Seventeen patients with advanced cancer pain, treated with chronic epidural morphine, were studied. Minimum plasma and CSF morphine concentrations (CSSmin) were determined at pharmacokinetic steady state. A linear relationship was found between epidural morphine dose and concentrations obtained in plasma (r = 0.92) as well as CSF (r = 0.90). The line for best fit was much steeper for CSF than for plasma. The CSF/plasma concentration gradient of morphine at CSSmin was 132 +/- 31 (mean +/- S.E.M.). Large interindividual variations of morphine concentrations in CSF were found. It is suggested that the variations are due to substantial differences in transdural morphine diffusion between individuals. No correlation was found between pain relief, evaluated with a visual analog scale, and CSF morphine concentrations at pharmacokinetic steady state, when calculated in 9 patients. Mean duration of treatment was 104 days (range 14-366) and the daily dose was increased from 18 +/- 2 to 87 +/- 31 mg/day (mean +/- S.E.M.). A total of 39 epidural catheters were inserted in 14 patients. The catheters were patent for 2-223 days with a mean of 38 days. When re-examined later during treatment, 2 out of 8 patients demonstrated decreased CSF morphine concentrations in spite of increased doses given. One patient with extremely high dose demand is reported on separately and data supporting the concept of a combined spinal and systemic brain morphine effect in such cases are presented. Side effects were not a major problem but the possibility of infectious complications should be considered during chronic epidural morphine therapy.