This study examines the behavioral effects and potential neurotoxicity of sufentanil, alfentanil, and morphine after chronic daily epidural (15-day) and intrathecal (28-day) administration in dogs. Dogs were chronically implanted with a lumbar intrathecal or epidural catheter and received daily injections for 28 or 15 days, respectively, of saline or one of three mu agonists: sufentanil (intrathecal 5, 25, or 50 micrograms/0.5 ml; epidural 10, 50, or 100 micrograms/2.0 ml), alfentanil (intrathecal 40 or 400 micrograms/0.5 ml; epidural 80 or 800 micrograms/2.0 ml), or morphine (intrathecal 0.5 or 5 mg/0.5 ml; epidural 1 or 10 mg/2.0 ml). Dogs were examined for antinociception (skin twitch) and neurobehavioral changes. When the animals were killed, cisternal cerebrospinal fluid was taken for clinical chemistry, and after perfusion fixation, spinal cord tissue was taken for histologic analysis. Bolus intrathecal and epidural injections of sufentanil, alfentanil, and morphine produced dose dependent antinociception, bradycardia, an initial tachypnea followed by a decrease in respiratory rate, hypothermia and somnolence. The order of potency was sufentanil > alfentanil > morphine on all measures. Over the extended period of drug delivery, a loss of response (tolerance) was observed on all measures. No abnormal morphologic or histologic effects were found when comparing the drug and dose groups. An inflammatory reaction secondary to the catheter was found in all animals. Intrathecal, but not epidural, catheters resulted in significant increases in cerebrospinal fluid protein and cell counts in vehicle animals. Values in drug treated animals did not differ significantly from the respective vehicle controls. A rapid systemic redistribution of all three drugs was observed. No differences were found in the pharmacokinetic parameters measured at day 1 and at the day of killing for any route. This large-animal model demonstrates the expected pharmacologic potency of these three agents and tolerance development. Based on cerebrospinal fluid and systematic histopathologic analyses, these three spinally administered agents showed no evidence of neurotoxicity over the range of doses/concentrations employed when given by the intrathecal or epidural route as compared to vehicle controls. Consideration of the toxicokinetics in this canine model suggests that it provides an appropriate test of the safety of these agents in concentrations which exceed those employed for daily intermittent epidural and intrathecal drug delivery in humans.