Background: Bacterial vaginosis has been associated with preterm birth. In clinical trials, the treatment of bacterial vaginosis in pregnant women who previously had a preterm delivery reduced the risk of recurrence. Methods: To determine whether treating women in a general obstetrical population who have asymptomatic bacterial vaginosis (as diagnosed on the basis of vaginal Gram's staining and pH) prevents preterm delivery, we randomly assigned 1953 women who were 16 to less than 24 weeks pregnant to receive two 2-g doses of metronidazole or placebo. The diagnostic studies were repeated and a second treatment was administered to all the women at 24 to less than 30 weeks' gestation. The primary outcome was the rate of delivery before 37 weeks' gestation. Results: Bacterial vaginosis resolved in 657 of 845 women who had follow-up Gram's staining in the metronidazole group (77.8 percent) and 321 of 859 women in the placebo group (37.4 percent). Data on the time and characteristics of delivery were available for 953 women in the metronidazole group and 966 in the placebo group. Preterm delivery occurred in 116 women in the metronidazole group (12.2 percent) and 121 women in the placebo group (12.5 percent) (relative risk, 1.0; 95 percent confidence interval, 0.8 to 1.2). Treatment did not prevent preterm deliveries that resulted from spontaneous labor (5.1 percent in the metronidazole group vs. 5.7 percent in the placebo group) or spontaneous rupture of the membranes (4.2 percent vs. 3.7 percent), nor did it prevent delivery before 32 weeks (2.3 percent vs. 2.7 percent). Treatment with metronidazole did not reduce the occurrence of preterm labor, intraamniotic or postpartum infections, neonatal sepsis, or admission of the infant to the neonatal intensive care unit. Conclusions: The treatment of asymptomatic bacterial vaginosis in pregnant women does not reduce the occurrence of preterm delivery or other adverse perinatal outcomes. (C)2000, Massachusetts Medical Society.