OBJECTIVE: To characterise the plasma cortisol profile and adreno-cortial responsiveness (short Synacthen test) of patients in septic shock. DESIGN: Retrospective assessment using case-notes and ICU charts. SETTING: University teaching hospital ICU. PATIENTS: 68 septic shock patients with plasma cortisol and/or short Synacthen test measured at ICU-admission or onset of shock post ICU-admission. Patients were identified from a total population of 155 patients who had PCL and/or SST measured over a 4.5 year period. INTERVENTION: None. MEASUREMENTS AND RESULTS: Patients with septic shock had a plasma cortisol ranging from 210-8900 nmol/l and mortality of 56%. There were 22 (32%) below (low) and 46 (68%) above (high) a 'critical' plasma cortisol of 500 nmol/l. Using stepwise logistic regression, mortality was adequately predicted by and increased with, increasing plasma cortisol and onset of shock remote from ICU-admission. Short Synacthen tests were available in 33 patients: 11 responders (cortisol increment > 200 nmol/l above baseline 30 min after 0.25 mg intravenous Synacthen) and 22 hypo-responders. Mortality in patients was adequately predicted by and increased with a decrease in cortisol increment post-Synacthen. Thirteen patients (plasma cortisol 606 +/- [SD] 297 nmol/l) had complete haemodynamic profiles before inotropic therapy; no relationship was demonstrated between plasma cortisol and circulatory variables. Follow-up revealed no cases of Addison's disease. CONCLUSIONS: In septic shock, 'hypocortisolaemia' is not uncommon and does not predict a high mortality; adrenocortical hypo-responsiveness may be associated with poor outcome.