As a result of randomized assignment, 24 premature infants with symptomatic patent ductus arteriosus (PDA) were either treated by 0,2 mg I.V. indomethacin just after diagnosis (12) or left untreated for 48 hours (12). Clinical, radiological, and biological data before trial were remarkably similar. In the 12 "treated" infants, the DA closed within 4 to 26 hours of the first dose in 8 cases. In all 12 "non-treated" infants, PDA persisted for at least 48 h, then 10 infants were treated ad libitum by I.V. (5) or oral (5) indomethacin. Three deaths (late) did not appear to be associated with either PDA or treatment. Two surgical closures were required (1 in each group). Four spontaneous closures occurred after more or less delay: 2 in the 2 never treated infants, and 2 after a double failure of I.V. or oral indomethacin. Treatment (even when unsuccessful) was usually followed by some water retention and blood dilution, sometimes of striking magnitude but always transient. To date, I.V. indomethacin can be given earlier than orally and offers slightly better chances for closure of PDA, thereby reducing the needs for surgical closure which should be reserved for badly tolerated failures of pharmacologic closure.