The immune status of the host is critical in determining the risk/benefit associated with corticosteroids therapy. Corticosteroids are likely to compound the immuno-paresis in immune-paresed patients increasing the risk of acquired infections. Classic teaching suggests that tissue injury from trauma and surgery results in a systemic inflammatory response syndrome (SIRS) with “unbridled inflammation” which after a few days/weeks evolves into an immuno-paretic phase known as the compensated anti-inflammatory response syndrome (CARS) [5–9]. However, multiple reports over the last two decades have indicated that the proliferative response to T cell mitogens is significantly impaired in patients and experimental animals immediately after traumatic or thermal injury [10–14]. The T-cell dysfunction after traumatic stress is characterized by a decrease in T-cell proliferation, an aberrant cytokine profile, decreased T-cell monocyte interactions and attenuated expression of the T-cell Receptor Complex (TCR). Furthermore, surgical stress induces a shift in the T-helper (Th)1/Th2 balance resulting in impaired cell mediated immunity [15–17]. While the Th1 cytokines may be increased following trauma and surgery, these cytokines do not reach the levels seen in patients with sepsis and unlike patients with sepsis, the Th2 response predominates. As corticosteroids are likely to compound the immuno-paresis following trauma and stress these agents are probably best avoided in the surgical patient who becomes septic. This hypothesis is supported by the failure of corticosteroids to improve outcome in the Corticosteroid Therapy of Septic Shock Study (CORTICUS) study where the majority of patients were surgical patients [18]. It would therefore appear illogical to give septic post-surgical patient corticosteroids as this is only likely to compound the immunosuppressive state and increase the risk of secondary infections (which is exactly what the CORTISUS study demonstrated). Furthermore, it should be noted that the incidence of CIRCI is very low in surgical/trauma patients. In a 5 year retrospective study of 2,100 trauma patients admitted to an ICU the incidence of CIRCI was only 3.3 % [19]. Similarly, in an analysis of 1,795 intubated trauma patients, 82 (4.5 %) were diagnosed with adrenal insufficiency [20]. Fann and colleagues performed statistical modeling to predict adrenal insufficiency in trauma patients [21]. In this study 3.3 % of patients admitted to the ICU were diagnosed with adrenal insufficiency.

Boomer and colleagues performed cytokine secretion assays and immunophenotyping of cell surface receptor-ligand expression profiles from postmortem spleen and lung tissue samples from 40 patients who died from sepsis and 29 brain-dead controls [22]. In this study, patients who died in the ICU following sepsis compared with patients who died of non-sepsis etiologies had biochemical, flow cytometric, and immunohistochemical findings consistent with severe immunosuppression. In patients with sepsis, the initial pro-inflammatory response is followed by three distinct clinical pathways, namely i) homeostasis is restored with return to a “normal immune status” ii) patients may develop a prolonged a pro-inflammatory response with ongoing tissue injury iii) while other patients may progress to a state of immuno-suppression (CARS). It is important for clinicians to be able to accurately determine the patients’ immune status before instituting immunomodulating interventions. It is likely that patients who receive corticosteroids late in the course of their disease and have progressed to CARS will suffer adverse sequela from such therapy (see timing below). Future research in sepsis will need to focus on developing tools that can dynamically and in real-time characterize the patient’s immune response to allow targeted immune therapy.

Since steroids enhance local immune defences but reduce global NF-kappa B expression and cause a predominant TH2 immunosuppressive state, steroids are likely to be beneficial early in the course of the disease but likely to compound the immunosuppression when given later in the course of sepsis. The time dependent initiation of the use of corticosteroids has not been taken into consideration in those studies (and meta-analyses) which have analyzed the benefits/risk of steroids in sepsis. In the study by Annane et al., the window for enrollment into the study was initially 3 h and then it was increased to 8 h [23]. In the CORTICUS study, the initial time frame of 24 h, increased to 72 h [18]. This time dependent effect was demonstrated by Park et al. who in a retrospective analysis of 178 patients with septic shock found that corticosteroids were only of benefit if given within 6 h after the onset of septic shock-related hypotension [24]. Similarly, Katsenos et al. demonstrated that in patients receiving hydrocortisone for septic shock initiation of therapy within 9 h was associated with improved survival [25]. Furthermore, ex-vivo mononuclear stimulation studies demonstrated attenuated TNF-α release only in those patients who received early corticosteroid therapy.

The effect of glucocorticoids on immune suppression is critically dose dependent. It is well know from the organ transplant experience that high-dose corticosteroids effectively abolish T-cell mediated immune responsiveness and are very effective in preventing/treating graft rejection. However, while stress-doses of corticosteroids inhibit systemic inflammation with decreased transcription of pro-inflammatory mediators, they maintain innate and acquired immune responsiveness and do not increase the risk of secondary infections [26–28]. Lim et al. demonstrated that the effect of corticosteroids on macrophage function was dose dependent [29]. Low doses enhanced macrophage function whereas high doses strongly depressed macrophage function.

It is important to recognize that patients with ARDS and many with sepsis have prolonged immune dysregulation requiring a more prolonged course of therapy [30]. Two longitudinal studies in patients with severe community acquired pneumonia found high levels of circulating inflammatory cytokines 3 weeks after clinical resolution of sepsis [31, 32]. Trials from the 1980s which investigated short-term (24–48 h) massive glucocorticoid doses (up to 40,000 mg/hydrocortisone eq./day) were associated with an increased risk of side effects, and no clear outcomes benefit [33, 34]. Recent studies, which investigated the use of low dose (stress dose) corticosteroids given over a more prolonged period have shown clinical benefit in terms of reduction in mortality with an increase in pressor free, ventilator free and ICU free days [33, 34].

Corticosteroids should never be stopped abruptly; this will lead to a “rebound” of inflammatory mediators with an increased likelihood of hypotension and/or rebound inflammation (lung injury). There is ample evidence that early removal of glucocorticoid treatment may lead to rebound inflammation and an exaggerated cytokine response to endotoxin [26, 35–42]. Experimental work has shown that short-term exposure of alveolar macrophages or animals to dexamethasone is followed by enhanced inflammatory cytokine response to endotoxin [43, 44]. Similarly, normal human subjects pretreated with hydrocortisone had significantly higher TNF-α and IL-6 response after endotoxin challenge compared to controls [45]. Two potential mechanisms may explain rebound inflammation: homologous down-regulation and GC-induced adrenal insufficiency. Glucocorticoid treatment down-regulates the GR levels in most cell types, thereby decreasing the efficacy of the treatment [46]. Down-regulation takes place at both the transcriptional and translational level, and hormone treatment decreases receptor half-life by approximately 50 % [46]. In experimental animals, overexpression of GRs improves resistance to endotoxin-mediated septic shock while GR blockade increases mortality [47].

Not all patients with sepsis/ARDS treated with corticosteroids respond to this treatment. Genetic polymorphisms of a number of genes may explain this finding. Gessner demonstrated that hydrocortisone failed to abolish NF-кB protein nuclear translocation in deletion allele carriers of the NFKB promoter polymorphism (-94ins/delATTG) [48]. In addition, these authors demonstrated that patients with this polymorphism receiving hydrocortisone had a much greater 30-day-mortality (57.6 %) than the other genotypes (24.4 %; HR: 3.18, 95 % CI: 1.61–6.28; p = 0.001). It is likely that other polymorphisms including those of the glucocorticoid receptor (GR) may influence the clinical response to glucocorticoids [49].

Deceased concentration or abnormal function of the GR may underlie the observation of variability of cortisol sensitivity amongst patients. Cortisol diffuses rapidly across cell membranes binding to the GR. Two isoforms of the GR have been isolated, namely GR-α and GR-β. The GR-β isoform fails to bind cortisol and activate gene expression and thus functions as a negative inhibitor of GR-α [50]. Through the association and disassociation of chaperone molecules the glucocorticoid-GR-α complex moves into the nucleus where it binds as a homodimer to DNA sequences called glucocorticoid-responsive elements (GRE’s) located in the promoter regions of target genes which then activate or repress transcription of the associated genes.

Guerrero et al. demonstrated increased expression of the GR-β isoform in patients with sepsis [51]. In a sheep model of ALI induced by Escherichia coli endotoxin, Liu et al. demonstrated decreased nuclear GRα binding capacity [52]. In an ex vivo model Meduri and colleagues compared the cytoplasmic to nuclear density of the GR-complex in patients with ARDS who were improvers with those of non-improvers [53]. These authors demonstrated a markedly reduced nuclear density of the GR-complex in non-improvers while the cytoplasmic density was similar between improvers and non-improvers. This study suggests glucocorticoid resistance due to diminished nuclear translocation of the GR-complex. Siebig et al. demonstrated deceased cytosolic receptor levels in critically ill patients as compared to control subjects [54]. Similarly, van den Akker noted that children with sepsis or septic shock had depressed levels of glucocorticoid receptor mRNA in their neutrophils.

Over the last three decades approximately 20 randomized controlled trials (RCTs) have been conducted evaluating the role of glucocorticoids in patients with sepsis, severe sepsis, septic shock and ARDS. Varying doses (37.5–40,000 mg/hydrocortisone eq./day), dosing strategies (single bolus/repeat boluses/continuous infusion/dose taper) and duration of therapy (1–32 days) were used in these studies [33, 55]. Despite multiple guidelines and over 20 meta-analyses, the use of glucocorticoids in patients with sepsis remains extremely controversial with conflicting recommendations. Furthermore, while there are large geographic variations in the prescription of glucocorticoids for sepsis up to 50 % of ICU patients receive such therapy [56]. Currently a number of large multicenter RCT’s are being conducted, which should hopefully resolve this issue. While it is difficult to make strong evidence based recommendations at this time, an evidence based review of the literature allows one to make the following conclusions [57, 58]: