The prognostic value of basal and corticotropin-stimulated cortisol concentration in patients with sepsis remains a controversial issue. In a retrospective cohort study, 82 consecutive patients with septic shock underwent a short corticotropin test performed more than 24 h after the onset of vasopressor therapy. Forty-one (50%) patients died within 28 days after the onset of septic shock. The mean (SD) basal cortisol level was 22.7 (10.6) microg/dL. With threshold values of 7 and 9 microg/dL maximal increases in cortisol level, 28 (34%) and 31 (38%) patients were, respectively, classified as nonresponders to the short corticotropin test. On multivariate analysis, a cortisol level >20 microg/dL (P = 0.0002), a maximal response to corticotropin <9 microg/dL (P = 0.044), abnormal lactate values (P = 0.0098), and positive blood cultures (P = 0.004) were independent predictors of 28-day mortality. In conclusion, high basal cortisol and low increase on corticotropin stimulation are predictors of a poor outcome in late septic shock. The underlying mechanisms of these prognostic patterns remain to be elucidated.