The aim of this study was to compare the incidence of technical complications of epidural catheters with subcutaneous injection ports to percutaneous epidural catheters without ports, fixed only by adhesive dressing. We reviewed 149 patients who received 250 epidural catheters for treatment of cancer pain during a 31/2-yr period from January 1, 1989, to June 30, 1992. Of the 250 catheters, 52 were provided with subcutaneous injection ports and 198 were percutaneous catheters. Of the 198 percutaneous catheters, 41 were tunneled for a short distance; the remainder entered the skin at the dorsal midline. In the percutaneous group 21% of the catheters became dislodged. In the injection port group, there were no catheter dislodgements. The overall incidence of infections was similar in both groups (13.6%). When we indexed the infection rate to catheter-days, the number of infections per 1000 catheter-days in the injection port group was half that of the percutaneous group (2.86 infections versus 5.97 for percutaneous catheters). No injection port became infected during the first 70 days of treatment, whereas in the percutaneous group infections occurred as early as the first week. Within the percutaneous group the complication rate in the tunneled epidural catheters was as high as in the nontunneled. We conclude that injection ports reduce the complication rate of epidural catheters, particularly catheter dislodgement and early infections.