Oakeshott et al failed to document an association between bacterial vaginosis in early pregnancy and sub-sequent preterm birth in their 37 community centre-based study.1 They further suggest that the relative risk of preterm birth in women with vaginosis may have been overestimated in hospital-based studies due to patient selection. Since we lack population-based prevalence estimates of bacterial vaginosis,2 the distinction between community- versus hospital-based risks may indeed be valid as has been suggested before.3 We are, however, concerned about the take-home messages sent out to the general practitioner.1
The first reason is that this community -based sample seems to enjoy a preterm birth rate of 4.9%, and among black women an even more favourable preterm birth prevalence of 1.1%, considering the reference population (England) has a prematurity risk of at least one out of seven pregnancies.4
One can only speculate, though, why this sample selectively drawn from a London community was at apparently lower risk. Although the authors explain the differential risk by putting emphasis on studying a low-risk community-based cohort — as allegedly opposed to hospital-attending women — it must be acknowledged that their sample may not be quite representative of the community it was drawn from. Indeed, the authors actually recruited 1216 women from 37 centres over a 2-year period, suggesting that within each centre, less than two patients a month on average volunteered to enroll in the study.
Secondly, even if there was no genuine association between bacterial vaginosis at <10 weeks' gestation and preterm birth, the study may lack the power to substantiate this. Indeed, the 95% confidence interval on the relative risk of preterm birth stretches from 0.4 to 2.2 and contains the typical bacterial vaginosis risk estimate of 2, as recently shown in a systematic review on this subject.5 In fact, for typical relative risk estimates of 1.5 and 2.0, this study has a power of 22.8% and 55.1%, respectively, to detect a significant effect at the 5% level. To document an overall relative risk of 2.0 with a power of 80%, at least 224 women with and 1494 women without bacterial vaginosis should have been included. Similarly, this community-based sample comprised 88 black Caribbean and black African women, accounting for merely two preterm births, and therefore this study did not allow risk stratification for ethnicity.
We therefore believe that methodological concerns prevent any firm conclusions being drawn from the study by Oakeshott et al.
- © British Journal of General Practice, 2004.
