We prospectively investigated noninvasive selective brain cooling (SBC) in patients with severe traumatic brain injury.Sixty-six in-patients were randomized into three groups. In one group, brain temperature was maintained at 33 -35°C by cooling the head and neck (SBC); in a second group, mild systemic hypothermia (MSH; rectal temperature 33 -35°C) was produced with a cooling blanket; and a control group was not exposed to hypothermia. Natural rewarming began after 3 days. Mean intracranial pressure 24, 48 or 72 h after injury was significantly lower in the SBC group than in the control group. Mean serum superoxide dismutase levels on Days 3 and 7 after injury in the SBC and MSH groups were significantly higher than in the control group. The percentage of patients with a good neurological outcome 2 years after injury was 72.7%, 57.1% and 34.8% in the SBC, MSH and control groups, respectively. Complications were managed without severe sequelae. Non-invasive SBC was safe and effective.