Forty-six patients with severe nonpenetrating brain injury [Glasgow Coma Scale (GCS) 4-7] were randomized to standard management at 37 degrees C (n = 22) and to standard management with systemic hypothermia to 32 to 33 degrees C (n = 24). The two groups were balanced in terms of age (Wilcoxon's rank sum test, p > 0.95), randomizing GCS (chi-square test, p = 0.54), and primary diagnosis. Cooling was begun within 6 h of injury by use of cooling blankets. Metocurine and morphine were given hourly during induction and maintenance of hypothermia. Rewarming was at a rate of 1 degree C per 4 h beginning 48 h after intravascular temperature had reached 33 degrees C. Muscle relaxants and sedation were continued until core temperature reached 35 degrees C. There were no cardiac or coagulopathy-related complications. Seizure incidence was lower in the hypothermia group (Fisher's exact text, p = 0.019). Sepsis was seen more commonly in the hypothermia group, but difference was not statistically significant (chi-square test). Mean Glasgow Outcome Scale (GOS) score at 3 months after injury showed an absolute increase of 16% (i.e., 36.4-52.2%) in the number of patients in the Good Recovery/Moderate Disability (GR/MD) category as compared with Severe Disability/Vegetative/Dead (SD/V/D) (chi-square test, p > 0.287). Based on evidence of improved neurologic outcome with minimal toxicity, we believe that phase III testing of moderate systemic hypothermia in patients with severe head injury is warranted.