Chronic intrathecal drug infusion for the treatment of neurological diseases, such as spasticity and chronic pain, has become an accepted method of therapy in recent years. Concurrent pharmacokinetic studies have shown that the cisternal cerebrospinal fluid (CSF) drug level is considerably lower than the lumbar CSF level during continuous infusion into the lumbar subarachnoid space. One factor that makes analysis of this decline in drug level difficult to quantify is that it is only feasible to sample CSF at the two extremes of the spinal subarachnoid space. Using a radionuclide technique, we have examined the distribution along the spinal canal of a hydrophilic compound, indium-111 diethylenetriamine pentaacetic acid, that was delivered over 72 hours into the lumbar subarachnoid space in five patients with implanted drug pumps. Over a 20-cm distance of the thoracic cord, radionuclide counts decreased gradually so that the indium-111 diethylenetriamine pentaacetic acid concentration surrounding the cord at the T2 vertebral level was 43% of that at the T12 level in four patients. Therefore, it appears that even with a hydrophilic compound, which minimizes spinal cord capillary losses, there is still a considerable reduction of CSF drug concentration along the spinal canal. The clinical implication of this gradual decline in drug level is that for intrathecal infusion of relatively hydrophilic compounds there may not be any advantage in placing the catheter tip at more rostral locations, such as at the midthoracic or cervical cord.