Women in labour receiving epidural analgesia with 15 ml bupivacaine 0.1% and 2 microg.ml(-1) fentanyl followed by 10-15-ml top-ups as required, who needed Caesarean section, were randomly allocated to receive 20 ml levobupivacaine 0.5% over 3 min with either 75 microg fentanyl (1.5 ml) or 1.5 ml saline. Further top-ups or inhaled or intravenous supplementation were given for breakthrough pain. Time to onset (loss of cold sensation to T4 and touch sensation to T5 bilaterally), quality of analgesia and side-effects were recorded. The study was stopped after 112 patients had been randomly assigned, due to a unit protocol change, from midwife-administered top-ups to patient-controlled epidural analgesia. Data from 51 patients given fentanyl and 54 given saline were available for analysis. There were no significant differences in onset times or supplementation between the groups, but there was more intra-operative nausea/vomiting with fentanyl (53%) than with saline (18%; p = 0.004). We found no advantage of adding fentanyl to epidural levobupivacaine when extending epidural analgesia in women already receiving epidural fentanyl during labour and there was an increased incidence of intra-operative nausea and vomiting. Power analysis suggested the same conclusion even had the study proceeded to completion.