The differential effects of intravenous versus epidural administration of short-acting, lipid-soluble opioids is controversial. This study was undertaken to compare these two routes of administration using the mixed agonist-antagonist opioid, butorphanol. Forty-five women undergoing elective cesarean delivery at term under epidural lidocaine anesthesia were randomized to receive a single bolus of either epidural or intravenous butorphanol 2 mg or saline control for postoperative analgesia. At precisely 60 min after the last dose of epidural local anesthetic, all patients received a simultaneous epidural and intravenous injection in a randomized, double-blinded fashion. The intravenous group received butorphanol intravenous and saline epidurally; the epidural group received saline intravenous and butorphanol epidurally; and a control group received saline via both routes. When additional analgesia was requested, all patients received patient-controlled analgesia (PCA) with intravenous morphine (2-mg demand dose, 7-min lockout interval). Analgesia was quantitated using a visual analogue scale and subsequent PCA morphine requirements. The interval from study drug injection until first request for PCA use was equivalent for the intravenous and epidural groups (89 +/- 9 and 83 +/- 8 min, respectively) and significantly longer than in control group (39 +/- 4 min, P less than 0.001, intravenous and epidural vs. control). Analgesia was equivalent in the intravenous and epidural groups at all observation points, and pain scores were significantly lower than control for the first 120 min after study drug injection.(ABSTRACT TRUNCATED AT 250 WORDS)