BACKGROUND: Patent ductus arteriosus (PDA) with significant left to right shunt in preterm infants increases morbidity and mortality. Early closure of the ductus arteriosus may be achieved pharmacologically using cyclooxygenase inhibitors, or by surgery. The efficacy of both treatment modalities is well established. However, the preferred initial treatment of a symptomatic PDA in a preterm infant, surgical ligation or trial of indomethacin, has not been well established. OBJECTIVES: To compare the effect of surgical ligation of PDA versus medical treatment with cyclooxygenase inhibitors (using indomethacin, ibuprofen, or mefenamic acid), each used as the initial treatment, on neonatal mortality in preterm infants with a symptomatic PDA. SEARCH STRATEGY: The standard search strategy of the Cochrane Neonatal Review Group was used. This included search of electronic databases: Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, Issue 4, 2002), MEDLINE (1966 - December 2002), CINAHL (1982 - December 2002), EMBASE (1980 - December 2002); and hand search of abstracts of Pediatric Academic Societies annual meetings published in Pediatric Research (1990 - April 2002). No language restrictions were applied. SELECTION CRITERIA: All trials 1) using randomized or quasi-randomized patient allocation, 2) in preterm infants < 37 weeks gestational age or low-birth-weight infants (< 2500 grams) with symptomatic PDA in the neonatal period (< 28 days) and 3) comparing surgical ligation with medical treatment with cyclooxygenase inhibitors, each used as the initial treatment for closure of PDA. DATA COLLECTION AND ANALYSIS: Assessment of methodological quality and extraction of data for included trials was undertaken independently by the authors. RevMan 4.1 was used for analysis of the data. MAIN RESULTS: Only one study, trial B in the report of Gersony 1983, was found eligible. The trial compared the effect of surgical ligation of PDA versus medical treatment with indomethacin, each used as the primary treatment. No trials comparing surgery to other cyclooxygenase inhibitors (ibuprofen, mefenamic acid) were found. Trial B of Gersony 1983 enrolled 154 infants. The study found no statistically significant difference between surgical closure and indomethacin treatment in mortality during hospital stay, chronic lung disease, other bleeding, necrotizing enterocolitis, sepsis, creatinine level, or intraventricular hemorrhage. There was a statistically significant increase in the surgical group in incidence of pneumothorax [RR 2.68 (95% CI 1.45, 4.93); RD 0.25 (95% CI 0.11, 0.38); NNH 4 (95% CI 3, 9)] and retinopathy of prematurity grade III and IV [RR 3.80 (95% CI 1.12, 12.93); RD 0.11 (95% CI 0.02, 0.20), NNH 9 (95% CI 5, 50] compared to the indomethacin group. There was as expected a statistically significant decrease in failure of ductal closure rate in the surgical group as compared to the indomethacin group: [RR 0.04 (95% CI 0.01, 0.27); RD -0.32 (95% CI -0.43, -0.21), NNT 3 (95% CI 2, 4)]. REVIEWER'S CONCLUSIONS: The data regarding net benefit/harm are insufficient to make a conclusion as to whether surgical ligation or medical treatment with indomethacin is preferred as initial treatment for symptomatic PDA in preterm infants.