Objective To determine whether fetal fibronectin detected in cervicovaginal secretions of patients with symptoms suggestive of preterm labour is a predictor for preterm birth. Design A blind prospective observational study. Setting A teaching hospital. Participants One hundred and forty‐one women presenting to the delivery suite between 24 and 37 weeks of gestation (112 were between 24 and 34 weeks of gestation) with symptoms suggestive of preterm labour, with intact membranes and cervical dilatation of < 2 cm. Intervention Specimens of cervicovaginal secretions were collected from the ectocervix and posterior fornix. Fetal fibronectin levels were measured by an ELISA and samples were considered positive if the level of fetal fibronectin present was ≥ 0.05 μg/ml. Main outcome measures Gestation at birth and sampling‐birth interval. Results The preterm birth (before 3 7 weeks of gestation) rate in the population studied was 19±1%. Fetal fibronectin predicted preterm birth with sensitivity of 63%, specificity of 95±6%, positive predictive value of 77±3%, and negative predictive value of 91±6%. Analysis of the data from women at less than 34 weeks of gestation showed similar results. A negative test accurately excluded (97±9%) the chance of subsequent birth during the three weeks interval following sampling. Conclusions The presence of fetal fibronectin in the cervicovaginal secretions of women admitted with symptoms suggestive of preterm labour indicates a significant risk for subsequent preterm birth. The absence of fetal fibronectin in this group is a very strong indication that subsequent preterm birth is unlikely to occur. Fetal fibronectin test, if combined with clinical findings, has a potentially important role in clinical management of women with symptoms suggestive of preterm labour.