Twenty patients undergoing thoracotomy were given 2, 4, or 6 mg morphine epidurally in a double-blind, randomized study for postoperative analgesia. Administration at T12-L1 or L1-L2 resulted in a dose-related analgetic duration (514 +/- 118 min, 778 +/- 207 min, and 938 +/- 155 min; means +/- SEM, respectively, for the groups). For the three groups, peak plasma morphine concentrations (range 19-34 ng/ml) were reached within 15 min. The plasma curves had a similar appearance as after an intramuscular injection and pharmacokinetic calculations showed an elimination half-life (mean +/- SEM) of 173 +/- 24 min, 200 +/- 60 min, and 213 +/- 57 min for the groups, respectively. The morphine concentrations in the CSF were considerably higher compared with plasma (45-100 times the plasma concentration at 1 h, 100-250 times at 3 h, and 125-175 times at 5 h) but the elimination half-life of morphine in the cerebrospinal fluid (CSF) was similar to that in plasma. The lumbar approach was used with similar efficacy as reported for thoracic administration. Side effects were few and nonsignificant. The authors conclude that epidural morphine administration results in a dose-dependent analgesia, as well as concentrations in the CSF that are considerably higher than in plasma. With similar elimination half-lives for morphine in CSF and plasma, the long analgetic duration probably depends on the locally high morphine concentrations achieved. For safety purposes, one may use the lumbar approach to the epidural space even for thoracic pain without reducing the efficacy.