OBJECTIVE:: To assess the predictive role of ultrasonographic cervical length for preterm delivery in women with threatened preterm labor. METHODS:: A retrospective cohort study of women with singleton pregnancies who presented with preterm labor before 34 weeks of gestation and underwent ultrasonographic measurement of cervical length in a tertiary medical center. Women with cervical cerclage, cervical dilatation greater than 3 cm at presentation, and pregnancies complicated by placental abruption, clinical chorioamnionitis, stillbirth, or major fetal anomalies were excluded. The predictive accuracy of cervical length for preterm delivery was analyzed using both fixed thresholds and outcome-specific thresholds, which are associated with a detection rate of 90%, false-positive rate of 10%, or the inflexion point of the receiver operator characteristic curve. RESULTS:: Between 2007 and 2012, 1,077 women presented with preterm labor and met the study criteria. The correlation between cervical length and the time interval to delivery was significant but weak (r=0.293, P<.001). Cervical length was independently associated with the risk of preterm delivery at less than 37, 35, and 32 weeks of gestation and within 14 and 7 days from presentation (a 4-7% decrease in the risk for each additional millimeter of cervical length) as well as with the time interval between presentation and delivery (each additional 2 mm was associated with an increase of 1 day). Overall, the accuracy of cervical length in predicting preterm delivery was relatively poor. CONCLUSION:: Although cervical length is an independent predictor of preterm delivery in women with preterm labor, its predictive accuracy as a single measure is relatively limited.