Stress results in activation of the hypothalamicpituitaryadrenal axis with increased circulating cortisol. It has been argued that a syndrome of relative adrenal insufficiency is common in critically ill patients. Patients who fail to increase their cortisol by >250 nmol/30 min following the administration of 250 μg ACTH in the short synacthen test (SST) have been reported to have a higher mortality (JAMA 2000, 283 10381045). Here we have retrospectively analysed the results of SSTs carried out on the day of admission in 273 consecutive patients with severe sepsis admitted to a single centre critical care unit over a 30-month period. The Intensive Care National Audit Research database was interrogated for the patients and the following variables recorded: age and gender, date of admission, severity of illness (APACHE II score), length of ICU/hospital admission, survival at discharge from hospital. Compared with survivors, non-survivors had higher APACHE II score on admission (median±S.E.M. 20.3 vs 17.4, P=0.01) and higher baseline cortisol levels (494±34 vs 356±19 nmol/l, P=0.0001) while ACTH-stimulated cortisol levels did not differ significantly (626±46 vs 539±201 nmol/l, P=0.054). The increase in cortisol (DeltaF) was significantly higher in survivors than in non-survivors (175 vs 153 nmol/l, P=0.0001). About 87/171 survivors and 43/102 non-survivors failed the cut-off of 550 nmol/l for stimulated cortisol (P=NS). Baseline or stimulated cortisol levels did not differ between those with severe (APACHEII>25) and less severe disease. However, baseline cortisol levels did correlate with the APACHEII score (r=0.165, P=0.007) and patients with a baseline cortisol ≥415 nmol/l, suggested as a delineator of compromised adrenal function in critical illness (NEJM 2003, 348 727734), had higher mean APACHEII score (19.6 vs 17.5, P=0.014). Also, the survival rate in patients with baseline cortisol ≥415 nmol/l was significantly lower (36 vs 64%, P=0.0001). Patients with a DeltaF above or below 250 nmol/l did not differ with regard to their APACHEII score or their survival rate (P=0.086). Logistic regression model analysis showed that baseline cortisol and APACHEII score are important predictors of survival, but not age, sex or DeltaF. The odds ratio (OR) of dying in the ICU for patients with baseline cortisol >415 nmol/l was 2.3 (95%CI: 1.353.9, P=0.002) as compared to patients with cortisol <415 nmol/l. Thus our findings do not support previous reports suggesting the use of the SST to determine which sepsis patients might benefit from hydrocortisone treatment.