Bench-to-bedside review: Immunoglobulin therapy for sepsis - biological plausibility from a critical care perspective

Sepsis represents a dysregulated host response to infection, the extent of which determines the severity of organ dysfunction and subsequent outcome. All trialled immunomodulatory strategies to date have resulted in either outright failure or inconsistent degrees of success. Intravenous immunoglobulin (IVIg) therapy falls into the latter category with opinion still divided as to its utility. This article provides a narrative review of the biological rationale for using IVIg in sepsis. A literature search was conducted using the PubMed database (1966 to February 2011). The strategy included the following text terms and combinations of these: IVIg, intravenous immune globulin, intravenous immunoglobulin, immunoglobulin, immunoglobulin therapy, pentaglobin, sepsis, inflammation, immune modulation, apoptosis. Preclinical and extrapolated clinical data of IVIg therapy in sepsis suggests improved bacterial clearance, inhibitory effects upon upstream mediators of the host response (for example, the nuclear factor kappa B (NF-κB) transcription factor), scavenging of downstream inflammatory mediators (for example, cytokines), direct anti-inflammatory effects mediated via Fcγ receptors, and a potential ability to attenuate lymphocyte apoptosis and thus sepsis-related immunosuppression. Characterizing the trajectory of change in immunoglobulin levels during sepsis, understanding mechanisms contributing to these changes, and undertaking IVIg dose-finding studies should be performed prior to further large-scale interventional trials to enhance the likelihood of a successful outcome.

reso lu tion of infl ammation, accelerating recovery processes or, if eff ected early enough, by primary targeting of 'upstream' mediators (such as signalosomes and infl amma somes) that trigger the excessive activation or suppres sion of 'downstream' mediators and multi-system pathways such as cytokines and the complement system. Particularly with regard to the latter strategy, it is unlikely that the patient with severe sepsis will present early enough for successful therapeutic administration of a drug modulating a single upstream pathway. Far greater utility is likely to be gained through a cocktail approach, or by using agents with multiple modes of action. Prime examples of multi-modal stand-alone agents for severe sepsis and septic shock are corticosteroids and polyvalent intravenous immunoglobulins (IVIg).
After a brief review of relevant sepsis biology, this article will focus upon immunoglobulins and their receptors, the potential benefi cial eff ects of IVIg therapy in modulating the host response to infection, and an overview of the possible reasons for the limited success to date of clinical trials.

Initiation of host response
Th e initial host response to infection involves over lapping, interlinked phases of innate pathogen and damage recognition. Microbial infection results in release of (i) pathogen-associated molecular patterns (PAMPs), that is, conserved molecular structures expressed by the microbe species, and (ii) damage-associated molecular patterns (DAMPs), that is, extracellular matrix compo nents and intracellular constituents (for example, mito chondria, DNA, S100 proteins) released due to local tissue damage or immune cell activation [14]. PAMPs and DAMPs are recognised as danger signals by pattern recognition receptors on the surface of immune, epi thelial, endothelial and parenchymal cells. Th is early innate response aims to limit systemic dissemination of infection, allowing slower though immunologically more potent and focused adaptive immune pathways to develop [15].

Host responses to 'danger signals'
Infection and/or tissue damage can trigger a dysregulated and systemic infl ammatory response through multi-point activation of genes transcribing for pro-infl ammatory mediators and receptors. Th ese act via infl ammasomes and signalosomes -'upstream' mediators of the host response [16,17]. Infl ammasomes are multimeric protein complexes generated in response to distress signals from PAMPs and DAMPs that act as primary initiators of the innate host response (for example, NLR-NOD proteins) [17][18][19]. Signalosomes are molecular complexes that mediate phosphorylation and poly-ubiquitination of inhibitory kinase complexes (for example, IκB), thereby releasing activated transcription factors that enter the nucleus and increase transcription of target genes involved in the infl ammatory response. Th ese include genes encoding downstream mediators such as cytokines, chemokines, adhesion factors, nitric oxide synthase, tissue factor and cyclo-oxygenase pathways [17,[20][21][22]. Th is complex 'downstream' response could be conceptualised as being generated and amplifi ed from an infl ammatory hub consisting of high mobility group B-1 protein (HMGB-1), complement factors, macrophage migration inhibitory factor, IL-17 and other mediators. Both upstream and downstream mediators and networks are interlinked, impairing cellular bioenergetic and metabolic function at multiple levels, and resulting in organ dysfunction [7,[23][24][25] (Figure 1). Th ese changes also aff ect innate immune cell function, thereby impairing bacterial clearance [26,27]. Th e host responses described above involve concomitant activation of pro-and anti-infl ammatory pathways. Th e balance of the host response shifts towards predominantly anti-infl ammatory pathways later on in critical illness. While this results in an overall immune anergy [28,29], some immune cell types remain hyperresponsive, underlying the complexity of the condition.
Emerging literature on viral reactivation following acute pro-infl ammatory critical illnesses provides further evidence that immunosuppression is a key sequela in sepsis and critical illness. Th is is likely related to T-cell defects, leading onto macrophage dysfunction [30,31]. Other causes for immune anergy include (but are not limited to) enhanced regulatory T-cell activity [32], activa tion of anti-infl ammatory phenotypes in infl ammatory cells [28,33], and activation of apoptotic pathways [34]. At present, anergy is considered primarily due to lymphocyte and dendritic cell loss.

Immunoglobulin physiology
Immunoglobulins (Ig) are glycoprotein molecules produced by plasma cells. B lymphocytes that are activated and propagated in a T-cell-dependent manner are the precursors of high-affi nity antibody-secreting plasma cells [35]. T-cell-independent pathways can also generate plasma cells, including those secreting naturally occurring antibodies.
Each Ig molecule monomer consists of identical heavy (50 to 70 kDa) and light chain pairs (23 kDa) held together by electrostatic forces and disulphide bonds. Each heavy chain consists of amino acid sequence regions (three to four constant, one variable) that fold into globular regions called domains. Within each variable region of heavy chains and light chains there are three hypervariable or complementarity-determining regions that determine antibody specifi city. Th e combined variable and constant regions of the heavy and light chains form the antigen-binding region on the Fab. Amino acid sequences in the remainder of the two constant regions of the heavy chains, the Fc, determine the immunoglobulin class and subclass, and therefore its functional capability.
Th e large diversity of antigenic epitopes are recognised by the variable region of the Ig molecules. Th is is a function of the adaptive immune system. Binding of Ig results in many diverse antigens being signalled through a small number of Ig isotypes. Based on their heavy chain characteristic, Ig isotypes are classifi ed into G, A, M, D and E [36,37].
Th e distribution of receptors on immune cells and their affi nity to IgG diff ers between FcγRs (Table 1) [38,[40][41][42][43]. FcγRs can be either activating or inhibitory depending on their inclusion or association with either the activating (immunoreceptor tyrosine-based activating motif (ITAM)) or inhibitory (ITIM) motifs in their cytoplasmic domains. Bacterial infection increases FcγR expression on innate and adaptive immune cells. FcγRI is the only high-affi nity receptor that can bind to circulating monomeric IgG, while all low-affi nity receptors only interact with immune complexes for signal transduction [44,45]. Interaction of PRRs with PAMPs/DAMPs initiates the cellular activation that characterises host response in sepsis syndromes. Infl ammasomes and signalosomes generated from these initiator pathways provide the feedback amplifi er loops perpetuating host response. This unregulated multi-system activation involves infl ammatory pathways, cytokines, coagulation, inducible nitric oxide pathways, the autonomic nervous system and the immune system. This is manifested biologically as microvascular failure, mitochondrial dysfunction and apoptotic changes -surrogates of severity of organ dysfunction in sepsis. HMGB, High mobility group box protein; iNO, inducible nitric oxide; MIF, macrophage migration inhibitory factor. All FcγRs except FcγRIIB are stimulatory (that is, associated with the ITAM cytoplasmic domain) and therefore activate immune cells following IgG-or pentraxinprotein or immune complex binding. Aggregation of ITAMs results in phosphorylation of ITAM tyrosines and stimulates multiple downstream activation pathways [46]. By contrast, FcγRIIB is associated with the ITIM cytoplasmic domain with phosphorylation of tyrosines in ITIMs resulting in attenuation of activation pathway activity [46].

Infec
The level of FcγRIIB-related activity compared to other FcγR activity (that is, those associated with ITAM domains) plays a key role in balancing the proand anti-inflammatory humoral pathways in sepsis [47,48]. It is biologically plausible that Ig modulates innate and adaptive immune effector activity essential for bacterial clearance by altering the balance between ITAM and ITIM activity via FcγRs; this equilibrium may be potentially influenced favourably with IVIg therapy [40].

Other receptors
Another receptor involved in IgG pathways is the neonatal FcR (FcRn). Th is belongs to the family of major histocompatibility (MHC) class I molecules but is not involved in antigen presentation. Its primary roles are to maintain constant IgG and albumin concentrations and to prolong the half-life of IgG and albumin through endosome-to-cell surface recycling. FcRn-mediated pathways are important in maintaining the serum retention of native and infused IgG preparations [42]. Other receptors for Ig molecules include the tripartite motif-containing (TRIM) protein family, some members of which appear to be particularly important in the response to viral infections [49].

Polyvalent intravenous immunoglobulins
IVIg is a blood product prepared from a pool of more than 1,000 donors (frequently more than 10,000 donors), thus providing a broad spectrum of opsonic and neutralizing IgG antibodies against a variety of microbial antigens and multiple epitopes. Opsonic and neutralizing IgG antibody content varies with each product batch, primarily due to diff erences in the local pathogen ecology of donor exposure. IgG and complement proteins are the principal classes of opsonins contributing to bacterial clear ance (amongst other opsonins such as C-reactive protein). Only one product, Pentaglobulin® (Biotest, Germany), is IgM-enriched. Th e principal manufacturing process in all current Ig preparations is cold ethanol fractionation with product-specifi c additional processes for manufacturing. Th e commonest processes for virus reduction include solvents/detergents, low pH (pH 4) incubation, nanofi ltration and chromatography.

The biological rationale for administering IVIg in sepsis
Th e biological rationale for IVIg therapy in sepsis can be summarized into four main categories: (i) its role in pathogen recognition, clearance and toxin scavenging, (ii) scavenging and inhibition of 'upstream mediator' gene transcription, (iii) scavenging and inhibition of infl ammatory 'downstream mediator' gene transcription, and (iv) non-apoptotic and anti-apoptotic immune cell eff ects.

Role in pathogen recognition, clearance and toxin scavenging
PAMPs are recognised by naturally occurring antibodies that can also act as innate immune receptors. IgG and the complement proteins are the principal opsonins for bacterial clearance. Th e classical pathway is activated by C1 complex interaction with Ig, acute phase proteins and various non-specifi c activators [50]. Th e C1q mole cule within the C1 complex contains a multimeric globular ligand detection domain with the ability to bind IgG and IgM Fc regions, and hence detect a large spec trum of antigens. Binding of C1q to IgG1 or IgM leads to potent activation of the classical complement pathway, thereby generating C4b2a (recently renamed as C4b2b), the classical pathway C3 convertase [50,51]. IgG com bines with C3b and this opsonisation facilitates phago cytosis [52].
Severe sepsis is associated with a decrease in circulating immunoglobulin levels [61][62][63][64]. Th ree-quarters of patients admitted with community-acquired pneumonia and shock had hypogammaglobulinaemia, of whom three quarters had low levels of IgG. Hypogammaglobulinaemic patients had a signifi cantly longer duration of shock and a higher incidence of severe lung injury [61]. Patients from the Score-Based Immunoglobulin G Th erapy of patients with sepsis (SBITs) study demonstrated a wide distribution in IgG levels, although in this study low levels did not carry any prognostic signifi cance [62,63]. Furthermore, in a recent observational study of patients enrolled within day 1 or 2 of presentation with septic shock, 61% had IgG levels below the lower limit for agematched reference values. Th is hypogammaglobulinaemia was transient and also had no prognostic signifi cance [64].

Role in scavenging toxins
Superantigen exotoxins released by staphylococci and streptococci activate T cells [65,66]. IVIg preparations contain inhibitory or neutralising IgG molecules against superantigens, and these inhibit superantigen-mediated T-cell and monocyte activation [67,68]. In addition, IVIg preparations have also been shown to inhibit superantigen-induced cytokine production and lymphocyte proliferation, that is, independent of the presence of neutralising antibodies [69]. Of note, toxin neutralisation is profoundly infl uenced not only by the antigen-binding activity of the antibodies within IVIg, but also by the Fc region of the IgG molecules, so the balance of IgG isotypes within the anti-toxin response is also critical [70]. IVIg therapy has been shown to be benefi cial in toxin-mediated bacterial diseases and shock syndromes [71,72], although the results are inconsistent [73]. IVIg preparations, in particular IgM-enriched preparations, contain antibodies against lipopolysaccharides of Escherichia coli, Pseudomonas aeruginosa and Klebsiella spp. [74].
Acquired hypogammaglobulinaemia may prevent optimal pathogen clearance and pathogen toxin scavenging, thereby perpetuating the sepsis response. As immunoglobulin levels in health vary signifi cantly, interpretation of single-time point determinations of immunoglobulin concentration in the context of sepsis pathobiology is a key challenge. By relating the temporal profi le of Ig concentrations to the trajectory and severity of sepsis, a high-risk sepsis cohort may be potentially identifi ed for stratifi ed IVIg intervention. However, the reasons underlying these temporal changes are currently unclear. Altered distribution due to endothelial dysfunction and capillary leak [17], an iatrogenic fl uid resuscitation-related increase in extravascular volume with dilution of Ig [64,75], decreased production and/or increased consumption may be implicated, as could altera tions in FcRn activity resulting in impaired recycling [76]. All the above knowledge gaps need addressing, ideally prior to further clinical trials.

Scavenging of 'upstream mediators' and inhibition of 'upstream mediator' gene transcription
NF-κB dependent signalling (signalosome) is a key mechanism for generating downstream host response mediators in sepsis and other infl ammatory diseases. Patients with hypogammaglobulinaemia [77], sepsis [17], and Kawasaki's disease [78] have NF-κB-mediated up-regulation of IL-1 and IL-1r activity. Th ese components of the IL-1 system decrease following IVIg dosing of 0.4 g/kg, secondary to a reduction in IL-1-mediated peripheral blood mononuclear cell activation, and by induction of IL-1 receptor antagonist (IL-1ra) [77]. Th e presence of neutralising antibodies in IVIg preparations may also be contributory. IVIg inhibit TNF-alpha-induced NF-κB activation on neutrophils while IgG1 blocks FcγRIIIA receptors on peripheral blood mono nuclear cells, further impairing their activation [79]. IVIg can also inhibit endothelial cell activation as demon strated by a decrease in markers such as adhesion molecules, endothelins, pro-infl ammatory cytokines (for example, IL-6) and inducible nitric oxide pathways [80,81]. In addition, naturally occurring antiidiotypic antibodies, auto-antibodies and immune proteins in IVIg preparations also contribute to its immunomodulatory properties [82].
Th ere is little direct evidence of IVIg eff ects on caspase signalling in sepsis. Caspases and calpain activation may contribute to myocardial dysfunction [83], pulmonary microvascular endothelial damage [84], and skeletal muscle and protein wasting in sepsis [85,86]; thus, inhibition of these pathways may be potentially benefi cial. In pemphigus, IVIg upregulated endogenous caspase and calpain inhibitors (FLIP and calpastatin, respectively) [87]. Extrapolating this evidence will help determine whether IVIg therapy in sepsis could potentially attenuate myocardial and pulmonary dysfunction.
HMGB-1 released into the circulation in sepsis syndromes is considered a key signalling molecule in the infl ammatory hub concept of severe sepsis, activating cell-to-cell signalling, procoagulant activity and late phase responses [7]. In addition, HMGB-1 is considered to have prognostic signifi cance [91] and is a possible therapeutic target in sepsis syndromes and other infl ammatory disorders [92]. In septic rats, high-dose IgG pre-treatment reduced HMGB-1 activity [93]. As IgG and IgM HMGB-1 antibodies are found in the serum samples of healthy individuals, an IVIg preparation should be able to limit HMGB-1-related activation of infl ammatory and coagulation pathways [94].
Bacterial clearance is the primary innate immune function of the complement system. Th is occurs via detection of PAMPs followed by recruitment and activation of proteases. Complement pathway activation results in opsoni sation of bacteria with C3b to facilitate phagocytosis, generation of membrane attack complex for bacterial cell lysis, and generation of pro-infl ammatory chemokine anaphylotoxins such as C3a and C5a, which are central mediators of the infl ammatory hub in sepsis [7,50,[97][98][99].
Notwithstanding its desired eff ect on bacterial clearance, there are numerous systemic and cellular adverse eff ects of unregulated complement system activation. In patients with severe sepsis this may cause impaired neutrophil chemotaxis and phagocytic function secondary to down-regulation of C5a receptor type 1 (CD88) expression, leading to reduced bacterial clearance [27,98,100]. Complement activation can also impair cardio myocyte function in sepsis. In an animal model this was prevented by C5a-blocking antibodies [101]. Furthermore, C5a is a key perpetuator of coagulation cascade activation [102], accelerated lymphocyte apoptosis, immuno paresis and autonomic nervous system dysfunction [7,50,98].
IVIg have complement-scavenging properties that attenuate these undesired eff ects of anaphylotoxins. Crucially, the anti-complement activity of IVIg does not aff ect bacterial clearance [103]. Th e Fab2 region of the Ig molecule interacts with and scavenges C3a and C5a, thereby reducing complement-mediated cytotoxicity [104]. Scavenging of C5a also reverses C5a-mediated upregulation of FcγIIIa receptors and down-regulation of FcγIIb receptors. Th e resulting high ratio of inhibitory FcγIIb to FcγIIIa on immune activator cells such as monocytes and macrophages is responsible for IVIginduced immunomodulation and contributes to its antiinfl ammatory eff ects [105]. As deregulated excessive C5a activity is likely to be a key molecular mechanism in sepsis [7,106], C5a scavenging by IVIg therapy should improve neutrophil [27] and myocardial function [101], as well as reducing coagulopathy [102], immune cell apoptosis [28,[107][108][109]] and autonomic nervous system dysfunction.

Immune cell eff ects Non-apoptotic
Dysregulation in the nitric oxide pathway, glucose metabolism and infl ammatory networks contribute to impaired neutrophil function in severe sepsis [110]. As these represent downstream mediators scavenged by IVIg, IVIg therapy could restore neutrophil function and improve bacterial clearance.
Dendritic cells act as intermediaries transducing the anti-infl ammatory eff ects of IVIg. Th e DC-SIGN recep tor (dendritic cell-specifi c ICAM3-grabbing non-integrin) acts as a major regulatory pathway [111]. IVIg can downregulate class II MHC expression by dendritic cells, directly inhibiting the classical CD3-T cell receptor pathway of T-cell activation [112]. Th e resulting reduction in pro-infl ammatory cytokine production and increas ing anti-infl ammatory cytokine production further contributes to the anti-infl ammatory and immuno modulatory activity of IVIg [113]. IVIg also inhibited invariant natural killer T-cell activation mediated through FcγRIIIA receptor eff ects [114]. IgG can determine the CD1 expression profi le of monocyte-derived dendritic cells as this is mediated, at least in part, by FCγIIA receptors. An Ig-rich milieu induced CD1d expression, whereas Ig depletion increased expression of CD1a, CD1b, and CD1c [115].
A relative IgG defi ciency in sepsis could potentially impair homeostatic T-cell regulation, with deleterious eff ects on host immune function [116][117][118]. Proliferation of activated T cells is regulated by suppressive CD4+CD25(hi) natural regulatory T cells, a pathway enhanced by IgG. In patients with common variable immunodefi ciency, low dose IVIg therapy directly activated B-cell proliferation independent of T-cell signalling. Th is eff ect could be benefi cial in sepsis by prevent ing late-onset immune anergy, potentially through reducing B cell loss [119]. Th us, IVIg therapy in Igdefi cient patients may potentially facilitate these benefi cial, lymphocyte-mediated immune responses orchestrated through dendritic cells.
Signifi cant B and T cell apoptosis reported in a humanised (innate and adaptive immune system) mouse model of severe sepsis has been replicated in patients with severe sepsis [123]. Excessive apoptosis in sepsis has been shown in both circulating and lymphoid organ lymphocytes [124][125][126]. In addition, lymphopenia has been associated with adverse outcomes in severe sepsis, although causality has yet to be shown [127].
Apoptotic pathways thus contribute signifi cantly to sepsis-induced immune 'anergy' via lymphocytes and dendritic cell loss. If IVIg can attenuate immune and non-immune cell apoptosis by inhibition of extrinsic pathway activity through its ability to target upstream and downstream mediators (for example, via NF-κB and C5a inhibition), this may prevent immune anergy and maintain the signifi cant role lymphocytes play in bacterial clearance. It may also moderate the organ dysfunction, including immune anergy [120,128,129].
Th e eff ects of IVIg on apoptotic pathways are inconsistent; some reports even suggest an increase in apoptosis [130,131]. Th is inconsistency probably relates to IVIg preparation, composition, disease state and dose. Further studies are needed before any claims of potential benefi t of IVIg therapy on immune cell apoptosis can be made.

Infl uence of IVIg preparations on effi cacy
Manufacturing processes have changed signifi cantly in the past two decades and will likely infl uence the pharmacodynamic and pharmacokinetic properties of the fi nal preparations. Current processes aim to maintain the physiological balance of the four IgG subclasses, which is important for both bacterial clearance and immunomodulation in severe sepsis [132][133][134]. Th e glyco sylation status of IgG in IVIg preparations can profoundly infl uence its anti-and pro-infl ammatory eff ects [135]. Studies of the degree and type of IgG glycosylation of commercial IVIg products demonstrate signifi cant inter-product diff erences (S Khan, WA Sewell et al., submitted). IVIg preparations have variable proand anti-apoptotic properties depending on the pharmacological composi tion of the IVIg preparation as these have varying levels of stimulating and inhibiting antibodies to Fas and Siglec receptors [136,137]. Which preparation is optimal may well be patient-and/or IVIg preparation-dependent. Th is area clearly needs further investigation.
No less than seven recent systematic reviews and meta-analyses have summarized these interventional trials yet have yielded confl icting results [153][154][155][156][157][158][159]. IVIg therapy is reported by most meta-analyses to be associated with an overall survival benefi t when compared with placebo or no intervention in adult patients with severe sepsis. Two meta-analyses [155,158] separately estimated treatment eff ects for IVIg and IVIgAM and found a strong treatment eff ect for IgMenriched IVIg (risk ratio (RR) = 0.66; 95% confi dence interval (CI) 0.51 to 0.85.) and a borderline signifi cant eff ect for IVIg (RR = 0.81; 95% CI 0.70 to 0.93) [158]. When analyses were restricted to studies at low risk of bias, neither IVIg nor IgM-enriched IVIg showed signifi cant benefi t at the 5% level (RR = 0.97; 95% CI 0.81 to 1.15; 5 trials, n = 945) [158]. Likewise, the metaanalyses restricted to 'high quality' trials report nonsignifi cant results with IVIg treatment [154,156,158]. Th e reasons for heterogeneity in treatment eff ects include dosage regimen, duration of therapy, trial quality, publication date and whether patients had septic shock or other forms of severe sepsis [153]. In seven studies (560 patients) that used either a total dose ≥1 g IVIg per kilogram body weight (RR = 0.61; 95% CI 0.40 to 0.94]), or provided IVIg therapy for ≥2 days (17 trials, n = 1,847, RR = 0.66; 95% CI 0.53 to 0.82) there was a strong association with survival benefi t [153].
It should also be stressed that IVIg therapy is not without side eff ects. Common complications reported include thromboembolic events, renal dysfunction, aseptic meningoencephalitis, and anaphylaxis or anaphylactoid reactions. IVIg are often dispensed as a 5% solution; the eff ects of inappropriate volume loading in critically ill patients could be detrimental. In addition, subclinical sepsis can be associated with IVIg infusion reactions in patients with antibody defi ciency [160].

Future research
Observational research is necessary (i) to characterize changes in Ig concentrations during the septic process and (ii) to delineate mechanisms contributing to any impact on outcome parameters (for example, duration, progression and severity of organ dysfunction, new organ dysfunctions during critical care stay, and fatality).
From the literature review presented, we feel the pleotropic eff ects of IVIg on the sepsis-induced host response are likely to be secondary to both suppression of synthesis and direct scavenging of upstream and downstream mediators of the host response, and complex yet unclarifi ed immunomodulatory eff ects mediated via Fcγ receptors. Th ese mechanisms require confi rmation with well-conducted pharmacodynamic studies to provide the rationale for use of a specifi ed dose and duration. Whether plasma Ig levels, or another variable, can be a useful theragnostic marker for identifying and optimally treating a septic cohort also requires delineation.
Pharmacokinetic studies of IVIg in sepsis are yet to be performed, and this is an important omission. Data for dosage selection in current practice are principally derived from studies in volunteers and in patients with primary immune defi ciencies and other indications for immunomodulation [161]. In severe sepsis, potential confounders include systemic infl ammation with fl uctuations in immune function, increased vascular permeability, massive trans-compartmental fl uid shifts and endothelial dysfunction. Existing pharmacokinetic studies [161] also do not address Ig clearance nor the serum Ig concentration to which dosing was targeted for modelling dosing calculations in sepsis.
Such observational studies will crucially underpin the design of an explanatory interventional trial by informing the hypothesis for justifying an IVIg intervention, that is, replacement of low Ig concentration to physiological levels versus immunomodulation. Th e dosing and frequency of IVIg administration may diff er signifi cantly depending on the underlying scientifi c rationale. A theragnostic marker(s) may identify a high-risk cohort and there may be a predefi ned value for an IgG cutoff . Th is explanatory trial should ideally precede any large, multicentre, interventional trial testing the effi cacy of IVIg in a welldefi ned critically ill population with sepsis.

Conclusion
Severe sepsis results in persistent excessive stimulation of multiple pro-infl ammatory cellular pathways leading to host tissue damage, amplifi cation and dysregulation of the immune response through further stimulation of the pattern recognition receptors. Th is destructive and selfamplifying response to infection is accompanied by a fall in serum Ig concentrations through mechanisms as yet unknown. Ig have many benefi cial eff ects, either as 1. At least 4 out of 9 components of sepsis criteria: temperature >38.5°C or <36°C; white blood cell count >12 × 10 9 l -1 or <3.5 × 10 9 l -1 ; heart rate >100 minute -1 ; respiratory rate >28 minute -1 or fraction of inspired oxygen (FiO 2 ) >0.21; mean arterial pressure <75 mmHg; cardiac index >4.5 l minute -1 m -2 or systemic vascular resistance <800 dyn s cm -5 ; platelet count <100 × 10 9 l -1 ; positive blood cultures; clinical evidence of sepsis (surgical or invasive procedure during the preceding 48 h or presence of an obvious septic focus).
2) Sepsis score 12 to 27 3) APACHE II score 20 to 35 natural, innate Ig or by inducing specifi c antibody through the adaptive immune response. It is logical to predict that replacement of serum Ig through infusion of IVIg would restore important Ig functions as described above. Th e failure to date to show benefi t may be a consequence of the diffi culty in providing meaningful data and the diff erences in preparations used. Stringently controlled studies are required, ideally against direct indicators of the patient's immune status.

Competing interests
Tha authors declare that they have no competing interests.