An anti-inflammatory cytokine profile on whole-blood stimulation in vitro is associated with fatal outcome of meningococcal disease. We investigated whether an anti-inflammatory cytokine profile in the circulation is associated with adverse outcome in other infectious diseases. We enrolled 464 consecutive patients (272 men, 192 women) who presented to hospital with fever (> or = 38.2 degrees C). On admission we measured plasma interleukin 10 (IL-10) and tumour necrosis factor alpha (TNF alpha), and collected clinical and microbiological data on the febrile illness, then followed up all patients for clinical outcome. In at least 399 of the 464 patients fever was caused by infection. 33 patients died after a median hospital stay of 11 days (interquartile range 3-20). Concentrations of IL-10 were significantly higher in non-survivors (median 169 pg/mL [IQR 83-530]) than in survivors (median 88 pg/mL [42-235], p=0.042). When dichotomised around the median, the mortality risk was two times higher in patients who had high concentrations of IL-10 than in those with low concentrations (relative risk 2.39 [95% CI 1.07-5.33]), in patients with low and high concentrations of TNF alpha. In the 406 patients without haemodynamic deterioration in the first 24 h, IL-10 was higher and TNF alpha lower in patients who died than in those who survived. The ratio of IL-10 to TNF alpha was higher in non-survivors (median 6.9 [3.0-21.0]) than in survivors (median 3.9 [2.0-7.0], p=0.040). This ratio was highest in patients who died without underlying disease (median 21.5 [5.0-25.0]). Age, sex, and duration of fever before admission did not explain the differences in IL-10 and TNF alpha. An anti-inflammatory cytokine profile of a high ratio of IL-10 to TNF alpha is associated with fatal outcome in febrile patients with community-acquired infection. Our findings caution against a widespread use of proinflammatory cytokine inhibition in patients with sepsis.