Corticosteroids in septic shock.

As the balance of evidence regarding corticosteroid treatment for septic shock shifts once again toward the negative, the study by Sprung et al.1 in this issue of the Journal elicits a strong feeling of déjà vu. Will the historical fate of high-dose corticosteroids, which were largely abandoned when the benefit observed in early studies could not be replicated in larger trials,2,3 now befall “physiologic-dose” corticosteroids? The rationale for therapy with corticosteroids at a physiologic dose (i.e., 200 to 300 mg of hydrocortisone per day) originated in the observations that patients with septic shock who had a reduced response to corticotropin (increase in total plasma cortisol, <9 μg per deciliter [248 nmol per liter]) were more likely to die4 and that the pressor response to norepinephrine may be improved by the administration of hydrocortisone.5 Although the validity of these observations appears to be increasingly doubtful as evidence accumulates that the standard corticotropin stimulation test is unreliable in critically ill patients,6,7 the findings have led to interest in treating such patients with corticosteroids. Encouraging results in small trials8,9 and then in a larger trial10 led to current recommendations to treat patients with septic shock with physiologic doses of hydrocortisone.11,12 The recommendations are based on five trials involving a total of 464 patients, of whom 265 (57.1%) died.13 Even though various treatment regimens were used, all five trials reported fewer deaths in patients who received corticosteroids. A meta-analysis of these trials suggested that the use of corticosteroids reduced mortality.13 In the face of such evidence, why did Sprung et al. conduct the Corticosteroid Therapy of Septic Shock (CORTICUS) study? As noted by the authors, the current recommendations are heavily dependent on one trial conducted by Annane et al.10 In that trial, patients were divided into “responders” and “nonresponders” on the basis of a corticotropin stimulation test; 229 of 299 patients (76.6%) did not have a response to corticotropin, a percentage that was much larger than the 40% the investigators expected. After statistical adjustment for baseline covariates, a significant reduction in the likelihood of death was observed in patients with no response to corticotropin who received corticosteroids. In contrast, crude estimates of in-hospital mortality were higher in patients who had a response to corticotropin. Two additional features of this trial bear mentioning. First, patients who were assigned to receive hydrocortisone also received fludrocortisone, although the importance of this factor is unknown. Second, 24% of the patients received etomidate, a short-acting intravenous anesthetic agent that selectively inhibits adrenal corticosteroid synthesis. Its use may have contributed to the unexpectedly high number of patients who did not have a response to corticotropin, and whether the trial results apply in health care systems in which etomidate is rarely used is unclear. Thus, the borderline result that was achieved only after statistical adjustment (combined with the unexpectedly high number of patients who did not have a response to corticotropin and the higher estimated mortality in those who did have a response to corticotropin) provide ample justification for the CORTICUS study. Patients who were enrolled in the CORTICUS study had septic shock and remained hypotensive or required treatment with vasopressors for at least 1 hour after adequate f luid resuscitation.