Minimum effective fentanyl concentrations (MEC) were determined in 230 ASA I-III patients undergoing a variety of elective surgical procedures under balanced anesthesia, and in 40 patients recovering from comparable operations and anesthesia during postoperative intravenous self-administration of fentanyl (demand dose 34.5 micrograms) by means of the On-Demand Analgesia Computer. Following induction of anesthesia with fentanyl 4 micrograms/kg, repetitive fentanyl reinjections (0.1-0.2 mg) were given intraoperatively whenever systolic blood pressure or pulse rate increased to more than 20% of preinduction values, resulting in an intraoperative fentanyl consumption of 4.2 +/- 1.2 micrograms/kg/h. Duration of postoperative patient-controlled analgesia (PCA) was 20.2 +/- 4.3 h during which time 15.5 +/- 12.9 demands per patient were registered, resulting in a postoperative fentanyl consumption of 0.46 +/- 0.35 micrograms/kg/h. MECs appeared to be log-normally distributed (intra-operative median 2.6 ng/ml, range 0.2-36.6; postoperative median 1.2 ng/ml, range 0.2-8.0). Mean intra- and interindividual variability in MEC was 37% intraoperatively or 27% postoperatively, and 68% or 63%, respectively. There was no obvious correlation between postoperative fentanyl consumption or individual mean MEC and analgesic efficacy, which was generally described as superior to conventional pain treatment. Individual MECs increased gradually during anesthesia (mean slope 0.0191 ng/ml/min) but decreased under PCA conditions (-0.0008 ng/ml/min); difference not significant. While the postoperative decrease could be explained by diminishing pain intensity during the observation period, the slight intraoperative increase is discussed as acute tolerance rather than as accumulation. It is concluded that repetitive fentanyl injections as indicated by clinical needs will not lead to relevant accumulation in serum, and that analgesic therapy should be individualized both intra- and postoperatively.