Intra-arterial carotid artery chemotherapy for malignant gliomas is limited by focal injuries to the eye and brain which may be caused by poor mixing of the drug with blood at the infusion site. This inadequate mixing can be eliminated in animal models with diastole-phased pulsatile infusion (DPPI) which creates 1-ml/sec spurts during the slow blood flow phase of diastole. Before treatment with intracarotid cisplatin, 10 patients with malignant gliomas were studied to determine whether intravascular streaming occurs after intracarotid infusion in humans, and if so, if it is reduced with DPPI. Regional cerebral blood flow (rCBF) studies were performed by intravenous injection of H2(15)O and positron emission tomography. This was followed by supra- or infraophthalmic internal carotid artery (ICA) injections of H2(15)O with either continuous infusion or DPPI. Local H2(15)O concentration in the brain was determined and the images of radiotracer distribution in the continuous infusion and DPPI studies were compared to the rCBF images. Intravascular streaming of the infusate was identified by a heterogeneous distribution of the infused H2(15)O in brain compared to rCBF. Extensive and variable intravascular streaming occurred in three patients who received infusions into the supraophthalmic segment of the ICA. Some brain areas received up to 11 times the expected radiotracer delivery, while other regions received as little as one-tenth. This streaming pattern was markedly reduced or eliminated by DPPI. In the five patients who received infraophthalmic infusions, a minimally heterogeneous distribution of the infusate was detected. The authors conclude that extensive intravascular streaming accompanies supraophthalmic ICA infusions in patients. The magnitude of streaming can be substantially reduced or eliminated with DPPI. Those who perform intra-arterial infusion should consider using DPPI to assure uniform drug delivery to brain.