Accumulating evidence has associated bacterial vaginosis (BV) with serious medical complications such as premature delivery. The present article synthesizes available data on the treatment of nonpregnant and pregnant women with BV to prevent preterm delivery. A literature search identified articles published since 1976 that evaluated treatment of BV. An intention-to-treat sensitivity analysis was performed, to better compare studies. Studies that evaluated therapy efficacy varied with regard to diagnostic criteria, patient characteristics, clinicians, and laboratories. Therapies varied in efficacy for cure 4 or more weeks after therapy, from 48% to 85%. Intervention studies to reduce BV-related adverse outcomes of pregnancy differed in populations studied, medication used, type of therapy (oral or intravaginal), and timing of treatment. The benefit of treating women at high risk with oral metronidazole has been shown in several studies; however, the effect of treating women without a history of premature delivery is unclear. The use of intravaginal clindamycin therapy, especially during the latter half of the second trimester and thereafter, appears to increase infections during the neonatal period.