Objective: To test the hypothesis that treatment with hypothermia affects the course of overwhelming acute respiratory failure associated with sepsis. Design: Concurrent-controlled, prospective study. Setting: Adult multidisciplinary ICU in a teaching hospital. Patients: Nineteen consecutive patients with septic ARDS mechanically ventilated and a P(A-a)O2 > 500 Torr during 36 h on greater-than-or-equal-to 10 cm H2O of PEEP. Interventions: Patients were assigned to receive conventional treatment (n = 10) or conventional treatment plus mild hypothermia (32-35-degrees-C) instituted as a last resort (n = 9). Results: Hypothermia (33.7 +/- 0.6-degrees-C) was associated with a reduction in mortality rate (67% vs. 100%, P < 0.05), P(A-a)O2 (P < 0.001), heart rate (P < 0.001), cardiac index (P < 0.01), and Q(S)/Q(T) (P < 0.01). There were no significant differences in oxygen consumption (Vo2) before (243 +/- 74 ml/min) and during treatment with hypothermia (246 +/- 87 ml/min) although 02 extraction increased during hypothermia (26 +/- 6 vs. 30 +/- 6%, P < 0.05). Conclusions: This study suggests that hypothermia was effective in improving oxygenation and survival in patients with severe ARDS associated with sepsis, even though VO2 was unchanged.