To identify factors affecting the efficacy of indomethacin in closing symptomatic patent ductus arteriosus (PDA), we studied the pharmacokinetics of intravenous indomethacin, 0.2 mg per kilogram of body weight, in 35 premature infants with symptomatic PDA. Most infants responded to indomethacin with ductus constriction. Indomethacin infusions that were ineffective (seven doses in six patients) were associated with significantly faster clearance, a shorter half-life, and lower plasma levels (p less than 0.05). Six infants had later reopening of the ductus. All six received indomethacin in the first postnatal week; they could not be distinguished from infants with permanent closure on the basis of indomethacin kinetics, but they were of low gestational age. There was a 20-fold variation in plasma indomethacin levels 24 hours after a dose. In view of this variation and the relation between plasma levels and ductus constriction, we suggest that measurement of the plasma indomethacin level could be of value in infants with no response to a first dose.