Objective: To assess the efficacy of treatment of bacterial vaginosis (BV) using metronidazole to reduce preterm labour in primigravidae and multigravidae with previous midtrimester abortion or preterm labour.

Design: Randomised controlled trial.

Setting: Tertiary academic hospital.

Method: Two different groups of patients were screened for BV at the first antenatal visit, namely primigravidae and high-risk multigravidae who had had a previous midtrimester abortion or preterm delivery. Patients where BV was diagnosed clinically or on Gram's stain of a smear taken from the posterior vaginal fornix, received either 400 mg metronidazole, or 100 mg vitamin C orally twice daily for 2 days. The Gram's stain was repeated after 4 weeks. If BV was found again, treatment with the same drug was repeated.

Outcome measures: Preterm delivery, birth weight and perinatal deaths.

Results: One thousand and five patients entered the study, but 40 were excluded for various reasons and 10 were lost to follow-up. There were 464 primigravidae, of whom 150 (32%) had BV. Except for the 5-minute Apgar score, no significant differences were found between primigravidae negative for BV and those who received either metronidazole or vitamin C. There were 491 high-risk multigravidae, of whom 127 (26%) had BV. The mean gestational age in the BV-negative group was 37 weeks, in contrast to 37.4 weeks in the vitamin C group and 35.6 weeks in the metronidazole group. Birth weights in these three groups were 2,752 g, 2,759 g and 2,475 g respectively, significantly less (P = 0.0109) in the metronidazole group in comparison with the BV-negative group. Delivery before 37 weeks occurred in 29% of high-risk multigravidae with no BV but in 24% of those who took vitamin C and in 43% who took metronidazole. Differences were significant between the BV-negative and metronidazole groups (P = 0.0231) and also between the metronidazole and vitamin C groups (P = 0.0274). Delivery before 28 weeks occurred in 4% of the high-risk multigravidae with no BV but in 10% of those with BV who took metronidazole. The difference was significant (P = 0.0430). Analysis for maximum likelihood estimates for preterm labour identified only previous preterm labour or midtrimester abortion as risk factors.

Conclusion: Metronidazole does not seem to reduce the prevalence of preterm labour when given for BV before 26 weeks' gestation.