Objectives: Assess the predictive values of bacterial vaginosis (BV) for preterm delivery (PD) and neonatal infection and compare them with standard markers of infection among women with preterm labour (PL).
Study design: Prospective blinded study in a tertiary referral centre in Paris. Women hospitalised for PL with intact membranes at a term between 24 and 34 weeks were included. Vaginal fluid, collected at inclusion was Gram-stained, scored, and interpreted according to Nugent's criteria.
Results: Out of 354 women tested, 254 had normal flora (72.3%), 76 intermediate (21.7%) and 24 BV (6.8%). A history of spontaneous miscarriage after 14 weeks was the only risk factor significantly associated with BV. BV was not significantly associated with PD<35 weeks or neonatal infection. Very preterm delivery (before 33 weeks) was significantly associated with the flora grade (P=0.02): women with normal, intermediate and abnormal flora, respectively had 27 (10.6%), 14 (18.4%) and 6 (25.0%) births before 33 weeks. Of the markers tested, the highest risk of very preterm delivery was associated with BV (odds ratio 2.95, 95% CI (1.1-0.8.1)) and CRP>20mg/dl (4.23 95% CI (1.8-9.7)). Predictive value of BV for preterm birth before 33 weeks were: sensitivity 12.8%, specificity 95.0%, positive predictive value 35.3%, and negative predictive value 84.3%.
Conclusions: The frequency of BV and its association with PD are probably very variable and must be interpreted differently from one population to another. While we found an association between BV results and delivery before 33 weeks, the predictive value of BV was disappointing. Although these findings reinforce the importance of a useful marker of subclinical infection, the usefulness of testing for BV in women with PL has not been demonstrated.