The dose, efficacy, and side effects of continuous intravenous infusion (CIVI) of morphine were compared with continuous subcutaneous infusion (CSCI) of morphine in patients with chronic cancer pain. Eligible patients were referred to the Palliative Care Program and were receiving a stable dose of CIVI of morphine. The design was a within-patient, one-way crossover; in which each patient provided data before and after a switch from CIVI to CSCI of morphine. "Rescue" doses were 50% of the hourly dose given every 2 hours as needed. Morphine was infused intravenously (i.v.) and subcutaneously (s.c.) via a McGaw/AccuPro Volumetric Infusion Pump. After baseline data, including side effects and pain assessment, were obtained, patients were evaluated twice daily for toxicity and analgesic efficacy. Those who had a stable CIVI dose for 48 consecutive hr were crossed over to the CSCI at the same dose as the intravenous (i.v.) phase. A stable dose was defined as no dose change, four or less rescue doses in the previous 24 hr, and a pain rating of none or mild. CIVI was considered equal to CSCI if these criteria were maintained for 96 consecutive hr. Fifty-seven patients were entered, and 40 were evaluable (15 women and 25 men). The median age was 67 (range 30-83 years). All 40 participants, after maintaining a stable dose throughout the i.v. phase, crossed to the s.c. phase and remained on s.c. for at least 48 hr. Thirty-two patients maintained a stable dose throughout the i.v. and s.c. phases. The mean stable i.v. dose (day 2) was 5.05 mg/hr, and the mean stable s.c. dose (day 4) was 5.7 mg/hr (P = 0.01). The mean number of rescue doses on day 2 was 0.83 per 24 hr versus 0.80 per 24 hours on day 4 (P = 0.6). The mean categorical pain score on day 2 was 0.83, and on day 4, 0.85 (P = 0.7). The mean visual analogue scale (VAS) on day 2 was 22.9 mm versus 17.6 mm on day 4 (P = 0.1). The mean incidence of side effects on day 2 was 1.7, and on day 4, 2.0 (P = 0.2). No patient was withdrawn or had a dose reduction due to unacceptable toxicity. There were two reports of local toxicity (mild erythema) at the SC needle insertion point, which required a site change. All of our 40 patients had adequate pain control with CIVI and CSCI morphine. Of the eight participants who were not maintained on the same i.v. and s.c. dose, all had adequate pain control and a similar side-effect profile on a higher s.c. morphine dose. These data suggest that the i.v. and s.c. routes are equianalgesic for most patients when administered as a continuous infusion. Pain control and side-effect profiles are quite similar and acceptable. s.c. morphine is an excellent alternative to i.v. morphine in both inpatients and outpatients requiring parenteral morphine for pain.