Objective  To evaluate the efficacy of metronidazole to reduce post-abortion complications among women with bacterial vaginosis.

Design  A randomised, double-blind placebo-controlled trial.

Setting  An American outpatient abortion facility between April 1999 and June 2000.

Sample  Women presenting for surgical abortion were screened for bacterial vaginosis using a pH and amines card test.

Methods  Women positive for elevated pH and amines on a self-collected vaginal discharge sample were randomised to 1000 mg oral metronidazole before abortion followed by 500 mg twice daily or placebo. All randomised women were also dispensed 100 mg doxycycline to take twice daily for seven days.

Main outcome measures  Data were collected by phone, daily diary or visit and scored from 0 to 7 for post-abortion complications. Intention-to-treat analyses were completed prior to unblinding.

Results  Of 1764 women screened by card test, 638 (36%) were positive. Of these, 393 were randomised. Follow up data were available for 253 (64%) of these women. A complication score of 3 or more occurred in 21% of women assigned to metronidazole, compared with 19% in those assigned placebo (RR 1.1, 95% CI 0.7–1.9). Among 153 women with Gram stain confirmation for bacterial vaginosis, there was a similar lack of benefit with treatment (RR 1.6, 95% CI 0.9–3.0). No individual symptom or sign was statistically different between groups even for second trimester procedures.

Conclusion  Among women undergoing abortion and diagnosed with bacterial vaginosis, oral metronidazole, in conjunction with doxycycline, did not reduce post-abortion complications.