We undertook a retrospective chart review of 37 neonates who received fentanyl by continuous infusion while undergoing extracorporeal membrane oxygenation (ECMO) between May 1986 and October 1988. We quantified the doses of all sedatives utilized, determined the Incidence of neonatal abstinence syndrome (NAS), and identified risk factors associated with NAS. We determined peak fentanyl Infusion rate, mean fentanyl infusion rate, total fentanyl dose, and duration of ECMO therapy. NAS was observed in 21 of 37 neonates (57%). In both the NAS and non-NAS neonates, mean infusion rate increased steadily during ECMO therapy, from a mean of 11.6 ± 6.9 (SD) μ.kg−1.h−1on day 1 to a mean of 52.5 ± 19.4 (SD) μ.kg−1.h−1by day 8. Total fentanyl dose and duration of ECMO were significantly greater in neonates with NAS. We found that neonates with a total dose > 1.6 mg/kg or an ECMO duration > 5 days had a significantly greater incidence of NAS (chi-squared test,P< 0.01 andP< 0.005; odds ratios = 7.0 and 13.9, respectively). With multiple logistic regression, ECMO duration was found to be the most powerful predictor of the occurrence of NAS. We also measured plasma fentanyl concentrations in a separate group of 5 neonates receiving fentanyl by continuous infusion for sedation. Fentanyl concentrations increased steadily during the period of infusion, suggesting the development of tolerance to the sedating effects. We conclude that continuous administration of fentanyl for sedation is associated with the uniform development of tolerance and a significant Incidence of dependence. Alternative approaches to sedation should be investigated.