A randomised controlled trial of metronidazole for the prevention of preterm birth in women positive for cervicovaginal fetal fibronectin: the PREMET study



Shennan et al. tested the hypothesis that oral metronidazole therapy at 24 weeks of gestation in women at high risk of preterm birth, who were also positive for fetal fibronectin, reduced their risk of recurrent preterm delivery.1 They concluded on the basis of their results that metronidazole did not reduce the risk of preterm delivery but actually increased the risk. These latter findings were unfortunately reported in the lay press as the latest medical evidence that metronidazole caused preterm births in pregnant women (http://www.telegraph.co.uk/news/main.jhtml?xml =/news/2006/01/13/nprem13.xml). We note that only 100 women (53 metronidazole, 47 placebo), approximately 24% of the actual number (420) required to achieve 80% statistical power, were recruited into the trial, therefore substantially weakening the strength of the trial to reach any conclusions. Their primary end point of preterm birth <30 weeks was not statistically significantly different, while the secondary outcome data on birth <37 weeks were only just statistically significant, with a very modest relative risk of 1.6.1 The trial was discontinued because of poor recruitment; yet, in the abstract section, this point was presented in such a way that readers may conclude that the trial was stopped because of the excess numbers of preterm delivery in women treated with metronidazole. Several larger trials that treated high-risk pregnant women with metronidazole at 24 weeks have not suggested that metronidazole at this gestation is harmful, neither did they consistently show a reduction in the risk of preterm delivery.2 On the other hand, metronidazole therapy for asymptomatic Trichomonas vaginalis in pregnancy (not tested for in this study) has been associated with an increased risk of preterm delivery.3,4

The aim of metronidazole therapy was presumably to eradicate genital tract ‘infection’, which is just one of a host of pathological processes that may disrupt the choriodecidual interface in pregnancy, leading to the release of fetal fibronectin into the cervicovaginal fluid. It is likely that in the case of infection, the inflammatory cascade may already be in progress by the time fetal fibronectin is detectable in the vagina and that antibiotic administration at this time is unlikely to be effective. We have reported that antibiotic therapy initiated early in the second trimester for bacterial vaginosis (BV) or intermediate was associated with a reduction in the risk of preterm delivery.5 However, detection of fibronectin is not a useful tool for the selection of women for antibiotic therapy early in pregnancy as this glycoprotein is present in the cervicovaginal fluid of almost all pregnant women before 20 weeks.

An unusual feature of this trial is the apparent lack of response of BV to metronidazole therapy, with BV resolving in only one of eight women in the metronidazole group. Furthermore, the usual stepwise spontaneous resolution of BV consistently observed as part of its natural history in pregnancy by other investigators was not replicated in this trial. The authors appear to offer no explanations for these unexpected results. Given the very small numbers of women with BV in the study (eight metronidazole, five placebo) we feel we must question whether the authors’ conclusions are justified.