The present report details the characteristics of the analgesic effects of morphine administered chronically by infusion pumps implanted in 53 patients suffering from terminal metastatic disease. The median postimplant survival time in these patients was 4 months. Patients (mean age 58 years) were characterized according to the duration of pain before pump implantation (mean 16 months), prior consumption of systemic opioids (mean one to six daily analgesic equivalents of morphine), and their response to a trial intrathecal dose of morphine (1 to 2 mg). The median infusion dose at 2 weeks was 3.8 mg/day. The analgesic index, calculated as (quality of pain relief x duration of pain relief in hours)/morphine dose in mg, that was observed after the trial dose of morphine was determined for each patient. A close correlation was observed between the acute (2-week) infusion dose necessary to produce pain relief and the analgesic index such that the infusion dose = -8.0 x log (analgesic index) + 17.1. By 16 weeks, the mean spinal morphine dose for the group had increased by a factor of about 2.5; however, significant variation in the dose incrementation was documented. The maximum increase was observed in patients with a low analgesic index, and this rapid incrementation was usually correlated with an unsatisfactory overall outcome. Evidence that long-term infusion continues to yield analgesia was evidenced in six cases where there was an unanticipated loss of drug infusion and a corresponding increase in parenteral narcotic consumption. These data indicate the long-term efficacy and safety of spinal opioid infusion in patients with terminal cancer, and emphasize the advantage of assessing the sensitivity of the patient to spinal opioids by a standardized trial injection prior to pump placement as a prognostic indication of outcome.