This study assessed the safety and efficacy of intrathecal baclofen in the treatment of intractable spasticity caused by spinal cord injury or multiple sclerosis. Twenty-three patients with severe chronic spasticity underwent bolus test dosing with 50, 75, or 100 micrograms of intrathecal baclofen administered by lumbar puncture. All patients were either refractory to oral baclofen at a dose of 120 mg/d or side effects were unacceptable at a lower dose. There was a significant decrease in tone and spasticity in all 23 patients. Nineteen patients underwent implantation of a programmable pump and intrathecal catheter designed to deliver baclofen directly to the spinal cord. Rigidity (tone) was decreased from a mean prebolus Ashworth score of 3.8 to a mean postbolus Ashworth score of 1.5 and spasms from a mean prebolus score of 3.5 to a mean postbolus score of 1.2 for a minimum of 4 hours. Patients have been observed for a mean of 16 months (range 2 to 34 months). Ashworth scores have remained reduced to an acceptable level (< or = 2 with periodic adjustment in dosage in all but three patients. There has been one pump malfunction and four catheter malfunctions; few serious medication and postoperative complications have occurred. There was one death caused by underlying disease, one patient voluntarily withdrew, and three patients developed tolerance to the extent that optimal control of spasticity tone could not be maintained. Although intrathecal baclofen is safe and effective in the majority of patients, three patients required > 1,000 micrograms/d with increasingly higher doses over time and exhibited a poor response.