Although experimental and pathological studies suggest an important role for ischemia in the majority of fatal cases of traumatic brain injury, ischemia has been a rare finding in most clinical studies of cerebral blood flow (CBF) in head-injured patients. The hypothesis of the present study was that cerebral ischemia occurs in the first few hours after injury, but that CBF measurements have not been performed early enough. Early measurements of CBF (by the 133Xe intravenous method) and arteriovenous oxygen difference (AVDO2) were obtained in 186 adult head-injured patients with a Glasgow Coma Scale score of 8 or less, and were correlated with neurological status and outcome. During the first 6 hours after injury, CBF was low (22.5 +/- 5.2 ml/100 gm/min) but increased significantly during the first 24 hours. The AVDO2 followed the opposite course; the decline of AVDO2 was most profound in patients with low motor scores, suggesting relative hyperemia after 24 hours. A significant correlation between motor score and CBF was found in the first 8 hours after injury (Spearman coefficient = 0.69, p less than 0.001), but as early as 12 hours postinjury this correlation was lost. A similar pattern was found for the relationship between CBF and outcome. Cerebral blood flow below the threshold for infarction (CBF less than or equal to 18 ml/100 gm/min) was found in one-third of the studies obtained within 6 hours, the incidence rapidly decreasing thereafter. A low CBF after 24 hours was not generally associated with a high AVDO2, and was probably a reflection of low oxidative metabolism rather than frank ischemia. In 24 patients, a CBF of 18 ml/100 gm/min or less was found at some point after injury; the mortality rate was significantly higher in this subgroup, and survivors did worse. In some cases, ischemia was successfully treated by reducing hyperventilation or inducing arterial hypertension. These results support the above hypothesis, and suggest that early ischemia after traumatic brain injury may be an important factor determining neurological outcome. Moreover, these data indicate that early hyperventilation or lowering of blood pressure to prevent brain edema may be harmful.