Clinical Trials of Immunomodulatory Therapies in Severe Sepsis and Septic Shock

Abstract

Sepsis remains one of the leading causes of mortality in critically ill patients. Increased insight into the complexities of this disease process has resulted in the targeting of various aspects of the inflammatory response as offering potential therapeutic benefits. There have been encouraging results in the past few years. Some of the tested agents have been shown to improve mortality rates in large randomized controlled trials involving patients with severe sepsis. In this article, we discuss the positive and negative results of trials in this field; some of the possible reasons for the negative results are examined, and directions for the future are suggested.

Recent years have seen vast amounts of research into the pathogenesis of sepsis, the results of which have led to the development of potential therapeutic strategies, and several agents have now been shown to decrease mortality rates in large, prospective, randomized clinical trials (RCT). In this review, we will focus on the immunomodulatory strategies that have been tested clinically (table 1). We will separate the various interventions into groups, although, in reality, these groups may well overlap, and it is unlikely that, for such a complex disease process, any single agent will be effective for all patients—or even at all times for the same patient. A combination of strategies—a “cocktail”—will more likely produce definitive results.

Table 1

Some immunomodulatory therapies tested in clinical trials.

Antitoxin Interventions

Many components of bacteria, viruses, and fungi, including endotoxins, peptidoglycans, lipoteichoic acids, and exotoxins, are able to trigger the inflammatory response. Of these components, the effects of endotoxins have been the most extensively studied [1].

Early approaches to endotoxin-directed therapy included the administration of nonspecific polyclonal immunoglobulin preparations [2], anti—core endotoxin antibodies that use human antiserum preparations obtained after injections of Escherichia coli J5 [3, 4], murine antibodies to the lipid A component of endotoxin (E5) [5–7], and human anti—lipid A antibodies (HA1A) [8, 9]. However, despite some encouraging results from early studies, none of the anti—lipid A strategies have been shown to be of benefit in large RCTs. Other anti-endotoxin antibodies are, however, being evaluated in preliminary clinical studies.

Another approach has been the use of bactericidal/permeability-increasing protein (BPI). This agent is released from activated neutrophils and binds to the lipid A component of gram-negative bacterial endotoxin. Initial results reported in children with meningococcal sepsis have been encouraging [10].

The binding of polymyxin B, which is effective at binding endotoxin but toxic when given systemically, to an insoluble polystyrene fiber creates a hemofiltration filter that may remove endotoxin more specifically than do classical filters. The results from limited clinical reports—primarily from Japan, where this system is already in use—have been encouraging [11–13]. The results of a European RCT are presently being analyzed (data on file; Toray Industries; Tokyo). Polymyxin B can also be combined with dextran and infused as an intravenously administered fluid [14, 15], but clinical trials that have used this preparation have not yielded very successful results (data on file; Novartis; Basel, Switzerland). Another extracorporeal system that uses albumin to eliminate endotoxin is being investigated [16].

Antimediator Interventions

Immunomodulating strategies have been aimed at many points in the sepsis response, and, as we continue to discover more about the mediators involved and their particular roles, further possible treatment targets are being revealed.

Corticosteroid therapy. Corticosteroids have a range of anti-inflammatory actions, including inhibition of the production of the key cytokines TNF and IL-1, inhibition of inducible nitric oxide synthase (iNOS), and a decreased release of other mediators, including platelet-activating factor (PAF). Corticosteroids might therefore be expected to be of benefit to patients with sepsis, and some early studies [17] did show reduced mortality rates for patients with sepsis treated with methylprednisolone or dexamethasone. However, later studies failed to show that high-dose steroid treatment had any benefit on patient outcome [18–20]. Two meta-analyses of the literature concluded that there was no evidence to support the use of steroids in patients with sepsis [21, 22], although it is possible that corticosteroid therapy may be beneficial for certain groups of patients—for example, those with severe typhoid fever [23]. More recent studies have suggested that the use of lower doses of hydrocortisone may have beneficial hemodynamic effects [24]; the concept is that relative adrenal insufficiency is present, which requires smaller doses of hydrocortisone than were used in the studies mentioned above. A multicenter RCT in France that involved 300 patients showed an improved survival rate for patients treated with 50 mg of hydrocortisone for 5 h [25].

Anti-TNF therapy. TNF is a key mediator in the sepsis response and produces many of the acute physiologic changes seen in sepsis when administered to human volunteers [26]. Murine anti-TNF monoclonal antibody preparations (CB0006, Celltech; BAYx1351, Nereliomomab, Bayer) were the first anti-TNF agents to be tested for treatment of human septic shock, but, despite some early positive results [27, 28], a large RCT (NORASEPT II) [29] found no beneficial effect on mortality, the duration of septic shock, or the resolution of sepsis-induced organ failure, even in a subgroup of patients who had elevated TNF levels on study entry.

MAK 195F (afelimomab; Knoll AG), a F(ab′)2 fragment of a murine monoclonal antibody, was developed to reduce the potential immunogenicity of TNF antibodies and to facilitate tissue penetration. Two multicenter RCTs, one in Europe and Israel [30] and the other in the United States and Canada [31], have investigated its effects by enrolling patients with sepsis, stratifying them according to their IL-6 levels, and randomizing them to receive either placebo or afelimomab for 3 days. The American study (MONARCS) [31] reported a 10% decrease in relative mortality risk and beneficial effects on organ dysfunction in the afelimomab-treated patients. Importantly, there were no significant differences between treatment and placebo arms in the incidence of secondary infection.

TNF acts via 2 distinct receptors, TNFR1 (p55) and TNFR2 (p75), which are regulated separately on the cell membrane. Soluble TNF receptors are produced by proteolytic cleavage from the cell-bound form, and they are upregulated in sepsis, probably acting as a negative control on TNF activity. One RCT [32] showed a trend toward reduced 28-day mortality rates for patients with severe sepsis who received the higher dose of soluble TNFR1 receptor; however, another RCT, which involved 1342 patients, was not able to duplicate these results [33].

Anti–IL-1 therapy. IL-1 acts synergistically with TNF [34] to produce the hemodynamic features of septic shock. Studies of patients with sepsis have focused on the IL-1 receptor antagonist (IL-1RA), a naturally occurring macrophage-produced protein that binds to IL-1 receptor. A phase 2 clinical trial of recombinant human IL-1RA that involved 99 patients with sepsis suggested that there is a dose-related survival benefit [35]. An initial multicenter RCT that involved 893 patients supported these findings, with suggestion of a dose-related increase in the duration of survival for patients who had sepsis with organ dysfunction and/or who had a predicted risk of mortality of ⩾24% [36]. However, a second, phase 3 trial was terminated after an interim analysis showed no significant differences in mortality rates [37]. The authors suggested several possible explanations for the apparent discrepancies between these 2 large trials [37], including subtle differences in the patient populations, which occurred despite apparently similar inclusion criteria, and the increasing incidence of gram-positive sepsis that may be less susceptible to antimediator therapy.

PAF antagonist therapy. Although an initial RCT that used the PAF antagonist BN 52021 suggested a reduced mortality rate in treated patients with gram-negative sepsis [38], a phase 3 trial involving 609 patients with gram-negative sepsis failed to confirm these findings [39]. Possible reasons for these different results include the fact that, although both studies were multicenter, the first [38] was limited to France, whereas the second [39] extended across Europe, increasing the potential for differences in patient management, antimicrobial susceptibility patterns, pathogen patterns, and interpretation of entry criteria [39]. In addition, in the period between the 2 studies, general supportive care and treatment of patients receiving intensive care had improved, which likely improved survival rates and made any further improvement that resulted from a new treatment strategy much harder to detect. A study of another PAF antagonist, BB-882 (lexipafant; British Biotechnology), which involved 152 patients, also did not find reduced mortality rates [40].

PAF is inactivated by the enzyme PAF acetylhydrolase (PAF-AH), which converts PAF to the inactive metabolite lyso-PAF, and it has been suggested that it confers protective antiinflammatory effects. Reduced PAF-AH activity in severely injured patients was associated with the increased development of multiple-organ failure. In a preliminary study of recombinant PAF-AH that involved 127 patients, the drug was well tolerated and there was evidence of improved oxygenation and reduced organ dysfunction [41]. A large multicenter study is under way [41].

Coagulation modulating therapy. Coagulation abnormalities in sepsis clinically result in a procoagulant state, which often manifests as disseminated intravascular coagulation. Antithrombin (AT) is a natural inhibitor of thrombin and of other serine proteases involved in coagulation. In patients with sepsis, AT levels are reduced, and low levels have been correlated with a poor outcome [42]. Administration of AT concentrates to patients with disseminated intravascular coagulation was suggested to have beneficial effects on hemostasis in small early studies [43, 44]; however, in a large, multicenter RCT that involved >2000 patients, AT administration was found to have had no effect on mortality of the subjects, with the possible exception of a subgroup of patients who did not receive heparin [45].

Protein C is a potent anticoagulant that inhibits factors Va and VIIa, activates fibrinolysis, and inhibits thrombin production. Small case series have demonstrated beneficial effects of protein C supplementation in patients with meningococcal sepsis [46–50]. Activated protein C, which has additional anti-inflammatory effects [51], was recently shown to have reduced the relative risk of death by 19% in a multicenter RCT that involved 1690 patients with severe sepsis [52]. This treatment effect was present regardless of patient age, number of organ failures, type of infection, or degree of protein C deficiency, and treatment was associated with reduced IL-6 levels.

In sepsis, tissue factor is produced by activated vascular endothelial cells and monocytes, which triggers the extrinsic coagulation pathway with factor VIIa. Tissue factor pathway inhibitor is a natural inhibitor of tissue factor, directly inhibiting activated factor X and indirectly inhibiting factor VIIa/tissue factor activity. Trials involving tissue factor pathway inhibitor have yielded encouraging results, but a just-recently completed phase 3 study did not show a reduction in the mortality rate (data on file; Chron; Emeryville, California).

Complement and contact system therapy. C1 inhibitor (C1-INH) is a naturally occurring inhibitor of both the classical complement pathway and of factor XII–mediated contact activation. In sepsis, elevated levels of inactive C1-INH are observed and have been associated with a worse patient outcome [53]. C1-INH concentrate has been used in pilot studies in patients with sepsis [54, 55] and may reduce the need for vasopressor agents.

Activation of the kallikrein-kinin system results in the release of bradykinin, a potent vasodilator that may be involved in the hemodynamic alterations seen in sepsis. A multicenter RCT that used deltibant (Bradycor; Cortech), a bradykinin antagonist, for treatment of patients with septic shock revealed that it had no overall effect on the 28-day mortality rate, although mortality rates were reduced in the subset of patients with gram-negative infection [56].

Arachidonic acid metabolite therapy. The cyclooxygenase and lipoxygenase pathways are activated in sepsis with the production of prostaglandins (PG) and leukotrienes. PGE1 and PGI2 (prostacyclin) have important anti-inflammatory effects, blocking macrophage activation and inhibiting the release of oxygen radicals and lysosomal enzymes. In clinical trials, however, administration of PGE1 failed to have any positive effect on the outcome for patients with acute respiratory distress syndrome (ARDS) [57]. More recently, the use of PGE1 bound to liposomes has been investigated. Use of this combination has resulted in an additive increase in intracellular cyclic adenosine monophosphate and a downregulation of CD18, which decreased neutrophil activation and adhesion [58]. Although an initial phase 2 clinical trial involving patients with ARDS had promising results [59], a larger multicenter RCT showed that TLC C-53 (The Liposome Company) had no effect on mortality rates or ventilator dependency [60].

Another approach to immunotherapy aimed at the arachidonic cascade has employed inhibitors of cyclooxygenase or thromboxane synthetase. In small early studies, patients treated with ibuprofen, a cyclooxygenase inhibitor, had fewer episodes of fever and a trend toward more-rapid reversal of shock [61, 62]. However, a later study that involved 455 patients with sepsis [63] found that ibuprofen had no effect on the development of shock or on survival rates. Ketoconazole, a thromboxane synthetase inhibitor, reduced the rates of development of ARDS and mortality in 1 study of patients with sepsis [64], but a larger trial in North America that involved 234 patients who had either acute lung injury or ARDS showed no beneficial effects on mortality rates or on duration of ventilation [65].

Antioxidant therapy. Clinical trials that have studied antioxidant agents have largely focused on N-acetylcysteine (NAC), an agent that restores cellular antioxidant potential by several mechanisms, including replenishing intracellular stores of glutathione and scavenging reactive oxygen species. In early studies, NAC was shown to have some beneficial effects on hemodynamic and oxygenation status [66–69]; however, other studies have failed to demonstrate any beneficial effects of treatment with NAC for patients with sepsis [70, 71], multiple-organ failure [72], or ARDS [73, 74].

One of the main scavenger systems for oxygen free radicals is the selenium-dependent glutathione peroxidase. Reduced selenium levels have been reported in patients with sepsis and are associated with increased rates of morbidity and mortality [75]. A pilot study reported an improved outcome and reduced incidence of acute renal failure in patients with sepsis who were given selenium replacement, as compared with those who received placebo [76].

Nitric oxide (NO) inhibition therapy. During the course of sepsis, increased amounts of NO are produced, and elevated levels of NO metabolites in patients with sepsis have been correlated with endotoxin levels [77] and with organ-failure scores [78]. However, although NO synthase (NOS) blockade restores blood pressure, it also lowers the cardiac index and increases pulmonary and systemic vascular resistance [79–82]. Selective NOS inhibitors that act against iNOS may be preferable, allowing the physiologic properties of constitutive NOS to be maintained [83, 84].

An alternative approach to NO blockade has been the use of methylene blue. Although it has some direct inhibitory action on NOS, methylene blue predominantly inhibits guanylyl cyclase and has therefore been proposed as a selective inhibitor of the hemodynamic effects of NO [85, 86]. In small, uncontrolled studies, short-term infusions of methylene blue have been shown to transiently increase arterial pressure, myocardial function, and oxygen delivery in patients who have septic shock [87–89].

Pentoxifylline therapy. Pentoxifylline is a phosphodiesterase inhibitor that elevates intracellular cyclic adenosine monophosphate levels and influences various aspects of the immune response, including inhibiting the release of TNF. Small clinical trials that used pentoxifylline have yielded conflicting data regarding effects on outcome [90–92].

Adhesion molecule therapy. Adhesion molecules (including integrins, intercellular adhesion molecules, vascular cell adhesion molecules, and E-selectin) are expressed on endothelial cell surfaces and mediate the interaction of the endothelium with leukocytes, resulting in the release of reactive oxygen species and arachidonic acid metabolites, which are key components of the inflammatory response and effectors of permeability alterations and organ damage. In a pilot trial that involved patients with septic shock [93], an anti–E-selectin antibody, CY1787, was well tolerated, and there were suggested beneficial effects on organ dysfunction and shock. However, a recent study that involved primates showed that an antibody to E- and S-selectin had no effect on E. coli–induced lung injury and decreased the duration of survival [94].

Hemofiltration. The use of various extracorporeal epuration techniques, including plasma exchange, hemofiltration, and plasmapheresis, to remove toxins and inflammatory mediators has been associated with improved hemodynamic status and outcome in animal models of sepsis [95]. In patients with sepsis, cytokines have been detected in the ultrafiltrate after hemofiltration [96, 97], but this is not always associated with a reduction in circulating cytokine levels [97–99]. Some small studies have reported an increase in arterial pressure with hemofiltration [97, 99], but others have found no effect on hemodynamics [98, 100].

Many questions regarding the use of hemofiltration in patients remain unanswered [101]. The type of membrane used is crucial, because different membranes have different convective and adsorptive clearances of inflammatory mediators [102]. Hemofiltration is relatively nonspecific and removes both pro- and anti-inflammatory mediators, as well as natural cytokine inhibitors [100]. As discussed above, the binding of polymyxin B or polymyxin B–dextran to an insoluble polystyrene fiber creates a hemofiltration membrane that may remove endotoxin more specifically than do classic filters [15, 103]. Filtration volume may also be of importance, because several experimental studies have demonstrated beneficial effects on hemodynamics and survival with high flows [95, 104, 105], whereas studies with low flows have not revealed beneficial effects [106, 107]. Clinical trials conducted in this area have been small and largely uncontrolled, and they are difficult to compare, because they include heterogeneous populations, different filter types, and varying ultrafiltration rates. To fully evaluate the possible effects of hemofiltration in patients with sepsis, a large RCT is needed, but this may be difficult to perform, in view of the many possible hemofiltration regimes.

Immunostimulation

Immunonutrition. Immunonutrition has received considerable attention in recent years for its potential to modulate the inflammatory response in critically ill patients [108, 109]. Various supplements have been proposed, including arginine, glutamine, nucleotides, and omega-3 fatty acids, and studies have been conducted that involve different groups of critically ill patients [110–113]. Several studies have reported reduced infection rates in critically ill patients treated with enteral immune-enhanced feeds [112–115].

IFN-γ therapy. In patients with sepsis, monocyte deactivation has been reported; monocyte deactivation is characterized by a reduction in human leukocyte antigen–DR expression and a reduced capacity to synthesize pro-inflammatory cytokines, and it is associated with an increased mortality rate [116]. IFN-γ is a major activator of monocytes, which can restore function in monocytes that have been deactivated by sepsis [117]. Clinical trials that used IFN-γ have concentrated on patients with major trauma or burn injuries. In a multicenter RCT that involved 213 trauma patients at high risk of infection, Polk et al. [118] reported no significant differences in infection rates or outcome between IFN-γ–treated and placebo-treated groups. Dries et al. [119] reported a decrease in the number of infection-related deaths among IFN-γ–treated patients in an RCT that involved 416 patients with severe trauma. However, in a multicenter RCT of 216 patients with major burns, Wasserman et al. [120] reported that IFN therapy did not significantly differ from placebo with regard to infectious complications, duration of stay in the hospital or intensive care unit, or patient outcome.

Granulocyte colony-stimulating factor (G-CSF) therapy. G-CSF is a naturally occurring glycoprotein that promotes the production, maturation, and function of neutrophils, and in animal studies, it has been shown to improve survival rates among animals with sepsis. G-CSF is widely used during chemotherapy in humans to prevent neutropenia and its associated increased risk of infection. Clinical trials of G-CSF have yielded conflicting results [121, 122]. In a study that involved 78 patients with burn-related sepsis, receipt of G-CSF treatment improved outcome [123], and, in patients with communityacquired pneumonia, there was a trend toward a reduced complication rate of pneumonia among subjects who received G-CSF [124]. In a follow-up study that involved 480 patients, a trend was seen toward reduced mortality rates among patients with pneumococcal pneumonia [125]. However, a multicenter trial of G-CSF treatment that involved 701 patients with pneumonia and sepsis showed no changes in mortality rates, organ dysfunction, or duration of ventilation in treated patients, as compared with placebo recipients (R.K. Root, personal communication). The timing of G-CSF administration may be crucial; prophylactic use has attained more consistent beneficial effects in animal models [126]. A multicenter phase 2 trial that evaluated the effect of prophylactic use of G-CSF on the incidence of nosocomial infections in patients with acute traumatic brain injury or cerebral hemorrhage [127] showed no beneficial effects on length of stay or incidence of nosocomial pneumonia, but there was a dose-dependent reduction in the frequency of bacteremia.

PGG-glucan therapy. PGG-glucan (poly-[1-6]-B-d-glucopyranosyl-[1-3]-B-d-glucopyranose) is derived from yeasts and promotes phagocytosis and intracellular killing of bacterial pathogens by leukocytes. After promising results from phase 1 and 2 studies [128, 129], a phase 3 study was conducted that involved 1249 patients undergoing gastrointestinal surgery [130]. PGG-glucan (1 dose given preoperatively and 3 given postoperatively) had no overall effect on mortality.

The Past: Why Have Trials Failed?

The results of clinical trials of immunomodulatory therapies have been disappointing more than encouraging, and many reasons, outlined below, have been put forward to explain the apparent “failures” [131–133].

The experimental agents are ineffective. There are certainly examples in which, in retrospect, we can say that the agents being tested were not able to do what they purported. The anti-endotoxin agents HA-1A and E5 were supposed to bind to the lipid A portion of endotoxin and neutralize endotoxin activity; in fact, in vitro testing showed that neither of these compounds were able to limit endotoxin activity or to reduce the release of IL-1 or TNF [134].

Doses of experimental agents are inadequate. With inadequate means of monitoring the effects of therapy on the immune response, it is difficult to establish dose-response curves, and initial doses often have to be estimated from animal models or from limited clinical trials.

Timing of intervention is inadequate. One difficulty in the use of anticytokine treatments is that it is not possible to predict the development of sepsis, and, thus, patients enrolled in clinical trials frequently already have well-established sepsis. Patients with louseborne relapsing fever pretreated with murine anti-TNF Fab have a reduced penicillin-induced Jarisch-Herxheimer reaction and associated increases in plasma IL-6 and IL-8 levels [135]. In other forms of sepsis, such pretreatment schedules are not readily applicable. TNF and IL-1 are released early in the course of sepsis; thus, there is a narrow window of opportunity for effective treatment. High-mobility group 1 protein is a mediator of sepsis that is released late—more than 8 h after endotoxin stimulation in cultured mouse macrophages—and elevated levels are associated with increased mortality rates [136]. Treatment targeted against such distal mediators may be more effective in patients with established sepsis.

Patient population is too heterogeneous. Patients with sepsis are a mixed group with diverse ages who present with diverse underlying conditions, and they have sepsis caused by various organisms and from different sites and origins. In addition to the innate heterogeneity of the population, terms such as “systemic inflammatory response syndrome” do little to reduce the “nonspecific septic” nature of the patients included in clinical trials. As we now understand much more clearly, the immune response varies among patients and in the same patient over time [137]. It is the balance between the pro- and anti-inflammatory aspects of the immune response that will determine, in large part, how a patient will respond to an immunomodulatory treatment. Thus, in studies with varying degrees of response, a single anti-inflammatory therapy, for example, will cause benefit in those patients with a predominantly pro-inflammatory response, but this may be negated by the harm done to patients who have a predominant anti-inflammatory response and who may rather have benefited from receipt of pro-inflammatory therapy. Indeed, many of the studies have shown beneficial effects in certain subgroups or in retrospective analyses [5, 32, 36, 38, 56, 138, 139]. Similarly, initial small pilot trials that have included perhaps more-select patient groups have often shown some benefit, which cannot be duplicated in the larger phase 3 trials [61, 140–143].

Single therapies may be ineffective. With the complexities of the immune response, it may be that single therapies are inadequate and that combinations of several agents will be necessary.

The Future: To Success!

Unless a common pathophysiologic mechanism is identified that occurs in all patients with sepsis, and unless an effective treatment becomes available that targets that pathway, improved patient classification for clinical trial inclusion will be the key to the future of effective antisepsis therapies. Developments in this field have focused on 2 main areas: genetic testing and immunologic modeling.

Genetic testing. Genetic aspects of the response to sepsis may be increasingly important, with some studies showing that several polymorphisms, including TNF-2 and IL-1RAA2, may be related to the susceptibility and outcome of septic shock [144–146]. Similarly, in meningococcal disease, presence of the homozygous 4G deletion polymorphism in the plasminogen-activator inhibitor 1 gene has been reported to be associated with an increased mortality rate [147]. Genetic testing could therefore identify early certain groups of patients who may benefit from targeted treatment—for example, a high TNF producer would likely benefit more from anti-TNF therapy than would a low TNF producer.

Immunologic monitoring. Immunologic monitoring, by use of cellular stimulation techniques to evaluate the production of cytokines by monocytes from patients with sepsis after stimulation with endotoxin, is being developed [117]. However, these techniques are cumbersome and involve in vitro analysis, which may not be directly applicable to circulating cells. Further work is needed to establish the full potential of this approach, but such techniques may help to characterize the degree of immune response and to guide immunomodulating therapies more effectively; thus, a patient with a predominantly pro-inflammatory response can be given anti-inflammatory treatment, whereas another patient with an anti-inflammatory profile may be given an immunostimulatory therapy.

Conclusion

Our knowledge of the pathophysiology of the inflammatory response in sepsis continues to expand—and, with it, the range of possible therapies aimed at modulating this response. Potential new therapies continue to be developed aimed at stimulating as well as inhibiting the inflammatory response. Although we should be encouraged and excited by the positive results from recent clinical studies, we cannot afford to sit back; improvements in clinical trial design can be made as a result of lessons learned from previous trial failures. In addition, many questions remain unanswered. For example, which patients should be treated and when? What doses should be used, and for how long? When can we consider combination therapies? Only when these questions have been addressed will we begin to significantly affect the considerable mortality associated with sepsis.