INTRODUCTION — Interest in the role of glucocorticoids in the pathophysiology of critical illness has existed since the early part of the 20th century, when investigators observed that the adrenal glands are crucial to survival under conditions of physiologic stress [1,2]. The subsequent development of cortisone for clinical use was followed by clinical studies exploring the potential therapeutic role of corticosteroids in the treatment of severe infections [3-7].

The following issues are discussed in this topic review:

• Effects of critical illness on the hypothalamic-pituitary-adrenal axis. (See 'Effects of critical illness' below.)

• Methods of evaluating adrenal reserve (random cortisol levels, synthetic adrenocorticotropic hormone [ACTH] stimulation tests, free cortisol levels) and the impact of each on prognosis. (See 'Assessing adrenal reserve' below.)

• The concept of relative adrenal insufficiency. (See 'Relative adrenal insufficiency' below.)

• Clinical evidence from trials evaluating the effect of corticosteroid therapy in patients with septic shock, plus practical aspects of corticosteroid administration. (See 'Corticosteroid therapy' below.)

• Recommendations regarding the role of corticosteroid therapy in patients with septic shock. (See 'Summary and Recommendations' below.)

Other aspects of the management of severe sepsis and septic shock are reviewed separately. (See "Management of severe sepsis and septic shock in adults".)

EFFECTS OF CRITICAL ILLNESS — The hypothalamus continuously integrates stimuli and secretes corticotropin releasing hormone (CRH) whenever homeostasis is threatened. CRH stimulates the release of adrenocorticotropic hormone (ACTH) from the anterior pituitary which, in turn, induces cortisol secretion from the adrenal glands. Cortisol is the effector molecule of the hypothalamic-pituitary-adrenal (HPA) axis.

Normal serum cortisol levels range between 5 to 24 mcg/dL, with significant variability depending on the time of day [8]. During physiological stress like major surgery, hypotension, or severe infection, the HPA axis is activated and diurnal variation is lost [9,10]. As a result, serum cortisol increases, reaching levels as high as 40 to 50 mcg/dL (graph 1) [9,11-14].

In addition to activation of the HPA axis by physiological stress, cortisol metabolism and function may be considerably altered by other aspects of critical illness. As an example, renal dysfunction may prolong the half-life of circulating cortisol. Plasma concentrations of cortisol binding globulin (CBG) and albumin, which bind >90 percent of circulation cortisol under normal circumstances, are frequently diminished, resulting in an increase in free cortisol, the physiologically active form [11,15,16]. Inflammatory cytokines may increase glucocorticoid receptor affinity, disable glucocorticoid inactivation, and increase the peripheral conversion of precursors to cortisol [11,17,18].

There are circumstances in which cortisol availability can not be increased sufficiently to meet demand. As an example, drugs like ketoconazole, phenytoin, or etomidate impair cortisol synthesis [11,19,20]. In an observational study of 62 patients, use of etomidate for intubation was associated with increased likelihood of having a poor response (≤9 mcg/dL) to ACTH stimulation 24 hours later (odds ratio 12, 95% CI 3-50) [21]. Head injury, central nervous system depressants, pituitary infarction, adrenal hemorrhage, infections, malignancy, and previous corticosteroid therapy can also impair the HPA axis [11]. Sometimes an etiology can not be identified.

ASSESSING ADRENAL RESERVE

Random serum cortisol — Serum cortisol levels vary widely in patients with septic shock [14,22-27]. This observation led multiple investigators to try to determine whether serum cortisol is predictive of mortality in patients with septic shock. Numerous studies found increased mortality associated with both low serum cortisol levels and high serum cortisol levels [28-33]. Other studies were unable to find a relationship between cortisol levels and mortality [23,25,34].

High-dose ACTH test — A high-dose adrenocorticotropic hormone (ACTH) stimulation test requires that a baseline serum cortisol be drawn and then 250 mcg of synthetic ACTH (cosyntropin) be administered intravenously. Serum cortisol levels are drawn 30 and 60 minutes later. (See "Evaluation of the response to ACTH in adrenal insufficiency", section on High-dose ACTH stimulation test.)

Studies have yielded variable results in patients with septic shock [23,26,34-36]. In a prospective inception cohort study, a high-dose ACTH stimulation test was performed on 189 patients with septic shock and the results correlated with 28-day mortality [36]. A baseline serum cortisol level >34 mcg/dL and a maximum increase in cortisol of ≤9 mcg/dL were identified as risk factors for death. Three prognostic groups were proposed (graph 2):

• Good prognosis (26 percent mortality): Baseline plasma cortisol ≤34 mcg/dL and a cortisol increase >9 mcg/dL in response to cosyntropin.
• Poor prognosis (82 percent mortality): Baseline plasma cortisol >34 mcg/dL and a cortisol increase ≤9 mcg/dL in response to cosyntropin.
• Intermediate prognosis (67 percent mortality): Baseline plasma cortisol >34 mcg/dL and a cortisol increase >9 mcg/dL in response to cosyntropin, or baseline plasma cortisol ≤34 mcg/dL and a cortisol increase ≤9 mcg/dL in response to cosyntropin.

The results from this study are supported by a more recent and larger retrospective cohort study, in which 477 patients with severe sepsis or septic shock underwent high-dose ACTH testing [37]. Nonsurvivors had a higher baseline cortisol level (30 versus 24 mcg/dL) and a smaller cortisol increase (6 versus 11 mcg/dL) than survivors. Patients with either a baseline cortisol level <15 mcg/dL or a cortisol increase ≤9 mcg/dL had a higher mortality, longer duration of shock, or shorter survival time.

Low-dose ACTH test — Some investigators believe that 250 mcg of cosyntropin is supraphysiologic and stimulates adrenal secretion of cortisol even when adrenal dysfunction exists, noting that even patients with impaired adrenal reserve can mount an apparently normal adrenal response to such a high dose of cosyntropin. Use of low-dose (1 mcg) cosyntropin has been proposed as an alternative [38]. (See "Evaluation of the response to ACTH in adrenal insufficiency", section on 'Low-dose ACTH stimulation test'.)

Two investigations have compared a low-dose (1 mcg) ACTH stimulation test to a high-dose (250 mcg) ACTH stimulation test [39,40].

The first study was a prospective cohort study of 59 patients with septic shock who were sequentially administered 1 mcg and 250 mcg of cosyntropin intravenously [39]. All of the patients then received hydrocortisone 100 mg intravenously every eight hours. Patients were designated as steroid responsive if they were able to maintain a mean arterial blood pressure >65 mmHg without norepinephrine infusion within 24 hours after starting hydrocortisone:

• Adrenal insufficiency (defined as post-cosyntropin serum cortisol of <18 mcg/dL) was detected in more patients by low-dose ACTH testing than high-dose ACTH testing (22 versus 8 percent).
• 54 percent of steroid responders had a diagnostic low-dose ACTH stimulation test, while 22 percent had a diagnostic high-dose ACTH stimulation test. This suggests that adrenal insufficiency detected by low-dose ACTH stimulation was superior to that detected by high-dose ACTH stimulation in predicting hemodynamic responsiveness to corticosteroids.
• Retrospective evaluation demonstrated that a random serum cortisol level <25 mcg/dL was the most accurate diagnostic threshold, predicting steroid responsiveness in 96 percent of the subjects.

The second study was similar, administering low-dose (1 mcg) and high-dose (250 mcg) cosyntropin consecutively to 46 patients with septic shock [40]. Serum cortisol levels were measured at baseline and after each stimulation test. A positive response to ACTH stimulation was defined as a serum cortisol increase >9 mcg/dL. Patients who responded to low-dose and high-dose ACTH were more likely to survive than those who did not respond to either dose. In addition, a response only to high-dose ACTH stimulation was associated with lower survival than a response to both low-dose and high-dose ACTH stimulation, suggesting that low-dose ACTH stimulation identified a subgroup of patients with inadequate adrenal reserve that would have been missed by high-dose ACTH stimulation alone.

Larger studies are needed before the significance of the findings from these two investigations can be fully understood.

Free cortisol — In critically ill patients, loss of cortisol binding globulin (CBG) results in decreased protein-bound cortisol and increased free cortisol. Therefore, for any given amount of serum total cortisol, there is a shift from inactive protein-bound cortisol to physiologically active free cortisol, suggesting that standard assays for plasma cortisol (which measure total plasma cortisol) underestimate HPA axis activity. It has been proposed that free cortisol more accurately reflects HPA axis activation [41].

This was illustrated by a study comparing total serum cortisol and free serum cortisol levels in 33 healthy volunteers to those in 60 critically ill patients (18 had sepsis) [41]. The critically ill patients had total serum cortisol concentrations that were two to three times higher and free cortisol levels that were 7 to 10 times higher than healthy volunteers. Although the total serum cortisol appeared lower in critically ill patients with hypoalbuminemia (albumin ≤2.5 g/dL) than in critically ill patients with preserved albumin concentrations, the free cortisol levels were similar.

The study has two major limitations. First, the critically ill patients did not include any patients with septic shock [42]. Second, the degree of hemodynamic instability at the time of plasma cortisol sampling was not reported; thus it is impossible to correlate the cortisol level with severity of illness or physiologic status [43]. The applicability of the study to routine clinical practice is uncertain because the free cortisol assay is not available at most clinical centers [44].

In another investigation, 74 patients had their total and free cortisol levels measured before and after the administration of 250 mcg of cosyntropin [45]:

• Pre-stimulation free cortisol levels were higher in patients with septic shock (186 nmol/L) than sepsis (29 nmol/L) or healthy controls (13 nmol/L).
• Pre-stimulation free cortisol levels reflected the severity of illness more closely than total cortisol levels.
• The calculated free cortisol level (based on direct measurements of total cortisol and the cortisol binding globulin [CBG] concentration) correlated well with the direct measurement of free cortisol.

Whether the free cortisol level provides useful prognostic information in critically ill patients has not been formally investigated.

RELATIVE ADRENAL INSUFFICIENCY — Cortisol is necessary to survive critical illness [1,2]. But, how much is enough?

Absolute adrenal insufficiency is rare among critically ill patients, with an incidence estimated to be ≤3 percent [46]. However, the following observations argue that suboptimal cortisol production may be common and associated with worse outcomes:

• A maximum increase of serum cortisol ≤9 mcg/dL following intravenous synthetic adrenocorticotropic hormone (ACTH) stimulation is associated with increased mortality from septic shock [36,40]

• Baseline and post-ACTH serum cortisol <23.7 mcg/dL predict responsiveness to exogenous steroids in septic shock, where responsiveness is defined by the ability to maintain a mean arterial blood pressure >65 mmHg without norepinephrine infusion within 24 hours after starting hydrocortisone [39]

Suboptimal cortisol production during septic shock has been termed "functional" or "relative" adrenal insufficiency [11]. There is no consensus about diagnostic criteria or indications for treatment. In addition, there is considerable disagreement over what cortisol level is "normal" or "appropriate" in septic shock, what constitutes an adequate response to ACTH, and what dose of synthetic ACTH should be used for stimulation testing.

Using an overnight metyrapone stimulation test as the "gold standard", an inception cohort study found that patients with septic shock who had a baseline total cortisol level <10 mcg/dL or a change in total cortisol <9 mcg/dL after high-dose ACTH stimulation, were likely to have adrenal insufficiency [47]. Conversely, patients with septic shock who had a baseline total cortisol level ≥44 mcg/dL or a change in total cortisol ≥16.8 mcg/dL, were unlikely to have adrenal insufficiency. The generalizability of these results has been questioned because the frequency of absolute adrenal insufficiency in this study was nearly four times that previously reported [48]. In addition, the values were not used to guide therapeutic decision making and, therefore, their impact on clinical outcome is unknown.

Until large, multicenter, randomized, placebo-controlled trials are complete, we believe that an increase of serum cortisol ≤9 mcg/dL following a high-dose (250 mcg) ACTH stimulation test is suggestive of relative adrenal insufficiency. In the absence of an ACTH stimulation test,we believe that a hemodynamic response to the administration of corticosteroids (successful discontinuation of vasoactive agents within 48 hours after the start of corticosteroids), may reflect relative adrenal insufficiency. Detection of relative adrenal insufficient appears to have prognostic, but not therapeutic, implications.

CORTICOSTEROID THERAPY — Interest in the possible therapeutic role of corticosteroids in severe infections has existed for at least 50 years [4-7,49,50].

Clinical evidence — In the 1980s, three randomized, double-blind, placebo-controlled studies were published [51-53]. None of the trials found a mortality benefit and only one trial noted decreased duration to shock resolution associated with the administration of corticosteroids. All of the trials administered high-dose corticosteroids for a short duration and used early endpoints.

In the 1990s, interest in corticosteroids as a therapy in septic shock was renewed, this time using smaller, more physiological doses for longer duration. Three small trials (approximately 40 patients each) compared corticosteroid administration to placebo in patients with septic shock and found decreased duration of shock associated with corticosteroid therapy [54-56]. These trials prompted larger randomized trials.

French trial — In a multicenter, double-blind trial conducted in France, 300 patients were randomly assigned to receive placebo or hydrocortisone (50 mg intravenously every six hours) plus fludrocortisone (50 mcg enterally once a day) within eight hours of the onset of septic shock [57]. Treatment continued for seven days. Based upon a high-dose (250 mcg) ACTH stimulation test, the patients were classified as having adequate adrenal reserve (maximum increase in serum cortisol of >9 mcg/dL) or inadequate adrenal reserve (maximum cortisol increase of ≤9 mcg/dL):

• Among all patients, hydrocortisone administration decreased 28-day mortality (55 versus 61 percent)
• Among patients with inadequate adrenal reserve, hydrocortisone administration decreased 28-day mortality (53 versus 63 percent), ICU mortality (58 versus 70 percent), and hospital mortality (61 versus 72 percent), as well as more vasopressor withdrawal (57 versus 40 percent)
• Among patients with adequate adrenal reserve, there was no difference in mortality or vasopressor withdrawal

The trial was criticized for its high placebo-group mortality, the statistical methods used to derive mortality, and the concomitant use of fludrocortisone, a potential confounder [58-60].

The association between corticosteroid therapy and improved mortality was supported by two subsequent meta-analyses [61,62]. However, no significant benefit was noted in a large randomized trial (CORTICUS) published after the meta-analyses.

CORTICUS — The Corticosteroid Therapy of Septic Shock (CORTICUS) trial was a multicenter, double-blind trial that randomly assigned 499 patients with septic shock to receive hydrocortisone (50 mg) or placebo intravenously every six hours for five days, followed by a tapering regimen [63]. Based upon a high-dose (250 mcg) ACTH stimulation test, the patients were classified as having inadequate adrenal reserve (maximum cortisol increase of ≤9 mcg/dL) or adequate adrenal reserve (maximum increase in serum cortisol of >9 mcg/dL).

Hydrocortisone administration did not improve 28-day mortality (35 versus 32 percent in the placebo group). This was also true in the two pre-defined subgroups — patients with inadequate adrenal reserve and patients with adequate adrenal reserve. The hydrocortisone group had faster reversal of shock among all patients (3.3 versus 5.8 days in the placebo group) and increased incidence of new infection that did not reach statistical significance (odds ratio 1.27, 95% CI 0.96-1.68). The trial was criticized for the lower than expected placebo-group mortality (32 percent versus the anticipated 50 percent).

Comparison — These randomized trials have important differences that may, at least in part, explain their conflicting results.

• The French trial enrolled patients within eight hours after the onset of shock and included patients with a greater severity of illness [57]. The mean Simplified Acute Physiology Score II (SAPS II) was 55.5 and septic shock was defined as a systolic blood pressure <90 mmHg for more than one hour despite adequate fluid resuscitation and vasopressor administration.
• The CORTICUS trial enrolled patients within 72 hours of the onset of shock and included patients with a lower severity of illness [63]. The mean SAPS II was 49 and septic shock was defined as a systolic blood pressure <90 mmHg despite adequate fluid resuscitation, or the need for vasopressor administration for more than one hour.

These differences imply that corticosteroid therapy is most likely to be beneficial to patients with severe septic shock. This was supported by a meta-analysis of 12 randomized trials, which included 1228 patients with severe sepsis or septic shock who received either prolonged low-dose corticosteroids (≤300 mg hydrocortisone or an equivalent for ≤5 days) or a control (placebo or standard therapy) [64]. Patients who received corticosteroid therapy had reduced 28-day mortality (37.5 versus 44.1 percent, relative risk 0.84, 95% CI 0.72-0.97), with meta-regression analysis demonstrating a trend toward greater impact of corticosteroid therapy among patients with a higher baseline risk of death (p=0.06).

Conclusions — Several conclusions can be based on the available evidence:

• Corticosteroid therapy is most likely to be beneficial to patients who have severe septic shock (which we define as a systolic blood pressure <90 mmHg for more than one hour despite adequate fluid resuscitation and vasopressor administration), especially if begun within eight hours of the onset of shock.
• There are no high or moderate quality data to suggest that corticosteroid therapy is beneficial to patients with less severe septic shock.
• Classification of adrenal reserve as adequate or inadequate (as determined by high-dose ACTH testing) fails to identify patients who are more likely to benefit from corticosteroid therapy. (See 'High-dose ACTH test' above.)

• Data suggesting that corticosteroid therapy may be harmful is of uncertain significance.

Administration — Hydrocortisone is a pharmacologic form of cortisol. Compared to hydrocortisone, other pharmacologic corticosteroids bind cortisol-binding globulin (CBG) poorly, resulting in more free, physiologically active corticosteroid and greater potency at any given dose. The different preparations vary widely in anti-inflammatory and mineralocorticoid potency.

Most studies evaluating the effect of corticosteroids in septic shock used hydrocortisone, although administration protocols and treatment durations varied. It is unknown whether the findings can be generalized to other synthetic corticosteroids. We typically administer 50 mg of hydrocortisone every six hours or 100 mg of hydrocortisone every eight hours.

There is no consensus regarding the optimal administration protocol or treatment duration, because no study has compared fixed-duration regimens to clinically-guided regimens, or tapering to abrupt cessation [65]. However, in one clinical study, abrupt cessation was associated with rebound of hemodynamic abnormalities and increased inflammatory markers [66].

Although one of the large randomized trials added fludrocortisone at a dose of 50 mcg per day for seven days, we do not typically add fludrocortisone because we believe that hydrocortisone alone has sufficient mineralocorticoid effect and because the absorption of the enterally administered drug is questionable in situations of compromised splanchnic perfusion [57].

SUMMARY AND RECOMMENDATIONS

Summary

• High-dose (250 mcg) synthetic adrenocorticotropic hormone (ACTH) stimulation appears to separate patients with septic shock into different prognostic groups. (See 'High-dose ACTH test' above.)

• Suboptimal cortisol production during septic shock has been termed "functional" or "relative" adrenal insufficiency. There is no consensus about diagnostic criteria. (See 'Relative adrenal insufficiency' above.)

• Corticosteroid therapy may be beneficial to patients who have severe septic shock (defined as a systolic blood pressure <90 mmHg for more than one hour despite adequate fluid resuscitation plus vasopressor administration), especially if begun within eight hours of the onset of shock. (See 'Clinical evidence' above.)

• There are no high or moderate quality data to suggest that corticosteroid therapy is beneficial to patients with less severe septic shock. (See 'Clinical evidence' above.)

• Classification of adrenal reserve as adequate or inadequate fails to identify patients who are more likely to benefit from corticosteroid therapy. (See 'Clinical evidence' above.)

Recommendations

• For patients with severe septic shock (which we define as a systolic blood pressure <90 mmHg for more than one hour despite both adequate fluid resuscitation and vasopressor administration), we suggest initiation of intravenous corticosteroid therapy within eight hours after the onset of shock (Grade 2B). Response to ACTH testing is not used to select patients for corticosteroid therapy. (See 'Clinical evidence' above.)

• We suggest that corticosteroids be administered for five to seven days (Grade 2C). A tapering course is optional; however, close observation of those patients whose steroid therapy is stopped without being tapered is warranted. (See 'Administration' above.)

• We suggest that fludrocortisone NOT be added to the corticosteroid regimen (Grade 2C). (See 'Administration' above.)