Among 3559 newborn infants with birth weight less than 1750 gm, 421 developing a hemodynamically significant patent ductus arteriosus were entered into a randomized trial to evaluate the role of indomethacin in the management of PDA. Indomethacin given concurrently with usual medical therapy at the time of diagnosis resulted in ductal closure in 79%, versus 35% with placebo (P less than 0.001). Indomethacin as backup to usual medical treatment resulted in similar closure rates. To assess overall effects through hospital discharge, three management strategies were compared. Although mortality did not differ significantly, infants given indomethacin only if usual therapy failed (strategy 2) had a lower incidence of bleeding than those to whom indomethacin was given with initial medical therapy (strategy 1) and lower rates of pneumothorax and retrolental fibroplasia than those to whom no indomethacin was administered, with surgery the only backup to medical therapy (strategy 3). Thus the administration of indomethacin only when medical treatment fails appears to be the preferable approach for the management of symptomatic PDA in premature infants.