of ARDS and LOS but no improvement in mortality. This study was confounded by the use of Chinese herbal preparations in both groups, the effect of which on AP is unclear. Finally, the case series administered variable doses of dexamethasone and Dextran colloid solution with improvements in symptoms however in the absence of a control group, it is difficult to draw definitive conclusions. Across all studies, the variable definitions of severity, treatment protocols and patient heterogeneity, as well as the lack of immunological outcomes, makes the findings from these studies largely exploratory.

Several studies have trialled pancreatic enzyme inhibitors including somatostatin, ulinastatin, octreotide and gabexate. These agents were proposed to reduce enzyme activation and minimise the initial tissue injury, which may therefore mitigate the subsequent inflammatory and immune response. Findings were variable with some evidence for improved immunological markers including increasing the CD4:CD8 ratio and reduced pro-inflammatory cytokine expression (TNFα, IL-6, IL-8), but importantly a reduction in mortality or length of stay was not demonstrated (188, 200, 203, 207).

Similarly, pentoxifylline, a non-selective phosphodiesterase inhibitor, has been trialled in AP with variable effect on improving serum cytokine levels but no clear evidence of mortality benefit (175). The trials in this field are limited by low participant numbers and lack of clarity regarding baseline physiological status of participants. Other confounding factors that varied across intervention studies included what was considered ‘standard care’, including the empirical administration of antibiotics and corticosteroids (188, 192, 208).

Lornoxicam, a non-selective cyclooxygenase inhibitor, was tested in patients with SAP compared to standard care in a randomised, non-blinded clinical trial (181). Patients were randomised to standard care (n=246) or lornoxicam plus standard care (n=88). The trial demonstrated a temporally significant reduction in TLR2 mRNA expression in the lornoxicam group. The treatment arm also demonstrated significantly reduced mortality and fewer gastrointestinal complications. These results are potentially of interest, however must be interpreted with caution as the study was underpowered for mortality (which also was not a reported primary or secondary outcome), it is unclear whether true randomisation occurred, participant baseline clinical and biochemical data were not reported (therefore differences in outcomes may be contributed to by cohort heterogeneity), and there was a significant difference in the cohort sizes which may confound the magnitude of effect.

Randomised trials have been conducted to assess the effects of lexipafant, antagonist to platelet activating factor, as well as activated protein C (APC) in improving mortality and immunological outcomes in AP. No significant improvements in immunological outcomes or overall mortality were observed for either therapy (189, 191).

A potential limitation of these drug therapies is that they target the early disease process; that is, reducing pancreatic secretions and initial inflammation rather than suppressing the subsequent immune responses. Given the systemic immune dysregulation has likely already onset by the time patients present to hospital, this probably explains why these pancreas-specific treatments have no benefits from an immunological outcome’s perspective.

There have been four RCTs of lactate-containing balanced crystalloid solutions with 0.9% sodium chloride (0.9% saline) in patients with AP (236–239). The outcomes of these studies include frequency and duration of SIRS, clinical outcomes (e.g. mortality and LOS), and inflammatory markers (e.g.CRP). Patients administered the lactate-containing crystalloid had trends towards reduced duration of SIRS, reduction in CRP and some improved clinical outcomes such as ICU LOS but no impact on mortality. It is hypothesised that lactate-based solutions may confer anti-inflammatory effects by raising pH in the pancreatic tissue. Furthermore, these studies administered high volume fluid resuscitation based on goal-directed therapy protocols. In other critical care cohorts such as sepsis, goal-directed fluid therapy has since been robustly shown to not improve outcomes (240) and high volume fluid therapies are associated with an increased trend to harm across numerous critical care cohorts (including the recent WATERFALL trial in AP (241)). Finally, these studies enrolled patients with all severities of AP, which further limits the ability to draw conclusions on the potential benefits of different crystalloids in this disease. While these study findings are of interest, the studies were underpowered and therefore these findings should be considered hypothesis-generating.

Two studies have examined the effect of IV fluid on inflammation in SAP by comparing 0.9% saline solution to various formulations of hydroxyethyl starch (HES). Compared to NS, patients with AP who received a starch-based solution had some temporal improvements in cytokine levels (174, 212). It is hypothesized that HES may play some role in attenuating the inflammatory response, in particular capillary leakage via NK-κB activation, and neutrophil adhesion and migration. However, starch-based solutions have been demonstrated to increase the risk of acute kidney injury in critical illness and therefore cannot be readily considered as a potential therapy (242).

Thirteen studies examined the effect of various nutritional approaches in SAP and their effect on the immune and/or inflammatory response. Overall, little insight into the immunopathology could be gained as many studies examined few time points, did not report specific cytokine levels, and primarily focused on non-immunological outcome measures such as organ failure and length of stay.

With respect to cytokines, total parenteral nutrition (TPN) adjuncts demonstrated some effect in reducing the inflammatory response compared to TPN alone e.g. fish oil was associated with reduced IL-8 and ICAM-1 (243), and increased IL-10 (204), however no difference was seen in sTNFR levels (194) or IL-6 (206). In comparing TPN and enteral nutrition (EN), EN was frequently demonstrated to be superior in respect to cytokine response - with reduced pro-inflammatory IL-6 (173, 202, 205), IL-1b (173), and TNFα (173, 205), and increased anti-inflammatory IL-11 and IL-10 (202, 205).

In regards to the cell-mediated response, TPN adjuncts such as glutamine increased lymphocyte count (194), and fish oil increased HLA-DR expression (204). HLA-DR expression also significantly increased with early EN compared to delayed EN (199).

In terms of gut permeability, multiple studies have demonstrated EN significantly reduces gut permeability compared to TPN via various measures including EndoCAb IgG and IgM antibodies (185, 209), endotoxin, and lactulose-mannitol ratios (205, 211). Supplementation of EN with glutamine was demonstrated to improve gut permeability and significantly reduce IL-6 (172) and endotoxaemia (198), however this did not translate into an improvement in infectious complications, morbidity or mortality. There is no evidence to suggest any immunological benefit from probiotic use in AP (197), and their use is actually associated with increased mortality in patients with SAP (244).

Finally, in terms of clinical outcomes, compared to TPN, EN was associated with lower rate of complications end organ dysfunction, as well as improved outcomes (173, 199, 202, 209).

Eleven studies investigated the effect of extracorporeal therapies in modulating the inflammatory response in patients with AP, predominantly those with severe disease. The aim of these therapies is to sequester inflammatory mediators from the plasma via dialysis, haemofiltration, or adsorption using specialized membranes.

Overall, it appears that extracorporeal therapies may be effective in reducing the plasma levels of inflammatory cytokines. Reduced cytokine levels were observed as early as within 6 hours of commencement of continuous veno-venous haemofiltration (177, 178), and high volume haemofiltration was associated with greater cytokine reduction than continuous filtration (176). In other studies, application of extracorporeal therapies was associated with improved clinical outcomes including reductions in mortality, improved parameters of respiratory function and improved haemodynamic outcomes (170, 183, 184). Wang et al. (208) randomised patients to standard care versus either haemofiltration or laparoscopic peritoneal lavage or haemofiltration plus peritoneal lavage. All intervention arms had improved clinical outcomes and maximal reduction in pro-inflammatory cytokines, however the greatest improvements were observed in the group who underwent both haemofiltration and lavage. This suggests an additional benefit may be derived from addressing the intra-abdominal inflammatory processes via lavage (i.e. removal of necrotic tissue, inflammatory mediators, and translocated intestinal flora). These findings are however confined to a single study and the potential benefits of this intervention would need to be balanced against the risks of significantly invasive surgery in an already critically ill patient population.

While these findings are promising, they were not consistent across the studies. Furthermore, methodological issues including lack of true randomisation, the absence of control groups, heterogenous study populations, variable definitions of disease severity, and low participant numbers make these observations hypothesis-generating rather than truly demonstrating therapeutic effect which would likely require further randomized interventional trials. Extracorporeal therapies are not without risk and these must be considered against any potential benefit. Inconsistency in the reporting of particularly outcomes, including the frequent measurement of non-standardised clinical outcomes such as ‘relief of abdominal distension’, makes comparison of results between studies challenging.