OBJECTIVES: We sought to determine the predictive accuracy for preterm delivery of transvaginal ultrasonography of the cervix between 14 and 24 weeks' gestation in high-risk patients and to determine whether cerclage prevents preterm delivery in patients with ultrasonographic cervical changes. STUDY DESIGN: Patients with asymptomatic singleton pregnancies at high risk for preterm delivery were followed prospectively from 14 weeks' to 23 weeks 6 days' gestation with transvaginal ultrasonography of the cervix. The subgroup of patients with either a cervical length of 25% or both was offered McDonald salvage cerclage, which was performed at the discretion of the patient and the obstetrician. The 2 groups (with and without cerclage) were compared for the primary outcome of preterm delivery at /=1 prior 14- to 34-week preterm deliveries. Of 63 (37. 5%) patients identified as having cervical changes, 23 (37%) had preterm delivery; of 105 patients with no cervical changes, 8 (8%) had preterm delivery (relative risk, 4.8; 95% confidence interval, 2. 3-10.1). The sensitivity, specificity, and positive and negative predictive values of either a short cervix of 25% or both were 74%, 70%, 37%, and 92%, respectively. Of 63 pregnancies in which there were cervical changes, 39 underwent cerclage and 24 did not. These 2 groups were similar for demographic characteristics, risk factors, and transvaginal ultrasonographic cervical length and funneling but dissimilar for gestational age at identification of cervical changes (18.3 vs 21.2 weeks' gestation in the groups with and without cerclage, respectively; P